一、车牌识别简介
随着科学技术的发展，人工智能技术在我们的生活中的应用越来越广泛，人脸识别、车牌识别、目标识别等众多场景已经落地应用，给我们的生活品质得到很大的提升，办事效率大大提高，同时也节约了大量的劳动力。
今天我们来讲一讲车牌识别这个任务，车牌识别技术经过多年的发展，技术路线也呈现多样化，我们来介绍一下其中的两种：
第一种，单个字符识别
首先，在地面上有传感器感应区域，当有车辆经过时自动进行拍照，然后对拍的照片进行预处理，变成灰度图像，去除噪声等，把一些干扰因素去掉，同时降低图像的大小，便于后期进行边缘提取；
其次，利用边缘提取技术，比如canny算子、sobel算子，把图像轮廓提取出来，根据车牌长宽比大概为1:2.5左右把大部分不疑似的去掉；然后，再利用图像灰度值在水平方向和竖直方向上的投影，将字符切割出来；
最后，利用模板匹配或者是已经训练好的深度学习模型进行识别，单独把每个字符识别后进行串联得到整张车牌号码。
这种方法的瓶颈在于怎么样把每一个字符准确地切割出来，这涉及到很多图像处理技术，比如高斯模糊、图像锐化、图像膨胀、伽马变化、仿射变换等等，目前对于英文字符和数字的识别准确度在96%左右，汉字识别的准确度在95%左右。
第二种办法，整张车牌识别
首先，与第一种车牌识别一样，利用传感器进行车辆感应，把整辆车进行拍照；
其次，利用训练好的车牌检测算法，比如YOLO算法或者其他的目标检测算法把车牌检测出来。为了提升车牌识别的准确度，可以利用图像预处理技术对图像进行处理，降低干扰项；
最后，利用已经训练好的深度学习模型对整张车牌进行识别。比如LPRNet、Darknet等
二、车牌识别技术实现
## 第一种方法 ——利用 pytesseract 进行识别
pytesseract是一个OCR识别工具，具体的 pytesseract 安装可百度安装，直接使用这个工具进行识别得到的准确度还有点低，达不到理想的效果，需要利用车牌数据进行训练才可以使用。
import cv2 as cv
from PIL import Image
import pytesseract as tess
def recoginse_text(image):
"""
步骤：
1、灰度，二值化处理
2、形态学操作去噪
3、识别
:param image:
:return:
"""
# 灰度 二值化
gray = cv.cvtColor(image,cv.COLOR_BGR2GRAY)
# 如果是白底黑字 建议 _INV
ret,binary = cv.threshold(gray,0,255,cv.THRESH_BINARY_INV| cv.THRESH_OTSU)
# 形态学操作 (根据需要设置参数（1，2）)
kernel = cv.getStructuringElement(cv.MORPH_RECT,(1,2)) #去除横向细线
morph1 = cv.morphologyEx(binary,cv.MORPH_OPEN,kernel)
kernel = cv.getStructuringElement(cv.MORPH_RECT, (2, 1)) #去除纵向细线
morph2 = cv.morphologyEx(morph1,cv.MORPH_OPEN,kernel)
cv.imshow("Morph",morph2)
# 黑底白字取非，变为白底黑字（便于pytesseract 识别）
cv.bitwise_not(morph2,morph2)
textImage = Image.fromarray(morph2)
# 图片转文字
text=tess.image_to_string(textImage)
print("识别结果：%s"%text)
def main():
# 读取需要识别的数字字母图片，并显示读到的原图
src = cv.imread(r'C:\Users\lenovo\Desktop\00A8CX87_5.jpg')
cv.imshow("src",src)
# 识别
recoginse_text(src)
cv.waitKey(0)
cv.destroyAllWindows()
if __name__=="__main__":
main()
## 第二种识别办法——PaddleOCR
PaddleOCR 是百度paddlepaddle下的OCR模块，可以用它来识别车牌，直接下载预训练模型进行识别，对于数字和子母都得到不错的准确率，但是针对中文识别的效果并不理想，需要自己的数据集进行再训练。
要想利用这个模型进行训练车牌识别，需要大量的车牌数据，人工收集不仅耗费时间长，要涉及每个省份的车牌需要比较大的成本。
另外的办法是利用图像处理技术进行数据生成，加上图像增强技术来制作数据集，这将在我们的第三章识别方法中讲到。
from paddleocr import PaddleOCR
import os
# ocr = PaddleOCR(use_gpu=False, use_angle_cls=True, lang="ch")
ocr = PaddleOCR(use_gpu=True, use_angle_cls=True, lang="ch")
path=r'E:\车牌数据'
results=[]
for file in os.listdir(path):
img_path =os.path.join(path,file)
result = ocr.ocr(img_path, cls=True)
if len(result)>0:
results.append([img_path,result[0][1][0]])
else:
results.append([img_path])
将结果导出到Excel文件夹
import pandas as pd
resultss=pd.DataFrame(results)
resultss.to_excel(r'C:\Users\lenovo\Desktop\11.xlsx')
## 第三种办法 ——LPRNet算法
由于要收集满足深度学习的数据需要花费大量的时间，有一种办法是通过数据生成的办法生成大量的模拟数据集。
紧接着，我们可以利用LPRNet算法进行车牌识别，从GitHub上下载有tensorflow、torch两个版本。
第一步，生成蓝色车牌数据集
第二步，根据算法模型要求将图像名称修改并放到指定的文件夹中
第三步，模型训练和测试
## (1) 蓝色车牌生成
import os
import cv2 as cv
import numpy as np
from math import *
from PIL import ImageFont
from PIL import Image
from PIL import ImageDraw
index = {"京": 0, "沪": 1, "津": 2, "渝": 3, "冀": 4, "晋": 5, "蒙": 6, "辽": 7, "吉": 8, "黑": 9,
"苏": 10, "浙": 11, "皖": 12, "闽": 13, "赣": 14, "鲁": 15, "豫": 16, "鄂": 17, "湘": 18, "粤": 19,
"桂": 20, "琼": 21, "川": 22, "贵": 23, "云": 24, "藏": 25, "陕": 26, "甘": 27, "青": 28, "宁": 29,
"新": 30, "0": 31, "1": 32, "2": 33, "3": 34, "4": 35, "5": 36, "6": 37, "7": 38, "8": 39,
"9": 40, "A": 41, "B": 42, "C": 43, "D": 44, "E": 45, "F": 46, "G": 47, "H": 48, "J": 49,
"K": 50, "L": 51, "M": 52, "N": 53, "P": 54, "Q": 55, "R": 56, "S": 57, "T": 58, "U": 59,
"V": 60, "W": 61, "X": 62, "Y": 63, "Z": 64}
chars = ["京", "沪", "津", "渝", "冀", "晋", "蒙", "辽", "吉", "黑",
"苏", "浙", "皖", "闽", "赣", "鲁", "豫", "鄂", "湘", "粤",
"桂", "琼", "川", "贵", "云", "藏", "陕", "甘", "青", "宁",
"新", "0", "1", "2", "3", "4", "5", "6", "7", "8",
"9", "A", "B", "C", "D", "E", "F", "G", "H", "J",
"K", "L", "M", "N", "P", "Q", "R", "S", "T", "U",
"V", "W", "X", "Y", "Z"]
def AddSmudginess(img, Smu):
"""
模糊处理
:param img: 输入图像
:param Smu: 模糊图像
:return: 添加模糊后的图像
"""
rows = r(Smu.shape[0] - 50)
cols = r(Smu.shape[1] - 50)
adder = Smu[rows:rows + 50, cols:cols + 50]
adder = cv.resize(adder, (50, 50))
