def diffImg(t0, t1, t2):
d1 = cv2.absdiff(t2, t1)
d2 = cv2.absdiff(t1, t0)
d_final = cv2.bitwise_and(d1, d2)
d_binary = cv2.threshold(d_final, 35, 255, cv2.THRESH_BINARY)[1]
d_blur = cv2.blur(d_binary, (15, 15))
return d_blur
def _colour_approach(person, background): # spliting images to red, green & blue components person_c = [person[:,:,0], person[:,:,1], person[:,:,2]] background_c = [background[:,:,0], background[:,:,1], background[:,:,2]] # subtracting images by component diff_c = [] diff_c.append(cv.absdiff(person_c[0], background_c[0])) diff_c.append(cv.absdiff(person_c[1], background_c[1])) diff_c.append(cv.absdiff(person_c[2], background_c[2])) # applying Gaussian blur to each component (reducing noise) diff_c[0] = cv.GaussianBlur(diff_c[0], (5, 5), 0) diff_c[1] = cv.GaussianBlur(diff_c[1], (5, 5), 0) diff_c[2] = cv.GaussianBlur(diff_c[2], (5, 5), 0) # merging components to a grey image # cv.add() is a saturated operation (250 + 10 = 260 => 255) diff = cv.add(cv.add(diff_c[0], diff_c[1]), diff_c[2]) # applying Gaussian blur again diff_b = cv.GaussianBlur(diff, (11, 11), 0) return diff_b
def diffImg(t0, t1, t2):
d1 = cv2.absdiff(t2, t1)
d2 = cv2.absdiff(t1, t0)
#cv2.imshow('d1',d1)
#cv2.imshow('d2',d2)
fixframe=cv2.bitwise_and(d1, d2)
return fixframe
def tantriggs(image):
# Convert to float
image = np.float32(image)
image = cv2.pow(image, GAMMA)
image = difference_of_gaussian(image)
# mean 1
tmp = cv2.pow(cv2.absdiff(image, 0), ALPHA)
mean = cv2.mean(tmp)[0]
image = cv2.divide(image, cv2.pow(mean, 1.0/ALPHA))
# mean 2
tmp = cv2.pow(cv2.min(cv2.absdiff(image, 0), TAU), ALPHA)
mean = cv2.mean(tmp)[0]
image = cv2.divide(image, cv2.pow(mean, 1.0/ALPHA))
# tanh
exp_x = cv2.exp(cv2.divide(image, TAU))
exp_negx = cv2.exp(cv2.divide(-image, TAU))
image = cv2.divide(cv2.subtract(exp_x, exp_negx), cv2.add(exp_x, exp_negx))
image = cv2.multiply(image, TAU)
image = cv2.normalize(image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8UC1)
return image
def diff(t1, t2, t3):
#compute diff images
d1 = cv2.absdiff(t1, t2)
d2 = cv2.absdiff(t2, t3)
diffImg = cv2.bitwise_and(d1, d2)
diffImg = cv2.cvtColor(diffImg, cv2.COLOR_RGB2GRAY)
return diffImg
def get_delta(self):
if len(self.frames) < 3:
return None
return cv2.bitwise_and(
cv2.absdiff(self.frames[2], self.frames[1]),
cv2.absdiff(self.frames[1], self.frames[0])
)
def processImage(t0, t1, t2): d1 = cv2.absdiff(t1, t0) d2 = cv2.absdiff(t2, t0) image = cv2.bitwise_and(d1, d2) image = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY) image = cv2.blur(image, (5,5)) value, image = cv2.threshold(image, 25, 255, cv2.THRESH_BINARY) #element = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (3,3)) #image = cv2.erode(image, element) image = cv2.blur(image, (5,5)) #image = cv2.normalize(image, 0., 1.) #mean, stddev = cv2.meanStdDev(image) #if stddev[0] > 40: # return image (minVal, maxVal, minLoc, maxLoc) = cv2.minMaxLoc(image) global cameraWidth, cameraHeight, x11 xDiv = (cameraWidth+1) / 7. yDiv = (cameraHeight+1) / 7. if maxVal > 100: if x11: image = cv2.cvtColor(image, cv2.COLOR_GRAY2RGB) cv2.rectangle(image, (maxLoc[0]-10,maxLoc[1]-10), (maxLoc[0]+20,maxLoc[1]+20), (0,0,255), 1) #print xLoc, yLoc global leftEye, rightEye leftEye.setPupilSmoothed(maxLoc[0] / xDiv, maxLoc[1] / yDiv) rightEye.setPupilSmoothed(maxLoc[0] / xDiv, maxLoc[1] / yDiv) return image
def diffImg(self, t0, t1, t2):
if not self.multiFrameDetection:
return cv.absdiff(t2, t1)
d1 = cv.absdiff(t2, t1)
d2 = cv.absdiff(t1, t2)
return cv.bitwise_and(d1, d2)
def frame_diff(f0, f1, f2):
d1 = cv2.absdiff(f0,f2)
d2 = cv2.absdiff(f1, f2)
result = cv2.bitwise_and(d1,d2)
ret, result = cv2.threshold(result, threshold, 255,cv2.THRESH_BINARY);
return result
def detect(capture, prev_images):
# Capture a new frame
new_frame = capture.grab_frame()
# Not enough frames: no detection, just store this one
if len(prev_images) < 2:
return None, None, new_frame, None
# Everything to grayscale
prev_images = [prev_images[1], prev_images[0], new_frame]
prev_images = [cv2.cvtColor(prev_frame, cv2.COLOR_BGR2GRAY)
for prev_frame in prev_images]
prev_frame, current_frame, next_frame = prev_images
# Diff
d1 = cv2.absdiff(prev_frame, next_frame)
d2 = cv2.absdiff(next_frame, current_frame)
motion = cv2.bitwise_and(d1, d2)
# Threshold & erode
cv2.threshold(motion, config.DIFF_THRESHOLD, 255, cv2.THRESH_BINARY,
dst=motion)
cv2.erode(motion, kernel_ero, dst=motion)
# Find and count changes
number_of_changes, location, std_dev = detect_motion(motion)
return number_of_changes, std_dev, new_frame, location
def flame_sub(im1,im2,im3,th,blur):
d1 = cv2.absdiff(im3, im2)
d2 = cv2.absdiff(im2, im1)
diff = cv2.bitwise_and(d1, d2)
# 差分が閾値より小さければFalse
return np.sum(th - diff) > diff.size
def scan(background_gray,cap):
for i in range(0,3) :
ret, frame2 = cap.read()
tracking = frame2
tracking = cv2.cvtColor(frame2, cv2.COLOR_BGR2GRAY)
cv2.absdiff(background_gray, tracking, tracking)
ret,tracking = cv2.threshold(tracking,35,255,cv2.THRESH_BINARY)
#post treatment
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(5,5))
erosion = cv2.erode(tracking,kernel, iterations=3)
dilatation = cv2.dilate(erosion,kernel, iterations=10)
tracking = dilatation
#blob detection
tracking_c =tracking
tracking_c, contours, hierarchy = cv2.findContours(tracking_c,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
if contours:
for cnt in contours:
x,y,w,h = cv2.boundingRect(cnt)
cv2.rectangle(frame2,(x,y),(x+w,y+h),(255,255,0),5)
frameBlobs = frame2
return tracking, contours,frameBlobs
def get_diffs_bw(detect_buffer):
temp_buff = detect_buffer
temp_buff_size = temp_buff.shape[0]
diff_buff = None
count = 0
#Store contents of detect_buffer into a temp array
#absdiff every third capture and throw it into a new temp array
#Repeat until temp array has a size of 2, then return the bitwise and.
while True:
if temp_buff_size == 2:
cummulativeFrames = cv2.bitwise_and(temp_buff[0], temp_buff[1])
break
else:
count = 0
while count < (temp_buff_size / 3):
diff1 = np.array([cv2.absdiff(temp_buff[(count*3)+2], temp_buff[(count*3)+1])])
diff2 = np.array([cv2.absdiff(temp_buff[(count*3)+1], temp_buff[(count*3)+0])])
if diff_buff == None:
diff_buff = np.concatenate((diff1,diff2),axis=0)
else:
diff_buff = np.concatenate((diff_buff,diff1,diff2),axis=0)
count+=1
temp_buff = diff_buff
diff_buff = None
temp_buff_size = temp_buff.shape[0]
return cummulativeFrames
def createOverlay(description):
if description == "before":
image = first_before
projection = cv2.cvtColor(cv2.absdiff(before_projection, first_before), cv2.COLOR_BGR2GRAY)
#invert
projection = cv2.bitwise_not(before_projection)
path = vidPath + "_2fps.AVI_"+str(before_start)+"_"+str(before_end) + "_before.jpg"
else if description == "after":
image = first_after
projection = cv2.cvtColor(cv2.absdiff(after_projection, first_after), cv2.COLOR_BGR2GRAY)
#invert
projection = cv2.bitwise_not(after_projection)
path = vidPath + "_2fps.AVI_"+str(after_start)+"_"+str(after_end) + "_after.jpg"
colorTrackImg = np.zeros(image.shape, np.uint8)
colorTrackImg.fill(255)
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
weight=0.5
colorTrackImg[:,:,1] = projection[:]
combinedTrackAndPhoto = cv2.addWeighted( colorTrackImg, weight, image, 1-weight, 0 )
cv2.imwrite(path, projection)
cv2.imwrite(vidPath + "_2fps.AVI_"+description+"_overlay.jpg", combinedTrackAndPhoto)
def diffImg(t0, t1, t2):
d1 = cv2.absdiff(t2, t1)
d2 = cv2.absdiff(t1, t0)
result = cv2.bitwise_and(d1, d2)
(value, result) = cv2.threshold(result, threshold, 255, cv2.THRESH_BINARY)
scalar = cv2.sumElems(result)
return scalar
def edge_detect(self, im, shift=1):
"""Return image filtered for hard color transitions.
Args:
im - Image to detect edges.
Keyword Args:
shift - Distance to shift image copies before calculating absolute
difference with original. (default: 1)
Note that this the (width, height) of returned image will be reduced by
(shift, shift).
