# -*- coding: utf-8 -*- # @Time : 19-11-19 22:25 # @Author : Miao Wenqiang # @Reference : None # @File : cut_twist_join.py # @IDE : PyCharm Community Edition """ 将身份证正反面从原始图片中切分出来。 需要的参数有: 1.图片所在路径。 输出结果为: 切分后的身份证正反面图片。 """ import os import cv2 import numpy as np def point_judge(center, bbox): """ 用于将矩形框的边界按顺序排列 :param center: 矩形中心的坐标[x, y] :param bbox: 矩形顶点坐标[[x1, y1], [x2, y2], [x3, y3], [x4, y4]] :return: 矩形顶点坐标,依次是 左下, 右下, 左上, 右上 """ left = [] right = [] for i in range(4): if bbox[i][0] > center[0]: # 只要是x坐标比中心点坐标大,一定是右边 right.append(bbox[i]) else: left.append(bbox[i]) if right[0][1] > right[1][1]: # 如果y点坐标大,则是右上 right_down = right[1] right_up = right[0] else: right_down = right[0] right_up = right[1] if left[0][1] > left[1][1]: # 如果y点坐标大,则是左上 left_down = left[1] left_up = left[0] else: left_down = left[0] left_up = left[1] return left_down, right_down, left_up, right_up def gray_and_fliter(img, image_name='1.jpg', save_path='./'): # 转为灰度图并滤波,后面两个参数调试用 """ 将图片灰度化,并滤波 :param img: 输入RGB图片 :param image_name: 输入图片名称,测试时使用 :param save_path: 滤波结果保存路径,测试时使用 :return: 灰度化、滤波后图片 """ # img = cv2.imread(image_path + image_name) # 读取图片 img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 转换为灰度图片 # cv2.imwrite(os.path.join(save_path, image_name + '_gray.jpg'), img_gray) # 保存,方便查看 img_blurred = cv2.filter2D(img_gray, -1, kernel=np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]], np.float32)) # 对图像进行滤波,是锐化操作 img_blurred = cv2.filter2D(img_blurred, -1, kernel=np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]], np.float32)) # cv2.imwrite(os.path.join(save_path, img_name + '_blurred.jpg'), img_blurred) # 锐化, 这里的卷积核可以更改 return img_blurred def gradient_and_binary(img_blurred, image_name='1.jpg', save_path='./'): # 将灰度图二值化,后面两个参数调试用 """ 求取梯度,二值化 :param img_blurred: 滤波后的图片 :param image_name: 图片名,测试用 :param save_path: 保存路径,测试用 :return: 二值化后的图片 """ gradX = cv2.Sobel(img_blurred, ddepth=cv2.CV_32F, dx=1, dy=0) gradY = cv2.Sobel(img_blurred, ddepth=cv2.CV_32F, dx=0, dy=1) img_gradient = cv2.subtract(gradX, gradY) img_gradient = cv2.convertScaleAbs(img_gradient) # sobel算子,计算梯度, 也可以用canny算子替代 # 这里改进成自适应阈值,貌似没用 img_thresh = cv2.adaptiveThreshold(img_gradient, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 3, -3) # cv2.imwrite(os.path.join(save_path, img_name + '_binary.jpg'), img_thresh) # 二值化 阈值未调整好 kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5, 5)) img_closed = cv2.morphologyEx(img_thresh, cv2.MORPH_CLOSE, kernel) img_closed = cv2.morphologyEx(img_closed, cv2.MORPH_OPEN, kernel) img_closed = cv2.erode(img_closed, None, iterations=9) img_closed = cv2.dilate(img_closed, None, iterations=9) # 腐蚀膨胀 # 这里调整了kernel大小(减小),腐蚀膨胀次数后(增大),出错的概率大幅减小 return img_closed def find_bbox(img, img_closed): # 寻找身份证正反面区域 """ 根据二值化结果判定并裁剪出身份证正反面区域 :param img: 原始RGB图片 :param img_closed: 二值化后的图片 :return: 身份证正反面区域 """ (contours, _) = cv2.findContours(img_closed.