def main():
args = get_args()
# Check for calib_images folder
if not os.path.exists('calib_images'):
print 'Please create a directory "calib_images"'
return
# Setup video display
video_disp = Display({'name': 'Video'})
# Setup controls
setup_trackbars('Controls') # , thresholds)
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame to start with
frame = video.next_frame()
global seek_callback_action
while True:
if play_or_pause == 'Play':
if not seek_callback_action:
frame = video.next_frame()
else:
frame = video.get_frame(cur_seek_pos * num_frames / 100)
seek_callback_action = False
if video.end_reached():
# Wait indefinitely if end of video reached
# Or until keypress and then exit
cv2.waitKey(0)
break
video_disp.refresh(frame)
cur_frame_num = video.get_cur_frame_num()
# Service the key events
# if s is pressed, save image
# if b is pressed, go back 1s
# if n is pressed, go ahead 1s
if video_disp.key_pressed('s'):
video_file = os.path.basename(args['video']).lower()
img_file_name = 'calib_images/{}_{}.png'.format(
video_file.strip('.mp4'), cur_frame_num)
if cv2.imwrite(img_file_name, frame):
print 'Saved', img_file_name
elif video_disp.key_pressed('n'):
seek_callback(
min((((cur_frame_num + 60) * 100) // num_frames), num_frames))
elif video_disp.key_pressed('b'):
seek_callback(max((((cur_frame_num - 60) * 100) // num_frames), 0))
# Add quitting event
if video_disp.can_quit():
break
def main():
args = get_args()
# Setup video displays
orig_video_disp = Display({'name': 'Original_Video'})
thresh_video_disp = Display({'name': 'Tresholded_Video'})
# Setup controls
setup_trackbars(controls_window_name)
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame to start with
frame = video.next_frame()
# To communicate with seek callback
global seek_callback_action
while True:
if play_or_pause == 'Play':
if not seek_callback_action:
frame = video.next_frame()
else:
frame = video.get_frame(cur_seek_pos * num_frames / 100)
seek_callback_action = False
if video.end_reached():
# Wait indefinitely if end of video reached
# Or until keypress and then exit
cv2.waitKey(0)
break
# Refresh original video display
orig_video_disp.refresh(frame)
# Get threshold values
h_min, h_max, s_min, s_max, v_min, v_max = get_thresholds(
controls_window_name)
# Convert image to HSV and apply threshold
frame_hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
frame_thresh = cv2.inRange(
frame_hsv, (h_min, s_min, v_min), (h_max, s_max, v_max))
# Refresh thresholded video display
thresh_video_disp.refresh(frame_thresh)
# Add quitting event
if orig_video_disp.can_quit() or thresh_video_disp.can_quit():
break
# On quit, save the thresholds
save_config = SaveConfig('new_thresholds', 'thresholds')
save_config.save(h_min=h_min, h_max=h_max, s_min=s_min,
s_max=s_max, v_min=v_min, v_max=v_max)
def scheme_1(file_name):
video = Video(file_name)
display = Display()
# Emulate a do-while loop
# with "while True" and breaking
# if condition fails after executing
while True:
frame = video.next_frame()
if video.end_reached():
break
# Do some operation
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# Refresh display with new image
display.refresh(gray)
if display.can_quit():
break
def main():
args = get_args()
# Read in configuration
load_config = LoadConfig('config/thresholds.npz', 'thresholds')
thresholds = load_config.load()
# Setup video displays
orig_video_disp = Display({'name': 'Original_Video'})
processed_video_disp = Display({'name': 'Processed_Video'})
# Setup controls
setup_trackbars(controls_window_name, thresholds)
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame to start with
frame = video.next_frame()
global seek_callback_action
while True:
if play_or_pause == 'Play':
if not seek_callback_action:
frame = video.next_frame()
else:
frame = video.get_frame(cur_seek_pos * num_frames / 100)
seek_callback_action = False
if video.end_reached():
# Wait indefinitely if end of video reached
# Or until keypress and then exit
cv2.waitKey(0)
break
# Refresh original video display
orig_video_disp.refresh(frame)
# Get threshold values
h_min, h_max, s_min, s_max, v_min, v_max, erode_size, dilate_size = get_params(
controls_window_name)
# Convert image to HSV and apply threshold
frame_hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
frame_thresh = cv2.inRange(frame_hsv, (h_min, s_min, v_min),
(h_max, s_max, v_max))
# Apply erosion
# Create a kernel first and then apply kernel
erode_kernel = cv2.getStructuringElement(
cv2.MORPH_ELLIPSE, (erode_size + 1, erode_size + 1))
frame_erode = cv2.erode(frame_thresh, erode_kernel)
# Apply dilate
# Create a kernel first and then apply kernel
dilate_kernel = cv2.getStructuringElement(
cv2.MORPH_ELLIPSE, (dilate_size + 1, dilate_size + 1))
frame_dilate = cv2.dilate(frame_erode, dilate_kernel)
# Refresh thresholded video display
processed_video_disp.refresh(frame_dilate)
# Add quitting event
if orig_video_disp.can_quit() or processed_video_disp.can_quit():
break
# On quit, save the params
save_config = SaveConfig('new_erode_dilate', 'erode_dilate')
save_config.save(dilate_size=dilate_size, erode_size=erode_size)
def main_worker(id, video_file, camera_model, K, D, R, T, measurements, quit_event):
# Setup video displays
video_disp = Display({'name': 'Camera_{}'.format(id)})
# Get input video
video = Video(video_file)
# Setup the undistortion stuff
if camera_model == 'P':
# Harcoded image size as
# this is a test script
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region
new_K = cv2.getOptimalNewCameraMatrix(K, D, img_size, 0.35)[0]