"""
x_dim, y_dim = im.shape[:2]
# Create three cropped images to simulate a pixel shift in two
# different directions. The direction of the shifts is orthagonal so
# that any direction of edges is detected.
im_crop = im[0 : x_dim - shift, 0 : y_dim - shift]
im_shift_up = im[0 : x_dim - shift, shift:y_dim]
im_shift_left = im[shift:x_dim, 0 : y_dim - shift]
vertical_detect = cv2.absdiff(im_crop, im_shift_up)
horizontal_detect = cv2.absdiff(im_crop, im_shift_left)
return cv2.add(vertical_detect, horizontal_detect)
def foreground(bg_img, raw_img):
#take the background and subtract somehow from the foreground
img = raw_img*1
raw_hsv = cv2.cvtColor(img,cv2.COLOR_BGR2HSV)
bg_hsv = cv2.cvtColor(bg_img,cv2.COLOR_BGR2HSV)
hmask = cv2.absdiff(raw_hsv[:,:,0], bg_hsv[:,:,0])
smask = cv2.absdiff(raw_hsv[:,:,1], bg_hsv[:,:,1])
vmask = cv2.absdiff(raw_hsv[:,:,2], bg_hsv[:,:,2])
ret,hmask_thresh = cv2.threshold(hmask,1.,1.,cv2.THRESH_BINARY)
ret,smask_thresh = cv2.threshold(smask,1.,1.,cv2.THRESH_BINARY)
ret, vmask_thresh = cv2.threshold(vmask,1.,1.,cv2.THRESH_BINARY)
hsv_mask = np.multiply(hmask_thresh, smask_thresh)
hsv_mask = np.multiply(hsv_mask, vmask_thresh)
hsv_mask = cv2.dilate(hsv_mask,(100,100) )
###Filter out colors with extreme values and no red for skin###
# ret, rmask = cv2.threshold(img[:,:,2],40,1., cv2.THRESH_BINARY)
# ret, r2mask = cv2.threshold(img[:,:,2],235.,1., cv2.THRESH_BINARY_INV)
# rb_mask = np.multiply(rmask, r2mask)
# img[:,:,0 ]= np.multiply(img[:,:,0], rb_mask)
# img[:,:,1 ]= np.multiply(img[:,:,1], rb_mask)
# img[:,:,2 ]= np.multiply(img[:,:,2], rb_mask)
# bmask = cv2.absdiff(img[:,:,0], bg_img[:,:,0])
# gmask = cv2.absdiff(img[:,:,1], bg_img[:,:,1])
rmask = cv2.absdiff(img[:,:,2], bg_img[:,:,2])
ret,rmask_thresh = cv2.threshold(rmask,20.,1.,cv2.THRESH_BINARY)
##Greyscale mask that kinda worked except for bright lighting
raw_gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
#raw_gray = cv2.GaussianBlur(raw_gray, (5,5), 2)
bg_gray = cv2.cvtColor(bg_img,cv2.COLOR_BGR2GRAY)
#bg_gray = cv2.GaussianBlur(bg_gray, (5,5), 5)
mask =cv2.absdiff(raw_gray, bg_gray)
ret,mask = cv2.threshold(mask,15.,1.,cv2.THRESH_BINARY)
###
###make changes here
mask = mask*1.0
mask = np.multiply(rmask_thresh, mask)
mask = np.multiply(hsv_mask, mask)
mask = cv2.dilate(mask,(100,100))
for i in range(4):
mask = cv2.erode(mask*255, (50,50))/255.
for i in range(5):
mask = cv2.dilate(mask*255., (50,50))/255.
fg_img = img*1.0
fg_img[:,:,0 ]= np.multiply(img[:,:,0], mask)
fg_img[:,:,1 ]= np.multiply(img[:,:,1], mask)
fg_img[:,:,2 ]= np.multiply(img[:,:,2], mask)
cv2.imshow("fg_img", np.array(fg_img, dtype= "uint8"))
return np.array(mask, dtype = "uint8")
def diffImg(t0, t1, t2): previous = 0 d1 = cv2.absdiff(t2, t1) d2 = cv2.absdiff(t1, t0) result = cv2.bitwise_and(d1, d2) return result
def differential_Image(d0, d1, d2): diff1 = cv2.absdiff(d2, d1) diff2 = cv2.absdiff(d1, d0) return cv2.bitwise_and(diff1, diff2)
def diffImg(t0, t1, t2): d1 = cv2.absdiff(t2, t1) d2 = cv2.absdiff(t1, t0) return cv2.bitwise_and(d1, d2)
def doConv(images):
total = []
previous = 0
binaries = []
#print "Size: ", len(images)
for i in range(len(images)):
img4 = cv2.cvtColor(images[i], cv2.COLOR_BGR2GRAY)
img4 = numpy.asarray(img4)
if i == 0:
previous = img4
total = img4
else:
newImage = cv2.absdiff(img4, previous)
binaries.append(newImage)
previous = img4
for i in range(len(binaries)):
#img4 = sg.convolve(numpy.asarray(images[i]), Ww[0], "valid")
img4 = numpy.asarray(binaries[i])
if(i==0):
total = img4
#total = img4
else:
#total = total + (img4*weights[i])
total = cv2.absdiff(img4 , total)
total = total*10
# total = cv2.cvtColor(total, cv2.COLOR_BGR2GRAY)
ret2,total = cv2.threshold(total,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
total = cv2.medianBlur(total,15)
return total
def binary_generator(training_element):
create_folder("binaryfeatures")
if (training_element == -1):
for j in xrange (0,folders):
create_folder("binaryfeatures/"+list_folder[j])
list_imgs=os.listdir("rgbfeatures/"+list_folder[j])
for i in xrange(0,(len(list_imgs)-1)):
img1 = cv2.imread("rgbfeatures/"+list_folder[j]+"/001.jpeg",0)
bgsub = img1
img2 = cv2.imread("rgbfeatures/"+list_folder[j]+"/"+list_imgs[i+1],0)
cv2.absdiff(img2, img1, bgsub)
bgsub_blur2 = cv2.GaussianBlur(bgsub,(5,5),0)
(thresh, im_bw) = cv2.threshold(bgsub_blur2, 55, 255, cv2.THRESH_BINARY)
a='{0:03}'.format(i+1)
cv2.imwrite("binaryfeatures/"+list_folder[j]+"/bw_"+a+".png",im_bw)
else:
list_imgs=os.listdir("rgbfeatures/"+list_folder[training_element])
create_folder("binaryfeatures/"+list_folder[training_element])
for i in xrange(0,(len(list_imgs)-1)):
img1 = cv2.imread("rgbfeatures/"+list_folder[training_element]+"/001.jpeg",0)
bgsub = img1
img2 = cv2.imread("rgbfeatures/"+list_folder[training_element]+"/"+list_imgs[i+1],0)
cv2.absdiff(img2, img1, bgsub)
bgsub_blur2 = cv2.GaussianBlur(bgsub,(5,5),0)
(thresh, im_bw) = cv2.threshold(bgsub_blur2, 55, 255, cv2.THRESH_BINARY)
a='{0:03}'.format(i+1)
cv2.imwrite("binaryfeatures/"+list_folder[training_element]+"/bw_"+a+".png",im_bw)
def update_roi(self, image, bInvertColor):
self.image = image
self.shape = image.shape
# Extract the ROI images.
if (self.roi is not None):
self.imgRoi = copy.deepcopy(image[self.roi[1]:self.roi[3], self.roi[0]:self.roi[2]])
# Color inversion.
if (bInvertColor):
self.imgRoiFg = 255-self.imgRoi
else:
self.imgRoiFg = self.imgRoi
# Background Subtraction.
if (self.params['gui'][self.name]['subtract_bg']):
with self.lockBackground:
if (self.imgRoiBackground is not None):
if (self.imgRoiBackground.shape==self.imgRoiFg.shape):
if (bInvertColor):
self.imgRoiFg = cv2.absdiff(self.imgRoiFg, 255-self.imgRoiBackground.astype(np.uint8))
else:
self.imgRoiFg = cv2.absdiff(self.imgRoiFg, self.imgRoiBackground.astype(np.uint8))
# Equalize the brightness/contrast.
if (self.bEqualizeHist):
if (self.imgRoiFg is not None):
self.imgRoiFg -= np.min(self.imgRoiFg)
max2 = np.max([1.0, np.max(self.imgRoiFg)])
self.imgRoiFg *= (255.0/float(max2))
def getCandidatesFromPPFrame(self,ppframe):
hsvSplit = cv2.split(ppframe);
#gray = cv2.bitwise_and(gray,mask)
# compute the absolute difference between the current frame and
frameDelta = cv2.absdiff(hsvSplit[0],self.background[0]) + cv2.absdiff(hsvSplit[1],self.background[1]) + cv2.absdiff(hsvSplit[2],self.background[2])
#thresh = cv2.threshold(frameDelta, 160, 0, cv2.THRESH_TOZERO_INV)[1]
thresh = cv2.threshold(frameDelta, self.threshold, 255, cv2.THRESH_BINARY)[1]
# dilate the thresholded image to fill in holes, then find contours
# on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
cv2.imshow("Thresh",thresh)
(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = [c for c in cnts if cv2.contourArea(c) > args["min_area"]]
cnts = sorted(cnts,key=lambda c: -cv2.contourArea(c))
def getCenter(c):
(x, y, w, h) = cv2.boundingRect(c)
return((x+w/2,y+h/2))
coordinates = [ getCenter(c) for c in cnts ]
#cv2.imshow("Thresh", thresh)
#cv2.imshow("Frame Delta", frameDelta)
return(cnts,coordinates)
def diffImg(t0, t1, t2):
'''Function to Obtain the difference between
Images
'''
d1 = cv2.absdiff(t2, t1)
d2 = cv2.absdiff(t1, t0)
return cv2.bitwise_and(d1, d2)
def _frame_diff(self, old_frame, new_frame):
if len(self.humans) == 0:
self.humans.append(self._find_human(old_frame))
self.humans.append(self._find_human(new_frame))
human_old_frame = self.humans[-2]
human_new_frame = self.humans[-1]
old_frame_copy = old_frame.copy()
new_frame_copy = new_frame.copy()
human = self._Human.union(human_old_frame, human_new_frame)
x_min, x_max, w = human.x_min, human.x_max, human.w
num_of_pixels = self.shape.height * self.shape.width
if w != 0:
x_l = int(x_min - w * 0.15 if x_min - w * 0.15 >= 0 else 0)
x_r = int(x_max + w * 0.15 if x_max + w * 0.15 < self.shape.width else self.shape.width)
for frame in [old_frame_copy, new_frame_copy]:
cv2.rectangle(frame, (x_l, 0), (x_r, self.shape.height), (0, 0, 0), cv2.FILLED)
abs_diff = cv2.absdiff(old_frame_copy, new_frame_copy)
num_of_pixels -= self.shape.height * (x_r - x_l)
else:
abs_diff = cv2.absdiff(old_frame, new_frame)
return float(abs_diff.sum() / (num_of_pixels or 1))
def diffImg(self, t0, t1, t2): """Uses absdiff to determine the difference between 2 images""" # Gets the difference between 2 images d1 = cv2.absdiff(t2, t1) d2 = cv2.absdiff(t1, t0) # Returns the bitwise_and between d1 and d2 return cv2.bitwise_and(d1, d2)
def diffImg3(self, img1, img2, img3):
gray1 = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
gray2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
gray3 = cv2.cvtColor(img3, cv2.COLOR_BGR2GRAY)
diff1 = cv2.absdiff(gray1, gray2)
diff2 = cv2.absdiff(gray2, gray3 )
return cv2.bitwise_and(diff1, diff2)
def detect_motion(t0, t1, t2):
"""
Calcula la diferencia absoluta entre dos pares de imagenes y
retorna la matriz resultante
"""
d1 = cv2.absdiff(t2, t1)
d2 = cv2.absdiff(t1, t0)
return cv2.bitwise_and(d1, d2)
def filter(self, cv_image):
b, g, r = cv2.split(cv_image.image)
is_grey = (cv2.absdiff(b, g).sum() == 0
and cv2.absdiff(b, r).sum() == 0)
return is_grey
def main():
# Load the parameters for the motion detection
try:
with open(PATH_MD_PARAMS) as f:
data = csv.DictReader(f)
for line in data:
params = line # Header is read too
except FileNotFoundError:
print("[ERROR] Motion detection parameters file not found.")