img = cv.resize(img,(50,50))
img = cv.bitwise_not(img)
img = cv.bitwise_and(adder, img)
img = cv.bitwise_not(img)
return img
def rot(img, angel, shape, max_angel):
"""
添加透视畸变
"""
size_o = [shape[1], shape[0]]
size = (shape[1]+ int(shape[0] * cos((float(max_angel ) / 180) * 3.14)), shape[0])
interval = abs(int(sin((float(angel) / 180) * 3.14) * shape[0]))
pts1 = np.float32([[0, 0], [0, size_o[1]], [size_o[0], 0], [size_o[0], size_o[1]]])
if angel > 0:
pts2 = np.float32([[interval, 0], [0, size[1]], [size[0], 0], [size[0] - interval, size_o[1]]])
else:
pts2 = np.float32([[0, 0], [interval, size[1]], [size[0] - interval, 0], [size[0], size_o[1]]])
M = cv.getPerspectiveTransform(pts1, pts2)
dst = cv.warpPerspective(img, M, size)
return dst
def rotRandrom(img, factor, size):
"""
添加放射畸变
:param img: 输入图像
:param factor: 畸变的参数
:param size: 图片目标尺寸
:return: 放射畸变后的图像
"""
shape = size
pts1 = np.float32([[0, 0], [0, shape[0]], [shape[1], 0], [shape[1], shape[0]]])
pts2 = np.float32([[r(factor), r(factor)], [r(factor), shape[0] - r(factor)], [shape[1] - r(factor), r(factor)],
[shape[1] - r(factor), shape[0] - r(factor)]])
M = cv.getPerspectiveTransform(pts1, pts2)
dst = cv.warpPerspective(img, M, size)
return dst
def tfactor(img):
"""
添加饱和度光照的噪声
"""
hsv = cv.cvtColor(img,cv.COLOR_BGR2HSV)
hsv[:, :, 0] = hsv[:, :, 0] * (0.8 + np.random.random() * 0.2)
hsv[:, :, 1] = hsv[:, :, 1] * (0.3 + np.random.random() * 0.7)
hsv[:, :, 2] = hsv[:, :, 2] * (0.2 + np.random.random() * 0.8)
img = cv.cvtColor(hsv, cv.COLOR_HSV2BGR)
return img
def random_envirment(img, noplate_bg):
"""
添加自然环境的噪声, noplate_bg为不含车牌的背景图
"""
bg_index = r(len(noplate_bg))
env = cv.imread(noplate_bg[bg_index])
env = cv.resize(env, (img.shape[1], img.shape[0]))
bak = (img == 0)
bak = bak.astype(np.uint8) * 255
inv = cv.bitwise_and(bak, env)
img = cv.bitwise_or(inv, img)
return img
def GenCh(f, val):
"""
生成中文字符
"""
img = Image.new("RGB", (45, 70), (255, 255, 255)) #白色
# img = Image.new("RGB", (45, 70), (0, 0, 0)) #黑色
draw = ImageDraw.Draw(img)
# draw.text((0, 3), val, (0, 0, 0), font=f)
draw.text((0, 3), val, (0, 0, 0), font=f)
img = img.resize((23, 70))
A = np.array(img)
return A
def GenCh1(f, val):
"""
生成英文字符
"""
img =Image.new("RGB", (23, 70), (255,255,255))#白色
# img =Image.new("RGB", (23, 70), (0, 0, 0))#黑色
draw = ImageDraw.Draw(img)
draw.text((0, 2), val, (0,125,125), font=f) # val.decode('utf-8')
A = np.array(img)
return A
def AddGauss(img, level):
"""
添加高斯模糊
"""
return cv.blur(img, (level * 2 + 1, level * 2 + 1))
def r(val):
return int(np.random.random() * val)
def AddNoiseSingleChannel(single):
"""
添加高斯噪声
"""
diff = 255 - single.max()
noise = np.random.normal(0, 1 + r(6), single.shape)
noise = (noise - noise.min()) / (noise.max() - noise.min())
noise *= diff
# noise= noise.astype(np.uint8)
dst = single + noise
return dst
def addNoise(img): # sdev = 0.5,avg=10
img[:, :, 0] = AddNoiseSingleChannel(img[:, :, 0])
img[:, :, 1] = AddNoiseSingleChannel(img[:, :, 1])
img[:, :, 2] = AddNoiseSingleChannel(img[:, :, 2])
return img
class GenPlate:
def __init__(self, fontCh, fontEng, NoPlates):
self.fontC = ImageFont.truetype(fontCh, 43, 0)
self.fontE = ImageFont.truetype(fontEng, 60, 0)
self.img = np.array(Image.new("RGB", (226, 70),(255, 255, 255)))
# self.img = np.array(Image.new("RGB", (226, 70),(0, 0, 0)))
self.bg = cv.resize(cv.imread(r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\generate_picture\images\blue.bmp"), (226, 70))
# template.bmp:车牌背景图
self.smu = cv.imread(r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\images\\smu2.jpg")
# smu2.jpg:模糊图像
self.noplates_path = []
for parent, parent_folder, filenames in os.walk(NoPlates):
for filename in filenames:
path = parent + "\\" + filename
self.noplates_path.append(path)
def draw(self, val):
offset = 2
self.img[0:70, offset+8:offset+8+23] = GenCh(self.fontC, val[0])
self.img[0:70, offset+8+23+6:offset+8+23+6+23] = GenCh1(self.fontE, val[1])
for i in range(5):
base = offset + 8 + 23 + 6 + 23 + 17 + i * 23 + i * 6
self.img[0:70, base:base+23] = GenCh1(self.fontE, val[i+2])
