车牌识别过程 图片导入 图片灰度 车牌定位 车牌识别 分析 车牌的大小固定，宽高比在2~5.5之间 这是一个十分重要的训练数据。
大致流程：
颜色查找（蓝色区域）高斯模糊灰度化二阈值处理边缘处理闭运算矩形筛选车牌切割车牌识别（调用百度aip实现）
调用百度aip请点击‘百度aip’进行注册获取对应的所需数据 颜色查找代码放在下方：
def color_find(img): #寻找蓝色,其他颜色均掩码 boundaries = [([86, 31, 4], [220, 88, 50])] #定义边界列表(lower[r, g, b], upper[r, g, b]) for (lower,upper) in boundaries: #遍历所有的边界 lower = np.array(lower, dtype = "uint8") upper = np.array(upper, dtype = "uint8") mask = cv2.inRange(img, lower, upper) #在指定边界内查找颜色并掩码 result_color_find = cv2.bitwise_and(img, img, mask = mask) return result_color_find
接下来是车牌识别的代码：
def pre_img(result_color_find,img): #对颜色处理后的车牌进行预处理 os.mkdir('license') img_gaus=cv2.GaussianBlur(result_color_find,(1,1),9) #高斯模糊 img_gray=cv2.cvtColor(img_gaus,cv2.COLOR_BGR2GRAY) #灰度化 ret,img_binary=cv2.threshold(img_gray,127,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU) #二阈值处理 img_canny=cv2.Canny(img_binary, 200, 255) #边缘处理 kernel=cv2.getStructuringElement(cv2.MORPH_RECT,(50,20)) img_close=cv2.morphologyEx(img_canny,cv2.MORPH_CLOSE,kernel) #闭运算 contours,hierarchy=cv2.findContours(img_close,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE) result_pre_img=[] for i in range(len(contours)): x,y,w,h=cv2.boundingRect(contours[i]) ratio=w/h if ratio > 2 and ratio < 5.5: #筛选出宽高比在2~5.5之间的矩形 cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),1) result=img[y:y+h,x:x+w] #预处理 pre_result=cv2.resize(result,(200,50)) result_pre_img.append(pre_result) for o in range(len(result_pre_img)): cv2.imwrite('license/license%d.jpg'%(o+1), result_pre_img[o]) return o+1 以上代码需调用： import cv2import numpy as npimport osimport shutil**
附上ui界面源码：
界面非原创设计，界面参考
import tkinter as tkfrom tkinter import filedialogfrom tkinter import *from tkinter import Button,Label,filedialogimport License_importimport cv2from PIL import Image,ImageTkfrom aip import AipOcrimport shutilimport ossrcPath=''api='填写你的百度aip'api_id='这里写id'api_key='这里写apikey值'secret_key='这里写secretkey值'license_dir='license/'client = AipOcr(api_id, api_key, secret_key)result=[]window=tk.Tk()window.title('车牌识别(LPR)')window.geometry('1000x600')canvas_pic = tk.Canvas(window, width=700, height=600)canvas_pic.pack(side='right')canvas_plate = tk.Canvas(window, width=200, height=50)canvas_plate.place(x=10, y=320)def resize( resize_width, resize_height, pil_image): w, h = pil_image.size f1 = 1.0*resize_width/w f2 = 1.0*resize_height/h factor = min([f1, f2]) width = int(w*factor) height = int(h*factor) return pil_image.resize((width, height), Image.ANTIALIAS)def printcoords(): File = filedialog.askopenfilename(parent=window, title='请选择一张图片') global srcPath srcPath=File src = Image.open(File) resizePic=resize(700,600,src) filename=ImageTk.PhotoImage(resizePic) canvas_pic.image=filename canvas_pic.create_image(350,300,image=filename)btn_choosePic = tk.Button(window, text='选择图片',command=printcoords).place(x=40, y=200)def location(): if os.path.exists('license'): shutil.rmtree('license') global pictureName img=cv2.imread(srcPath) result_file=License_import.color_find(img) for i in range(License_import.pre_img(result_file,img)): image = get_file_content(license_dir+'license%d.jpg'%(i+1)) src = Image.open(license_dir+'license%d.jpg'%(i+1)) text=client.basicGeneral(image) for j in text.get('words_result'): var_returnTxt.set(j['words']) resizePic = resize(200, 50, src) filename = ImageTk.PhotoImage(resizePic) canvas_plate.image = filename canvas_plate.create_image(100, 25, image=filename)btn_location = tk.Button(window, text='开始识别', command=location).place(x=40, y=250)tk.Label(window, text='车牌定位结果: ', width=18).place(x=20, y= 300)tk.Label(window, text='车牌识别结果: ', width=18).place(x=20, y= 450)tk.Label(window, text='车牌识别步骤:n' '1.请选择需要识别的图片n' '2.请单击开始识别按钮n' '3.车牌和识别结果将会被显示至下方nnn' '该识别软件只能识别单个车牌n' '如需识别多个车牌的图请裁剪图片后导入').place(x=0, y=18)var_returnTxt = tk.StringVar()var_returnTxt.set("等待识别......")tk.Label(window,textvariable=var_returnTxt,font=('Arial', 11),width=18, height=8).place(x=20, y=470)def get_file_content(filePath): with open(filePath, 'rb') as fp: return fp.read()if __name__ == '__main__': window.mainloop()
附上效果图：