def predict(self, path_image):
img_ = themvien(path_image, (TARGET_SIZE, TARGET_SIZE), 0)
emb = incpetion_predict(img_)
lab_img = cv2.cvtColor(img_, cv2.COLOR_BGR2LAB)
l_channel, a_channel, b_channel = cv2.split(lab_img)
img_resize = cv2.resize(img_, (RESIZE_SIZE, RESIZE_SIZE))
img_resize = np.array(img_resize, dtype = float)
x = rgb2lab(1.0/255*img_resize)[:, :, 0]
x = x.reshape(1, RESIZE_SIZE, RESIZE_SIZE, 1)
output = self.model_.predict([x, emb])
output *= 128
cur = np.zeros((RESIZE_SIZE, RESIZE_SIZE, 3))
cur[:, :, 0] = x[0][:, :, 0]
cur[:, :, 1:] = output[0]
result_name = str(uuid.uuid4()) + '.png'
imsave('./static/image/' + result_name, lab2rgb(cur))
l, a, b = processing('./static/image/' + result_name)
out = np.zeros((TARGET_SIZE, TARGET_SIZE, 3), dtype = np.uint8)
out[:, :, 0] = l_channel
out[:, :, 1] = a
out[:, :, 2] = b
out = cv2.cvtColor(out, cv2.COLOR_LAB2BGR)
cv2.imwrite('./static/image/' + result_name, out)
return result_name
def init(files, folder=''):
data = []
t_sizes = []
for file in files:
_, edges_parameter, lines_parameter = read_params_file(file,
folder=folder)
target = imageProcessing.processing(file,
folder=folder,
edges_parameter=edges_parameter,
lines_parameter=lines_parameter)
data.append((file, target))
t_sizes.append(target.size)
return max(t_sizes), data
def routine(frame, camera_mks_rm):
global rel_x_ratio, num_data
processed_frame, explain1 = processing(frame)
choosen_left, choosen_right, explain2 = sliding_window(processed_frame)
ret_left, linear_func_left = get_sliding_window_function(choosen_left)
ret_right, linear_func_right = get_sliding_window_function(
choosen_right)
if ret_left and ret_right:
steering_angle, explain2 = get_steering_angle_from_linear_function(
(linear_func_left + linear_func_right) / 2, explain2)
# left = linear_func_left(480)
# right = linear_func_right(480)
# print(left, right, right - left)
# steering_angle = filter_deg.get_moving_average(steering_angle)
elif ret_left:
steering_angle, explain2 = get_steering_angle_from_linear_function(
linear_func_left, explain2, rel_x_ratio=1.2)
# steering_angle = filter_deg.get_moving_average(steering_angle)
elif ret_right:
steering_angle, explain2 = get_steering_angle_from_linear_function(
linear_func_right, explain2, rel_x_ratio=0.8)
# steering_angle = filter_deg.get_moving_average(steering_angle)
else:
steering_angle = filter_deg2.prev_lpf
# steering_angle = filter_deg.prev_avg
steering_angle = filter_deg2.get_lpf(steering_angle)[0]
_, explain2 = tfcp.get_segment(explain2, camera_mks_rm,
linear_func_left, linear_func_right)
# ret_line, _ = tfcp.get_line(explain2, num_data, camera_mks_rm)
# Merge Explain
vertical_line = np.zeros((explain1.shape[0], 5, 3), dtype=np.uint8)
explain_merge = np.hstack((explain2, vertical_line, explain1))
return steering_angle, explain_merge
def routine(frame):
global rel_x_ratio
processed_frame, explain1 = processing(frame)
choosen_left, choosen_right, explain2 = sliding_window(processed_frame)
ret_left, linear_func_left = get_sliding_window_function(choosen_left)
ret_right, linear_func_right = get_sliding_window_function(choosen_right)
if ret_left and ret_right:
steering_angle, explain2 = get_steering_angle_from_linear_function((linear_func_left + linear_func_right)/2, explain2)
# left = linear_func_left(480)
# right = linear_func_right(480)
# print(left, right, right - left)
## distance ~= 435.662 PIXEL or 0.845m
## 515.58 PIXEL / m
## Offset from lidar to ROI : 0.400 m
# steering_angle = filter_deg.get_moving_average(steering_angle)
elif ret_left:
steering_angle, explain2 = get_steering_angle_from_linear_function(linear_func_left, explain2, rel_x_ratio=1.2)
# steering_angle = filter_deg.get_moving_average(steering_angle)
elif ret_right:
steering_angle, explain2 = get_steering_angle_from_linear_function(linear_func_right, explain2, rel_x_ratio=0.8)
# steering_angle = filter_deg.get_moving_average(steering_angle)
else:
steering_angle = filter_deg2.prev_lpf
# steering_angle = filter_deg.prev_avg
steering_angle = filter_deg2.get_lpf(steering_angle)[0]
# Merge Explain
vertical_line = np.zeros((explain1.shape[0], 5, 3), dtype=np.uint8)
explain_merge = np.hstack((explain2, vertical_line, explain1))
return steering_angle, explain_merge
def start(self):
class_name = 'QWidget'
title_name = '金山打字通 2016'
self.hwnd = gui.FindWindow(class_name,title_name)
#print(self.hwnd)
if self.hwnd:
#gui.CloseWindow(self.hwnd)
left, top, right, bottom = self.gui_control()
if (left<0 or top<0 or right<0 or bottom<0):
raise Exception("Please don't hide the window!")