return self.img
def generate(self, text):
if len(text) == 7:
fg = self.draw(text) # decode(encoding="utf-8")
fg = cv.bitwise_not(fg)
com = cv.bitwise_or(fg, self.bg)
com = rot(com, r(30)-10, com.shape,30)
com = rotRandrom(com, 5, (com.shape[1], com.shape[0]))
com = tfactor(com)
com = random_envirment(com, self.noplates_path)
com = AddGauss(com, 1+r(4))
com = addNoise(com)
return com
@staticmethod
def genPlateString(pos, val):
"""
生成车牌string，存为图片
生成车牌list，存为label
"""
plateStr = ""
plateList=[]
box = [0, 0, 0, 0, 0, 0, 0]
if pos != -1:
box[pos] = 1
for unit, cpos in zip(box, range(len(box))):
if unit == 1:
plateStr += val
plateList.append(val)
else:
if cpos == 0:
plateStr += chars[r(31)]
plateList.append(plateStr)
elif cpos == 1:
plateStr += chars[41 + r(24)]
plateList.append(plateStr)
else:
plateStr += chars[31 + r(34)]
plateList.append(plateStr)
plate = [plateList[0]]
b = [plateList[i][-1] for i in range(len(plateList))]
plate.extend(b[1:7])
return plateStr, plate
@staticmethod
def genBatch(batchsize, outputPath, size):
"""
将生成的车牌图片写入文件夹，对应的label写入label.txt
:param batchsize: 批次大小
:param outputPath: 输出图像的保存路径
:param size: 输出图像的尺寸
:return: None
"""
if not os.path.exists(outputPath):
os.mkdir(outputPath)
outfile = open(r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\label_blue.txt',
'w', encoding='utf-8')
for i in range(batchsize):
plateStr, plate = G.genPlateString(-1, -1)
img = G.generate(plateStr)
img = cv.resize(img, size)
content = "".join(plate)
cv.imwrite(outputPath + "\\" + str(i).zfill(2) + ".jpg", img)
outfile.write("./train/"+str(i).zfill(2)+".jpg"+'\t'+content +"\n")
G = GenPlate(r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\font\platech.ttf",
r'E:\AI_projects\vehicle license plate recognition\car_recognition\data\font\platechar.ttf',
r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\NoPlates")
G.genBatch(1000, r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_blue', (272, 72))
需要生成多少张车牌数据可通过修改最后一行代码的数字即可
## (2) 生产黄色车牌
生成车牌的步骤是把字符写到纯净的背景图片上，蓝色的车牌是把白色的字写到蓝色底的车牌上，可以直接把字符写到底牌上去，但是黄色底的车牌不能简单把黑色字写上去，写上去的结果是字符被隐藏起来，需要特别处理。
正确的做法是：
第一步，构造一张车牌大小的纯黑色车牌，往上面写白色字符
第二步，将黑白色进行反转，把这牌的字符和背景进行翻转
第三步，将一张没有字符的黄色底牌和反转后的车牌进行结合，然后经过高斯模糊、改变光线、旋转角度等各种随机组合模拟出于生活中拍摄处理的图像。
#黄色车牌生成
import os
import cv2 as cv
import numpy as np
from math import *
from PIL import ImageFont
from PIL import Image
from PIL import ImageDraw
index = {"京": 0, "沪": 1, "津": 2, "渝": 3, "冀": 4, "晋": 5, "蒙": 6, "辽": 7, "吉": 8, "黑": 9,
"苏": 10, "浙": 11, "皖": 12, "闽": 13, "赣": 14, "鲁": 15, "豫": 16, "鄂": 17, "湘": 18, "粤": 19,
"桂": 20, "琼": 21, "川": 22, "贵": 23, "云": 24, "藏": 25, "陕": 26, "甘": 27, "青": 28, "宁": 29,
"新": 30, "0": 31, "1": 32, "2": 33, "3": 34, "4": 35, "5": 36, "6": 37, "7": 38, "8": 39,
"9": 40, "A": 41, "B": 42, "C": 43, "D": 44, "E": 45, "F": 46, "G": 47, "H": 48, "J": 49,
"K": 50, "L": 51, "M": 52, "N": 53, "P": 54, "Q": 55, "R": 56, "S": 57, "T": 58, "U": 59,
"V": 60, "W": 61, "X": 62, "Y": 63, "Z": 64}
chars = ["京", "沪", "津", "渝", "冀", "晋", "蒙", "辽", "吉", "黑",
"苏", "浙", "皖", "闽", "赣", "鲁", "豫", "鄂", "湘", "粤",
"桂", "琼", "川", "贵", "云", "藏", "陕", "甘", "青", "宁",
"新", "0", "1", "2", "3", "4", "5", "6", "7", "8",
"9", "A", "B", "C", "D", "E", "F", "G", "H", "J",
"K", "L", "M", "N", "P", "Q", "R", "S", "T", "U",
"V", "W", "X", "Y", "Z"]
def AddSmudginess(img, Smu):
"""
模糊处理
:param img: 输入图像
:param Smu: 模糊图像
:return: 添加模糊后的图像
"""
rows = r(Smu.shape[0] - 50)
cols = r(Smu.shape[1] - 50)
adder = Smu[rows:rows + 50, cols:cols + 50]
adder = cv.resize(adder, (50, 50))
img = cv.resize(img,(50,50))
img = cv.bitwise_not(img)
img = cv.bitwise_and(adder, img)
img = cv.bitwise_not(img)
return img
def rot(img, angel, shape, max_angel):
"""
添加透视畸变
"""
size_o = [shape[1], shape[0]]
size = (shape[1]+ int(shape[0] * cos((float(max_angel ) / 180) * 3.14)), shape[0])
interval = abs(int(sin((float(angel) / 180) * 3.14) * shape[0]))
pts1 = np.float32([[0, 0], [0, size_o[1]], [size_o[0], 0], [size_o[0], size_o[1]]])
if angel > 0:
pts2 = np.float32([[interval, 0], [0, size[1]], [size[0], 0], [size[0] - interval, size_o[1]]])
