如何使用OpenCV-Python识别扫描身份证图像的四边形边界

以下是可直接运行的实现方案，可根据实际扫描场景调整参数优化准确率：

• 步骤1：图像预处理，降噪并提取边缘

import cv2
import numpy as np

# 读取扫描图像
img = cv2.imread("scan_id_card.jpg")
# 统一缩放尺寸，降低后续计算量
scale = 500 / img.shape[0]
img_resize = cv2.resize(img, (0, 0), fx=scale, fy=scale)
# 转灰度+高斯模糊降噪，避免误检边缘
gray = cv2.cvtColor(img_resize, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (5, 5), 0)
# Canny边缘检测
edged = cv2.Canny(blur, threshold1=75, threshold2=200)

• 步骤2：轮廓筛选，定位身份证四边形外轮廓

# 提取所有外层轮廓
contours, _ = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# 按轮廓面积从大到小排序，取面积最大的前5个轮廓排查
contours = sorted(contours, key=cv2.contourArea, reverse=True)[:5]
card_contour = None

for c in contours:
    perimeter = cv2.arcLength(c, closed=True)
    # 轮廓多边形近似，0.02*perimeter为近似精度，可根据图像清晰度调整
    approx = cv2.approxPolyDP(c, epsilon=0.02*perimeter, closed=True)
    # 身份证为矩形，近似后为4个顶点
    if len(approx) == 4:
        card_contour = approx
        break

• 步骤3：透视变换，提取校正后的身份证区域

# 顶点坐标重排序：左上、右上、右下、左下
def order_points(pts):
    rect = np.zeros((4, 2), dtype="float32")
    s = pts.sum(axis=1)
    rect[0] = pts[np.argmin(s)]
    rect[2] = pts[np.argmax(s)]
    diff = np.diff(pts, axis=1)
    rect[1] = pts[np.argmin(diff)]
    rect[3] = pts[np.argmax(diff)]
    return rect

# 透视变换实现图像校正
def four_point_transform(origin_img, pts, scale):
    rect = order_points(pts.reshape(4, 2)) / scale
    (tl, tr, br, bl) = rect
    # 计算校正后图像的宽高
    max_width = max(int(np.sqrt(((br[0] - bl[0])**2) + ((br[1] - bl[1])**2))),
                    int(np.sqrt(((tr[0] - tl[0])**2) + ((tr[1] - tl[1])**2))))
    max_height = max(int(np.sqrt(((tr[0] - br[0])**2) + ((tr[1] - br[1])**2))),
                     int(np.sqrt(((tl[0] - bl[0])**2) + ((tl[1] - bl[1])**2))))
    # 目标坐标映射
    dst = np.array([[0, 0], [max_width-1, 0], [max_width-1, max_height-1], [0, max_height-1]], dtype="float32")
    # 计算变换矩阵并执行透视变换
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(origin_img, M, (max_width, max_height))
    return warped

if card_contour is not None:
    result_id_card = four_point_transform(img, card_contour, scale)
    cv2.imwrite("extracted_id_card.jpg", result_id_card)

• 优化适配建议
  • 若扫描图像存在严重反光，可先执行自适应阈值处理或直方图均衡化，再做边缘检测
  • 若背景干扰较多，可先通过HSV颜色空间筛选身份证的底色范围，缩小检测区域
  • 可加入宽高比校验：二代身份证宽高比约为1.58:1，符合该比例的轮廓再判定为身份证，可大幅降低误检率
  • 图像清晰度较低时，可适当调大轮廓近似的精度参数，避免识别到过多细碎边缘

内容的提问来源于stack exchange，提问作者Buoy Rina