raise
params['iou'] = float(params['iou']) / 2
params['gaussWindow'] = int(params['gaussWindow'])
params['residualConnections'] = int(params['residualConnections'])
params['dilationIterations'] = int(params['dilationIterations'])
params['sigma'] = float(params['sigma'])
print("[INFO] Motion detection params:\n", params)
# Load the CNN
try:
loaded_model = transfer_learning.load_model(PATH_ARCHITECTURE,
PATH_WEIGHTS)
except FileNotFoundError:
print("[ERROR] CNN model|weights file not found.")
raise
# Run through the video
first_bbox = min(bbox_heli_ground_truth.keys())
last_bbox = max(bbox_heli_ground_truth.keys())
print("[INFO] Using bbox frames {} to {}".format(first_bbox, last_bbox))
# Single core example
timing = {
'Read frame': 0,
'Convert to grayscale': 0,
'Stabilize': 0,
'Double Gauss': 0,
'Abs diff': [],
'Thresholding': 0,
'Dilation': 0,
'Count boxes': 0,
'Finalize': 0
}
nb_bbox = [] # Stores bbox data for a sim
# Get ready to store residualConnections frames over and over
# Create Tracker
Tracker = OPENCV_OBJECT_TRACKERS["csrt"]()
flag_tracker_active = False
#fps = FPS().start()
# Create an extractor to get the contours later on
extractor = extract.extractor()
# Create a deque for frame accumulation
previous_gray_frame = collections.deque(
maxlen=params['residualConnections'])
# Track the results
Y_prediction = []
Y_test = []
# Count global number of bboxes
counter_bbox = 0
# Start re-encoding
if REENCODE:
reencoded_video = cv2.VideoWriter(PATH_REENCODED_VIDEO, FOURCC, 25,
(1920, 1080))
t0 = time.perf_counter() # Mostly used for first frame display
for frame_number in range(NB_FRAMES):
current_frame = VIDEO_STREAM.read()[1]
t1 = time.perf_counter()
# Check the status of the tracker
#print("[INFO] Frame {}\tTracker status: {}".format(frame_number, flag_tracker_active))
flag_tracker_active = False
if flag_tracker_active: # The tracker is ON
flag_success, bbox_roi = Tracker.update(current_frame)
bbox_roi = [int(value) for value in bbox_roi
] # The update returns floating point values...
if frame_number % CNN_CHECK_STRIDE == 0:
print("[INFO] Verifying tracker at frame", frame_number)
# Time to verify that the Tracker is still locked on the helico
prediction, crop = infer_bbox(loaded_model, current_frame,
bbox_roi, METHOD)
if prediction == 1: # All good, keep going
flag_quit_program = display_frame(current_frame, bbox_roi,
frame_number,
flag_tracker_active)
#fps.update()
if flag_quit_program:
return
continue
elif prediction == 0: # There is actually no helico in the tracked frame!
Tracker = OPENCV_OBJECT_TRACKERS["csrt"]() # Reset tracker
flag_tracker_active = False
previous_gray_frame = collections.deque(
maxlen=params['residualConnections'])
pass # Engage the motion detection algo below
else:
print(
"[ERROR] The model is supposed to be a binary classifier"
)
raise
else: # Between checks from the CNN, go to the next frame
flag_quit_program = display_frame(current_frame, bbox_roi,
frame_number,
flag_tracker_active)
#fps.update()
if flag_quit_program:
return
continue
else: # The tracker is OFF
#if frame_number%CNN_CHECK_STRIDE:
#continue
# Engage the motion detection algo below
pass
# Switch the gray space
current_gray_frame = cv2.cvtColor(current_frame.copy(),
cv2.COLOR_RGB2GRAY)
#print(len(previous_gray_frame))
# Populate the deque with the params['residualConnections'] next gray frames
if len(previous_gray_frame) < params['residualConnections']:
previous_gray_frame.append(current_gray_frame)
continue
t4 = time.perf_counter()
# Gaussian blur
current_gauss_frame, previous_gauss_frame = gaussian_blur(
[current_gray_frame, previous_gray_frame[0]],
params['gaussWindow'])
t5 = time.perf_counter()
# Differentation
diff_frame = cv2.absdiff(current_gauss_frame, previous_gauss_frame)
t7 = time.perf_counter()
# Canny
canny_frame = canny_contours(diff_frame, params['sigma'])
t7 = time.perf_counter()
# Morphological transformations
morph_frame = cv2.dilate(canny_frame,
None,
iterations=params['dilationIterations'])
#diff_frame = cv2.morphologyEx(diff_frame, cv2.MORPH_CLOSE, None, iterations=params['dilationIterations'])
# Get the contours sorted by area (largest first)
contours = extractor.image_contour(morph_frame,
sorting='area',
min_area=MIN_AREA)
t8 = time.perf_counter()
# Process the bbox that came out of the contours
large_box = 0
counter_bbox_heli = 0
if first_bbox <= frame_number <= last_bbox:
(x_gt, y_gt, w_gt,
h_gt) = bbox_heli_ground_truth[frame_number] # Ground Truth data
else:
(x_gt, y_gt, w_gt, h_gt) = (1919, 1079, 1, 1)
counter_failed_detection = 0
(x, y, w, h) = (0, 0, 0, 0)
success = []
for contour in contours:
c = contour[0]
# A. Filter out useless BBs
# 1. if the contour is too small or too large, ignore it
if cv2.contourArea(c) < MIN_AREA:
continue
if cv2.contourArea(c) > MAX_AREA:
continue
# compute the bounding box for the contour, draw it on the current_frame,
# and update the text
(x, y, w, h) = cv2.boundingRect(c)
# 2. Box partially out of the frame
if x < 0 or x + w > FRAME_WIDTH or y < 0 or y + h > FRAME_HEIGHT:
continue
# 3. Box center is not in the PADDING area next to the edges of the frame
if not (PADDING < x + w // 2 < FRAME_WIDTH - PADDING
and PADDING < y + h // 2 < FRAME_HEIGHT - PADDING):
continue
# B. Classify BBs - a large_box is a potential bbox_heli_ground_truth
large_box += 1
counter_bbox += 1 # global counter
# Infer bbox
prediction, crop = infer_bbox(loaded_model, current_frame,
(x, y, w, h), METHOD)
# Determine the label for this box based on the IoU with the ground truth one.
converted_bbox = bbox.xywh_to_x1y1x2y2((x, y, w, h))
converted_gt_bbox = bbox.xywh_to_x1y1x2y2((x_gt, y_gt, w_gt, h_gt))
label = 1 if bbox.intersection_over_union(
converted_bbox, converted_gt_bbox) >= params['iou'] else 0
# Append both results to their respective lists
Y_prediction.append(prediction)
Y_test.append(label)
#prediction = 0
name = 'Helico' if prediction else 'Motion'
color = COLOR['RED'] if prediction else COLOR['BLUE']
cv2.putText(current_frame, name, (x, y - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
cv2.rectangle(current_frame, (x, y), (x + w, y + h), color, 2)
# Update the deque
previous_gray_frame.append(current_gray_frame)
# Add the total number of bboxes detected so far
text_bboxes = 'Number of detected bboxes: {}'.format(counter_bbox)
cv2.putText(current_frame, text_bboxes, (10, 80),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLOR['WHITE'], 2)
# Add the current accuracy for the CNN
#accuracy = 1-np.sum(np.abs(np.array(Y_prediction)-np.array(Y_test)))/len(Y_test)
if len(Y_test):
conf_mx = confusion_matrix(Y_test, Y_prediction)
if conf_mx.shape == (2, 2):
accuracy = (conf_mx[0, 0] + conf_mx[1, 1]) / np.sum(conf_mx)
else:
conf_mx = np.zeros((2, 2))
accuracy = 1
else:
conf_mx = np.zeros((2, 2))
accuracy = 1
text_accuracy = 'Cumulative accuracy: {:.1f} % TN: {} TP: {} FN: {} FP: {}'.format(
100 * accuracy, conf_mx[0, 0], conf_mx[1, 1], conf_mx[1, 0],
conf_mx[0, 1])
cv2.putText(current_frame, text_accuracy, (10, 60),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLOR['WHITE'], 2)
# Update tracker state
tracker_state = 'ON' if flag_tracker_active else 'OFF'
text_tracker = 'Tracker: {}'.format(tracker_state)
cv2.putText(current_frame, text_tracker, (10, 40),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLOR['WHITE'], 2)
# Update FPS and frame count
fps = 1 / (time.perf_counter() - t0)
text_count = 'FPS: {:.1f} Frame number: {}'.format(fps, frame_number)
cv2.putText(current_frame, text_count, (10, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLOR['WHITE'], 2)
t0 = time.perf_counter()
# Re-encode the frame
if REENCODE:
reencoded_video.write(current_frame)
# Display frame
cv2.imshow(VIDEO_STREAM_PATH, imutils.resize(current_frame,
width=1000))
#cv2.imshow("Diff_frame", diff_frame)
#cv2.imshow("Canny frame", canny_frame)
key = cv2.waitKey(1) & 0xFF
if key == ord('q'):
#print(Y_prediction)
#print(Y_test)
if REENCODE:
reencoded_video.release()
return
"""
print("[INFO] Inferred bbox: ", prediction)
if prediction == 1: # First helico detected!
#Tracker.init(current_frame, (x, y, w, h))
#flag_tracker_active = True
#success.append([prediction, crop])
#break
#continue
elif prediction == 0:
counter_failed_detection += 1
if counter_failed_detection >= MAX_FAILED_INFERENCES:
#break # Not time for another attempt. Improve the motion detection.
#continue
else:
print("[ERROR] The model is supposed to be a binary classifier")
print("[INFO] Length success:", len(success))
if len(success):
fig, ax = plt.subplots(1, max(2, len(success)))
print(len(success))
for index, res in enumerate(success):
print(index)
ax[index].imshow(cv2.cvtColor(res[1], cv2.COLOR_BGR2RGB))
ax[index].set_title("Heli")
ax[index].axis('off')
plt.show()
"""
#fps.update()
"""[Is display_frame still useful?]