else:
pts2 = np.float32([[0, 0], [interval, size[1]], [size[0] - interval, 0], [size[0], size_o[1]]])
M = cv.getPerspectiveTransform(pts1, pts2)
dst = cv.warpPerspective(img, M, size)
return dst
def rotRandrom(img, factor, size):
"""
添加放射畸变
:param img: 输入图像
:param factor: 畸变的参数
:param size: 图片目标尺寸
:return: 放射畸变后的图像
"""
shape = size
pts1 = np.float32([[0, 0], [0, shape[0]], [shape[1], 0], [shape[1], shape[0]]])
pts2 = np.float32([[r(factor), r(factor)], [r(factor), shape[0] - r(factor)], [shape[1] - r(factor), r(factor)],
[shape[1] - r(factor), shape[0] - r(factor)]])
M = cv.getPerspectiveTransform(pts1, pts2)
dst = cv.warpPerspective(img, M, size)
return dst
def tfactor(img):
"""
添加饱和度光照的噪声
"""
hsv = cv.cvtColor(img,cv.COLOR_BGR2HSV)
hsv[:, :, 0] = hsv[:, :, 0] * (0.8 + np.random.random() * 0.2)
hsv[:, :, 1] = hsv[:, :, 1] * (0.3 + np.random.random() * 0.7)
hsv[:, :, 2] = hsv[:, :, 2] * (0.2 + np.random.random() * 0.8)
img = cv.cvtColor(hsv, cv.COLOR_HSV2BGR)
return img
def random_envirment(img, noplate_bg):
"""
添加自然环境的噪声, noplate_bg为不含车牌的背景图
"""
bg_index = r(len(noplate_bg))
env = cv.imread(noplate_bg[bg_index])
env = cv.resize(env, (img.shape[1], img.shape[0]))
bak = (img == 0)
bak = bak.astype(np.uint8) * 255
inv = cv.bitwise_and(bak, env)
img = cv.bitwise_or(inv, img)
return img
def GenCh(f, val):
"""
生成中文字符
"""
img = Image.new("RGB", (45, 70), (255, 255, 255)) #白色
# img = Image.new("RGB", (45, 70), (0, 0, 0)) #黑色
draw = ImageDraw.Draw(img)
# draw.text((0, 3), val, (0, 0, 0), font=f)
draw.text((0, 3), val, (0, 0, 0), font=f)
img = img.resize((23, 70))
A = np.array(img)
return A
def GenCh1(f, val):
"""
生成英文字符
"""
img =Image.new("RGB", (23, 70), (255,255,255))#白色
# img =Image.new("RGB", (23, 70), (0, 0, 0))#黑色
draw = ImageDraw.Draw(img)
draw.text((0, 2), val, (0,0,0), font=f) # val.decode('utf-8')
A = np.array(img)
return A
def AddGauss(img, level):
"""
添加高斯模糊
"""
return cv.blur(img, (level * 2 + 1, level * 2 + 1))
def r(val):
return int(np.random.random() * val)
def AddNoiseSingleChannel(single):
"""
添加高斯噪声
"""
diff = 255 - single.max()
noise = np.random.normal(0, 1 + r(6), single.shape)
noise = (noise - noise.min()) / (noise.max() - noise.min())
noise *= diff
# noise= noise.astype(np.uint8)
dst = single + noise
return dst
def addNoise(img): # sdev = 0.5,avg=10
img[:, :, 0] = AddNoiseSingleChannel(img[:, :, 0])
img[:, :, 1] = AddNoiseSingleChannel(img[:, :, 1])
img[:, :, 2] = AddNoiseSingleChannel(img[:, :, 2])
return img
class GenPlate:
def __init__(self, fontCh, fontEng, NoPlates):
self.fontC = ImageFont.truetype(fontCh, 43, 0)
self.fontE = ImageFont.truetype(fontEng, 60, 0)
self.img = np.array(Image.new("RGB", (226, 70),(255, 255, 255)))
# self.img = np.array(Image.new("RGB", (226, 70),(0, 0, 0)))
self.bg = cv.resize(cv.imread(r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\generate_picture\images\black.jpg"), (226, 70))
# template.bmp:车牌背景图
self.smu = cv.imread(r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\images\\smu2.jpg")
# smu2.jpg:模糊图像
self.noplates_path = []
for parent, parent_folder, filenames in os.walk(NoPlates):
for filename in filenames:
path = parent + "\\" + filename
self.noplates_path.append(path)
def draw(self, val):
offset = 2
self.img[0:70, offset+8:offset+8+23] = GenCh(self.fontC, val[0])
self.img[0:70, offset+8+23+6:offset+8+23+6+23] = GenCh1(self.fontE, val[1])
for i in range(5):
base = offset + 8 + 23 + 6 + 23 + 17 + i * 23 + i * 6
self.img[0:70, base:base+23] = GenCh1(self.fontE, val[i+2])
return self.img
def generate(self, text):
if len(text) == 7:
fg = self.draw(text) # decode(encoding="utf-8")
fg = cv.bitwise_not(fg)
com = cv.bitwise_or(fg, self.bg)
# com = rot(com, r(60)-30, com.shape,30)
# com = rotRandrom(com, 10, (com.shape[1], com.shape[0]))
# com = tfactor(com)
# com = random_envirment(com, self.noplates_path)
# com = AddGauss(com, 1+r(4))
# com = addNoise(com)
return com
@staticmethod
def genPlateString(pos, val):
"""
生成车牌string，存为图片
生成车牌list，存为label
"""
plateStr = ""
