def run_image_processing_v2_depth(test_all, depth_cal, show=True):
# Prepare Calibration Values
empt_img, cali_img, crop, depth_shape, empt_sort = depth_cal
#Prepare Test Images
test, test_depth, test_ir = test_all
test_img2 = vt.crop_out(test_depth, crop)
test_img = vt.clean_image(test_img2)
if show:
plt.figure()
plt.imshow(test_depth)
plt.show()
normclean = vt.create_normalised_image(test_img, empt_img, cali_img)
if show:
plt.figure("Normclean")
plt.imshow((normclean * 255).astype('uint8'))
plt.show()
normclean_uint8 = (normclean * 255).astype('uint8')
edged, img, cnts, hierarchy = vt.extract_contours(normclean_uint8,
min_thresh=20,
max_thresh=90,
dilate=2,
erode=1)
new_cnts, new_h = kv.extract_depth_contours(cnts,
hierarchy,
normclean * 255,
minsize=200,
show=show)
if len(new_cnts) > 0:
family = kv.create_family(new_cnts, new_h)
sorted_family = kv.sort_family(family,
normclean,
show=False,
debug=False)
if show:
plt.figure("OBJECT SEGMENTATION")
mask = np.zeros_like(normclean)
for member in sorted_family:
if member['generation'] == 0:
obj_mask = vt.create_contour_mask([member['contour'][0]],
normclean)
mask = mask + obj_mask
plt.imshow(mask)
plt.show()
print "Depth Done"
else:
sorted_family = None
return normclean, sorted_family
def run_image_processing_v2_rgb(test_all, rgb_cal, depth_cal, show=True):
# Prepare the images
test, test_depth, test_ir = test_all
# Prepare rgb values
empt_rgb_img, rgb_crop = rgb_cal
test = cv2.resize(test, (0, 0), fx=0.5, fy=0.5)
test_rgb_large_img = vt.crop_out(test, rgb_crop)
depth_shape = depth_cal[3]
test_rgbx_img = vt.resize_image(test_rgb_large_img, shape=depth_shape)
rgbnormclean = vt.create_normalised_image(test_rgbx_img, empt_rgb_img)
if show:
plt.figure("RGBNORMCLEAN")
plt.imshow((rgbnormclean * 255).astype('uint8'))
plt.show()
rgbnormclean_uint8 = (rgbnormclean * 255).astype('uint8')
rgb_edged, rgb_img, rgb_cnts, rgb_hierarchy = vt.extract_contours(
rgbnormclean_uint8,
min_thresh=10,
max_thresh=100,
dilate=3,
erode=1,
cnt_mode=cv2.RETR_TREE)
rgb_family = kv.create_family(rgb_cnts, rgb_hierarchy)
if show:
plt.figure("OBJECT SEGMENTATION")
mask = np.zeros_like(rgbnormclean)
for member in rgb_family:
if member['generation'] == 0:
obj_mask = vt.create_contour_mask([member['contour']],
rgbnormclean)
mask = mask + obj_mask
plt.imshow(mask)
plt.show()
return rgbnormclean, rgb_family, test_rgbx_img
def run_image_processing(test, empt, cali, adjust=True, show=True):
test, test_depth, test_ir = kv.prepare_im_array(test)
empt, empt_depth, empt_ir = kv.prepare_im_array(empt)
cali, cali_depth, cali_ir = kv.prepare_im_array(cali)
empt_ir = cv2.resize(empt_ir, (0, 0), fx=2, fy=2)
CAL_PARAM = {'thresh': [85, 220], 'radius': [10, 16]}
BLACK = [270, 760, 230, 670]
copy_empt_ir = copy.copy(empt_ir)
#Find crop
while True:
empt_ir2 = vt.convert2gray(copy_empt_ir * 255, "uint8")
empt_ir3 = vt.black_out(empt_ir2, BLACK)
empt_circles, empt_cimg = vt.find_circles(empt_ir3,
2,
param=CAL_PARAM,
blur=1,
show=False)
empt_sort, crop = vt.sort_circles3(empt_circles)
empt_ir_img = vt.crop_out(empt_ir, crop)
if adjust:
plt.figure("CIRCLES")
plt.imshow(empt_cimg)
plt.figure("CROPPED EMPTY IMAGE")
plt.imshow(empt_ir_img)
plt.show()
check = raw_input("Continue?: ")
if check == "yes":
break
else:
cal_check = raw_input("Change calibration?: ")
if cal_check == "yes":
print CAL_PARAM
t1 = int(raw_input("Thresh 1: "))
t2 = int(raw_input("Thresh 2: "))
r1 = int(raw_input("Radius 1: "))
r2 = int(raw_input("Radius 2: "))
CAL_PARAM = {'thresh': [t1, t2], 'radius': [r1, r2]}
print "New CAL_PARAM: ", CAL_PARAM
black_check = raw_input("Change black crop?: ")
if black_check == "yes":
print BLACK
y1 = int(raw_input("y1: "))
y2 = int(raw_input("y2: "))
x1 = int(raw_input("x1: "))
