用Opencv给韦小宝的身份证透视变换

上一节我们说了 opencv 的安装教程，这次我们来实战一波～

有时候我们项目上需要身份证识别，但是有的人拍的身份证不是倒的，就是歪的。让后台去识别很麻烦，经常识别不准。但有了 opencv 的透视变换，你再怎么歪，我也能帮你矫正过来。

透视变换原理

透视变换(Perspective Transformation)是将成像投影到一个新的视平面(Viewing Plane)，也称作投影映射(Projective Mapping)。如图所示

JAVA 实现

首先获取身份证四个点的坐标，https://uutool.cn/img-coord/ 这个网站可以在线获取图片坐标，实际在项目的时候，这四个点的坐标可以让前端传递给你。

通过工具我们获取到了四个点的坐标，分别是 [376,133]，[205,422]，[29,316]， [194,26]

import org.opencv.core.*;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import org.opencv.utils.Converters;

import java.util.List;

public class PerspectiveTransform {

    public static void main(String[] args) {
        String srcImgPath = "/home/IdeaProjects/opencv-java/src/main/resources/2020-09-13_11-55.png";

        String x1 = "104,278";
        String x2 = "523,49";
        String x3 = "652,235";
        String x4 = "223,491";
        String destImgPath = "/home/IdeaProjects/opencv-java/src/main/resources/test5.png";

        perspectiveTransform(srcImgPath,x1,x2,x3,x4,destImgPath);
    }

    private static void perspectiveTransform(String srcImgPath,String x1, String x2,
                                             String x3,String x4, String destImgPath){
        System.loadLibrary(Core.NATIVE_LIBRARY_NAME);

        Mat src = Imgcodecs.imread(srcImgPath);
        // 读取图像到矩阵中,取灰度图像
        if (src.empty()) {
            return;
        }
        try {
            Mat dst = new Mat();

            List<Point> listSrcs = java.util.Arrays.asList(getPoint(x1), getPoint(x2),
                    getPoint(x3), getPoint(x4));
            Mat srcPoints = Converters.vector_Point_to_Mat(listSrcs, CvType.CV_32F);

            List<Point> listDsts = java.util.Arrays.asList(new Point(0, 0), new Point(1011, 0),
                    new Point(1011, 638), new Point(0, 638));

            Mat dstPoints = Converters.vector_Point_to_Mat(listDsts, CvType.CV_32F);

            Mat perspectiveMmat = Imgproc.getPerspectiveTransform(srcPoints, dstPoints);

            Size size = new Size(new Point(1011, 638));
            Imgproc.warpPerspective(src, dst, perspectiveMmat, size, Imgproc.INTER_LINEAR);

            Mat gray = gray(dst);
            Imgcodecs.imwrite(destImgPath, gray);
        } catch (Exception e) {
            e.printStackTrace();
        }
    }

    private static Point getPoint(String points) {
        if (points == null || "".equals(points)) {
            throw new NullPointerException("坐标参数为空");
        }

        String[] split = points.split(",");
        if (split.length == 0) {
            throw new NullPointerException("坐标参数为空");
        }
        return new Point(Double.valueOf(split[0]), Double.valueOf(split[1]));
    }


    /**
     * 作用：灰度化
     *
     * @param src Mat矩阵图像
     * @return
     */
    public static Mat gray(Mat src) {
        Mat gray = new Mat();
        if (src.channels() == 3) {
            Imgproc.cvtColor(src, gray, Imgproc.COLOR_BGR2GRAY);
            src = gray;
        } else {
            System.out.println("The Image File Is Not The RGB File!");
        }
        return src;
    }
}

透视变化+灰度化 后的:

本项目所有代码地址：https://gitee.com/feiling0825/opencv-java

觉得写的不错话，还是希望能给个Star的

参考

详解 OpenCV 透视变换原理 及 实例

Java基于opencv—透视变换矫正图像