class Gestures():
def __init__(self):
#initialize the webcam
self.camera = Webcam()
#initialize optical debug mode
self.opticalDebugMode = False
#initialize the video window
cv2.namedWindow(WIN_NAME, cv2.CV_WINDOW_AUTOSIZE)
#start collecting frames for processing
self.init_frames()
#initialize Algorithm instance
self.alg = Algorithm()
def start(self):
"""Runs image processing loop"""
while True:
self.update_frames()
self.alg.calc_scroll(self.dir)
key = cv2.waitKey(4)
if key == 27:
#Quit if the user presses ESC
self.stop_gui()
break
if key == 104:
#Toggle hand tracking debug window
self.tracker.toggle_debug()
if key == 111:
#Toggle optical flow debug window
self.toggle_optical()
def init_frames(self):
"""Initiates camera and 3 initial frames for processing"""
#Get 3 successive frames for difference calculation
self.img0, self.frame0 = self.camera.get_frame_bgr_and_gray()
self.img1, self.frame1 = self.camera.get_frame_bgr_and_gray()
self.img2, self.frame2 = self.camera.get_frame_bgr_and_gray()
if USE_HANDTRACKING:
self.init_hand_tracking()
def init_hand_tracking(self):
"""Initiates hand tracker and 3 initial bounding point sets"""
#Initialize hand tracker
self.tracker = HandTracking()
#Get 3 successive sets of features (feature: array of points)
#Each feature array is based on a frame's bounding boxes
#and indicates the areas for which optical flow should be calculated
self.mpts0 = self.tracker.getBoundingMidpoints(self.img0)
self.mpts1 = self.tracker.getBoundingMidpoints(self.img1)
self.mpts2 = self.tracker.getBoundingMidpoints(self.img2)
def update_frames(self):
"""Updates frames for next loop of processing"""
#Store old frames
self.img0 = self.img1
self.frame0 = self.frame1
#Get new rgb and grayscale frames
self.img1, self.frame1 = self.camera.get_frame_bgr_and_gray()
#Update bounding arrays
if USE_HANDTRACKING:
self.mpts0 = self.mpts1
self.mpts1 = self.tracker.getBoundingMidpoints(self.img1)
self.dir = self.get_direction_vector()
self.tracker.update_dir_data(self.dir)
if self.opticalDebugMode: #Show optical flow field
flow = cv2.calcOpticalFlowFarneback(self.frame0, self.frame1,
PYR_SCALE, LEVELS, WINSIZE,
ITER, POLY_N, POLY_SIGMA,
FLAGS)
image = self.alg.create_flow(
self.frame1, flow, 10) #create the flow overlay for display
self.show_image(image)
def get_direction_vector(self):
"""Calculates the vector between the midpoints of bounding boxes."""
if len(self.mpts0) is not len(self.mpts1):
return
diffs = []
for i in range(len(self.mpts0)):
mp0 = self.mpts0[i]
mp1 = self.mpts1[i]
diff = [a - b for a, b in zip(mp1, mp0)]
dist = np.sqrt(diff[0]**2 + diff[1]**2)
diffs.append((diff, dist))
if len(diffs) == 0:
return None
return max(diffs, key=lambda a: a[1])
def toggle_optical(self):
"""Toggle the optical flow debug window"""
self.opticalDebugMode = not self.opticalDebugMode
if self.opticalDebugMode:
#initialize the video window
cv2.namedWindow(WIN_NAME, cv2.CV_WINDOW_AUTOSIZE)
else:
self.stop_gui()
def show_image(self, img):
"""Show a GUI with the webcam feed for debugging purposes"""
cv2.imshow(WIN_NAME, img)
def stop_gui(self):
"""Stop the webcam"""
cv2.destroyWindow(WIN_NAME)