python+cv2+zbar实现二维码识别

安装库

安装 pyzbar：https://pypi.org/project/zbar/

sudo apt-get install libzbar0
pip install pyzbar

安装 opencv

pip install opencv-python

完整代码

import cv2
from pyzbar.pyzbar import decode
import numpy as np

def IsQrRate(rate):
    return np.logical_and(rate>0.15, rate<1.9)

#横向黑白比例判断
def IsQrColorRateX(image, flag):
    # 默认处理的是二值化之后的图
    nr = int(image.shape[0] / 2)
    nc = image.shape[1]
    bar = image[nr,:]
    edge = np.where(bar[:-1]!=bar[1:])[0]
    edge = np.array([-1] + list(edge) + [nc])
    count = edge[1:] - edge[:-1]   # 每个区域的长度
    if len(count)<5:   # 至少有5个区域，考虑到切割图象边界处可能会出现7个区域
        return False
    
    ## 横向黑白比例1:1:3:1:1
    maxCount = np.max(count)
    maxIdx = np.argmax(count)
    if maxIdx<2 or maxIdx>len(count)-3:
        return False
    rate = np.concatenate([count[maxIdx-2:maxIdx],count[maxIdx+1:maxIdx+3]],axis=0)/maxCount
    rate = IsQrRate(rate)
    
    return False not in rate

#纵向黑白比例判断
def IsQrColorRateY(image, flag):
    # 默认处理的是二值化之后的图
    nc = int(image.shape[1] / 2)
    nr = image.shape[0]
    bar = image[:,nc]
    edge = np.where(bar[:-1]!=bar[1:])[0]
    edge = np.array([-1] + list(edge) + [nr])
    count = edge[1:] - edge[:-1]   # 每个区域的长度
    if len(count)<5:   # 至少有5个区域，考虑到切割图象边界处可能会出现7个区域
        return False
    
    ## 纵向黑白比例1:1:3:1:1
    maxCount = np.max(count)
    maxIdx = np.argmax(count)
    if maxIdx<2 or maxIdx>len(count)-3:
        return False
    rate = np.concatenate([count[maxIdx-2:maxIdx],count[maxIdx+1:maxIdx+3]],axis=0)/maxCount
    rate = IsQrRate(rate)
    
    return False not in rate

# 判断是横纵两个方向比例
def IsQrColorRate(image, flag):
    x = IsQrColorRateX(image, flag)
    if not x:
        return False
    y = IsQrColorRateY(image, flag)
    return y

# 二维码切割
def CropImage(img, rotatedRect):
    points = cv2.boxPoints(rotatedRect)
    R = np.sum(points**2, axis=1)
    topLeftIdx = np.argmin(R)
    topLeftR = np.min(R)
    x1,y1 = points[(topLeftIdx + 1) % 4] - points[topLeftIdx]
    x2,y2 = points[(topLeftIdx + 3) % 4] - points[topLeftIdx]
    w = np.sqrt(x1*x1+y1*y1).astype(np.int32)
    h = np.sqrt(x2*x2+y2*y2).astype(np.int32)
    if img.ndim == 3:
        crop = np.zeros((h,w,3),dtype=np.uint8)
    else:
        crop = np.zeros((h,w),dtype=np.uint8)
    for j in range(h):
        for i in range(w):
            kx,ky = i/w, j/h
            x = int(points[topLeftIdx][0]+ kx*x1 + ky*x2)
            y = int(points[topLeftIdx][1]+ kx*y1 + ky*y2)
            crop[j, i] = img[y, x]
    return crop
    
# 判断是否为二维码定位角
def IsQrPoint(contour, img, i):
    rotatedRect = cv2.minAreaRect(contour)
    # 最小大小限定
    if rotatedRect[1][0]<10 or rotatedRect[1][1]<10:
        return False
    # 将二维码从图中抠出来
    imgCrop = CropImage(img, rotatedRect)
    flag = i
    # 黑白比例1:1:3:1:1
    result = IsQrColorRate(imgCrop, flag)
    return result