# Then calculate new image size according to the scaling
# Unfortunately the Python API doesn't directly provide the
# the new image size. They forgot?
new_img_size = (int(img_size[0] + (new_K[0, 2] - K[0, 2])), int(
img_size[1] + (new_K[1, 2] - K[1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1, map2 = cv2.initUndistortRectifyMap(
K, D, None, new_K, new_img_size, cv2.CV_16SC2)
elif camera_model == 'F':
# Harcoded image size
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region. The alpha
# parameter in pinhole model is equivalent to balance parameter here.
new_K = cv2.fisheye.estimateNewCameraMatrixForUndistortRectify(
K, D, img_size, np.eye(3), balance=1)
# Then calculate new image size according to the scaling
# Well if they forgot this in pinhole Python API,
# can't complain about Fisheye model. Note the reversed
# indexing here too.
new_img_size = (int(img_size[0] + (new_K[0, 2] - K[0, 2])), int(
img_size[1] + (new_K[1, 2] - K[1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1, map2 = cv2.fisheye.initUndistortRectifyMap(
K, D, np.eye(3), new_K, new_img_size_1, cv2.CV_16SC2)
# Set up foreground and background separation
fgbg = cv2.createBackgroundSubtractorMOG2()
# Averaging kernel that will be used in opening
kernel = np.ones((6, 6), np.uint8)
# Code commented out because not using
# confidence currently, but could be
# used again with changes later
# # Will be used for histogram comparison
# # (Confidence measure)
# ball_image_file = 'ball_image.jpg'
# ball_image = cv2.imread(ball_image_file)
# 2D ball detection and 3D ball tracking setup
ball_position_frame = None
ball_wc = [0, 0, 0]
while not video.end_reached() and not quit_event.value:
# Get each frame
frame = video.next_frame()
# Undistort the current frame
img_undistorted = cv2.remap(
frame, map1, map2, cv2.INTER_LINEAR, cv2.BORDER_CONSTANT)
# Convert to HSV and threshold range of ball
img_hsv = cv2.cvtColor(img_undistorted, cv2.COLOR_BGR2HSV)
mask = cv2.inRange(img_hsv, np.array(
(15, 190, 200)), np.array((25, 255, 255)))
# Foreground and background separation mask
fgmask = fgbg.apply(img_undistorted)
mask_color_bgs = cv2.bitwise_and(mask, mask, mask=fgmask)
masked_and_opened = cv2.morphologyEx(
mask_color_bgs, cv2.MORPH_OPEN, kernel)
# Hough transform to detect ball (circle)
circles = cv2.HoughCircles(masked_and_opened, cv2.HOUGH_GRADIENT, dp=3,
minDist=2500, param1=300, param2=5, minRadius=3, maxRadius=30)
if circles is not None:
# Make indexing easier and
# convert everything to int
circles = circles[0, :]
circles = np.round(circles).astype("int")
# Take only the first
# (and hopefully largest)
# circle detected
x, y, r = circles[0]
ball_position_frame = [x - r, y - r, 2 * r, 2 * r]
else:
ball_position_frame = None
# Determine the correct ball radius
mask_ball_radius = cv2.bitwise_and(fgmask, fgmask, mask=cv2.inRange(
img_hsv, np.array((10, 150, 180)), np.array((40, 255, 255))))
if ball_position_frame:
x1, y1, w1, h1 = ball_position_frame
ball_crop_temp = mask_ball_radius[(
y1 + h1 // 2 - 50):(y1 + h1 // 2 + 50), (x1 + w1 // 2 - 50):(x1 + w1 // 2 + 50)]
height, width = ball_crop_temp.shape
if height and width:
# Successfully cropped image
ball_crop = ball_crop_temp
cnts = cv2.findContours(
ball_crop.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
center = None
if len(cnts) > 0:
# contour detected
c = max(cnts, key=cv2.contourArea)
ellipse = cv2.fitEllipse(c)
width = min(ellipse[1])
ball_position_frame = [
ball_position_frame[0], ball_position_frame[1], 2 * width, 2 * width]
# Code commented out because not using
# confidence currently, but could be
# used again with changes later
# # Calculate confidence
# confidence = histogram_comparison(ball_image, img_undistorted, ball_position_frame)
# print confidence
if ball_position_frame:
x1, y1, w1, h1 = ball_position_frame
pixels_per_mm = (
K[0, 0] + K[1, 1]) / 2 / DEFAULT_FOCAL_LENGTH
z = PING_PONG_DIAMETER * \
DEFAULT_FOCAL_LENGTH / (w1 / pixels_per_mm)
x = ((x1 - K[0, 2]) /
pixels_per_mm) * z / DEFAULT_FOCAL_LENGTH
y = ((y1 - K[1, 2]) /
pixels_per_mm) * z / DEFAULT_FOCAL_LENGTH
ball_cc = np.array([x, y, z]) / 1000
ball_wc = np.dot(R.T, ball_cc - T.ravel())
# Push measurements to be processed/visualized
measurement = {
'id': id,
'frame_num': video.get_cur_frame_num(),
'ball_ic': ball_position_frame,
'ball_wc': ball_wc
}
measurements.put(measurement)
# Update video display
video_disp.refresh(img_undistorted)
# Add quitting event
if video_disp.can_quit():