# Display the result from motion detection
#print("[INFO] Display frame")
#flag_quit_program = display_frame(current_frame, (x, y, w, h), frame_number, flag_tracker_active)
print()
if flag_quit_program:
return
"""
"""[For later]
# Classify bboxes based on their IOU with ground truth
converted_current_bbox = bbox.xywh_to_x1y1x2y2(bbox_crop)
converted_ground_truth_bbox = bbox.xywh_to_x1y1x2y2((x_gt, y_gt, w_gt, h_gt))
if bbox.intersection_over_union(converted_current_bbox, converted_ground_truth_bbox) >= IOU:
counter_bbox_heli += 1
"""
#fps.stop()
conf_mx = confusion_matrix(Y_test, Y_prediction)
print("[INFO] Confusion Matrix:\n", conf_mx)
#print("[INFO] Final accuracy: {:.1f}".format(100*accuracy))
#plt.figure()
plt.matshow(conf_mx, cmap=plt.cm.gray)
plt.title("Confusion matrix on {} | Accuracy: {:.1f}%".format(
FOLDER_NAME, 100 * accuracy))
plt.savefig("Confusion_matrix_" + FOLDER_NAME, tight_layout=False)
plt.show()
if REENCODE:
reencoded_video.release()
def diffImg(t0, t1, t2): # Function to calculate difference between images.
d1 = cv2.absdiff(t2, t1)
d2 = cv2.absdiff(t1, t0)
return cv2.bitwise_and(d1, d2)
def Begin(Type):
avg = None
count = 0
flag = 0
cnt = 0
first = None
detectflag = None
while True:
if Type == 0:
frame = vs.read()
elif Type == 1:
if vs.more() == False:
break
frame = vs.read()
if first is None:
frame = imutils.resize(frame, width=500)
r = selectRoi(frame)
frameroi = frame[int(r[1]):int(r[1] + r[3]),
int(r[0]):int(r[0] + r[2])]
grayf = cv2.cvtColor(frameroi, cv2.COLOR_BGR2GRAY)
grayfb = cv2.GaussianBlur(grayf, (21, 21), 0)
first = 1
text = "Unoccupied"
frame = imutils.resize(frame, width=500)
frameroi = frame[int(r[1]):int(r[1] + r[3]),
int(r[0]):int(r[0] + r[2])]
gray = cv2.cvtColor(frameroi, cv2.COLOR_BGR2GRAY)
grayb = cv2.GaussianBlur(gray, (5, 5), 0)
if avg is None:
print("[INFO] starting background model...")
avg = grayb.copy().astype("float")
continue
cv2.accumulateWeighted(grayb, avg, 0.6)
frameDelta = cv2.absdiff(grayb, cv2.convertScaleAbs(avg))
thresh = cv2.threshold(frameDelta, 10, 255, cv2.THRESH_BINARY)[1]
thresh = cv2.dilate(thresh, None, iterations=2)
cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if imutils.is_cv2() else cnts[1]
for c in cnts:
(x, y, w, h) = cv2.boundingRect(c)
if w < 20 or h < 20:
continue
cv2.rectangle(frameroi, (x, y), (x + w, y + h), (0, 255, 0), 2)
text = "Occupied"
if len(cnts) != 0:
cnt += 1
flag = 1
count = 0
if len(cnts) == 0:
count += 1
if count >= 100:
if flag == 0:
grayf = gray
cnt = 0
else:
flag = 0
if cnt >= 50:
grayfb = cv2.GaussianBlur(grayf, (21, 21), 0)
grayb = cv2.GaussianBlur(gray, (21, 21), 0)
frameD = cv2.absdiff(grayb, grayfb)
threshd = cv2.threshold(frameD, 35, 255,
cv2.THRESH_BINARY)[1]
threshd = cv2.dilate(threshd, None, iterations=2)
#cv2.imshow("thred",threshd)
cntsd = cv2.findContours(threshd.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cntsd = cntsd[0] if imutils.is_cv2() else cntsd[1]
fram = frameroi.copy()
#cv2.drawContours(fram, cntsd, -1, (0,255,0), 3)
for c in cntsd:
(x, y, w, h) = cv2.boundingRect(c)
area = w * h
if area < 81:
continue
cv2.rectangle(fram, (x, y), (x + w, y + h),
(0, 255, 0), 2)
image1 = grayf[int(y):int(y + h), int(x):int(x + w)]
image2 = gray[int(y):int(y + h), int(x):int(x + w)]
s = ssim(image1, image2)
if s < 0.4:
if detectflag == None:
print("object is detected")
detectflag = 1
cv2.rectangle(frameroi, (x, y), (x + w, y + h),
(0, 255, 0), 2)
continue
if area > 2500:
c = compare(image1, image2)
if c >= 3:
if detectflag == None:
print("object is detected")
detectflag = 1
cv2.rectangle(frameroi, (x, y), (x + w, y + h),
(0, 255, 0), 2)
#cv2.imshow("Security", frame
detectflag = None
cv2.imshow("Security", frameroi)
#cv2.imshow("Sec", fram)
cv2.waitKey(1)
cv2.imshow("Security Feed", frameroi)
cv2.imshow("thresh", thresh)
key = cv2.waitKey(1) & 0xFF
if key == ord("q"):
break
while True:
# a += 1
check, frame = video.read()
status = 0
# print(frame)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
if first_name is None:
first_frame = gray
continue
status = 1
delta_frame = cv2.absdiff(first_frame, gray)
thresh_delta = cv2.threshold(delta_frame, 30, 255, cv2.THRESH_BINARY)[1]
thresh_delta = cv2.dilate(thresh_delta, None, iterations=0)
(_, cnts, _) = cv2.findContours(thresh_delta.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
for contour in cnts:
if cv2.contourArea(contour) < 1000:
continue
(x, y, w, h) = cv2.boundingRect(contour)
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 3)
status_list.append(status)
status_list = status_list[-2:]
def process_frame(self, frame_num, frame_img):
# type: (int, numpy.ndarray) -> bool, Optional[int]
"""
Returns:
"""
# Similar to ThresholdDetector, but using the HSV colour space DIFFERENCE instead
# of single-frame RGB/grayscale intensity (thus cannot detect slow fades with this method).
cut_list = []
metric_keys = self._metric_keys
_unused = ''
if self.last_frame is not None:
# Change in average of HSV (hsv), (h)ue only, (s)aturation only, (l)uminance only.
delta_hsv_avg, delta_h, delta_s, delta_v = 0.0, 0.0, 0.0, 0.0
if (self.stats_manager is not None
and self.stats_manager.metrics_exist(
frame_num, metric_keys)):
delta_hsv_avg, delta_h, delta_s, delta_v = self.stats_manager.get_metrics(
frame_num, metric_keys)
else:
curr_hsv = cv2.cvtColor(frame_img, cv2.COLOR_BGR2HSV)
curr_hsv = curr_hsv.astype(numpy.int16)
if self.last_hsv is None:
last_hsv = cv2.cvtColor(self.last_frame, cv2.COLOR_BGR2HSV)
last_hsv = last_hsv.astype(numpy.int16)
else:
last_hsv = self.last_hsv
# Image math is faster with cv2
absdiff = cv2.absdiff(curr_hsv, last_hsv)[:3]
delta_h, delta_s, delta_v = cv2.mean(absdiff)[:3]
delta_hsv_avg = cv2.mean(
numpy.array([delta_h, delta_s, delta_v]))[0]
if self.stats_manager is not None:
self.stats_manager.set_metrics(
frame_num, {
metric_keys[0]: delta_hsv_avg,
metric_keys[1]: delta_h,
metric_keys[2]: delta_s,
metric_keys[3]: delta_v
})
self.last_hsv = curr_hsv
if delta_hsv_avg >= self.threshold:
if self.last_scene_cut is None or (
(frame_num - self.last_scene_cut) >= self.min_scene_len):
cut_list.append(frame_num)
self.last_scene_cut = frame_num
if self.last_frame is not None and self.last_frame is not _unused:
del self.last_frame
# If we have the next frame computed, don't copy the current frame
# into last_frame since we won't use it on the next call anyways.
if (self.stats_manager is not None and
self.stats_manager.metrics_exist(frame_num + 1, metric_keys)):
self.last_frame = _unused
else:
self.last_frame = frame_img.copy()
return cut_list
frame = cv2.resize(frame, dim)
gray_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# gray_frame = np.dstack([gray_frame, gray_frame, gray_frame])
gray_frame = cv2.GaussianBlur(gray_frame, (21, 21), 0)
# if first_frame is None:
# first_frame = gray_frame
# continue
if avg is None:
avg = gray_frame.copy().astype("float")
continue
cv2.accumulateWeighted(gray_frame, avg, 0.5)
frame_delta = cv2.absdiff(gray_frame, cv2.convertScaleAbs(avg))
# frame_delta = cv2.absdiff(first_frame, gray_frame)
thresh = cv2.threshold(frame_delta, 20, 255, cv2.THRESH_BINARY)[1]
thresh = cv2.dilate(thresh, None, iterations=5)
cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)
for c in cnts:
if cv2.contourArea(c) < args["min_area"]:
continue
(x, y, w, h) = cv2.boundingRect(c)
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 0, 255), 2)
text = "Occupied"
def DetectObjectFromImage(beforeImage, afterImage, beforeGrayImage,
afterGrayImage):
resizeRate = GetContour.SquareDetectAndReturnRateAsSquare(beforeGrayImage)
beforeImage = CustomOpenCV.ResizeImageAsRate(beforeImage, resizeRate)
beforeGrayImage = CustomOpenCV.ResizeImageAsRate(beforeGrayImage,
resizeRate)
afterImage = CustomOpenCV.ResizeImageAsRate(afterImage, resizeRate)
afterGrayImage = CustomOpenCV.ResizeImageAsRate(afterGrayImage, resizeRate)
#squareContourData = DetectBackgroundSquare.DetectBackgroundSquareFromImage(beforeImage) #형광색 인식으로 점 4개 찾는 함수
# in mac
# this function is not working and falling loop.
squareContourData = DetectBlackBoardContourFromOriginImage(beforeGrayImage)
# 굴곡진 큰 사각형 정사각형으로 보정
perspectiveUpdatedBeforeImage = ImageMatrixMove.ImageMatrixMove(
beforeImage, squareContourData)
perspectiveUpdatedAfterImage = ImageMatrixMove.ImageMatrixMove(
afterImage, squareContourData)
perspectiveUpdatedBeforeImage = CustomOpenCV.ResizeImageAsWidth(
perspectiveUpdatedBeforeImage, DefineManager.IMAGE_WIDTH)
perspectiveUpdatedAfterImage = CustomOpenCV.ResizeImageAsWidth(
perspectiveUpdatedAfterImage, DefineManager.IMAGE_WIDTH)
# Resize image as shape [ rateHeight, DefineManager.IMAGE_WIDTH ]
#CustomOpenCV.ShowImagesWithName([perspectiveUpdatedBeforeImage, perspectiveUpdatedAfterImage],
# ["perspectiveUpdatedBeforeImage", "perspectiveUpdatedAfterImage"])
perspectiveUpdatedBeforeGrayImage = cv2.cvtColor(
perspectiveUpdatedBeforeImage, cv2.COLOR_BGR2GRAY)
perspectiveUpdatedAfterGrayImage = cv2.cvtColor(
perspectiveUpdatedAfterImage, cv2.COLOR_BGR2GRAY)
morphologyKernel = np.ones(
(Setting.DefineManager.MORPHOLOGY_MASK_SIZE + 1,
Setting.DefineManager.MORPHOLOGY_MASK_SIZE + 1), np.uint8)
perspectiveUpdatedBeforeMorphologyGrayImage = cv2.morphologyEx(
perspectiveUpdatedBeforeGrayImage, cv2.MORPH_OPEN, morphologyKernel)
perspectiveUpdatedAfterMorphologyGrayImage = cv2.morphologyEx(
perspectiveUpdatedAfterGrayImage, cv2.MORPH_OPEN, morphologyKernel)
# Reduce image noise
beforeThresholdedBlackBoardImage = cv2.adaptiveThreshold(
perspectiveUpdatedBeforeMorphologyGrayImage,
Setting.DefineManager.SET_IMAGE_WHITE_COLOR,
cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY,
Setting.DefineManager.NEIGHBORHOOD_MASK_SIZE, 10)
afterThresholdedBlackBoardImage = cv2.adaptiveThreshold(
perspectiveUpdatedAfterMorphologyGrayImage,
Setting.DefineManager.SET_IMAGE_WHITE_COLOR,
cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY,
Setting.DefineManager.NEIGHBORHOOD_MASK_SIZE, 10)
# Adaptive Threshold Image
#CustomOpenCV.ShowImagesWithName([beforeThresholdedBlackBoardImage, afterThresholdedBlackBoardImage], ['beforeThresholdedBlackBoardImage', 'afterThresholdedBlackBoardImage'])
differenceBasedOnThreshImage = cv2.absdiff(
beforeThresholdedBlackBoardImage, afterThresholdedBlackBoardImage)
differenceBasedOnThreshImage[
differenceBasedOnThreshImage > Setting.DefineManager.
