bfinder.background_init = 60
bfinder.threshold = 10
bfinder.gaussian_blur = 1
bfinder.skip_filter_contours = False
# Parameter for croping
xmargin = [0, 0]
ymargin = [0, 0]
# Build up an expected histogram
ninitframes = 50
bin_edges = np.arange(1, 255)
ref_cumsum = np.zeros(bin_edges.shape[0] - 1)
for _ in range(bfinder.background_init):
# grab the current frame and initialize the occupied/unoccupied
frame = btgen.randomblob(background, color=blob_brightness[0])[0]
bfinder.run(frame)
for _ in range(ninitframes):
# grab the current frame and initialize the occupied/unoccupied
frame = btgen.randomblob(background, color=blob_brightness[0])[0]
bfinder.run(frame)
#
image = bfinder.original_image
image = image.astype(np.float)
image[bfinder.processed_image == 0] = np.nan
croped_images = crop(image, bfinder.filtered_contours, xmargin, ymargin)
hist = btfilt.image2hist(croped_images[0], bin_edges)
cumsum = btfilt.hist2cumsum(hist)
ref_cumsum += cumsum / ninitframes
frame_i = 0
if not grabbed:
break
# Calculate normed difference
frame = frame[..., 0]
diff = frame - background
diff = np.abs(diff)
diff -= diff.min()
diff /= diff.max()
diff *= 255
diff = 255 - diff.astype(np.uint8)
# for display
image = bfimshow.bw2color(diff.copy())
#Find contours
bmarker.run(diff)
# Assign center of mass ellipses
ellipses = list()
for con in bmarker.contours:
moments = cv2.moments(con)
if moments['m00'] <= 0:
continue
cx = moments['m10'] / moments['m00']
cy = moments['m01'] / moments['m00']
ellipse = Ellipse(x=cx, y=cy, height=5, width=5)
ellipses.append(ellipse)
btio.append(file, frame_i, ellipses)
# display images when asked
if (time.time() - t_start) > refresh_time:
while True:
frame_i += 1
# grab the current frame and initialize the occupied/unoccupied
(grabbed, frame) = camera.read()
# if the frame could not be grabbed, then we have reached the end
# of the video
if not grabbed:
break
# Apply background
frame=frame[...,0]
diff = background.apply(frame.copy(),
learningRate=0)#frame-background
#Find largest contours
bbee.run(diff)
area = 0
cont = []
for con in bbee.contours:
moments = cv2.moments(con)
if moments['m00'] <=0:
continue
if moments['m00']>area:
area = moments['m00']
cont = con
cx_bee = moments['m10']/moments['m00']
cy_bee = moments['m01']/moments['m00']
if len(cont) < 5:
continue
ell = Ellipse()
class TestBlobFinder(unittest.TestCase):
def __init__(self, *args, **kwargs):
super(TestBlobFinder, self).__init__(*args, **kwargs)
self.run_ntimes_oneblob()
def run_ntimes_oneblob(self):
"""
Run the BlobFinder on one blob
"""
self.background = background()
self.mask = mask(self.background)
blob = randomblob(self.background)[0]
self.bfinder = BlobFinder(self.mask)
# init background
for _ in range(10):
blob = randomblob(self.background)[0]
self.bfinder.run(blob)
self.blobs_center = list()
self.blobs_found = list()
for frame_i in range(10):
blob, blob_center, _, _ = \
randomblob(self.background)
self.bfinder.run(blob)
self.blobs_center.append(blob_center)
self.blobs_found.append(self.bfinder.filtered_contours)
self.blob = blob
def test_setget_image(self):
myim = self.bfinder.original_image
self.assertTrue(np.allclose(myim, self.blob))
def test_masked(self):
myim = self.bfinder.masked_image
testimage = cv2.bitwise_and(self.blob, self.blob, mask=self.mask)
self.assertTrue(np.allclose(myim, testimage))
def test_segmented(self):
myim = self.bfinder.segmented_image
self.assertTrue(np.allclose(np.unique(myim), [0, 255]))
def test_oneblob_numbers(self):
condition = True
for frame_i in range(len(self.blobs_center)):
contours = self.blobs_found[frame_i]
if len(contours) > 1:
print('Too many blobs detected')
condition = False
continue
elif len(contours) < 1:
print('Too few blobs detected')
condition = False
continue
self.assertTrue(condition)
def test_oneblob_positions(self):
condition = True
for frame_i in range(len(self.blobs_center)):
contours = self.blobs_found[frame_i]
contours = contours[0]
px_error = np.sqrt(
np.sum(contours.center - self.blobs_center[frame_i])**2)
if px_error > 1:
print('Pixel error too large : {}'.format(px_error))
condition = False
self.assertTrue(condition)