break
# Setting this will signal
# the other parallel process
# to exit too.
quit_event.value = 1
def main():
args = get_args()
# Read in intrinsic calibration
load_config = LoadConfig(
'config/intrinsic_calib_{}.npz'.format(args['model'].lower()), 'calib')
calib = load_config.load()
# Read in extrinsic calibration
load_config_e = LoadConfig('config/extrinsic_calib_p_camera_1.npz',
'extrinsics')
extrinsics = load_config_e.load()
# Setup video display
video_disp = Display({'name': 'Video'})
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame; to see
# if video framework works
frame = video.next_frame()
shared_var = Array('d', [0, 0, 0])
visu = Process(target=visualize_table, args=(shared_var, ))
visu.start()
# Setup the undistortion stuff
if args['model'].upper() == 'P':
# Harcoded image size as
# this is a test script
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region
new_calib_matrix, _ = cv2.getOptimalNewCameraMatrix(
calib['camera_matrix'], calib['dist_coeffs'], img_size, 0.35)
# Then calculate new image size according to the scaling
# Unfortunately the Python API doesn't directly provide the
# the new image size. They forgot?
new_img_size = (
int(img_size[0] +
(new_calib_matrix[0, 2] - calib['camera_matrix'][0, 2])),
int(img_size[1] +
(new_calib_matrix[1, 2] - calib['camera_matrix'][1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1, map2 = cv2.initUndistortRectifyMap(calib['camera_matrix'],
calib['dist_coeffs'], None,
new_calib_matrix,
new_img_size, cv2.CV_16SC2)
elif args['model'].upper() == 'F':
# Harcoded image size
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region. The alpha
# parameter in pinhole model is equivalent to balance parameter here.
new_calib_matrix = cv2.fisheye.estimateNewCameraMatrixForUndistortRectify(
calib['camera_matrix'],
calib['dist_coeffs'],
img_size,
np.eye(3),
balance=1)
# Then calculate new image size according to the scaling
# Well if they forgot this in pinhole Python API,
# can't complain about Fisheye model. Note the reversed
# indexing here too.
new_img_size = (
int(img_size[0] +
(new_calib_matrix[0, 2] - calib['camera_matrix'][0, 2])),
int(img_size[1] +
(new_calib_matrix[1, 2] - calib['camera_matrix'][1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1, map2 = cv2.fisheye.initUndistortRectifyMap(
calib['camera_matrix'], calib['dist_coeffs'], np.eye(3),
new_calib_matrix, new_img_size, cv2.CV_16SC2)
# STUFF
corr_threshold = -1
radius_change_threshold = 5
ball_image_file = 'ball_image.jpg'
# will be used for histogram comparison
ball_image = cv2.imread(ball_image_file)
fgbg2 = cv2.createBackgroundSubtractorMOG2()
kernel = np.ones((6, 6), np.uint8)
ball_position_frame2 = None
prev_frame2 = None
# Get the rotation matrix
R = cv2.Rodrigues(extrinsics['rvec'])[0]
while not video.end_reached():
frame = video.next_frame()
img_undistorted = cv2.remap(frame, map1, map2, cv2.INTER_LINEAR,
cv2.BORDER_CONSTANT)
# STUFF
img_hsv = cv2.cvtColor(img_undistorted, cv2.COLOR_BGR2HSV)
mask2 = cv2.inRange(img_hsv, np.array((15, 190, 200)),
np.array((25, 255, 255)))
fgmask2 = fgbg2.apply(img_undistorted)
mask2_color_bgs = cv2.bitwise_and(mask2, mask2, mask=fgmask2)
frame2_hsv_bgs = cv2.bitwise_and(img_hsv,
img_hsv,
mask=mask2_color_bgs)
frame_hsv = img_hsv
frame_gray = cv2.cvtColor(img_undistorted, cv2.COLOR_BGR2GRAY)
mask = cv2.inRange(frame_hsv, np.array((10, 150, 150)),
np.array((40, 255, 255)))
open_mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)
frame_thresholded_opened_gray = cv2.bitwise_and(frame_gray,
frame_gray,
mask=open_mask)
frame_thresholded_opened_gray_smoothed = cv2.GaussianBlur(
frame_thresholded_opened_gray, (11, 11), 0)
# opening
a = cv2.inRange(frame_thresholded_opened_gray_smoothed, 10, 256)
b = cv2.morphologyEx(mask2_color_bgs, cv2.MORPH_OPEN, kernel)
circles = cv2.HoughCircles(b,
cv2.HOUGH_GRADIENT,
dp=3,
minDist=2500,
param1=300,
param2=5,
minRadius=3,
maxRadius=30)
if circles is not None:
# convert the (x, y) coordinates and radius of the circles to integers
circles = np.round(circles[0, :]).astype("int")
x, y, r = circles[0]
ball_position_frame2 = [x - r, y - r, 2 * r, 2 * r]