x2 = int(raw_input("x2: "))
BLACK = [y1, y2, x1, x2]
print "New BLACK: ", BLACK
else:
break
#get normclean
crop = [i / 2 for i in crop]
empt_img = vt.crop_out(empt_depth, crop)
test_img = vt.crop_out(test_depth, crop)
cali_img = vt.crop_out(cali_depth, crop)
test_img = vt.clean_image(test_img)
empt_img = vt.clean_image(empt_img)
cali_img = vt.clean_image(cali_img)
#plt.figure()
#plt.imshow(test_img)
#plt.show()
#plt.figure()
#plt.imshow(empt_img)
#plt.show()
#plt.figure()
#plt.imshow(cali_img)
#plt.show()
normclean = vt.create_normalised_image(test_img, empt_img, cali_img)
if show:
plt.figure("Normclean")
plt.imshow((normclean * 255).astype('uint8'))
plt.show()
normclean_uint8 = (normclean * 255).astype('uint8')
edged, img, cnts, hierarchy = vt.extract_contours(normclean_uint8,
min_thresh=20,
max_thresh=90,
dilate=2,
erode=1)
new_cnts, new_h = kv.extract_depth_contours(cnts,
hierarchy,
normclean * 255,
minsize=200,
show=False)
family = kv.create_family(new_cnts, new_h)
sorted_family = kv.sort_family(family, normclean, show=False, debug=False)
if show:
plt.figure("OBJECT SEGMENTATION")
mask = np.zeros_like(normclean)
for member in sorted_family:
if member['generation'] == 0:
obj_mask = vt.create_contour_mask([member['contour'][0]],
normclean)
mask = mask + obj_mask
plt.imshow(mask)
plt.show()
print "Depth Done"
test = cv2.resize(test, (0, 0), fx=0.5, fy=0.5)
####################################### RGB ##########################################
#Find crop
empt = cv2.resize(empt, (0, 0), fx=0.5, fy=0.5)
RGB_CAL_PARAM = {'thresh': [85, 220], 'radius': [5, 8]}
RGB_BLACK = [300, 630, 100, 430]
while True:
empt_2 = vt.convert2gray(empt * 255, "uint8")
empt_3 = vt.black_out(empt_2, RGB_BLACK)
empt_rgb_circles, empt_rgb_cimg = vt.find_circles(empt_3,
2,
param=RGB_CAL_PARAM,
blur=3,
show=False)
empt_rgb_sort, rgb_crop = vt.sort_circles3(empt_rgb_circles)
empt_rgb_large_img = vt.crop_out(empt, rgb_crop)
if adjust:
plt.figure("CROPPED EMPTY RGB LARGE IMAGE")
plt.imshow(empt_rgb_large_img)
plt.show()
rgb_check = raw_input("Continue?: ")
if rgb_check == "yes":
break
else:
rgb_cal_check = raw_input("Change calibration?: ")
if rgb_cal_check == "yes":
print CAL_PARAM
t1 = raw_input("Thresh 1: ")
t2 = raw_input("Thresh 2: ")
r1 = raw_input("Radius 1: ")
r2 = raw_input("Radius 2: ")
RGB_CAL_PARAM = {'thresh': [t1, t2], 'radius': [r1, r2]}
rgb_black_check = raw_input("Change black crop?: ")
if rgb_black_check == "yes":
print RGB_BLACK
y1 = raw_input("y1: ")
y2 = raw_input("y2: ")
x1 = raw_input("x1: ")
x2 = raw_input("x2: ")
RGB_BLACK = [y1, y2, x1, x2]
else:
break
#get normclean
test_rgb_large_img = vt.crop_out(test, rgb_crop)
depth_shape = (np.shape(test_img)[1], np.shape(test_img)[0])
empt_rgb_img = vt.resize_image(empt_rgb_large_img, shape=depth_shape)
test_rgb_img = vt.resize_image(test_rgb_large_img, shape=depth_shape)
if show:
plt.figure()
plt.imshow(empt_rgb_img) #
plt.show()
#ergb_img = vt.clean_image(ergb_img)
rgbnormclean = vt.create_normalised_image(test_rgb_img, empt_rgb_img)
if show:
plt.figure("RGBNORMCLEAN")
plt.imshow((rgbnormclean * 255).astype('uint8'))
plt.show()
rgbnormclean_uint8 = (rgbnormclean * 255).astype('uint8')
rgb_edged, rgb_img, rgb_cnts, rgb_hierarchy = vt.extract_contours(
rgbnormclean_uint8,
min_thresh=10,
max_thresh=100,
dilate=3,
erode=1,
cnt_mode=cv2.RETR_TREE)
rgb_family = kv.create_family(rgb_cnts, rgb_hierarchy)
if show:
plt.figure("OBJECT SEGMENTATION")
mask = np.zeros_like(rgbnormclean)
for member in rgb_family:
if member['generation'] == 0:
obj_mask = vt.create_contour_mask([member['contour']],
rgbnormclean)
mask = mask + obj_mask
plt.imshow(mask)
plt.show()
return normclean, sorted_family, rgbnormclean, rgb_family, test_rgb_img