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE) # 求出框的个数 # 这里opencv如果版本不对(4.0或以上)会报错,只需把(contours, _)改成 (_, contours, _) contours = sorted(contours, key=cv2.contourArea, reverse=True) # 按照面积大小排序 countours_res = [] for i in range(0, len(contours)): area = cv2.contourArea(contours[i]) # 计算面积 if (area <= 0.4 * img.shape[0] * img.shape[1]) and (area >= 0.05 * img.shape[0] * img.shape[1]): # 人为设定,身份证正反面框的大小不会超过整张图片大小的0.4,不会小于0.05(这个参数随便设置的) rect = cv2.minAreaRect(contours[i]) # 最小外接矩,返回值有中心点坐标,矩形宽高,倾斜角度三个参数 box = cv2.boxPoints(rect) left_down, right_down, left_up, right_up = point_judge([int(rect[0][0]), int(rect[0][1])], box) src = np.float32([left_down, right_down, left_up, right_up]) # 这里注意必须对应 dst = np.float32([[0, 0], [int(max(rect[1][0], rect[1][1])), 0], [0, int(min(rect[1][0], rect[1][1]))], [int(max(rect[1][0], rect[1][1])), int(min(rect[1][0], rect[1][1]))]]) # rect中的宽高不清楚是个怎么机制,但是对于身份证,肯定是宽大于高,因此加个判定 m = cv2.getPerspectiveTransform(src, dst) # 得到投影变换矩阵 result = cv2.warpPerspective(img, m, (int(max(rect[1][0], rect[1][1])), int(min(rect[1][0], rect[1][1]))), flags=cv2.INTER_CUBIC) # 投影变换 countours_res.append(result) return countours_res # 返回身份证区域 def find_cut_line(img_closed_original): # 对于正反面粘连情况的处理,求取最小点作为中线 """ 根据规则,强行将粘连的区域切分 :param img_closed_original: 二值化图片 :return: 处理后的二值化图片 """ img_closed = img_closed_original.copy() img_closed = img_closed // 250 #print(img_closed.shape) width_sum = img_closed.sum(axis=1) # 沿宽度方向求和,统计宽度方向白点个数 start_region_flag = 0 start_region_index = 0 # 身份证起始点高度值 end_region_index = 0 # 身份证结束点高度值 for i in range(img_closed_original.shape[0]): # 1000是原始图片高度值,当然, 这里也可以用 img_closed_original.shape[0]替代 if start_region_flag == 0 and width_sum[i] > 330: start_region_flag = 1 start_region_index = i # 判定第一个白点个数大于330的是身份证区域的起始点 if width_sum[i] > 330: end_region_index = i # 只要白点个数大于330,便认为是身份证区域,更新结束点 # 身份证区域中白点最少的高度值,认为这是正反面的交点 # argsort函数中,只取width_sum中判定区域开始和结束的部分,因此结果要加上开始点的高度值 min_line_position = start_region_index + np.argsort(width_sum[start_region_index:end_region_index])[0] img_closed_original[min_line_position][:] = 0 for i in range(1, 11): # 参数可变,分割10个点 temp_line_position = start_region_index + np.argsort(width_sum[start_region_index:end_region_index])[i] if abs(temp_line_position - min_line_position) < 30: # 限定范围,在最小点距离【-30, 30】的区域内 img_closed_original[temp_line_position][:] = 0 # 强制变为0 return img_closed_original def cut_part_img(img, cut_percent): """ # 从宽度和高度两个方向,裁剪身份证边缘 :param img: 身份证区域 :param cut_percent: 裁剪的比例 :return: 裁剪后的身份证区域 """ height, width, _ = img.shape height_num = int(height * cut_percent) # 需要裁剪的高度值 h_start = 0 + height_num // 2 # 左右等比例切分 h_end = height - height_num // 2 - 1 width_num = int(width * cut_percent) # 需要裁剪的宽度值 w_start = 0 + width_num // 2 w_end = width - width_num // 2 - 1 return img[h_start:h_end, w_start:w_end] # 返回裁剪后的图片 def preprocess_cut_one_img(img_path, img_name, save_path='./save_imgs/', problem_path='./problem_save/'): # 处理一张图片 """ 裁剪出一张图片中的身份证正反面区域 :param img_path: 图片所在路径 :param img_name: 图片名称 :param save_path: 结果保存路径 测试用 :param problem_path: 出错图片中间结果保存 测试用 :return: 身份证正反面图片 """ img_path_name = os.path.join(img_path, img_name) if not os.path.exists(img_path_name): # 判断图片是否存在 print('img {name} is not exits'.format(name=img_path_name)) return 1, [] # 图片不存在,直接返回,报错加一 img = cv2.imread(img_path_name) # 读取图片 img_blurred = gray_and_fliter(img, img_name) # 灰度化并滤波 img_t = cv2.filter2D(img, -1, kernel=np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]], np.float32)) # 对图像进行锐化 img_binary = gradient_and_binary(img_blurred) # 二值化 res_bbox = find_bbox(img_t, img_binary) # 切分正反面 if len(res_bbox) != 2: # 异常处理 print('Error happened when cut img {name}, try exception cut program '.format(name=img_path_name)) # cv2.imwrite(os.path.join(problem_path, img_name.split('.')[0] + '_blurred.jpg'), img_blurred) # cv2.imwrite(os.path.join(problem_path, img_name.split('.')[0] + '_binary.jpg'), img_binary) # cv2.imwrite(os.path.join(problem_path, img_name), img) # 调试用,保存中间处理结果 img_binary = find_cut_line(img_binary) # 强制分割正反面 res_bbox = find_bbox(img_t, img_binary) if len(res_bbox) != 2: # 纠正失败 print('Failed to cut img {name}, exception program end'.format(name=img_path_name)) return 1, None else: # 纠正成功 print('Correctly cut img {name}, exception program end'.format(name=img_path_name)) return 0, res_bbox else: # 裁剪过程正常 # cv2.imwrite(os.path.join(save_path, img_name.split('.')[0] + '_0.jpg'), cut_part_img(res_bbox[0], 0.0)) # cv2.imwrite(os.path.join(save_path, img_name.split('.')[0] + '_1.jpg'), cut_part_img(res_bbox[1], 0.0)) # cv2.imwrite(os.path.join(save_path, img_name.split('.')[0] + '_original.jpg'), img) return 0, res_bbox def process_img(img_path, save_path, problem_path): """ 切分一个目录下的所有图片 :param img_path: 图片所在路径 :param save_path: 结果保存路径 :param problem_path: 问题图片保存路径 :return: None """ if not os.path.exists(img_path): # 判断图片路径是否存在 print('img path {name} is not exits, program break.'.format(name=img_path)) return if not os.path.exists(save_path): # 保存路径不存在,则创建路径 os.makedirs(save_path) if not os.path.exists(problem_path): # 保存路径不存在,则创建路径 os.makedirs(problem_path) img_names = os.listdir(img_path) error_count = 0 error_names = [] for img_name in img_names: error_temp, res_bbox = preprocess_cut_one_img(img_path, img_name, save_path, problem_path) error_count += error_temp if error_temp == 0: cv2.imwrite(os.path.join(save_path, img_name.split('.')[0] + '_0.jpg'), cut_part_img(res_bbox[0], 0.0)) cv2.imwrite(os.path.join(save_path, img_name.split('.')[0] + '_1.jpg'), cut_part_img(res_bbox[1], 0.0)) else: error_names.append(img_name) print('total error number is: ', error_count) print('error images mame :') for error_img_name in error_names: print(error_img_name) return if __name__ == '__main__': origin_img_path = './problem_imgs/' cutted_save_path = './res_imgs/' cut_problem_path = './temp_imgs/' #process_img(img_path=origin_img_path, save_path=cutted_save_path, problem_path=cut_problem_path)