# loop over the (x, y) coordinates and radius of the circles
else:
ball_position_frame2 = None
frame2 = img_undistorted
mask2_ball_radius = cv2.bitwise_and(fgmask2,
fgmask2,
mask=cv2.inRange(
img_hsv,
np.array((10, 150, 180)),
np.array((40, 255, 255))))
if ball_position_frame2 != None:
x2, y2, w2, h2 = ball_position_frame2
ball_crop_temp = mask2_ball_radius[(y2 + h2 / 2 -
30):(y2 + h2 / 2 + 30),
(x2 + w2 / 2 -
30):(x2 + w2 / 2 + 30)]
ball_crop_color = frame2[(y2 + h2 / 2 - 30):(y2 + h2 / 2 + 30),
(x2 + w2 / 2 - 30):(x2 + w2 / 2 + 30)]
height, width = ball_crop_temp.shape
else:
ball_crop_temp = []
height = 0
width = 0
if height != 0 and width != 0:
ball_crop = ball_crop_temp
cnts = cv2.findContours(ball_crop.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)[-2]
center = None
if len(cnts) > 0 and ball_position_frame2:
c = max(cnts, key=cv2.contourArea)
rect = cv2.minAreaRect(c)
width, height = rect[1]
box = cv2.boxPoints(rect)
box = np.int0(box)
cv2.drawContours(ball_crop_color, [box], 0, (0, 0, 255), 2)
ball_position_frame2 = [
ball_position_frame2[0], ball_position_frame2[1],
min(width, height),
min(width, height)
]
prev_frame2 = frame2
# print ball_position_frame2
if ball_position_frame2:
x2, y2, w2, h2 = ball_position_frame2
# x = (x2 - 960) / PIXELS_PER_MM
# y = (y2 - 540) / PIXELS_PER_MM
pixels_per_mm = (new_calib_matrix[0, 0] +
new_calib_matrix[1, 1]) / 2 / FOCAL_LENGTH
z = PING_PONG_DIAMETER * FOCAL_LENGTH / (w2 / pixels_per_mm)
x = ((x2 - new_calib_matrix[0, 2]) /
pixels_per_mm) * z / FOCAL_LENGTH
y = ((y2 - new_calib_matrix[1, 2]) /
pixels_per_mm) * z / FOCAL_LENGTH
ball_cc = np.array([x, y, z]) / 1000
#ball_wc = np.dot(R.T, extrinsics['tvec'].ravel() - ball_cc)
ball_wc = np.dot(R.T, ball_cc - extrinsics['tvec'].ravel())
print 'Ball IC', np.array([x2, y2]), 'Dia', w2
print 'Ball CC', ball_cc
print 'Ball WC', ball_wc
shared_var[0] = ball_wc[0]
shared_var[1] = ball_wc[1]
shared_var[2] = ball_wc[2]
# Update GUI with new image
video_disp.refresh(frame2)
#print "Pixels", ball_position_frame2
# Add quitting event
if video_disp.can_quit():
break
global main_process_end_reached
main_process_end_reached = True
visu.join()
def main():
args = get_args()
# Read in intrinsic calibration
load_config_1 = LoadConfig(
'config/intrinsic_calib_{}_camera_1.npz'.format(args['model'].lower()),
'calib_camera_1')
intrinsic_1 = load_config_1.load()
K_1 = intrinsic_1['camera_matrix']
D_1 = intrinsic_1['dist_coeffs']
load_config_2 = LoadConfig(
'config/intrinsic_calib_{}_camera_2.npz'.format(args['model'].lower()),
'calib_camera_2')
intrinsic_2 = load_config_2.load()
K_2 = intrinsic_2['camera_matrix']
D_2 = intrinsic_2['dist_coeffs']
# Read in extrinsic calibration
load_config_e_1 = LoadConfig(
'config/extrinsic_calib_{}_camera_1.npz'.format(args['model'].lower()),
'extrinsic_camera_1')
extrinsic_1 = load_config_e_1.load()
R_1 = cv2.Rodrigues(extrinsic_1['rvec'])[0]
T_1 = extrinsic_1['tvec']
load_config_e_2 = LoadConfig(
'config/extrinsic_calib_{}_camera_2.npz'.format(args['model'].lower()),
'extrinsic_camera_2')
extrinsic_2 = load_config_e_2.load()
R_2 = cv2.Rodrigues(extrinsic_2['rvec'])[0]
T_2 = extrinsic_2['tvec']
# Setup video displays
video_disp_1 = Display({'name': 'Camera_1'})
video_disp_2 = Display({'name': 'Camera_2'})
# Get input video
video_1 = Video(args['video_1'])
video_2 = Video(args['video_2'])
# Get the first frame; to see
# if video framework works
frame_1 = video_1.next_frame()
frame_2 = video_2.next_frame()
# Original code was to multiprocess, but
# found MACOSX doesn't like forking processes
# with GUIs. However, retaining Array from
# multiproc for future.
shared_var = Array('d', [0, 0, 0])
end_reached = Value('b', False)
# visu = Process(target=visualize_table, args=(shared_var,))
# visu.start()
visu = Thread(target=visualize_table, args=(end_reached, shared_var))
visu.daemon = True
visu.start()
# Setup the undistortion stuff
if args['model'].upper() == 'P':
# Harcoded image size as
# this is a test script
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region
new_K_1 = cv2.getOptimalNewCameraMatrix(K_1, D_1, img_size, 0.35)[0]
new_K_2 = cv2.getOptimalNewCameraMatrix(K_2, D_2, img_size, 0.35)[0]