plateList=[]
box = [0, 0, 0, 0, 0, 0, 0]
if pos != -1:
box[pos] = 1
for unit, cpos in zip(box, range(len(box))):
if unit == 1:
plateStr += val
plateList.append(val)
else:
if cpos == 0:
plateStr += chars[r(31)]
plateList.append(plateStr)
elif cpos == 1:
plateStr += chars[41 + r(24)]
plateList.append(plateStr)
else:
plateStr += chars[31 + r(34)]
plateList.append(plateStr)
plate = [plateList[0]]
b = [plateList[i][-1] for i in range(len(plateList))]
plate.extend(b[1:7])
return plateStr, plate
@staticmethod
def genBatch(batchsize, outputPath, size):
"""
将生成的车牌图片写入文件夹，对应的label写入label.txt
:param batchsize: 批次大小
:param outputPath: 输出图像的保存路径
:param size: 输出图像的尺寸
:return: None
"""
if not os.path.exists(outputPath):
os.mkdir(outputPath)
outfile = open(r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\label_yellow.txt',
'w', encoding='utf-8')
for i in range(batchsize):
plateStr, plate = G.genPlateString(-1, -1)
img = G.generate(plateStr)
img = cv.resize(img, size)
content = "".join(plate)
cv.imwrite(outputPath + "\\" + str(i).zfill(2) + ".jpg", img)
outfile.write("./train/"+str(i).zfill(2)+".jpg"+'\t'+content +"\n")
G = GenPlate(r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\font\platech.ttf",
r'E:\AI_projects\vehicle license plate recognition\car_recognition\data\font\platechar.ttf',
r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\NoPlates")
G.genBatch(1000, r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_yellow', (272, 72))
字符和背景反转
#将车牌黑色反转
import cv2
import random
def addNoise(img): # sdev = 0.5,avg=10
img[:, :, 0] = AddNoiseSingleChannel(img[:, :, 0])
img[:, :, 1] = AddNoiseSingleChannel(img[:, :, 1])
img[:, :, 2] = AddNoiseSingleChannel(img[:, :, 2])
return img
sample_cp=cv.resize(cv2.imread(r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\generate_picture\images\yellow.bmp"), (272, 72))
path=r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_yellow'
for file in os.listdir(path):
if file.endswith('.jpg'):
image=os.path.join(path,file)
img=cv2.imread(image)
for i in range(72):
for j in range(272):
if img[i][j][0]>50:
img[i][j][0]=0 #黑色
img[i][j][1]=0
img[i][j][2]=0
else:
img[i][j][0]=255 #白色
img[i][j][1]=255
img[i][j][2]=255
for i in range(72):
for j in range(272):
if img[i][j][0]<50 and img[i][j][1]<50 and img[i][j][2]<50:
img[i][j][0]=img[i][j][0]
img[i][j][1]=img[i][j][1]
img[i][j][2]=img[i][j][2]
else:
img[i][j][0]=sample_cp[i][j][0] #+random.randint(0,9)
img[i][j][1]=sample_cp[i][j][1] #+random.randint(0,5)
img[i][j][2]=sample_cp[i][j][2] #+random.randint(0,10)
addNoise(img)
images=os.path.join(path,file)
cv2.imwrite(images,img)
字符和底牌相结合，并进行畸变、模糊、旋转等各种方法，构造出多种多样的车牌图像。
import cv2
import numpy as np
from math import cos,sin
import os
def rot(img, angel, shape, max_angel):
"""
添加透视畸变
"""
size_o = [shape[1], shape[0]]
size = (shape[1]+ int(shape[0] * cos((float(max_angel ) / 180) * 3.14)), shape[0])
interval = abs(int(sin((float(angel) / 180) * 3.14) * shape[0]))
pts1 = np.float32([[0, 0], [0, size_o[1]], [size_o[0], 0], [size_o[0], size_o[1]]])
if angel > 0:
pts2 = np.float32([[interval, 0], [0, size[1]], [size[0], 0], [size[0] - interval, size_o[1]]])
else:
pts2 = np.float32([[0, 0], [interval, size[1]], [size[0] - interval, 0], [size[0], size_o[1]]])
M = cv2.getPerspectiveTransform(pts1, pts2)
dst = cv2.warpPerspective(img, M, size)
return dst
def rotRandrom(img, factor, size):
"""
添加放射畸变
:param img: 输入图像
:param factor: 畸变的参数
:param size: 图片目标尺寸
:return: 放射畸变后的图像
"""
shape = size
pts1 = np.float32([[0, 0], [0, shape[0]], [shape[1], 0], [shape[1], shape[0]]])
pts2 = np.float32([[r(factor), r(factor)], [r(factor), shape[0] - r(factor)], [shape[1] - r(factor), r(factor)],
[shape[1] - r(factor), shape[0] - r(factor)]])
M = cv2.getPerspectiveTransform(pts1, pts2)
dst = cv2.warpPerspective(img, M, size)
return dst
def r(val):
return int(np.random.random() * val)
def tfactor(img):
"""
添加饱和度光照的噪声
"""
hsv = cv2.cvtColor(img,cv2.COLOR_BGR2HSV)
hsv[:, :, 0] = hsv[:, :, 0] * (0.8 + np.random.random() * 0.2)
