def removeSmallContours(orig_image, fgmask, minContourArea): # initialize a mask of all 255s which will be used in removing small contours mask = np.ones(orig_image.shape[:2], dtype="uint8") * 255 contours, hier = cv2.findContours(fgmask,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE) # find contours with area < minContourArea for cnt in contours: if cv2.contourArea(cnt) < minContourArea: cv2.drawContours(mask, [cnt], -1, 0, -1) # apply fgmask on original image to extract foreground orig_masked = cv2.bitwise_and(orig_image, orig_image, mask=fgmask) # apply mask on orig_masked to hide small contours orig_masked = cv2.bitwise_and(orig_masked, orig_masked, mask=mask) # apply median blur to smoothen the image orig_masked = cv2.medianBlur(orig_masked, 5) return orig_masked, contours
def getBigContours(counter, bkgSubtr):
if config.logger.isEnabledFor(logging.DEBUG):
config.logger.debug("Finding foreground object..."),
while(1):
# print ".",
# sys.stdout.flush()
frame = counter.getNextFrame()
if frame is None:
if config.logger.isEnabledFor(logging.DEBUG):
config.logger.debug("No more frames to analyze, exiting")
sys.exit()
height, width = frame.shape[:2]
frameArea = width * height
# apply background subtraction
fgmask = bkgSubtr.getForegroundMask(frame)
# initialize a mask of all 255s which will be used in removing small contours
mask = np.ones(frame.shape[:2], dtype="uint8") * 255
contours, hier = cv2.findContours(fgmask,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
# find contour with area > contourThreshArea
contourFound = 0
boundingBoxes = {}
for cnt in contours:
ctourArea = cv2.contourArea(cnt)
if config.contourAreaThresh < ctourArea < (0.7 * frameArea):
if config.logger.isEnabledFor(logging.DEBUG):
config.logger.debug("Passing contour area: %s", ctourArea)
(x,y,w,h) = cv2.boundingRect(cnt)
if config.logger.isEnabledFor(logging.DEBUG):
config.logger.debug("bb: %s", (x,y,w,h))
newX = x + int(w * config.bbRatio)
newY = y + int(h * config.bbRatio)
newXmax = newX + int(w * config.bbRatio)
newYmax = newY + int(h * config.bbRatio)
key = str(newX) + "|" + str(newY) + "|" + str(newXmax - newX) + "|" + str(newYmax - newY)
boundingBoxes[key] = (newX, newY, newXmax - newX, newYmax - newY)
if config.logger.isEnabledFor(logging.DEBUG):
config.logger.debug("new bb: %s", (newX, newY, newXmax - newX, newYmax - newY))
contourFound = 1
else: # mask the unwanted contours
cv2.drawContours(mask, [cnt], -1, 0, -1)
if contourFound == 1:
# apply fgmask on original image to extract foreground
orig_masked = cv2.bitwise_and(frame, frame, mask=fgmask)
# apply mask on orig_masked to hide small contours
orig_masked = cv2.bitwise_and(orig_masked, orig_masked, mask=mask)
# apply median blur to smoothen the image
orig_masked = cv2.medianBlur(orig_masked, 5)
# all the coordinates are assuming top left corner as origin (0,0) and lower area positive
# cv2.circle(orig_masked,(x,y), 10, (0,255,0), -1)
# cv2.circle(orig_masked,(newX,newY), 10, (0,0,255), -1)
# cv2.rectangle(orig_masked,(x, y),(x+w, y+h),(0,255,0),3)
# cv2.rectangle(orig_masked,(newX, newY),(newXmax, newYmax),(255,0,0),3)
return (orig_masked, boundingBoxes)
# target contour not found in the video
if config.logger.isEnabledFor(logging.DEBUG):
config.logger.debug("no contour with area > contourAreaThresh found")
return (None, None)