EACH_IMAGE_DIFFERENCE_THRESHOLD] = Setting.DefineManager.SET_IMAGE_WHITE_COLOR
# Detect each image difference from Threshold Image
#CustomOpenCV.ShowImagesWithName([differenceBasedOnThreshImage], ["differenceBasedOnThreshImage"])
objectFoundedImage = GetContour.GetObjectImage(
perspectiveUpdatedBeforeImage, perspectiveUpdatedAfterImage)
humanDetectedContour, contourLineDrawImage = GetContour.GetContour(
objectFoundedImage, perspectiveUpdatedAfterImage)
faceMinY, faceMaxY = GetContour.DetectFaceAndGetY(
perspectiveUpdatedAfterImage)
navelPoint, faceRate, maxY, minY = GetContour.FindNavel(
humanDetectedContour, faceMaxY, contourLineDrawImage)
height = maxY - minY
importantPoint = GetContour.AngleAsDealWithPointFromContours(
humanDetectedContour, contourLineDrawImage)
beforeDrawImage = np.copy(perspectiveUpdatedBeforeImage)
afterDrawImage = np.copy(perspectiveUpdatedAfterImage)
functionParameter = []
for index in range(len(importantPoint)):
xArray = []
yArray = []
for point in importantPoint[index]:
x, y = point.ravel()
xArray.append(x)
yArray.append(y)
xArray = np.asarray(xArray)
yArray = np.asarray(yArray)
if xArray.shape[0] > 0:
# ax + b = y (a, b 를 받아옴)
functionCharacteristic = sp.polyfit(
xArray, yArray, DefineManager.FUNCTION_DIMENSION)
functionParameter.append(functionCharacteristic)
yRegressionArray = sp.polyval(functionCharacteristic, xArray)
err = np.sqrt(
sum((yArray - yRegressionArray)**2) / yArray.shape[0])
pointA, pointB = GetContour.GetStartAndEndPointsFromLine(
functionCharacteristic, xArray)
cv2.line(beforeDrawImage, pointA, pointB,
DefineManager.RGB_COLOR_GREEN, 1)
cv2.line(afterDrawImage, pointA, pointB,
DefineManager.RGB_COLOR_GREEN, 1)
CustomOpenCV.ShowImagesWithName([beforeDrawImage, afterDrawImage])
return [
perspectiveUpdatedBeforeImage, perspectiveUpdatedAfterImage, height,
navelPoint, humanDetectedContour, functionParameter, beforeDrawImage,
faceRate
]
graybg = cv2.cvtColor(bg, cv2.COLOR_BGR2GRAY)
graybg_sink = cv2.cvtColor(bg_sink, cv2.COLOR_BGR2GRAY)
# Step 2 : medianBlur
median = cv2.medianBlur(gray, 7)
median_sink = cv2.medianBlur(gray_sink, 7)
medianbg = cv2.medianBlur(graybg, 7)
medianbg_sink = cv2.medianBlur(graybg_sink, 7)
# Step 3 : find lastframe
if i % 6 == 0:
lastframe = median
lastframe_sink = median_sink
#print "lastframe dealed"
raw.truncate(0)
continue
# Step 4 : absolute diff between lastframe and current frame
deltab = cv2.absdiff(medianbg, median)
delta = cv2.absdiff(lastframe, median)
deltab_sink = cv2.absdiff(medianbg_sink, median_sink)
delta_sink = cv2.absdiff(lastframe_sink, median_sink)
# 5 sets of accumulation between frames
if i % 6 == 1:
accu_img = delta
accu_imgb = deltab
accu_img_sink = delta_sink
accu_imgb_sink = deltab_sink
elif i % 6 == 2:
accu_img = cv2.addWeighted(delta, 0.5, accu_img, 0.5, 0)
accu_imgb = cv2.addWeighted(deltab, 0.5, accu_imgb, 0.5, 0)
accu_img_sink = cv2.addWeighted(delta_sink, 0.5, accu_img_sink,
0.5, 0)
accu_imgb_sink = cv2.addWeighted(deltab_sink, 0.5, accu_imgb_sink,
def detectWithCam(self):
self.logger.debug("Starting to detect Motion")
#variables for TensorFlow human detection
model_path = os.path.join(self.TOPDIR, "client/objectDetect/ssd_mobilenet_v1_coco_2017_11_17/frozen_inference_graph.pb")
detector = detectorAPI(self.ENVIRON, path_to_ckpt=model_path)
# define feed from camera
camera = cv2.VideoCapture(0)
time.sleep(1)
# initialize variables used by the motion sensing
firstFrame = None
lastAlert = datetime.datetime.today()
frames = 0
# loop over the frames of the video feed and detect motion
while True:
# if we are busy processing a job then skip motion until we are done
#if self.ENVIRON["listen"] == False:
if busyCheck(self.ENVIRON, self.logger) == True:
continue
# grab the current frame and initialize the occupied/unoccupied text
self.logger.debug("Getting another frame. ENVIRON listen = %s" % self.ENVIRON["listen"])
(grabbed, frame) = camera.read()
frames += 1
# resize the frame, convert it to grayscale, and blur it
frame = imutils.resize(frame, width=500)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
# if the first frame is None, initialize it
if firstFrame is None:
firstFrame = gray
continue
# compute the absolute difference between the current frame and first frame
frameDelta = cv2.absdiff(firstFrame, gray)
thresh = cv2.threshold(frameDelta, 25, 255, cv2.THRESH_BINARY)[1]
# Update the reference frame
firstFrame = gray
# dilate the thresholded image to fill in holes, then find contours on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# loop over the contours
for c in cnts:
# if the contour is too small, ignore it
if cv2.contourArea(c) < self.min_area:
continue
#motion detected, see if a person was detected by objectDetect
#------------------------------------------------------------
objDict = detector.objectCount(frame)
if 'person' in objDict:
lastAlert = self.detectionEvent(lastAlert, camera)
# check the ENVIRON when frame count reaches check point
if frames > self.framesCheck:
self.logger.debug("Checking to see if we should stop detecting motion")
frames = 0
if not self.ENVIRON["motion"]:
self.logger.debug("Time to stop detecting motion")
# cleanup the camera quit function
camera.release()
break
def match_card(value_image,
suit_image,
is_it_table_card,
VALUE_DIFFERENCE_LIMIT=830,
SUIT_DIFFERENCE_LIMIT=330):
#All calculated diffrences amounts are lower than this large number.
best_value_difference_amount = 1000000
best_suit_difference_amount = 1000000
best_value_match_name = "Unknown"
best_suit_match_name = "Unknown"
for value_name in [
'Ace', 'Two', 'Three', 'Four', 'Five', 'Six', 'Seven', 'Eight',
'Nine', 'Ten', 'Jack', 'Queen', 'King'
]:
if is_it_table_card == True:
image_path = os.path.abspath(os.path.dirname(__file__)) +\
"/Source Card Images for Celeb/Table Cards/%s.png"%value_name
elif is_it_table_card == False:
image_path = os.path.abspath(os.path.dirname(__file__)) +\
"/Source Card Images for Celeb/My Cards/%s.png"%value_name
value_source_image = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)
if isinstance(value_source_image, type(None)):
raise Exception("Unable to read %s.png value source image" %
value_name)
#comparing 2 images. value_image and value_source_image sizes should be equal otherwise it errors
difference_image = cv2.absdiff(value_image, value_source_image)
difference_amount = int(np.sum(difference_image) / 255)
if difference_amount < best_value_difference_amount:
#best_value_difference_image = difference_image
best_value_difference_amount = difference_amount
best_value_name = value_name
for suit_name in ['Spade', 'Heart', 'Club', 'Diamond']:
if is_it_table_card == True:
image_path = os.path.abspath(os.path.dirname(__file__)) +\
"/Source Card Images for Celeb/Table Cards/%s.png"%suit_name
elif is_it_table_card == False:
image_path = os.path.abspath(os.path.dirname(__file__)) +\
"/Source Card Images for Celeb/My Cards/%s.png"%suit_name
suit_source_image = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)
if isinstance(suit_source_image, type(None)):
raise Exception("Unable to read %s.png suit source image" %
suit_name)
#comparing 2 images. value_image and value_source_image sizes should be equal otherwise it errors
difference_image = cv2.absdiff(suit_image, suit_source_image)
difference_amount = int(np.sum(difference_image) / 255)
if difference_amount < best_suit_difference_amount:
#best_suit_difference_image = difference_image
best_suit_difference_amount = difference_amount
best_suit_name = suit_name
if (best_value_difference_amount < VALUE_DIFFERENCE_LIMIT):
best_value_match_name = best_value_name
if (best_suit_difference_amount < SUIT_DIFFERENCE_LIMIT):
best_suit_match_name = best_suit_name
return best_value_match_name, best_suit_match_name, best_value_difference_amount, best_suit_difference_amount
while True:
check, frame = video.read(
) #check function will correspond to activeness of camera and video.read() would capture the image and store the data in a numpy array
#print(check)
# print(frame) #just printing hte array of image(numpyarray)
gray = cv2.cvtColor(
frame, cv2.COLOR_BGR2GRAY
) #we are simply converting our image in a black & white model so that face detection could become easier
gray = cv2.GaussianBlur(
gray, (21, 21), 0
) #this would help our program to detect any object easily...for more info google it...