# Then calculate new image size according to the scaling
# Unfortunately the Python API doesn't directly provide the
# the new image size. They forgot?
new_img_size_1 = (int(img_size[0] + (new_K_1[0, 2] - K_1[0, 2])),
int(img_size[1] + (new_K_1[1, 2] - K_1[1, 2])))
new_img_size_2 = (int(img_size[0] + (new_K_2[0, 2] - K_2[0, 2])),
int(img_size[1] + (new_K_2[1, 2] - K_2[1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1_1, map2_1 = cv2.initUndistortRectifyMap(K_1, D_1, None, new_K_1,
new_img_size_1,
cv2.CV_16SC2)
map1_2, map2_2 = cv2.initUndistortRectifyMap(K_2, D_2, None, new_K_2,
new_img_size_2,
cv2.CV_16SC2)
elif args['model'].upper() == 'F':
# Harcoded image size
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region. The alpha
# parameter in pinhole model is equivalent to balance parameter here.
new_K_1 = cv2.fisheye.estimateNewCameraMatrixForUndistortRectify(
K_1, D_1, img_size, np.eye(3), balance=1)
new_K_2 = cv2.fisheye.estimateNewCameraMatrixForUndistortRectify(
K_2, D_2, img_size, np.eye(3), balance=1)
# Then calculate new image size according to the scaling
# Well if they forgot this in pinhole Python API,
# can't complain about Fisheye model. Note the reversed
# indexing here too.
new_img_size_1 = (int(img_size[0] + (new_K_1[0, 2] - K_1[0, 2])),
int(img_size[1] + (new_K_1[1, 2] - K_1[1, 2])))
new_img_size_2 = (int(img_size[0] + (new_K_2[0, 2] - K_2[0, 2])),
int(img_size[1] + (new_K_2[1, 2] - K_2[1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1_1, map2_1 = cv2.fisheye.initUndistortRectifyMap(
K_1, D_1, np.eye(3), new_K_1, new_img_size_1, cv2.CV_16SC2)
map1_2, map2_2 = cv2.fisheye.initUndistortRectifyMap(
K_2, D_2, np.eye(3), new_K_2, new_img_size_2, cv2.CV_16SC2)
# STUFF
corr_threshold = -1
radius_change_threshold = 5
ball_image_file = 'ball_image.jpg'
# will be used for histogram comparison
ball_image = cv2.imread(ball_image_file)
fgbg1 = cv2.createBackgroundSubtractorMOG2()
fgbg2 = cv2.createBackgroundSubtractorMOG2()
kernel = np.ones((6, 6), np.uint8)
ball_position_frame1 = None
ball_position_frame2 = None
prev_frame1 = None
prev_frame2 = None
ball_wc = [0, 0, 0]
while not video_1.end_reached() and not video_2.end_reached():
frame_1 = video_1.next_frame()
frame_2 = video_2.next_frame()
img_undistorted_1 = cv2.remap(frame_1, map1_1, map2_1,
cv2.INTER_LINEAR, cv2.BORDER_CONSTANT)
img_undistorted_2 = cv2.remap(frame_2, map1_2, map2_2,
cv2.INTER_LINEAR, cv2.BORDER_CONSTANT)
# STUFF1
frame_1_hsv = cv2.cvtColor(img_undistorted_1, cv2.COLOR_BGR2HSV)
mask1 = cv2.inRange(frame_1_hsv, np.array((15, 190, 200)),
np.array((25, 255, 255)))
fgmask1 = fgbg1.apply(img_undistorted_1)
mask1_color_bgs = cv2.bitwise_and(mask1, mask1, mask=fgmask1)
frame1_hsv_bgs = cv2.bitwise_and(frame_1_hsv,
frame_1_hsv,
mask=mask1_color_bgs)
# opening
b1 = cv2.morphologyEx(mask1_color_bgs, cv2.MORPH_OPEN, kernel)
circles1 = cv2.HoughCircles(b1,
cv2.HOUGH_GRADIENT,
dp=3,
minDist=2500,
param1=300,
param2=5,
minRadius=3,
maxRadius=30)
if circles1 is not None:
# convert the (x, y) coordinates and radius of the circles to integers
circles1 = np.round(circles1[0, :]).astype("int")
x, y, r = circles1[0]
ball_position_frame1 = [x - r, y - r, 2 * r, 2 * r]
# loop over the (x, y) coordinates and radius of the circles
else:
ball_position_frame1 = None
mask_ball_radius1 = cv2.bitwise_and(fgmask1,
fgmask1,
mask=cv2.inRange(
frame_1_hsv,
