opencv信用卡卡号识别

一.程序预处理

在正式进行卡号识别编写程序时，我们首先需要知道我们需要使用到的函数。
我们需要用到图像的二值化处理，轮廓筛选，形态学变化，数字截取与排序。
**二值化处理：**图像的二值化类型分为全局阈值和自适应阈值，这里我们用到的时自适应阈值，全局阈值无法得到我们想要的效果。

adapt=cv2.adaptiveThreshold(src,maxVal,adaptiveMethod,type,blockSize,c)

上面为自定义阈值的模板，第一个参数时图片，阈值的最大值，自适应阈值，处理类型，后面是两个常数。
下面的代码如下：

# -*- coding =utf-8 -*-
# @Time :2022/7/1 8:51
# @Author :CHINA
# @File :卡号识别.py
# @software:PyCharm
import cv2
import numpy as np
card_gray=cv2.imread("F:\\7.1\\card .jpg",cv2.IMREAD_GRAYSCALE)
card_gray4=cv2.resize(card_gray,(2*card_gray.shape[1],2*card_gray.shape[0]))
adapt=cv2.adaptiveThreshold(card_gray4,255,cv2.ADAPTIVE_THRESH_MEAN_C,cv2.THRESH_BINARY_INV,13,3)
con,hiera=cv2.findContours(adapt,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
for i in range(len(con)):
    if cv2.contourArea(con[i])<160:
        adapt=cv2.drawContours(adapt,con,i,(0,0,0),-1)
ker=np.ones((15,15),dtype=np.uint8)
blackhat=cv2.morphologyEx(adapt,cv2.MORPH_BLACKHAT,ker)

ker=np.ones((3,3),dtype=np.uint8)
opening=cv2.morphologyEx(blackhat,cv2.MORPH_OPEN,ker)
con,hiera=cv2.findContours(opening,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
for i in range(len(con)):
    x,y,w,h=cv2.boundingRect(con[i])
    ratio=float(w)/h
    Area=w*h
    if Area<1800 or Area>6000:
        opening = cv2.drawContours(opening, con, i, (0, 0, 0), -1)
    else:
        if ratio>0.7 or ratio<0.57:
            opening = cv2.drawContours(opening, con, i, (0, 0, 0), -1)
cv2.imshow("adapt4",opening)
ker=np.ones((3,3),dtype=np.uint8)
opening=cv2.morphologyEx(opening,cv2.MORPH_DILATE,ker)
cv2.imshow("adapt5",opening)
model=cv2.imread("F:\\7.1\\model.jpg",cv2.IMREAD_GRAYSCALE)
ret,model=cv2.threshold(model,200,255,cv2.THRESH_BINARY)
cv2.imshow("adapt6",model)
def squence(image,width,height):
    con, hiera = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    n=len(con)
    Reactbox0=np.ones((n,4),dtype=int)
    for i in range(n):
        Reactbox0[i]=cv2.boundingRect(con[i])
    Reactbox= np.ones((n, 4), dtype=int)
    for i in range (n):
        p=0
        for j in range(n):
            if Reactbox0[i][0]>Reactbox0[j][0]:
                p=p+1
        Reactbox[p]=Reactbox0[i]
    Imgbox=[[]for i in range (n)]
    for i in range (n):
        x,y,w,h=Reactbox[i]
        ROI=image[y:y+h,x:x+w]
        ROI = cv2.resize(ROI,(width,height))
        thresh,ROI=cv2.threshold(ROI,200,255,cv2.THRESH_BINARY)
        Imgbox[i]=ROI

    return Reactbox,Imgbox
Reactbox_Temp,Imgbox_Temp=squence(model,50,80)
print(Reactbox_Temp)
cv2.imshow("adapt7",Imgbox_Temp[1])
Reactbox,Imgbox=squence(opening,50,80)
for i in range (len(Imgbox)):
    score=np.zeros(len(Imgbox_Temp),dtype=int)
    for j in range(len(Imgbox_Temp)):
        score[j]=cv2.matchTemplate(Imgbox[i],Imgbox_Temp[j],cv2.TM_SQDIFF)
    min_val,max_val,min_indx,max_indx=cv2.minMaxLoc(score)
    print(min_indx[1])
cv2.waitKey(0)

这是运行后的结果。