A judgment method by feature matching through the positioning area of the QRCode

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

# 读取目标图像和模板图像
def show_image(img):
    cv2.namedWindow("1",cv2.WINDOW_AUTOSIZE )
    cv2.imshow("1",img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

# 1. 生成标准回字形模板（7x7）
def create_qrcode_template(size=7):
    """创建二维码定位点的回字形模板（黑=0，白=255）"""
    template = np.ones((size, size), dtype=np.uint8) * 255  # 初始全白
    # 外层黑框（第0行、第6行，第0列、第6列）
    template[0, :] = 0
    template[-1, :] = 0
    template[:, 0] = 0
    template[:, -1] = 0
    # 内层黑方块（第2-4行，第2-4列）
    template[2:5, 2:5] = 0

    return template

def exist_rect(candidates,score_x,score_y,score_w):
    for (x, y, w, h, score) in candidates:
        if abs(score_x -x )< 10 and abs(score_y - y) < 10 and 0.8 <= score_w / w <= 1.2:
            return True
         
    return False

# 2. 多尺度模板匹配
def match_template_multi_scale(img_gray, template, min_scale=3, max_scale=7.0, step=0.1):
    """在多个尺度下进行模板匹配，返回高相似度区域"""
    h, w = template.shape[:2]
    candidates = []
    
    # 遍历不同尺度
    scale = min_scale
    while scale <= max_scale:
        # 缩放模板
        scaled_h = int(h * scale)
        scaled_w = int(w * scale)
        if scaled_h < 7 or scaled_w < 7:  # 最小不小于原始模板尺寸（避免信息丢失）
            scale += step
            continue
        scaled_template = cv2.resize(template, (scaled_w, scaled_h))
        #show_image(scaled_template)
        
        # 模板匹配（TM_CCOEFF_NORMED 输出归一化相关系数，值越接近1越匹配）
        result = cv2.matchTemplate(img_gray, scaled_template, cv2.TM_CCOEFF_NORMED)
        
        # 筛选匹配得分高于阈值的区域（阈值可根据图像质量调整）
        threshold = 0.7
        locations = np.where(result >= threshold)
        for pt in zip(*locations[::-1]):  # (x, y) 坐标
            # 记录匹配区域（左上角坐标、宽高、得分）
            score = result[pt[1], pt[0]]
            if not exist_rect(candidates,pt[0], pt[1], scaled_w):
                candidates.append((pt[0], pt[1], scaled_w, scaled_h, score))
        scale += step
    
    # 按得分排序，保留前N个高得分区域（去重）
    candidates.sort(key=lambda x: x[4], reverse=True)
    return candidates[:10] # 取前10个候选

# 3. 验证候选区域是否为正方形（定位点应为正方形）
def is_square(candidate, ratio_threshold=0.1):
    x, y, w, h, _ = candidate
    aspect_ratio = w / h
    return 1 - ratio_threshold <= aspect_ratio <= 1 + ratio_threshold

# 4. 主函数：定位二维码的三个定位点
def find_qrcode_position_points(img):
    img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    #增强图像对比度 
    clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
    gray_c = clahe.apply(img_gray)

    #高斯滤波
    gray_c = cv2.GaussianBlur(gray_c, (3, 3), 0)    # 二值化（增强黑白对比）
    _, img_bin = cv2.threshold(gray_c, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

    
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
    img_bin = cv2.morphologyEx(img_bin, cv2.MORPH_CLOSE, kernel)

    #show_image(img_bin)
    # 创建模板
    template = create_qrcode_template()
    
    # 多尺度匹配
    candidates = match_template_multi_scale(img_bin, template)
    
    # 筛选正方形候选区域
    valid_candidates = [c for c in candidates if is_square(c)]

    # 绘制结果
    result_img = img.copy()
    for (x, y, w, h, score) in valid_candidates[:3]:  # 取前3个最可能的定位点
        cv2.rectangle(result_img, (x, y), (x + w, y + h), (0, 255, 0), 2)
        cv2.putText(result_img, f"{score:.2f}", (x, y - 10), 
                    cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1)
        # range_img = result_img[y:  y +h, x : x+w]
        # show_image(range_img)
    
    return result_img, valid_candidates[:3]

# 测试
def main():
    folder_path = './imgs9'
    #模板定位点信息
   #bar_range = (1377, 138, 1536, 293)
    bar_range = (152, 192, 352, 350)
    paper_num = 0
    error_paper_num = 0
    less_than_3 = 0

    # 递归遍历文件夹
    for root, dirs, files in os.walk(folder_path): 
        for filename in files:
            if filename.lower().endswith('.jpg'):
                paper_num += 1

                file_path = os.path.join(root, filename)
                ori_img = cv2.imread(file_path, cv2.IMREAD_ANYCOLOR)  # 目标图像
                if len(ori_img.shape) == 3 and ori_img.shape[2] == 3:
                    gray_img = cv2.cvtColor(ori_img, cv2.COLOR_BGR2GRAY)
                elif len(ori_img.shape) == 2:
                    gray_img = ori_img.copy()
                else:
                    print('error: bad image')
                    return
     
                ori_bar_area = ori_img[bar_range[1]:bar_range[3], bar_range[0]:bar_range[2]]

                # height, width = ori_bar_area.shape[:2]
                # if height > 300 or width > 300:
                #     # 缩小图像：用INTER_AREA
                #     resized = cv2.resize(ori_bar_area, (300, 300), interpolation=cv2.INTER_AREA)
                # else:
                #     # 放大图像：用INTER_CUBIC（更清晰）
                #     resized = cv2.resize(ori_bar_area, (300, 300), interpolation=cv2.INTER_CUBIC)
                
                result_img, points = find_qrcode_position_points(ori_bar_area)

                if len(points) > 0:
                    print('识别到特征点个数：' , len(points))
                    if(len(points)) < 2:
                        less_than_3 += 1
                        show_image(result_img)
                    #show_image(result_img)
                    pass
                else:
                    error_paper_num += 1
                    # show_image(result_img)
                    # show_image(ori_img)


                print('识别总份数： ', paper_num)
                print('特征点低于3的份数 ', less_than_3)
                print('未识别到二维码份数： ', error_paper_num)
                print('二维码识别率： ', str(1 - error_paper_num / paper_num))
                print()   

                # cv2.imshow("QR Code Position Points", result_img)
                # cv2.waitKey(0)
                # cv2.destroyAllWindows()

if __name__ == "__main__":
    main()