hsv[:, :, 1] = hsv[:, :, 1] * (0.3 + np.random.random() * 0.7)
hsv[:, :, 2] = hsv[:, :, 2] * (0.2 + np.random.random() * 0.8)
img = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)
return img
def random_envirment(img, noplate_bg):
"""
添加自然环境的噪声, noplate_bg为不含车牌的背景图
"""
bg_index = r(len(noplate_bg))
env = cv2.imread(noplate_bg[bg_index])
env = cv2.resize(env, (img.shape[1], img.shape[0]))
bak = (img == 0)
bak = bak.astype(np.uint8) * 255
inv = cv2.bitwise_and(bak, env)
img = cv2.bitwise_or(inv, img)
return img
def AddGauss(img, level):
return cv2.blur(img, (level * 2 + 1, level * 2 + 1))
def AddNoiseSingleChannel(single):
"""
添加高斯噪声
"""
diff = 255 - single.max()
noise = np.random.normal(0, 1 + r(6), single.shape)
noise = (noise - noise.min()) / (noise.max() - noise.min())
noise *= diff
# noise= noise.astype(np.uint8)
dst = single + noise
return dst
def addNoise(img): # sdev = 0.5,avg=10
img[:, :, 0] = AddNoiseSingleChannel(img[:, :, 0])
img[:, :, 1] = AddNoiseSingleChannel(img[:, :, 1])
img[:, :, 2] = AddNoiseSingleChannel(img[:, :, 2])
return img
path0=r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\generate_picture\NoPlates"
noplates_path=[]
for file in os.listdir(path0):
noplates_path.append(os.path.join(path0,file))
# bg=cv2.resize(cv2.imread(r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\generate_picture\images\new.jpg"), (272, 72))
# fg=cv2.imread(r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\picture_new\00.jpg')
# com=fg
# com = rot(com, r(30)-10, com.shape,30)
# com = rotRandrom(com, 5, (com.shape[1], com.shape[0]))
# com = tfactor(com)
# # com = random_envirment(com,noplates_path)
# # com = AddGauss(com, 1+r(4))
# com = addNoise(com)
# cv2.imshow('com',com)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
# fg = cv2.bitwise_not(fg)
# com = cv2.bitwise_or(fg, bg)
path=r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_yellow"
path1=r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_yellow"
for file in os.listdir(path):
if file.endswith('.jpg'):
fg=cv2.imread(os.path.join(path,file))
com=fg
com = rot(com, r(30)-10, com.shape,30)
com = rotRandrom(com, 5, (com.shape[1], com.shape[0]))
com = tfactor(com)
com = random_envirment(com,noplates_path)
com = AddGauss(com, 1+r(4))
com = addNoise(com)
images=os.path.join(path1,file)
cv2.imwrite(images,com)
## (3) 生产新能源车牌
新能源车牌的生成办法与黄色车牌的生成办法是一样的，但是新能源车牌长度比蓝牌、黄牌多一位，而且指定第三个字符为D或者F，但是我们我们在生成时不考虑，只是修改代码让程序生成同样长度的车牌即可。
#黄色车牌生成
import os
import cv2 as cv
import numpy as np
from math import *
from PIL import ImageFont
from PIL import Image
from PIL import ImageDraw
index = {"京": 0, "沪": 1, "津": 2, "渝": 3, "冀": 4, "晋": 5, "蒙": 6, "辽": 7, "吉": 8, "黑": 9,
"苏": 10, "浙": 11, "皖": 12, "闽": 13, "赣": 14, "鲁": 15, "豫": 16, "鄂": 17, "湘": 18, "粤": 19,
"桂": 20, "琼": 21, "川": 22, "贵": 23, "云": 24, "藏": 25, "陕": 26, "甘": 27, "青": 28, "宁": 29,
"新": 30, "0": 31, "1": 32, "2": 33, "3": 34, "4": 35, "5": 36, "6": 37, "7": 38, "8": 39,
"9": 40, "A": 41, "B": 42, "C": 43, "D": 44, "E": 45, "F": 46, "G": 47, "H": 48, "J": 49,
"K": 50, "L": 51, "M": 52, "N": 53, "P": 54, "Q": 55, "R": 56, "S": 57, "T": 58, "U": 59,
"V": 60, "W": 61, "X": 62, "Y": 63, "Z": 64}
chars = ["京", "沪", "津", "渝", "冀", "晋", "蒙", "辽", "吉", "黑",
"苏", "浙", "皖", "闽", "赣", "鲁", "豫", "鄂", "湘", "粤",
"桂", "琼", "川", "贵", "云", "藏", "陕", "甘", "青", "宁",
"新", "0", "1", "2", "3", "4", "5", "6", "7", "8",
"9", "A", "B", "C", "D", "E", "F", "G", "H", "J",
"K", "L", "M", "N", "P", "Q", "R", "S", "T", "U",
"V", "W", "X", "Y", "Z"]
def AddSmudginess(img, Smu):
"""
模糊处理
:param img: 输入图像
:param Smu: 模糊图像
:return: 添加模糊后的图像
"""
rows = r(Smu.shape[0] - 20) #50
cols = r(Smu.shape[1] - 20)
adder = Smu[rows:rows + 20, cols:cols + 20]
adder = cv.resize(adder, (20, 20))
img = cv.resize(img,(20,20))
img = cv.bitwise_not(img)
img = cv.bitwise_and(adder, img)
img = cv.bitwise_not(img)
return img
def rot(img, angel, shape, max_angel):
"""
添加透视畸变
"""
size_o = [shape[1], shape[0]]
size = (shape[1]+ int(shape[0] * cos((float(max_angel ) / 180) * 3.14)), shape[0])
interval = abs(int(sin((float(angel) / 180) * 3.14) * shape[0]))