if first_frame is None:
first_frame = gray
continue # consider the background.... the first loop will collect background details and store in first_frame and so untill unless we get a backgruond this loop print no frame
delta_frame = cv2.absdiff(
first_frame, gray
) # this is anotherimage created by difference in background ..... for suppose some object enters the space so numpy array would change and hence a diffrence array would be created giving us a new anotherimage
thresh_frame = cv2.threshold(
delta_frame, 30, 255, cv2.THRESH_BINARY
)[1] #this is a threshold set up...If the diffrence betn the pixels of any certain region is more than 30 show it white (255 is code of white if we"ll give 252 green will appoear)..and [1] is beacause threshold gives two values amd here we need the second...
thresh_frame = cv2.dilate(thresh_frame, None, iterations=2)
(_, cnts, _) = cv2.findContours(thresh_frame.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
for contours in cnts:
if cv2.contourArea(contours) < 1000:
continue
(x, y, w, h) = cv2.boundingRect(contours)
cv2.rectangle(frame, (x, w), (x + w, y + h), (0, 255, 0), 3)
cv2.imshow("Back ground", delta_frame)
cv2.imshow(
"Capture", gray
colorizer = rs.colorizer()
colorized_depth = np.asanyarray(
colorizer.colorize(depth_frame).get_data())
# Create alignment primitive with color as its target stream:
align = rs.align(rs.stream.color)
frameset = align.process(frameset)
# Update color and depth frames:
aligned_depth_frame = frameset.get_depth_frame()
colorized_depth = np.asanyarray(
colorizer.colorize(aligned_depth_frame).get_data())
### motion detector
d = cv2.absdiff(color_init, color)
gray = cv2.cvtColor(d, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (5, 5), 0)
_, th = cv2.threshold(blur, 20, 255, cv2.THRESH_BINARY)
dilated = cv2.dilate(th, np.ones((3, 3), np.uint8), iterations=3)
(c, _) = cv2.findContours(dilated, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
# cv2.drawContours(color, c, -1, (0, 255, 0), 2)
color_init = color
depth = np.asanyarray(aligned_depth_frame.get_data())
for contour in c:
if cv2.contourArea(contour) < 1500:
continue
(x, y, w, h) = cv2.boundingRect(contour)
import cv2
import numpy as np
from threading import Thread
from playsound import playsound
cap = cv2.VideoCapture(1)
k = True
key_threshold = 1200
ret, t0 = cap.read()
ret, t1 = cap.read()
t0 = cv2.cvtColor(t0, cv2.COLOR_BGR2GRAY)
while True:
t1 = cv2.cvtColor(t1, cv2.COLOR_BGR2GRAY)
diff = cv2.absdiff(t1, t0)
ret, threshold_image = cv2.threshold(diff, 25, 255, cv2.THRESH_BINARY)
key1 = cv2.countNonZero(threshold_image[355:410, 0:35])
key2 = cv2.countNonZero(threshold_image[355:410, 40:85])
key3 = cv2.countNonZero(threshold_image[355:410, 90:130])
key4 = cv2.countNonZero(threshold_image[355:410, 135:183])
key5 = cv2.countNonZero(threshold_image[355:410, 183:230])
key6 = cv2.countNonZero(threshold_image[355:410, 230:275])
key7 = cv2.countNonZero(threshold_image[355:410, 275:325])
key8 = cv2.countNonZero(threshold_image[355:410, 325:370])
key9 = cv2.countNonZero(threshold_image[355:410, 370:420])
key10 = cv2.countNonZero(threshold_image[355:410, 420:470])
key11 = cv2.countNonZero(threshold_image[355:410, 470:515])
key12 = cv2.countNonZero(threshold_image[355:410, 515:560])
key13 = cv2.countNonZero(threshold_image[355:410, 560:610])
key14 = cv2.countNonZero(threshold_image[355:410, 610:638])
def main(textOut, textIn):
det.__int__()
camera = cv2.VideoCapture("ch12_20190703090318.mp4")
firstFrame = None
# loop over the frames of the video
while True:
# grab the current frame
(grabbed, frame) = camera.read()
# if the frame could not be grabbed, then we have reached the end
# of the video
if not grabbed:
break
# resize the frame, convert it to grayscale, and blur it
frame = imutils.resize(frame, width=width)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
cv2.imshow('', gray)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
#cv2.imshow('',gray)
# if the first frame is None, initialize it and treat it as the background
if firstFrame is None:
firstFrame = gray
continue
# computing the absolute difference between the current frame and background(first frame)
frameDelta = cv2.absdiff(firstFrame, gray)
thresh = cv2.threshold(frameDelta, 25, 255, cv2.THRESH_BINARY)[1]
# dilating the thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)[0]
#print (cnts)
# loop over the contours
xset = 0
rects = []
start = 0
for c in cnts:
#rects.append(c.astype("int"))
xset += 1
#print(c)
#To track the moevement of the object
up = 0
down = 0
movin = 0
# if the contour is too small, ignore it
#if cv2.contourArea(c) < 12000:
#continue
# compute the bounding box for the contour, draw it on the frame,
cv2.line(frame, (450, 300), (700, 300), (0, 0, 255), 2) # red line
(x, y, w, h) = cv2.boundingRect(c)
#print(cv2.boundingRect(c))
rects.append(cv2.boundingRect(c))
#print(rects[0])
#print(enumerate(rects))
#start = 1
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)
#cv2.line(frame, (450, 300), (700, 300), (250, 0, 1), 2) # blue line
# cv2.line(frame, (450,300), (700,300), (0, 0, 255), 2) # red line
rectagleCenterPont = ((x + x + w) // 2, (y + y + h) // 2)
cv2.circle(frame, rectagleCenterPont, 1, (0, 0, 255), 5)
'''print(rectagleCenterPont)
if(rectagleCenterPont[1]<300):
down = 1
movin = 0
else:
up = 1
movin = 1
print(down, up, movin)
#check for in and out'''
_, up, down = det.update(rects, up, down)
textIn += down
textOut += up
# show the frame and record if the user presses a key
cv2.imshow("Thresh", thresh)
cv2.imshow("Frame Delta", frameDelta)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cv2.putText(frame, "In: {}".format(str(textIn)), (10, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
cv2.putText(frame, "Out: {}".format(str(textOut)), (10, 70),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
cv2.putText(frame,
datetime.datetime.now().strftime("%A %d %B %Y %I:%M:%S%p"),
(10, frame.shape[0] - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.35,
(0, 0, 255), 1)
cv2.imshow("Security Feed", frame)
#close any open windows
camera.release()
cv2.destroyAllWindows()
cv2.VideoWriter_fourcc('D', 'I', 'V', 'X'), fps,
size)
while True:
check, frame = video.read()
status = -1
text = 'Unoccupied'
grayImg = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
grayImg = cv2.GaussianBlur(grayImg, (21, 21), 0)
if first_frame is None:
first_frame = grayImg
else:
pass
deltaFrame = cv2.absdiff(first_frame, grayImg)
threshFrame = cv2.threshold(deltaFrame, 30, 255, cv2.THRESH_BINARY)[1]
threshFrame = cv2.dilate(threshFrame, None, iterations=2)
(cnts, _) = cv2.findContours(threshFrame.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
for contour in cnts:
if cv2.contourArea(contour) < 10000:
continue
status = 1
(x, y, w, h) = cv2.boundingRect(contour)
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 1)
text = 'Occupied'
def diff_img(self):
f0, f1, f2 = self.three_frames()
d1 = cv2.absdiff(f1, f0)
d2 = cv2.absdiff(f2, f1)
self.diff = cv2.bitwise_and(d1, d2)
#
# print(avg.shape)
#print(avg_show.shape)
cur_back = avg
if (flag == 0):
#buf_back[:] = 0 #no need as buf_back is initialized to be zero matrix
flag = 10 # pour eliminer le cas ou nframe=1
if (flag == 10 and count[i - 1] >= nframe):
buf_back[
i -
1] = cur_back # doub maya3mil hekom el nframe , yimchi ya3ti lil buf_back el deffirence mabin el back w awil frame 5dheha ba3d el back
flag = 20
sub = cv2.absdiff(
cur_back, buf_back[i - 1]
) # voila lahna nhoto la difference entre le background wil frame ali 9e3din nitraitiw fiha
img_show = cv2.resize(img, (400, 400))
#img_show = img_show.astype(int)
#print(img_show.shape)
#print(img_show)
#time.sleep(1)
cv2.imshow("img", img_show) # affichage de l'image originale
gray_show = cv2.resize(gray, (400, 400))
#gray_show = gray_show.astype(int)
cv2.imshow("gray", gray_show)
#print(cur_back)
cur_back_show = cv2.resize(cur_back, (400, 400))
frame = imutils.resize(frame, width=500)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
# if the average frame is None, initialize it
if avg is None:
print ("[INFO] starting background model...")
avg = gray.copy().astype("float")
rawCapture.truncate(0)
continue
# accumulate the weighted average between the current frame and
# previous frames, then compute the difference between the current
# frame and running average
cv2.accumulateWeighted(gray, avg, 0.5)
frameDelta = cv2.absdiff(gray, cv2.convertScaleAbs(avg))
# threshold the delta image, dilate the thresholded image to fill
# in holes, then find contours on thresholded image
thresh = cv2.threshold(frameDelta, 5, 255,
cv2.THRESH_BINARY)[1]
thresh = cv2.dilate(thresh, None, iterations=2)
#(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
_, cnts, _= cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
# loop over the contours
for c in cnts:
# if the contour is too small, ignore it
if cv2.contourArea(c) < 5000:
continue
def video_thread(self):
# open midi port
self.out_port = mido.open_output('Output',
client_name='Motion Detector (OUT)')
logging.info('Output port: {}'.format(self.out_port))
camera = cv2.VideoCapture(0)
time.sleep(0.25)
if self.conf.C_DISPLAY_VIDEO == 1:
cv2.namedWindow("M2M Motion", cv2.WINDOW_NORMAL)
# initialize the first frame in the video stream
previousFrame = None
gray = None
# loop over the frames of the video
while self.gRun:
(grabbed, frame) = camera.read()
# if the frame could not be grabbed, then we have reached the end
# of the video
if not grabbed:
break
# save the previous frame and grab a new
previousFrame = gray
# resize the frame, convert it to grayscale, and blur it
frame = imutils.resize(frame, width=500)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (7, 7), 0)
# skip diff if this was the first frame
if previousFrame is None:
continue
# compute the absolute difference between the current frame and
# previous frame
frameDelta = cv2.absdiff(previousFrame, gray)
thresh = cv2.threshold(frameDelta, 25, 255, cv2.THRESH_BINARY)[1]
# dilate the thresholded image to fill in holes, then find contours
# on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
# compress image array to one int
currentChange = sum(sum(thresh))
# update the highest found if needed
if currentChange >= self.gHighestSeenChange:
self.gHighestSeenChange = currentChange
# calucate the amount of change and make it into a MIDI (0-127)
percent = float(currentChange) / float(self.gHighestSeenChange)
self.gMidiChange = int(percent * 127)
# send a MIDI message based on timing
if self.conf.C_TRIGGER_BY_TIMING == 1:
if self.gSync == 0:
self.gSync = self.conf.C_VIDEO_FPS / self.conf.C_MIDI_MPS
logging.debug("Sending " + str(self.gMidiChange))
cc = Message('control_change',
channel=13,
control=1,
value=int(self.gMidiChange))
self.out_port.send(cc)
else:
self.gSync = self.gSync - 1
# slowly readjust the highest found
if self.conf.C_READJUST_AMOUNT != 0 and self.gHighestSeenChange >= int(
self.conf.C_READJUST_AMOUNT):
self.gHighestSeenChange = self.gHighestSeenChange - self.conf.C_READJUST_AMOUNT
# show display if needed
if self.conf.C_DISPLAY_VIDEO == 1:
cv2.putText(thresh,
"Movement in MIDI: {}".format(self.gMidiChange),
(10, 20), cv2.FONT_HERSHEY_SIMPLEX, 0.5,
(255, 255, 255), 2)
cv2.imshow("M2M Motion", thresh)
# check for keyboard input
cv2.waitKey(1) & 0xFF
ms = 1000 / self.conf.C_VIDEO_FPS
time.sleep(ms / 1000.0) # 1000.0 because we want a float
# cleanup the camera and close any open windows
camera.release()
cv2.destroyAllWindows()
logging.info("Leaving Video thread.")
