利用opencv对拍摄图片进行文字识别

python 复制代码

import cv2 as cv
import numpy as np
import pytesseract
from PIL import Image

img = cv.imread('test.jpg')
rows, cols, _ = img.shape
img = cv.resize(img, (int(cols/2), int(rows/2)))
img = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
nrows, ncols = img.shape
print(cols, ncols, rows, nrows)
gray_blurred = cv.GaussianBlur(img, (5, 5), 0)

flag = 200

lines = []
while len(lines) != 4:
    # 使用Canny边缘检测
    edges = cv.Canny(gray_blurred, 50, 150, apertureSize=3)
    lines = cv.HoughLines(edges, 1, np.pi / 180, flag)
    if lines is None:
        lines = []
    if flag < 80:
        raise Exception('未找到合适的边缘处理参数')
    flag -= 5
print(flag)
nlines = []
# 如果找到了直线，使用它们来计算仿射变换矩阵
if lines is not None:
    for rho, theta in lines[:, 0]:
        a = np.cos(theta)
        b = np.sin(theta)
        x0 = a * rho
        y0 = b * rho
        x1 = int(x0 + 1000 * (-b))
        y1 = int(y0 + 1000 * (a))
        x2 = int(x0 - 1000 * (-b))
        y2 = int(y0 - 1000 * (a))
        cv.line(img, (x1, y1), (x2, y2), (0, 0, 255), 2)
        nlines.append([(x1, y1), (x2, y2)])
points = []
for i in range(len(nlines) - 1):
    for j in range(i + 1, len(nlines)):
        line = nlines[i]
        x1, y1 = line[0]
        x2, y2 = line[1]
        line1 = nlines[j]
        x3, y3 = line1[0]
        x4, y4 = line1[1]
        try:
            u = ((x4-x3)*(y1-y3) - (y4-y3)*(x1-x3)) / ((y4-y3)*(x2-x1) - (x4-x3)*(y2-y1))
        except Exception as e:
            continue
        x = x1 + u * (x2 - x1)
        y = y1 + u * (y2 - y1)
        if x > 0 and y > 0 and x < ncols and y < nrows:
            points.append((x, y))
pytesseract.pytesseract.tesseract_cmd = r'D:\Program Files\Tesseract-OCR\tesseract.exe'
center = (int(ncols/2), int(nrows/2))
pstmap = {}
for point in points:
    x, y = point
    cx, cy = center
    if x < cx and y < cy:
        pstmap['lt'] = point
    elif x > cx and y < cy:
        pstmap['rt'] = point
    elif x > cx and y > cy:
        pstmap['rb'] = point
    else:
        pstmap['lb'] = point

pst1 = np.float32([pstmap['lt'], pstmap['rt'], pstmap['rb'], pstmap['lb']])
pst2 = np.float32([[0, 0], [ncols, 0], [ncols, nrows], [0, nrows]])
M = cv.getPerspectiveTransform(pst1, pst2)
dst = cv.warpPerspective(img, M, (ncols, nrows))

x1, y1 = 0, 0
def mouse_callback(event, x, y, flags, param):
    global x1, y1
    if event == cv.EVENT_LBUTTONDOWN:
        x1, y1 = x, y
    elif event == cv.EVENT_LBUTTONUP:
        x2, y2 = x, y
        wimg = dst[y1:y2, x1:x2]
        _, wimg = cv.threshold(wimg, 80, 255, cv.THRESH_BINARY)
        wimg = cv.bitwise_not(wimg)
        cv.imwrite('test_dst.jpg', wimg)
        image = Image.open('test_dst.jpg')
        # 打印选定区域的坐标
        print(f"({x1}, {y1}) -> ({x2}, {y2})")
        print(pytesseract.image_to_string(image, lang='chi_sim'))
cv.namedWindow('dst')
cv.setMouseCallback("dst", mouse_callback)
cv.imshow('img', img)
cv.imshow('dst', dst)
print(dst[2])
cv.waitKey(0)
cv.destroyAllWindows()

方法：

首先读取图片，因为我手机拍摄图片尺寸太大，所以进行了缩放
对图片进行高斯模糊，方便进行边缘处理
从高到低适配不同的阈值检测图片内容边缘
通过反向霍夫变换获取确定边缘直线的四个点
通过直线两两相交确定四个定点
进行透视变换
添加鼠标事件，监测鼠标选定区域
鼠标选定区域后，裁剪图片，对图片进行二值化处理，我这里做了文字黑白反转
利用pytesseract对裁剪后的图片进行文字识别

注意事项：

选择的文字区域会影响识别成功率，如果文字区域紧贴文字，可能会失败，盲猜影响了特征提取
图片尺寸大小会影响边缘检测，不缩放图片时，阈值调整不当的话，很容易生成N条边缘直线，阈值怎么选定请了解霍夫变换的原理。

识别效果（加了二值化处理的准确度会很好）：