np.array((10, 150, 180)),
np.array((40, 255, 255))))
# determine the correct radius
if ball_position_frame1 != None:
x1, y1, w1, h1 = ball_position_frame1
ball_crop_temp1 = mask_ball_radius1[(y1 + h1 // 2 -
30):(y1 + h1 // 2 + 30),
(x1 + w1 // 2 -
30):(x1 + w1 // 2 + 30)]
height, width = ball_crop_temp1.shape
if height != 0 and width != 0:
# successfully cropped image
ball_crop1 = ball_crop_temp1
cnts = cv2.findContours(ball_crop1.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)[-2]
center = None
if len(cnts) > 0:
# contour detected
c = max(cnts, key=cv2.contourArea)
rect = cv2.minAreaRect(c)
width, height = rect[1]
ball_position_frame1 = [
ball_position_frame1[0], ball_position_frame1[1],
min(width, height),
min(width, height)
]
prev_frame1 = img_undistorted_1
if ball_position_frame1:
x1, y1, w1, h1 = ball_position_frame1
pixels_per_mm = (K_1[0, 0] + K_1[1, 1]) / 2 / FOCAL_LENGTH
z = PING_PONG_DIAMETER * FOCAL_LENGTH / (w1 / pixels_per_mm)
x = ((x1 - K_1[0, 2]) / pixels_per_mm) * z / FOCAL_LENGTH
y = ((y1 - K_1[1, 2]) / pixels_per_mm) * z / FOCAL_LENGTH
ball_cc1 = np.array([x, y, z]) / 1000
ball_wc1 = np.dot(R_1.T, ball_cc1 - T_1.ravel())
# STUFF2
frame_2_hsv = cv2.cvtColor(img_undistorted_2, cv2.COLOR_BGR2HSV)
mask2 = cv2.inRange(frame_2_hsv, np.array((15, 190, 200)),
np.array((25, 255, 255)))
fgmask2 = fgbg2.apply(img_undistorted_2)
mask2_color_bgs = cv2.bitwise_and(mask2, mask2, mask=fgmask2)
frame2_hsv_bgs = cv2.bitwise_and(frame_2_hsv,
frame_2_hsv,
mask=mask2_color_bgs)
# opening
b2 = cv2.morphologyEx(mask2_color_bgs, cv2.MORPH_OPEN, kernel)
circles2 = cv2.HoughCircles(b2,
cv2.HOUGH_GRADIENT,
dp=3,
minDist=2500,
param1=300,
param2=5,
minRadius=3,
maxRadius=30)
if circles2 is not None:
# convert the (x, y) coordinates and radius of the circles to integers
circles2 = np.round(circles2[0, :]).astype("int")
x, y, r = circles2[0]
ball_position_frame2 = [x - r, y - r, 2 * r, 2 * r]
# loop over the (x, y) coordinates and radius of the circles
else:
ball_position_frame2 = None
mask_ball_radius2 = cv2.bitwise_and(fgmask2,
fgmask2,
mask=cv2.inRange(
frame_2_hsv,
np.array((10, 150, 180)),
np.array((40, 255, 255))))
# determine the correct radius
if ball_position_frame2 != None:
x2, y2, w2, h2 = ball_position_frame2
ball_crop_temp2 = mask_ball_radius2[(y2 + h2 // 2 -
30):(y2 + h2 // 2 + 30),
(x2 + w2 // 2 -
30):(x2 + w2 // 2 + 30)]
height, width = ball_crop_temp2.shape
if height != 0 and width != 0:
# successfully cropped image
ball_crop2 = ball_crop_temp2
cnts = cv2.findContours(ball_crop2.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)[-2]
center = None
if len(cnts) > 0:
# contour detected
c = max(cnts, key=cv2.contourArea)
rect = cv2.minAreaRect(c)
width, height = rect[1]
ball_position_frame2 = [
ball_position_frame2[0], ball_position_frame2[1],
min(width, height),
min(width, height)
]
prev_frame2 = img_undistorted_2
if ball_position_frame2:
x2, y2, w2, h2 = ball_position_frame2
pixels_per_mm = (K_2[0, 0] + K_2[1, 1]) / 2 / FOCAL_LENGTH
z = PING_PONG_DIAMETER * FOCAL_LENGTH / (w2 / pixels_per_mm)
x = ((x2 - K_2[0, 2]) / pixels_per_mm) * z / FOCAL_LENGTH
y = ((y2 - K_2[1, 2]) / pixels_per_mm) * z / FOCAL_LENGTH
ball_cc2 = np.array([x, y, z]) / 1000
ball_wc2 = np.dot(R_2.T, ball_cc2 - T_2.ravel())
# Additional rotation for absolute coordinates
R = np.array([[-1, 0, 0], [0, -1, 0], [0, 0, 1]])
ball_wc2 = np.dot(R, ball_wc2)