def CamMovement(qc):
global Cam_Run
firstFrame = None
hits = 0 # counter for us to cycle over files
CHNG_THRESH = 65 # Change Threshold used to be 25
HR_Cam = qc.get()
cc = qc.get()
i = 0
vs = [] # init VS array
while i < cam_count:
vs.append(cv2.VideoCapture(i))
if not vs[i].isOpened():
print('Could not open webcam #' + str(i) + ' \n')
vs[i].release()
vs.pop(i)
i = i - 1
break
i = i + 1
signal.signal(signal.SIGTERM, sigterm_cam)
signal.signal(signal.SIGINT, sigterm_cam)
while Cam_Run:
# grab the current frame and initialize the occupied/unoccupied
retval, frame = vs[HR_Cam].read()
text = "Unoccupied"
# if the frame could not be grabbed, then we have reached the end
# of the video
if frame is None:
break
# resize the frame, convert it to grayscale, and blur it
frame = imutils.resize(frame, width=500)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
# if the first frame is None, initialize it
if firstFrame is None:
firstFrame = gray
continue
# compute the absolute difference between the current frame and
# first frame
frameDelta = cv2.absdiff(firstFrame, gray)
thresh = cv2.threshold(frameDelta, CHNG_THRESH, 255,
cv2.THRESH_BINARY)[1]
# dilate the thresholded image to fill in holes, then find contours
# on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if imutils.is_cv2() else cnts[1]
# loop over the contours
Caption = "Empty"
for c in cnts:
# compute the bounding box for the contour, draw it on the frame,
# and update the text
(x, y, w, h) = cv2.boundingRect(c)
if (w > 10) and (h > 10): # trying to eliminate tiny changes
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)
text = "Occupied"
Caption = text + ' !'
# draw the text and timestamp on the frame
cv2.putText(frame, "Room Status: {}".format(Caption), (10, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
cv2.putText(frame,
datetime.datetime.now().strftime("%A %d %B %Y %I:%M:%S%p"),
(10, frame.shape[0] - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.35,
(0, 0, 255), 1)
# show the frame
if text == "Occupied":
cv2.imwrite(str(hits) + '_Security' + '.png', frame)
for x in range(cc):
if x != HR_Cam:
retval, frame = vs[x].read()
Caption = str(x)
cv2.putText(frame, "Camera: {}".format(Caption), (10, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
cv2.putText(
frame,
datetime.datetime.now().strftime(
"%A %d %B %Y %I:%M:%S%p"),
(10, frame.shape[0] - 10), cv2.FONT_HERSHEY_SIMPLEX,
0.35, (0, 0, 255), 1)
cv2.imwrite(
str(hits) + '_Cam' + str(x) + '_' + '.png', frame)
hits = hits + 1
qc.put([True, datetime.datetime.now()])
sleep(4) # give it 4 secs before you grab more frames
if hits > 19: # recycle videos so as not to eat space
hits = 0
i = 0
while i < (cam_count - 1):
vs[i].release()
vs.pop(i)
i = i + 1
def_wt = INITIAL_AVERAGE_WEIGHT
else:
def_wt = DEFAULT_AVERAGE_WEIGHT
cv2.accumulateWeighted(grayFrame, avg, def_wt)
# export averaged background for use in next video feed run
# if frame_no > int(total_frames * 0.975):
if frame_no > int(200):
grayOp = cv2.cvtColor(cv2.convertScaleAbs(avg), cv2.COLOR_GRAY2BGR)
backOut = loc + "/backgrounds/" + camera + "_bg.jpg"
cv2.imwrite(backOut, grayOp)
# Compute the grayscale difference between the current grayscale frame and
# the average of the scene.
differenceFrame = cv2.absdiff(grayFrame, cv2.convertScaleAbs(avg))
# blur the difference image
differenceFrame = cv2.GaussianBlur(differenceFrame, (5, 5), 0)
# cv2.imshow("difference", differenceFrame)
diffout = cv2.cvtColor(differenceFrame, cv2.COLOR_GRAY2BGR)
diffop.write(diffout)
# get estimated otsu threshold level
retval, _ = cv2.threshold(differenceFrame, 0, 255,
cv2.THRESH_BINARY + cv2.THRESH_OTSU)
# add to list of threshold levels
t_retval.append(retval)
# apply threshold based on average threshold value
if frame_no < 10:
ret2, thresholdImage = cv2.threshold(differenceFrame,
def diff(present_frame, last_frame):
'''
Target:
predict the position of moving object by diff the 2 continious frames and some cv operation
we only care about the left-most motion blob, which indicate how far the hand/object reach
***when the box or coor(xmin,ymin) occurs on the left side for ***many times/
we can turn the state to ***hand in.
Agrs:
inputs:
present frame
last_frame
outputs:
bbox:(xmin,ymin,xmax,ymax) of most-left motion blob
To do: copy the present_gray to last_gray, in oder to get rid of color conversion
'''
# var to save the xmin coor
xmin = inWidth
ymin = None
xmax = None
ymax = None
present_gray = cv.cvtColor(present_frame, cv.COLOR_BGR2GRAY)
present_gray = cv.GaussianBlur(present_gray, (21, 21), 0)
last_gray = cv.cvtColor(last_frame, cv.COLOR_BGR2GRAY)
last_gray = cv.GaussianBlur(last_gray, (21, 21), 0)
frame_delta = cv.absdiff(last_gray, present_gray)
frame_delta = cv.threshold(
frame_delta, 25, 255,
cv.THRESH_BINARY)[1] # the threshold should be modified
# img show the diff img without operation
# cv.imshow('diff_1', frame_delta)
#operation, need to be modified
# frame_delta = cv.erode(frame_delta, cv.getStructuringElement(cv.MORPH_ELLIPSE, (3, 3)), iterations=2)
# frame_delta = cv.dilate(frame_delta, cv.getStructuringElement(cv.MORPH_ELLIPSE, (8, 3)), iterations=2)
frame_delta = cv.erode(frame_delta, None, iterations=2)
frame_delta = cv.erode(frame_delta, None, iterations=2)
frame_delta = cv.erode(frame_delta, None, iterations=2)
frame_delta = cv.erode(frame_delta, None, iterations=1)
frame_delta = cv.dilate(frame_delta, None, iterations=10)
# img show the diff img with operation
# cv.imshow("diff_2", frame_delta)
# find be connected components and draw the rect bbox
(_, cnts, _) = cv.findContours(frame_delta.copy(), cv.RETR_CCOMP,
cv.CHAIN_APPROX_SIMPLE)
for c in cnts:
if cv.contourArea(c) < 20: # need to be tested
continue
(x, y, w, h) = cv.boundingRect(c)
#print(cv.contourArea(c))
if x < xmin:
xmin = x
ymin = y
xmax = x + w
ymax = y + h
# cv.rectangle(frame_delta, (x, y), (x + w, y + h), (0, 255, 0), 5) # when comment these drawing func, gain a higher speed
# img show the diff img with operation and rect bbox
# cv.imshow("diff_3", frame_delta)
return xmin, ymin, xmax, ymax
import cv2
import numpy as np
I = cv2.imread('input/in000301.jpg')
I = cv2.cvtColor(I, cv2.COLOR_BGR2GRAY)
I = cv2.threshold(I, 100, 255, cv2.THRESH_BINARY)
I = I[1]
for i in range(1, 150):
I_prev = I
I_clr = cv2.imread('input/in000' + str(300 + i) + '.jpg')
I = cv2.cvtColor(I_clr, cv2.COLOR_BGR2GRAY)
I_diff = cv2.absdiff(I, I_prev)
I_diff = cv2.threshold(I_diff, 40, 255, cv2.THRESH_BINARY)
I_diff = I_diff[1]
I_diff = cv2.medianBlur(I_diff, 3)
I_diff = cv2.dilate(I_diff,
cv2.getStructuringElement(cv2.MORPH_RECT, (7, 7)),
iterations=3)
I_diff = cv2.erode(I_diff,
cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3)))
retval, labels, stats, centroids = cv2.connectedComponentsWithStats(I_diff)
# cv2.imshow("Labels", np.uint8(labels / stats.shape[0]*255))
if (stats.shape[0] > 1): # czy sa jakies obiekty
pi, p = max(enumerate(stats[1:, 4]), key=(lambda x: x[1]))
pi = pi + 1
# wyrysownie bbox
cv2.rectangle(
I_clr, (stats[pi, 0], stats[pi, 1]),
(stats[pi, 0] + stats[pi, 2], stats[pi, 1] + stats[pi, 3]),
(255, 0, 0), 2)
import cv2
import numpy as np
cap = cv2.VideoCapture('k.avi')
_, frame1 = cap.read()
_, frame2 = cap.read()
while True:
diff = cv2.absdiff(frame1, frame2)
gray = cv2.cvtColor(diff, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (5, 5), 0)
_, thresh = cv2.threshold(blur, 20, 255, cv2.THRESH_BINARY)
dilated = cv2.dilate(thresh, None, iterations=3)
contours, _ = cv2.findContours(dilated, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
# cv2.drawContours(frame1,contours,-1,(0,244,0),2)
for cont in contours:
(x, y, w, h) = cv2.boundingRect(cont)
if cv2.contourArea(cont) < 700:
continue
cv2.rectangle(frame1, (x, y), (x + w, y + h), (124, 233, 44), 2)
print(frame1.shape)
cv2.putText(frame1, "Status: {}".format('Movement'), (10, 20),
cv2.FONT_HERSHEY_COMPLEX, 1, (0, 200, 0), 2)
cv2.putText(frame1, "Made by KAMI_360 Using Open-cv/python",
(190, 520), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 0),
2)
cv2.putText(frame1, "KAMI_360 Kindly Alert", (10, 70),
cv2.FONT_HERSHEY_COMPLEX_SMALL, 2, (0, 0, 0), 1)
cv2.imshow('Detected', frame1)
frame1 = frame2
ret, frame2 = cap.read()
if cv2.waitKey(5) == 27:
def test_pair(transform, jpg, mat, out):
"""
jpg = filename
mat = filename
"""
data = scipy.io.loadmat(mat)
regions = data['regions'].flatten()
max_type = 0
for r in regions:
max_type = max(max_type, r['type'][0][0][0][0])
r_vals = {}
for t in np.arange(max_type):
r_vals[t + 1] = np.array([], 'float32')
g_vals = copy.deepcopy(r_vals)
b_vals = copy.deepcopy(r_vals)
h_vals = copy.deepcopy(r_vals)
s_vals = copy.deepcopy(r_vals)
v_vals = copy.deepcopy(r_vals)
result_stats = {
'average_abs_err': [],
'total_pixels': 0,
'total_error': 0,
'total_regions': 0,
'r_vals': r_vals,
'g_vals': g_vals,
'b_vals': b_vals,
'h_vals': h_vals,
's_vals': s_vals,
'v_vals': v_vals
}
for r in regions:
logger.info('region')
x = r['x'][0][0].flatten()
y = r['y'][0][0].flatten()
mask = r['mask'][0][0]
mask3 = cv2.merge([mask, mask, mask])
print 'x', x
print 'y', y
print 'mask shape', mask.shape
# type in 1- based / matlab-based indices from the list of region types (i.e road, white,
# yellow, red, or what ever types were annotated)
print 'type', r['type'][0][0][0][0]
# color in [r,g,b] where [r,g,b]are between 0 and 1
print 'color', r['color'][0]
t = r['type'][0][0][0][0]
# print 'guy look here'
region_color = r['color'][0]
region_color = region_color[0][0]
rval = region_color[0] * 255.