left += 45
top += 155
right -= 50
bottom -= 130
rect = (left, top, right, bottom)
img = ImageGrab.grab().crop(rect)
'''
plt.imshow(img)
plt.show()
'''
if self.isSave:
for j in range(5):
for i in range(66):
img_ready = img.crop((10.4*i,63.8*j,10.4*(i+1),63.8*j+18))
plt.imshow(img_ready)
plt.xticks([]), plt.yticks([])
plt.show()
img_ready.save(PICTURE_PATH.format(self.title, j, i))
char = input("Please enter the char: ")
with open(LABEL_PATH,'a') as f:
f.write("./pictures/{}.{}.{}.jpg {}\n".format(self.title, j, i, ord(char)))
elif self.isGray:
for j in range(5):
for i in range(66):
img_croped = img.crop((10.4*i,63.8*j,10.4*(i+1),63.8*j+18))
'''
plt.imshow(img_croped)
plt.xticks([]), plt.yticks([])
plt.show()
'''
img_processed = image_processing.processing(img_croped,False)
img_ready = pre_pic.pre_pic(img_processed,isApp=True)
if j==0 and i==0:
imgs = img_ready
else:
imgs = np.r_[imgs,img_ready]
#print(imgs)
yield self.hwnd,imgs
else:
for j in range(5):
for i in range(66):
img_croped = img.crop((10.4*i,63.8*j,10.4*(i+1),63.8*j+18))
'''
plt.imshow(img_croped)
plt.xticks([]), plt.yticks([])
plt.show()
'''
img_processed = image_processing.processing(img_croped,False)
img_ready = pre_pic.pre_pic(img_processed,isApp=True,isGray=False)
if j==0 and i==0:
imgs = img_ready
else:
imgs = np.r_[imgs,img_ready]
#print(imgs)
yield self.hwnd,imgs
else:
raise Exception("金山打字通2016 窗口捕获失败!")
minVal = 15_000
maxVal = 20_000
aperture_Size = 7
L2_gradient = True
rho = 7
threshold = 2
min_line_length = 2
max_line_gap = 4
# parameters for Target detection
edges_parameter = imageProcessing.edges_parm(minVal, maxVal, aperture_Size, L2_gradient)
lines_parameter = imageProcessing.lines_parm(rho, threshold, min_line_length, max_line_gap)
target = imageProcessing.processing(file, folder, edges_parameter, lines_parameter)
print('IND_SIZE: ' + str(target.size))
print('Height: ' + str(target.height))
print('Width: '+ str(target.width))
viewer.show_individual(target)
viewer.save_image(viewer.draw_individual(target), name='tmp/' + file + '_size_' +str(target.size))
# make params-file for input image
data = []
data.append(file)
data.append('edges_parameter')
data.append('minVal= ' + str(minVal))
data.append('maxVal= ' + str(maxVal))
data.append('aperture_Size= ' + str(aperture_Size))
def routine(frame, POS):
global filter_coef_left, filter_coef_right, heading_deg
frame = calibrate_image(frame)
processed_frame, explain1 = processing(frame)
processed_height, processed_width = processed_frame.shape[:2]
processed_height2 = processed_height // 3 * 2
choosen_left, choosen_right, explain2 = sliding_window(processed_frame)
func_coef_left = get_linear_function(choosen_left)
func_coef_right = get_linear_function(choosen_right)
if func_coef_left and func_coef_right:
for coef in [func_coef_left, func_coef_right]:
func = np.poly1d(coef)
new_ys = np.linspace(0,
processed_frame.shape[0],
num=processed_frame.shape[0],
endpoint=True)
new_xs = func(new_ys)
for x, y in zip(new_xs, new_ys):
if 0 < y < processed_frame.shape[
0] and 0 < x < processed_frame.shape[1]:
x, y = int(x), int(y)
cv.circle(explain2, (x, y), 3, (0, 255, 255), -1)
for coef in [(func_coef_left + func_coef_right) / 2]:
func = np.poly1d(coef)
new_ys = np.linspace(0,
processed_frame.shape[0],
num=processed_frame.shape[0],
endpoint=True)
new_xs = func(new_ys)
for x, y in zip(new_xs, new_ys):
if 0 < y < processed_frame.shape[
0] and 0 < x < processed_frame.shape[1]:
x, y = int(x), int(y)
cv.circle(explain2, (x, y), 3, (0, 0, 255), -1)
heading_src = (processed_width // 2, processed_height)
heading_dst = (func(processed_height2), processed_height2)
heading_sub = (heading_src[0] - heading_dst[0],
heading_src[1] - heading_dst[1])
heading_rad = np.arctan2(heading_sub[1], heading_sub[0])
heading_deg = np.degrees(heading_rad)
# heading_deg = filter_deg.LPF(np.array([heading_deg]))[0]
heading_deg = filter_deg2.get_lpf(heading_deg)
steering_deg = heading_deg - 90
heading_rad = np.radians(heading_deg)