# Combine STUFF1 and STUFF2
# If positions from either available,
# update as predicted from either.
# But if both available, predict the average
if ball_position_frame1 and ball_position_frame2:
ball_wc = (ball_wc1 + ball_wc2) / 2
elif ball_position_frame1:
ball_wc = ball_wc1
elif ball_position_frame2:
ball_wc = ball_wc2
# print 'Ball IC', np.array([x2, y2]), 'Dia', w2
# print 'Ball CC', ball_cc2
# print 'Ball WC', ball_wc2
shared_var[0] = ball_wc[0]
shared_var[1] = ball_wc[1]
shared_var[2] = ball_wc[2]
# Update GUI with new image
video_disp_1.refresh(img_undistorted_1)
video_disp_2.refresh(img_undistorted_2)
# Add quitting event
if video_disp_2.can_quit() or video_disp_1.can_quit():
break
end_reached.value = True
visu.join()
def main():
args = get_args()
# Read in configuration
load_config = LoadConfig('new_calib_{}.npz'.format(args['model'].lower()),
'calib')
calib = load_config.load()
# Setup video displays
video_disp = Display({'name': 'Video'})
# Setup controls
setup_trackbars('Controls') # , thresholds)
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame to start with
frame = video.next_frame()
global seek_callback_action
while True:
if play_or_pause == 'Play':
if not seek_callback_action:
frame = video.next_frame()
else:
frame = video.get_frame(cur_seek_pos * num_frames / 100)
seek_callback_action = False
if video.end_reached():
# Wait indefinitely if end of video reached
# Or until keypress and then exit
cv2.waitKey(0)
break
# Undistort according to pinhole model
if args['model'].upper() == 'P':
# Make sure distortion coeffecients
# follow pinhole model
if calib['dist_coeffs'].shape[1] != 5:
print 'Input configuration probably not pinhole'
return
# Harcoded image size as
# this is a test script
img_size = (1920, 1080)
# First create scaled intrinsics because we will undistort
# into region beyond original image region
new_calib_matrix, _ = cv2.getOptimalNewCameraMatrix(
calib['camera_matrix'], calib['dist_coeffs'], img_size, 0.35)
# Then calculate new image size according to the scaling
# Unfortunately the Python API doesn't directly provide the
# the new image size. They forgot?
new_img_size = (
int(img_size[0] +
(new_calib_matrix[0, 2] - calib['camera_matrix'][0, 2])),
int(img_size[1] +
(new_calib_matrix[1, 2] - calib['camera_matrix'][1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1, map2 = cv2.initUndistortRectifyMap(calib['camera_matrix'],
calib['dist_coeffs'],
None, new_calib_matrix,
new_img_size,
cv2.CV_16SC2)
img_undistorted = cv2.remap(frame, map1, map2, cv2.INTER_LINEAR,
cv2.BORDER_CONSTANT)
# Undistort according to fisheye model
elif args['model'].upper() == 'F':
# Make sure distortion coeffecients
# follow fisheye model
if calib['dist_coeffs'].shape[0] != 4:
print 'Input configuration probably not fisheye'
return
# Harcoded image size as
# this is a test script.
# As already ranted before
# someone messed with the image
# size indexing and reversed it.
img_size = (1920, 1080)
# Also, the basic undistortion DOES NOT work
# with the fisheye module
# img_undistorted = cv2.fisheye.undistortImage(
# frame, calib['camera_matrix'], calib['dist_coeffs'])
# First create scaled intrinsics because we will undistort
# into region beyond original image region. The alpha
# parameter in pinhole model is equivalent to balance parameter here.
new_calib_matrix = cv2.fisheye.estimateNewCameraMatrixForUndistortRectify(
calib['camera_matrix'],
calib['dist_coeffs'],
img_size,
np.eye(3),
balance=1)
# Then calculate new image size according to the scaling
# Well if they forgot this in pinhole Python API,
# can't complain about Fisheye model. Note the reversed
# indexing here too.
new_img_size = (
int(img_size[0] +
(new_calib_matrix[0, 2] - calib['camera_matrix'][0, 2])),
int(img_size[1] +
(new_calib_matrix[1, 2] - calib['camera_matrix'][1, 2])))
# Standard routine of creating a new rectification
# map for the given intrinsics and mapping each
# pixel onto the new map with linear interpolation
map1, map2 = cv2.fisheye.initUndistortRectifyMap(
calib['camera_matrix'], calib['dist_coeffs'], np.eye(3),
new_calib_matrix, new_img_size, cv2.CV_16SC2)
img_undistorted = cv2.remap(frame, map1, map2, cv2.INTER_LINEAR,
cv2.BORDER_CONSTANT)
# Update GUI with new image
video_disp.refresh(img_undistorted)
# Service the s key to save image
if video_disp.key_pressed('s'):
cur_frame_num = video.get_cur_frame_num()
orig_img_file_name = 'image_for_markers_orig.png'
undistorted_img_file_name = 'image_for_markers_undistorted.png'
if cv2.imwrite(orig_img_file_name, frame):
print 'Saved original {} at frame {}'.format(
orig_img_file_name, cur_frame_num)
if cv2.imwrite(undistorted_img_file_name, img_undistorted):
print 'Saved undistorted {} at frame {}'.format(
undistorted_img_file_name, cur_frame_num)
# Add quitting event
if video_disp.can_quit():
break
def main():
args = get_args()
# Read in configuration
# load_config = LoadConfig('config/thresholds.npz', 'thresholds')
# thresholds = load_config.load()
# Setup video displays
video_disp = Display({'name': 'Video'})
# Setup controls
setup_trackbars('Controls') # , thresholds)