gval = region_color[1] * 255.
bval = region_color[2] * 255.
image = dtu.image_cv_from_jpg_fn(jpg)
transformed = transform(image)
[b2, g2, r2] = cv2.split(transformed)
thsv = cv2.cvtColor(transformed, cv2.COLOR_BGR2HSV)
[h2, s2, v2] = cv2.split(thsv)
r2_ = r2[mask.nonzero()]
g2_ = g2[mask.nonzero()]
b2_ = b2[mask.nonzero()]
h2_ = h2[mask.nonzero()]
s2_ = s2[mask.nonzero()]
v2_ = v2[mask.nonzero()]
result_stats['r_vals'][t] = np.concatenate(
(result_stats['r_vals'][t], r2_), 0)
result_stats['g_vals'][t] = np.concatenate(
(result_stats['g_vals'][t], g2_), 0)
result_stats['b_vals'][t] = np.concatenate(
(result_stats['b_vals'][t], b2_), 0)
result_stats['h_vals'][t] = np.concatenate(
(result_stats['h_vals'][t], h2_), 0)
result_stats['s_vals'][t] = np.concatenate(
(result_stats['s_vals'][t], s2_), 0)
result_stats['v_vals'][t] = np.concatenate(
(result_stats['v_vals'][t], v2_), 0)
absdiff_img = cv2.absdiff(transformed, np.array([bval, gval, rval,
0.]))
masked_diff = cv2.multiply(np.array(absdiff_img, 'float32'),
np.array(mask3, 'float32'))
num_pixels = cv2.sumElems(mask)[0]
region_error = cv2.sumElems(cv2.sumElems(masked_diff))[0]
avg_abs_err = region_error / (num_pixels + 1.)
print 'Average abs. error', avg_abs_err
result_stats['average_abs_err'].append(avg_abs_err)
result_stats[
'total_pixels'] = result_stats['total_pixels'] + num_pixels
result_stats[
'total_error'] = result_stats['total_error'] + region_error
result_stats['total_regions'] = result_stats['total_regions'] + 1
# XXX: to finish
return result_stats
def cameraLoop(camera):
# initialize the first frame in the video stream
firstFrame = None
testOccupied = "false"
firstOccupied = 0
testOccupiedCounter = 0;
# loop over the frames of the video
while True:
# grab the current frame and initialize the occupied/unoccupied
# text
(grabbed, frame) = camera.read()
text = "Unoccupied"
frame = frame[25:500, 110:475]
# resize the frame, convert it to grayscale, and blur it
frame = imutils.resize(frame, width=500)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
# if the first frame is None, initialize it
if firstFrame is None:
firstFrame = gray
continue
# compute the absolute difference between the current frame and
# first frame
frameDelta = cv2.absdiff(firstFrame, gray)
thresh = cv2.threshold(frameDelta, 50, 255, cv2.THRESH_BINARY)[1]
# dilate the thresholded image to fill in holes, then find contours
# on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
# loop over the contours
for c in cnts:
# if the contour is too small, ignore it
if cv2.contourArea(c) < 1000:
continue
# compute the bounding box for the contour, draw it on the frame,
# and update the text
# (x, y, w, h) = cv2.boundingRect(c)
# cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)
text = "Occupied"
# draw the text and timestamp on the frame
# cv2.putText(frame, "Room Status: {}".format(text), (10, 20),
# cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
# show the frame and record if the user presses a key
cv2.imshow("Security Feed", frame)
key = cv2.waitKey(1) & 0xFF
# if the `q` key is pressed, break from the loop
if key == ord("q"):
camera.release()
cv2.destroyAllWindows()
exit()
if (firstOccupied == 0) & (text == "Occupied") & (testOccupied != "testing"):
firstOccupied = 1;
if firstOccupied == 1:
testOccupied = "testing"
testOccupiedCounter = 0
firstOccupiedFrame = gray
firstOccupied = 0;
if testOccupied == "testing":
frameDelta = cv2.absdiff(firstOccupiedFrame, gray)
thresh = cv2.threshold(frameDelta, 100, 255, cv2.THRESH_BINARY)[1]
# dilate the thresholded image to fill in holes, then find contours
# on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
temptestOccupiedCheck = "true"
# loop over the contours
for c in cnts:
# if the contour is too small, ignore it
if cv2.contourArea(c) > 500:
temptestOccupiedCheck = "false"
break
if temptestOccupiedCheck == "false":
testOccupiedCounter = 0;
testOccupied = "false";
firstOccupied = 0;
else:
testOccupiedCounter = testOccupiedCounter + 1
# if text== "Occupied":
# stablecounter = stablecounter+1
if (text == "Occupied") & (testOccupiedCounter == 50):
cv2.imshow("test", frame)
cv2.imwrite("test.jpg", frame)
with open("test.jpg", "rb") as imageFile:
f = imageFile.read()
b = bytearray(f)
imageFile.close();
data = visionAPIImg(b)
digit = isRecycable(data)
print digit
arduino.write(str(digit))
break
baseimage_change = datetime.datetime.now() + datetime.timedelta(minutes=.5)
print('starting endtime: ', endTime)
while checker:
status = 0
check, frame = video.read()
gray_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray_frame = cv2.GaussianBlur(gray_frame, (25, 25), 0)
if baseline_image is None or datetime.datetime.now() >= baseimage_change:
baseline_image = gray_frame
baseimage_change = datetime.datetime.now() + datetime.timedelta(
minutes=.5)
print('baseline_frame_changed')
continue
delta = cv2.absdiff(baseline_image, gray_frame)
threshold = cv2.threshold(delta, 25, 255, cv2.THRESH_BINARY)[1]
(contours, _) = cv2.findContours(threshold, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
font = cv2.FONT_HERSHEY_SIMPLEX
if not contours:
cv2.putText(frame, 'no motion detected!', (0, 50), font, 1,
(200, 255, 155), 2, cv2.LINE_AA)
if datetime.datetime.now() >= endTime:
print('No motion detected for last one minute: ',
datetime.datetime.now())
print('[info]:storing in database')
if status == 1:
times.append(datetime.datetime.now())
checker = False
break
#se nao foi possivel obter frame, nada mais deve ser feito
if not grabbed:
break
#converte frame para escala de cinza e aplica efeito blur (para realcar os contornos)
FrameGray = cv2.cvtColor(Frame, cv2.COLOR_BGR2GRAY)
FrameGray = cv2.GaussianBlur(FrameGray, (21, 21), 0)
#como a comparacao eh feita entre duas imagens subsequentes, se o primeiro frame eh nulo (ou seja, primeira "passada" no loop), este eh inicializado
if PrimeiroFrame is None:
PrimeiroFrame = FrameGray
continue
#ontem diferenca absoluta entre frame inicial e frame atual (subtracao de background)
#alem disso, faz a binarizacao do frame com background subtraido
FrameDelta = cv2.absdiff(PrimeiroFrame, FrameGray)
FrameThresh = cv2.threshold(FrameDelta, ThresholdBinarizacao, 255, cv2.THRESH_BINARY)[1]
#faz a dilatacao do frame binarizado, com finalidade de elimunar "buracos" / zonas brancas dentro de contornos detectados.
#Dessa forma, objetos detectados serao considerados uma "massa" de cor preta
#Alem disso, encontra os contornos apos dilatacao.
FrameThresh = cv2.dilate(FrameThresh, None, iterations=2)
cnts, _ = cv2.findContours(FrameThresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
QtdeContornos = 0
#desenha linhas de referencia
CoordenadaYLinhaEntrada = int((height / 2)-OffsetLinhasRef)
CoordenadaYLinhaSaida = int((height / 2)+OffsetLinhasRef)
cv2.line(Frame, (0,CoordenadaYLinhaEntrada), (width,CoordenadaYLinhaEntrada), (255, 0, 0), 2)
cv2.line(Frame, (0,CoordenadaYLinhaSaida), (width,CoordenadaYLinhaSaida), (0, 0, 255), 2)
break
# resize the frame to 640, convert it to grayscale, and blur it
frame = imutils.resize(frame, width=640)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (21, 21), 0)
# if the first frame is None, initialize it
if anchorFrame is None:
anchorFrame = gray
continue
# compute the absolute difference between the current frame and anchor frame
frameDelta = cv2.absdiff(
anchorFrame, gray
) # delta frames is empty for the first frame, so delta start at number 2
# update anchor frame every 10 frames
if countframe % 10 == 0:
anchorFrame = gray
thresh = cv2.threshold(frameDelta, 10, 255, cv2.THRESH_BINARY)[
1] # frame difference intensity set to be 10 or larger
# dilate the thresholded image to fill in holes, then find contours on thresholded image
thresh = cv2.dilate(thresh, None, iterations=2)
#(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
tmp1, cnts, tmp2 = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
def image_diff_dwt(lhs_image, rhs_image) -> int:
_, (lhs_LH, lhs_HL, lhs_HH) = pywt.dwt2(lhs_image, 'haar')
_, (rhs_LH, rhs_HL, rhs_HH) = pywt.dwt2(rhs_image, 'haar')
d1 = cv2.absdiff(lhs_LH, rhs_LH).sum()
d2 = cv2.absdiff(lhs_HL, rhs_HL).sum()
d3 = cv2.absdiff(lhs_HH, rhs_HH).sum()
return d1 + d2 + d3