gradient = np.tan(heading_rad)
heading_dst = ((processed_height2 - processed_height) / gradient +
processed_width // 2, processed_height2)
heading_src = tuple(rint(heading_src))
heading_dst = tuple(rint(heading_dst))
cv.circle(explain2, heading_src, 3, (0, 0, 255), -1)
cv.circle(explain2, heading_dst, 3, (0, 0, 255), -1)
cv.line(explain2, heading_src, heading_dst, (0, 0, 255), 1)
else:
# heading_rad = np.radians(heading_deg)
# gradient = np.tan(heading_rad) + 1e-10
# heading_src = (processed_width//2, processed_height)
# heading_dst = ((processed_height2 - heading_src[1])/gradient + heading_src[0], processed_height2)
# heading_src = tuple(rint(heading_src))
# heading_dst = tuple(rint(heading_dst))
# cv.circle(explain2, heading_src, 3, (0, 0, 255), -1)
# cv.circle(explain2, heading_dst, 3, (0, 0, 255), -1)
# cv.line(explain2, heading_src, heading_dst, (0, 0, 255), 1)
if func_coef_left:
func = np.poly1d(func_coef_left)
new_ys = np.linspace(0,
processed_frame.shape[0],
num=processed_frame.shape[0],
endpoint=True)
new_xs = func(new_ys)
left_src = (func(processed_height), processed_height)
left_dst = (func(processed_height2), processed_height2)
left_sub = (left_src[0] - left_dst[0],
left_src[1] - left_dst[1])
left_rad = np.arctan2(left_sub[1], left_sub[0])
heading_deg = np.degrees(left_rad)
heading_deg = filter_deg2.get_lpf(heading_deg)
steering_deg = heading_deg - 90
heading_rad = np.radians(heading_deg)
gradient = np.tan(heading_rad) + 1e-10
heading_src = (processed_width // 2, processed_height)
heading_dst = (
(processed_height2 - processed_height) / gradient +
processed_width // 2, processed_height2)
heading_src = tuple(rint(heading_src))
heading_dst = tuple(rint(heading_dst))
cv.circle(explain2, heading_src, 3, (0, 0, 255), -1)
cv.circle(explain2, heading_dst, 3, (0, 0, 255), -1)
cv.line(explain2, heading_src, heading_dst, (0, 0, 255), 1)
elif func_coef_right:
func = np.poly1d(func_coef_right)
new_ys = np.linspace(0,
processed_frame.shape[0],
num=processed_frame.shape[0],
endpoint=True)
new_xs = func(new_ys)
right_src = (func(processed_height), processed_height)
right_dst = (func(processed_height2), processed_height2)
right_sub = (right_src[0] - right_dst[0],
right_src[1] - right_dst[1])
right_rad = np.arctan2(right_sub[1], right_sub[0])
heading_deg = np.degrees(right_rad)
heading_deg = filter_deg2.get_lpf(heading_deg)
steering_deg = heading_deg - 90
heading_rad = np.radians(heading_deg)
gradient = np.tan(heading_rad) + 1e-10
heading_src = (processed_width // 2, processed_height)
heading_dst = (
(processed_height2 - processed_height) / gradient +
processed_width // 2, processed_height2)
heading_src = tuple(rint(heading_src))
heading_dst = tuple(rint(heading_dst))
cv.circle(explain2, heading_src, 3, (0, 0, 255), -1)
cv.circle(explain2, heading_dst, 3, (0, 0, 255), -1)
cv.line(explain2, heading_src, heading_dst, (0, 0, 255), 1)
else:
new_ys = []
new_xs = []
for x, y in zip(new_xs, new_ys):
if 0 < y < processed_frame.shape[
0] and 0 < x < processed_frame.shape[1]:
x, y = int(x), int(y)
cv.circle(explain2, (x, y), 3, (255, 255, 0), -1)
print(heading_deg)
# Merge Explain
vertical_line = np.zeros((explain1.shape[0], 5, 3), dtype=np.uint8)
explain_merge = np.hstack((explain2, vertical_line, explain1))
cv.putText(explain_merge, "POS: {}".format(int(POS)), (0, 40),
cv.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
return explain_merge
image = image.reshape(784, )
image = np.expand_dims(
image, axis=0) # p转换成了2维,模型第一层定义了输入格式为2维: input_shape=(784,)
# 测试图片
pre = model1.predict(image)
return np.argmax(pre, axis=1)[0]
def recognition2(image):
"""
使用模型2识别
:param image: 待识别图片
:return: 识别结果
"""
# 转化为列表的array格式
image_list = []
image = get_feature(image)
image_list.append(image)
image_array = np.array(image_list)
# 测试图片
pre = model2.predict(image_array)
return np.argmax(pre, axis=1)[0]
if __name__ == '__main__':
p, o_image = processing("./image/1.png")
recognition1(o_image)
recognition2(o_image)