pts1 = np.float32([[0, 0], [0, size_o[1]], [size_o[0], 0], [size_o[0], size_o[1]]])
if angel > 0:
pts2 = np.float32([[interval, 0], [0, size[1]], [size[0], 0], [size[0] - interval, size_o[1]]])
else:
pts2 = np.float32([[0, 0], [interval, size[1]], [size[0] - interval, 0], [size[0], size_o[1]]])
M = cv.getPerspectiveTransform(pts1, pts2)
dst = cv.warpPerspective(img, M, size)
return dst
def rotRandrom(img, factor, size):
"""
添加放射畸变
:param img: 输入图像
:param factor: 畸变的参数
:param size: 图片目标尺寸
:return: 放射畸变后的图像
"""
shape = size
pts1 = np.float32([[0, 0], [0, shape[0]], [shape[1], 0], [shape[1], shape[0]]])
pts2 = np.float32([[r(factor), r(factor)], [r(factor), shape[0] - r(factor)], [shape[1] - r(factor), r(factor)],
[shape[1] - r(factor), shape[0] - r(factor)]])
M = cv.getPerspectiveTransform(pts1, pts2)
dst = cv.warpPerspective(img, M, size)
return dst
def tfactor(img):
"""
添加饱和度光照的噪声
"""
hsv = cv.cvtColor(img,cv.COLOR_BGR2HSV)
hsv[:, :, 0] = hsv[:, :, 0] * (0.8 + np.random.random() * 0.2)
hsv[:, :, 1] = hsv[:, :, 1] * (0.3 + np.random.random() * 0.7)
hsv[:, :, 2] = hsv[:, :, 2] * (0.2 + np.random.random() * 0.8)
img = cv.cvtColor(hsv, cv.COLOR_HSV2BGR)
return img
def random_envirment(img, noplate_bg):
"""
添加自然环境的噪声, noplate_bg为不含车牌的背景图
"""
bg_index = r(len(noplate_bg))
env = cv.imread(noplate_bg[bg_index])
env = cv.resize(env, (img.shape[1], img.shape[0]))
bak = (img == 0)
bak = bak.astype(np.uint8) * 255
inv = cv.bitwise_and(bak, env)
img = cv.bitwise_or(inv, img)
return img
def GenCh(f, val):
"""
生成中文字符
"""
img = Image.new("RGB", (45, 70), (255, 255, 255)) #白色
# img = Image.new("RGB", (45, 70), (0, 0, 0)) #黑色
draw = ImageDraw.Draw(img)
# draw.text((0, 3), val, (0, 0, 0), font=f)
draw.text((0, 2), val, (0, 0, 0), font=f)
img = img.resize((19, 70))
A = np.array(img)
return A
def GenCh1(f, val):
"""
生成英文字符
"""
img =Image.new("RGB", (19, 70), (255,255,255))#白色
# img =Image.new("RGB", (23, 70), (0, 0, 0))#黑色
draw = ImageDraw.Draw(img)
draw.text((0, 2), val, (0,0,0), font=f) # val.decode('utf-8')
A = np.array(img)
return A
def AddGauss(img, level):
"""
添加高斯模糊
"""
return cv.blur(img, (level * 2 + 1, level * 2 + 1))
def r(val):
return int(np.random.random() * val)
def AddNoiseSingleChannel(single):
"""
添加高斯噪声
"""
diff = 255 - single.max()
noise = np.random.normal(0, 1 + r(6), single.shape)
noise = (noise - noise.min()) / (noise.max() - noise.min())
noise *= diff
# noise= noise.astype(np.uint8)
dst = single + noise
return dst
def addNoise(img): # sdev = 0.5,avg=10
img[:, :, 0] = AddNoiseSingleChannel(img[:, :, 0])
img[:, :, 1] = AddNoiseSingleChannel(img[:, :, 1])
img[:, :, 2] = AddNoiseSingleChannel(img[:, :, 2])
return img
class GenPlate:
def __init__(self, fontCh, fontEng, NoPlates):
self.fontC = ImageFont.truetype(fontCh, 43, 0)
self.fontE = ImageFont.truetype(fontEng, 52, 0) #60--52
self.img = np.array(Image.new("RGB", (226, 70),(255, 255, 255)))
# self.img = np.array(Image.new("RGB", (226, 70),(0, 0, 0)))
self.bg = cv.resize(cv.imread(r"C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\generate_picture\images\black.jpg"), (226, 70))
# template.bmp:车牌背景图
self.smu = cv.imread(r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\images\\smu2.jpg")
# smu2.jpg:模糊图像
self.noplates_path = []
for parent, parent_folder, filenames in os.walk(NoPlates):
for filename in filenames:
path = parent + "\\" + filename
self.noplates_path.append(path)
def draw(self, val):
offset = 2
self.img[0:70, offset+6:offset+6+19] = GenCh(self.fontC, val[0])
self.img[0:70, offset+6+19+7:offset+6+19+7+19] = GenCh1(self.fontE, val[1])
for i in range(6):
base = offset + 6 + 19 + 7 + 19 + 17 + i * 19 + i * 7
self.img[0:70, base:base+19] = GenCh1(self.fontE, val[i+2]) #23--19 6--7 8--6
return self.img
def generate(self, text):
if len(text) == 8:
fg = self.draw(text) # decode(encoding="utf-8")
fg = cv.bitwise_not(fg)
com = cv.bitwise_or(fg, self.bg)
# com = rot(com, r(60)-30, com.shape,30)
# com = rotRandrom(com, 10, (com.shape[1], com.shape[0]))
# com = tfactor(com)
# com = random_envirment(com, self.noplates_path)
# com = AddGauss(com, 1+r(4))
# com = addNoise(com)
return com
