def display(self):
if self.img_src_path == None: # 还没选择图片就进行预测
self.can_pred1.create_text(32, 15, text='请选择图片', anchor='nw', font=('黑体', 28))
else:
img_src = cv2.imdecode(np.fromfile(self.img_src_path, dtype=np.uint8), -1) # 从中文路径读取时用
h, w = img_src.shape[0], img_src.shape[1]
if h * w <= 240 * 80 and 2 <= w / h <= 5: # 满足该条件说明可能整个图片就是一张车牌,无需定位,直接识别即可
lic = cv2.resize(img_src, dsize=(240, 80), interpolation=cv2.INTER_AREA)[:, :, :3] # 直接resize为(240,80)
img_src_copy, Lic_img = img_src, [lic]
else: # 否则就需通过unet对img_src原图预测,得到img_mask,实现车牌定位,然后进行识别
img_src, img_mask = unet_predict(self.unet, self.img_src_path)
img_src_copy, Lic_img = locate_and_correct(img_src, img_mask) # 利用core.py中的locate_and_correct函数进行车牌定位和矫正
Lic_pred = cnn_predict(self.cnn, Lic_img) # 利用cnn进行车牌的识别预测,Lic_pred中存的是元祖(车牌图片,识别结果)
if Lic_pred:
img = Image.fromarray(img_src_copy[:, :, ::-1]) # img_src_copy[:, :, ::-1]将BGR转为RGB
self.img_Tk = ImageTk.PhotoImage(img)
self.can_src.delete('all') # 显示前,先清空画板
self.can_src.create_image(258, 258, image=self.img_Tk, anchor='center') # img_src_copy上绘制出了定位的车牌轮廓,将其显示在画板上
for i, lic_pred in enumerate(Lic_pred):
if i == 0:
self.lic_Tk1 = ImageTk.PhotoImage(Image.fromarray(lic_pred[0][:, :, ::-1]))
self.can_lic1.create_image(5, 5, image=self.lic_Tk1, anchor='nw')
self.can_pred1.create_text(35, 15, text=lic_pred[1], anchor='nw', font=('黑体', 28))
elif i == 1:
self.lic_Tk2 = ImageTk.PhotoImage(Image.fromarray(lic_pred[0][:, :, ::-1]))
self.can_lic2.create_image(5, 5, image=self.lic_Tk2, anchor='nw')
self.can_pred2.create_text(40, 15, text=lic_pred[1], anchor='nw', font=('黑体', 28))
elif i == 2:
self.lic_Tk3 = ImageTk.PhotoImage(Image.fromarray(lic_pred[0][:, :, ::-1]))
self.can_lic3.create_image(5, 5, image=self.lic_Tk3, anchor='nw')
self.can_pred3.create_text(40, 15, text=lic_pred[1], anchor='nw', font=('黑体', 28))
else: # Lic_pred为空说明未能识别
self.can_pred1.create_text(47, 15, text='未能识别', anchor='nw', font=('黑体', 27))
lpn = provinces[gt[0]] + alphabets[gt[1]] + ads[gt[2]] + ads[
gt[3]] + ads[gt[4]] + ads[gt[5]] + ads[gt[6]]
# img = cv2.imread(img_src_path)
# img = cv2.resize(img,(512,512))
# cv2.imwrite(img_src_path,img)
img_src = cv2.imdecode(np.fromfile(img_src_path, dtype=np.uint8),
-1)
h, w = img_src.shape[0], img_src.shape[1]
if h * w <= 240 * 80 and 2 <= w / h <= 5: # 满足该条件说明可能整个图片就是一张车牌,无需定位,直接识别即可
lic = cv2.resize(
img_src, dsize=(240, 80),
interpolation=cv2.INTER_AREA)[:, :, :
3] # 直接resize为(240,80)
img_src_copy, Lic_img = img_src, [lic]
else: # 否则就需通过unet对img_src原图预测,得到img_mask,实现车牌定位,然后进行识别
img_src, img_mask = unet_predict(unet, img_src_path)
img_src_copy, Lic_img, prebox = locate_and_correct(
img_src,
img_mask) # 利用core.py中的locate_and_correct函数进行车牌定位和矫正
prebox = np.array(prebox)
if get_iou(prebox, gtbox) > 0.7:
iou_right += 1
Lic_pred = cnn_predict(
cnn, Lic_img) # 利用cnn进行车牌的识别预测,Lic_pred中存的是元祖(车牌图片,识别结果)
#print('print the result')
#print(Lic_pred)
if Lic_pred:
for i, lic_pred in enumerate(Lic_pred):
if i == 0:
text = lic_pred[1]
text = text[0:2] + text[3:]
import cv2
import numpy as np
from tensorflow import keras
from Unet import unet_predict
from core import locate, erode, dilate
import tensorflow as tf
print(tf.__version__)
UNet = tf.keras.models.load_model('unet.h5')
img_src, img_mask = unet_predict(UNet, '/home/sjq/81.jpg') #测试图片
# cv2.imshow('mask',img_mask)
img2gray = erode(img_mask)
# cv2.imshow('img2gray',img2gray)
ret, mask = cv2.threshold(img2gray, 200, 255, cv2.THRESH_BINARY)
# cv2.imshow('mask',mask)
cv2.imshow('img_mask',
cv2.bitwise_and(img_src, img_src, mask=cv2.bitwise_not(mask)))
cv2.waitKey(0)
gray2img = cv2.cvtColor(img2gray, cv2.COLOR_GRAY2BGR)
img_src_copy = locate(img_src, gray2img)
# cv2.imshow('show', img_mask)
cv2.imshow('plate', img_src_copy)
cv2.waitKey(0)
cv2.destroyAllWindows()