# 二维码解码
def QrDecode(img):
    string = decode(img)
    return string

def FindQr(imgSrc):
    qrPoint = []

    # 彩色图转灰度图
    imgGray = cv2.cvtColor(imgSrc,cv2.COLOR_BGR2GRAY)

    # 高斯平滑滤波
    imgBlur=cv2.GaussianBlur(imgGray,(3,3),0)

    # 腐蚀
    element1 = cv2.getStructuringElement(0,(3,3))
    imgSobel = cv2.erode(imgBlur, element1)

    # 二值化
    ret, imgBinary = cv2.threshold(imgSobel, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
    imgBinaryClone = np.copy(imgBinary)

    # 查找轮廓
    binary, contours, hierarchy = cv2.findContours(imgBinary, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
    imgSrcClone = np.copy(imgSrc)
    for cont in contours:
        x,y,w,h = cv2.boundingRect(cont)
        imgSrcClone = cv2.rectangle(imgSrcClone, (x,y), (x+w,y+h), color=(0,0,255), thickness=2)

    # 筛选定位角：黑色定位角满足父轮廓有两个子轮廓
    for i in range(len(contours)):
        k = i
        ic = 0
        parentIdx = i
        # 计算子轮廓深度
        while hierarchy[0][k][2] != -1:
            k = hierarchy[0][k][2]
            ic = ic + 1

        # 判断是否为定位角
        if ic>=2:
            isQr = IsQrPoint(contours[parentIdx], imgBinary, parentIdx)
            print(parentIdx, isQr)
            if isQr:
                qrPoint.append(contours[parentIdx])

    # 绘制二维码定位角
    for point in qrPoint:
        x,y,w,h = cv2.boundingRect(point)
        cv2.rectangle(imgSrc, (x,y), (x+w,y+h), (255,0,0), thickness=2)

    # 根据相邻三个定位角确定二维码整体位置框
    qrCenter = []
    state = [0]*len(qrPoint)
    qrBox = []
    qrCode = []
    # 计算定位角质心
    for i in range(len(qrPoint)):
        center = np.sum(qrPoint[i], axis=0) / qrPoint[i].shape[0]
        qrCenter.append(center)
    # 判断是否构成一个二维码
    for i in range(len(qrPoint)):
        if state[i]:
            continue
        for j in range(len(qrPoint)):
            if j==i or state[j]:
                continue
            for k in range(len(qrPoint)):
                if k==j or k==i or state[k]:
                    continue
                Dij = (np.sum((qrCenter[i]-qrCenter[j])**2))
                Dik = (np.sum((qrCenter[i]-qrCenter[k])**2))
                Djk = (np.sum((qrCenter[k]-qrCenter[j])**2))
                ratio = Dij / Dik
                ratio1 = (Dij + Dik) / Djk
                if ratio>0.6 and ratio<1.6 and ratio1>0.85 and ratio1<1.15:
                    state[i]=1
                    state[j]=1
                    state[k]=1
                    contour = np.concatenate([qrPoint[i],qrPoint[j],qrPoint[k]], axis=0)
                    rotatedRect = cv2.minAreaRect(contour)
                    box = cv2.boxPoints(rotatedRect)
                    qrBox.append(np.int0(box))
                    imgTemp = CropImage(imgGray, rotatedRect)
                    qrCodeTemp = QrDecode(imgTemp)[0].data
                    qrCode.append(qrCodeTemp)
    imgQR = cv2.drawContours(imgSrc, qrBox, -1, (0,0,255), thickness=2)
    #cv2.namedWindow('QR Code', 0)
    #cv2.resizeWindow('QR Code',1000, 600)
    #cv2.imshow("QR Code",imgQR)
    #cv2.waitKey(5000)
    
    return qrCode

    
if __name__ == '__main__':
    image = cv2.imread('./imgs/14.jpg')
    height,width,_ = image.shape
    print("%d X %d"%(height,width))
    qrcode = FindQr(image)
    print("Detected QR Code:")
    print(qrcode)