@staticmethod
def genPlateString(pos, val):
"""
生成车牌string，存为图片
生成车牌list，存为label
"""
plateStr = ""
plateList=[]
box = [0, 0, 0, 0, 0, 0, 0,0]
if pos != -1:
box[pos] = 1
for unit, cpos in zip(box, range(len(box))):
if unit == 1:
plateStr += val
plateList.append(val)
else:
if cpos == 0:
plateStr += chars[r(31)] #31
plateList.append(plateStr)
elif cpos == 1:
plateStr += chars[41 + r(24)] #41 24
plateList.append(plateStr)
else:
plateStr += chars[31 + r(34)] #31 34
plateList.append(plateStr)
plate = [plateList[0]]
b = [plateList[i][-1] for i in range(len(plateList))]
plate.extend(b[1:8])
return plateStr, plate
@staticmethod
def genBatch(batchsize, outputPath, size):
"""
将生成的车牌图片写入文件夹，对应的label写入label.txt
:param batchsize: 批次大小
:param outputPath: 输出图像的保存路径
:param size: 输出图像的尺寸
:return: None
"""
if not os.path.exists(outputPath):
os.mkdir(outputPath)
outfile = open(r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\label_new.txt',
'w', encoding='utf-8')
for i in range(batchsize):
plateStr, plate = G.genPlateString(-1, -1)
img = G.generate(plateStr)
img = cv.resize(img, size)
content = "".join(plate)
cv.imwrite(outputPath + "\\" + str(i).zfill(2) + ".jpg", img)
outfile.write("./train/"+str(i).zfill(2)+".jpg"+'\t'+content +"\n")
G = GenPlate(r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\font\platech.ttf",
r'E:\AI_projects\vehicle license plate recognition\car_recognition\data\font\platechar.ttf',
r"E:\AI_projects\vehicle license plate recognition\car_recognition\data\NoPlates")
G.genBatch(1000, r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_new', (272, 72))
以上为车牌生成部分代码，其他代码参考之前的部分
(4) 重命名各种车牌名称，使得它适合训练便签生成要求
#将以数字命名的图片转换为省份编号+车牌命名
import os
import pandas as pd
path=r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\label_yellow.txt'
num=[]
with open(path,'rb') as f:
lines=f.readlines()
for line in lines:
data=line.decode('utf-8')#.encode('gb2312')
data1=data.split('\t')[0]
data2=data.split('\t')[1].split('\r')[0]
num.append([data1,data2])
number=pd.DataFrame(num,columns=['p','j'])
CHARS = ['京', '沪', '津', '渝', '冀', '晋', '蒙', '辽', '吉', '黑',
'苏', '浙', '皖', '闽', '赣', '鲁', '豫', '鄂', '湘', '粤',
'桂', '琼', '川', '贵', '云', '藏', '陕', '甘', '青', '宁',
'新',
'0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'J', 'K',
'L', 'M', 'N', 'P', 'Q', 'R', 'S', 'T', 'U', 'V',
'W', 'X', 'Y', 'Z','_'
]
dicts = {'A01':'京','A02':'津','A03':'沪','B02':'蒙',
'S01':'皖','S02':'闽','S03':'粤','S04':'甘',
'S05': '贵', 'S06': '鄂', 'S07': '冀', 'S08': '黑', 'S09': '湘',
'S10': '豫', 'S12': '吉', 'S13': '苏', 'S14': '赣', 'S15': '辽',
'S17': '川', 'S18': '鲁', 'S22': '浙',
'S30':'渝', 'S31':'晋', 'S32':'桂', 'S33':'琼', 'S34':'云', 'S35':'藏',
'S36':'陕','S37':'青', 'S38':'宁', 'S39':'新'}
path_root=r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_yellow'
for k in range(len(number)):
for t in range(len(CHARS)):
pp=number['p'][k].split('/train/')[1]
if number['j'][k][0]==CHARS[t]:
original_name=os.path.join(path_root,pp)
d=list(dicts.keys())[t]
pd=number['j'][k]
dd=d+'_'+pd[1:]+'_0'+'.jpg'
post_name=os.path.join(path_root,dd)
if os.path.exists(original_name):
if not os.path.exists(post_name):
os.rename(original_name,post_name)
(5) 统计各个省份的车牌数量
主要是确认中文字符数量均衡
import os
path=r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\picture_real'
S=[]
for file in os.listdir(path):
a=file.split('_')[0]
S.append(a)
Set=set(S)
Dict={}
for item in Set:
Dict.update({item:S.count(item)})
print(Dict)
(6) 把车牌移动到训练和测试两个文件夹中
import os
import shutil
path_from=r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test_in_new'
path_to=r'C:\Users\lenovo\Desktop\Plate_Recognition-LPRnet-master\Plate_Recognition-LPRnet-master\test'
for file in os.listdir(path_from):
image=os.path.join(path_from,file)
# if not os.path.join(path_to,file):
shutil.move(image,path_to)
(7) 模型训练及测试
接下来进行模型训练。。。 （略）
总结
用第三种方法 LPRNet 来进行车牌识别这个项目是一个很不错的，涉及到图像处理，深度学习等各种图像相关的知识，完成这个项目训练后，项目工程能力得到较大提升，对于以后找工作和工作中遇到的问题处理都有比较大的帮助。
能认真看到这的小伙伴帮忙点个赞！