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame to start with
frame = video.next_frame()
global seek_callback_action
# Deck for storing calib images
calib_img_deck = deque(maxlen=MAX_NUM_IMAGES_FOR_CALIB)
# Deck for storing charuco info
charuco_corners_deck = deque(maxlen=MAX_NUM_IMAGES_FOR_CALIB)
charuco_ids_deck = deque(maxlen=MAX_NUM_IMAGES_FOR_CALIB)
skip_count = 0
test_camera_matrix = np.array([
[11096.77, 0, 540],
[0, 11096.77, 960],
[0, 0, 1]
])
while True:
if play_or_pause == 'Play':
if not seek_callback_action:
frame = video.next_frame()
else:
frame = video.get_frame(cur_seek_pos * num_frames / 100)
seek_callback_action = False
if video.end_reached():
# Wait indefinitely if end of video reached
# Or until keypress and then exit
cv2.waitKey(0)
break
corners, ids, rejected_img_points = cv2.aruco.detectMarkers(
frame, dictionary)
if ids is not None:
img_markers = cv2.aruco.drawDetectedMarkers(frame, corners, ids)
num_charuco, charuco_corners, charuco_ids = cv2.aruco.interpolateCornersCharuco(
corners, ids, frame, board, cameraMatrix=test_camera_matrix)
if charuco_corners is not None:
img_markers = cv2.aruco.drawDetectedCornersCharuco(
img_markers, charuco_corners, charuco_ids)
if ids.shape[0] == MAX_ARUCO_IDS \
and num_charuco == MAX_CHARUCO_IDS \
and skip_count % 15 == 0:
calib_img_deck.append(frame)
charuco_corners_deck.append(charuco_corners)
charuco_ids_deck.append(charuco_ids)
else:
img_markers = frame
cv2.putText(img_markers, '{}/{}'.format(len(calib_img_deck), MAX_NUM_IMAGES_FOR_CALIB),
(200, 100), cv2.FONT_HERSHEY_COMPLEX, 2, (0, 0, 255), 5)
video_disp.refresh(img_markers)
if video_disp.key_pressed('s'):
pass
skip_count = skip_count + 1
# Add quitting event
if video_disp.can_quit():
break
# On quit, save the params
# save_config = SaveConfig('new_erode_dilate', 'erode_dilate')
# save_config.save(dilate_size=dilate_size, erode_size=erode_size)
img_size = calib_img_deck[0].shape[:2]
# print charuco_ids_deck
# error, camera_matrix, dist_coeffs = cv2.aruco.calibrateCameraCharuco(
# charuco_corners_deck, charuco_ids_deck, board, img_size, test_camera_matrix, None)[:3]
objPoints, imgPoints = [board.chessboardCorners.reshape(
1, -1, 3)] * len(charuco_corners_deck), charuco_corners_deck
calibration_flags = cv2.fisheye.CALIB_USE_INTRINSIC_GUESS + \
cv2.fisheye.CALIB_FIX_PRINCIPAL_POINT + cv2.fisheye.CALIB_FIX_SKEW
error, camera_matrix, dist_coeffs = cv2.fisheye.calibrate(
objPoints, imgPoints, img_size, test_camera_matrix, np.zeros(4), flags=calibration_flags)[:3]
print error, camera_matrix
save_config = SaveConfig('new_calib', 'calib')
save_config.save(camera_matrix=camera_matrix, dist_coeffs=dist_coeffs)
def main():
args = get_args()
# Setup video displays
orig_video_disp = Display({'name': 'Original_Video'})
thresh_video_disp = Display({'name': 'Thresholded_Video'})
mean_shift_video_disp = Display({'name': 'Mean-Shift Tracking Video'})
# Setup controls
setup_trackbars(controls_window_name)
# Get input video
video = Video(args['video'])
num_frames = video.get_num_frames()
# Get the first frame to start with
frame = video.next_frame()
global seek_callback_action
# setup initial location of window
top, length, left, width = 450, 36, 1000, 43 # simply hardcoded the values
track_window = (left, top, width, length)
# set up the ROI for tracking
roi = frame[top:top + length, left:left + width]
hsv_roi = cv2.cvtColor(roi, cv2.COLOR_BGR2HSV)
h_min, h_max, s_min, s_max, v_min, v_max = get_thresholds(
controls_window_name)
mask = cv2.inRange(hsv_roi, np.array(
(h_min, s_min, v_min)), np.array((h_max, s_max, v_max)))
roi_hist = cv2.calcHist([hsv_roi], [0], mask, [180], [0, 180])
cv2.normalize(roi_hist, roi_hist, 0, 255, cv2.NORM_MINMAX)
# Setup the termination criteria, either 10 iteration or move by at least 1 pt
term_criteria = (cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 20, 1)
while True:
if play_or_pause == 'Play':
if not seek_callback_action:
frame = video.next_frame()
else:
frame = video.get_frame(cur_seek_pos * num_frames / 100)
seek_callback_action = False
if video.end_reached():
# Wait indefinitely if end of video reached
# Or until keypress and then exit
cv2.waitKey(0)
break
# Refresh original video display
orig_video_disp.refresh(frame)
# Get threshold values
h_min, h_max, s_min, s_max, v_min, v_max = get_thresholds(
controls_window_name)
# Convert image to HSV and apply threshold
frame_hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
frame_thresh = cv2.inRange(
frame_hsv, (h_min, s_min, v_min), (h_max, s_max, v_max))
# threshold image in hsv domain
frame_hsv_threshold = cv2.bitwise_and(
frame_hsv, frame_hsv, mask=frame_thresh)
# Refresh thresholded video display
thresh_video_disp.refresh(frame_hsv_threshold)
# Find the backprojection of the histogram
dst = cv2.calcBackProject([frame_hsv_threshold], [
0], roi_hist, [0, 180], 1)
# apply meanshift to get the new location
ret, track_window = cv2.meanShift(dst, track_window, term_criteria)
# Draw it on image
x, y, w, h = track_window
frame_mean_shift = cv2.rectangle(frame, (x, y), (x + w, y + h), 255, 2)
# Refresh mean shift tracking video display
mean_shift_video_disp.refresh(frame_mean_shift)