1. opencv概述
OpenCV是一个开源的计算机视觉库,它提供了一系列丰富的图像处理和计算机视觉算法,包括图像读取、显示、滤波、特征检测、目标跟踪等功能。
opencv官网:https://opencv.org/
opencv官网文档:https://docs.opencv.org/4.7.0/index.html
参考教程1:https://www.w3cschool.cn/opencv/
参考教程2:https://www.yiibai.com/opencv/opencv_adding_text.html
2. 安装opencv
2.1 下载opencv
opencv下载:https://opencv.org/releases/
这里我们使用4.7.0版本,下载到本地后,双击进行安装即可。
进入到opencv的安装目录:
build :基于window构建sources:开源,提供源码
进入到build\java 目录
x64与x86目录下是对应的.dll文件:代表给不同的系统使用,下面的代码会使用到.dll文件
opencv-460.jar给java操作openvc的程序包
2.2 准备文件
# 1. 特征分类器:windows 和 linux 中的配置文件都一样,随便用哪个都行
haarcascade_frontalface_alt.xml
# windows 路径 : opencv\build\etc\haarcascades
# linux 路径 : /usr/local/share/opencv4/haarcascades# 2. jar 包 - 也可以直接使用 javacv 中的 opencv 包
opencv-470.jar
# windows 路径 : {opencv安装目录}\opencv\build\java
# linux 路径 : /usr/local/share/java/opencv4# 3. 动态库
opencv_java470.dll (windows系统使用此文件)
# windows 路径 : {opencv安装目录}\opencv\build\java\{x64}/{x86} 跟据系统选择
libopencv_java470.so (linux系统使用此文件)
# linux 路径 : /usr/local/share/java/opencv4
3. 代码实现
3.1 pom.xml添加依赖
<!-- 版本的依赖与下载的opencv版本一致-->
<dependency><groupId>org.bytedeco</groupId><artifactId>opencv</artifactId><version>4.7.0-1.5.9</version></dependency>
或:
<dependency><groupId>org.bytedeco</groupId><artifactId>javacv-platform</artifactId><version>1.5.9</version></dependency>
或:
<dependency><groupId>org.openpnp</groupId><artifactId>opencv</artifactId><version>4.7.0-0</version></dependency>
以上三个依赖任选其一即可,项目打包后观察一下使用哪个依赖打包后的jar文件更小
ps:依赖包太大,优化参考:https://blog.csdn.net/u014644574/article/details/122067708
3.2 编写代码
ps:代码中存在加载.dll、haarcascade_frontalface_alt.xml文件,请确保文件地址正确
package com.testpro.test.opencv;import org.opencv.core.*;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import org.opencv.objdetect.CascadeClassifier;import java.util.Arrays;public class FaceCompare {// 初始化人脸探测器static CascadeClassifier faceDetector;private static final String PATH_PREFIX = "C:\\Users\\dev\\Desktop\\";static int i = 0;static {// 判断系统String os = System.getProperty("os.name");// 加载动态库if (os != null && os.toLowerCase().startsWith("windows")) {// Windows操作系统// todo windows 系统部署加载 .dll 文件 - 路径跟据自己存放位置更改【这里需要使用绝对路径】System.load("D:\\opencv\\opencv\\build\\java\\x64\\opencv_java470.dll");} else if (os != null && os.toLowerCase().startsWith("linux")) {// Linux操作系统// todo Linux 服务器部署加载 .so 文件 - 路径跟据自己存放位置更改【是否需要绝对路径有待验证,目前只在windows 系统实践过】System.load("/opt/face/libopencv_java440.so");}// 引入 特征分类器配置 文件:haarcascade_frontalface_alt.xml 文件路径// 此文件在opencv的安装目录build\etc\haarcascades下可以找到String property = "D:\\opencv\\opencv\\build\\etc\\haarcascades\\haarcascade_frontalface_alt.xml";System.out.println(property);faceDetector = new CascadeClassifier(property);}public static void main(String[] args) {// 图片路径不能包含中文String str1 = PATH_PREFIX + "3-1.jpg";String str2 = PATH_PREFIX + "3-2.jpg";long start = System.currentTimeMillis();double compareHist = compare_image(str1, str2);System.out.println("time:" + (System.currentTimeMillis() - start));System.out.println(compareHist);if (compareHist > 0.6) {System.out.println("人脸匹配");} else {System.out.println("人脸不匹配");}}// 灰度化人脸public static Mat conv_Mat(String img) {Mat image0 = Imgcodecs.imread(img);Mat image1 = new Mat();// 灰度化Imgproc.cvtColor(image0, image1, Imgproc.COLOR_BGR2GRAY);// 探测人脸MatOfRect faceDetections = new MatOfRect();faceDetector.detectMultiScale(image1, faceDetections);// rect中人脸图片的范围for (Rect rect : faceDetections.toArray()) {Mat face = new Mat(image1, rect);return face;}return null;}// 比较图片public static double compare_image(String img_1, String img_2) {Mat mat_1 = conv_Mat(img_1);Mat mat_2 = conv_Mat(img_2);Mat hist_1 = new Mat();Mat hist_2 = new Mat();//颜色范围MatOfFloat ranges = new MatOfFloat(0f, 256f);//直方图大小, 越大匹配越精确 (越慢)MatOfInt histSize = new MatOfInt(10000000);Imgproc.calcHist(Arrays.asList(mat_1), new MatOfInt(0), new Mat(), hist_1, histSize, ranges);Imgproc.calcHist(Arrays.asList(mat_2), new MatOfInt(0), new Mat(), hist_2, histSize, ranges);// CORREL 相关系数double res = Imgproc.compareHist(hist_1, hist_2, Imgproc.CV_COMP_CORREL);return res;}}
上述代码加载.dll文件也可使用以下方式:
ps:【不过以下方式需要将opencv安装目录下的build\java\x64\opencv_java470.dll文件复制到C:\Windows\System32目录下才可使用否则会报错】
// 使用此方法需将D:\opencv\opencv\build\java\x64\opencv_java470.dll文件复制到C:\Windows\System32目录下
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
如下:
4. 效果
5. 附:完整代码
包括:
从摄像头实时人脸识别,识别成功保存图片到本地
从本地视频文件中识别人脸
本地图片人脸识别,识别成功并保存人脸图片到本地
package com.testpro.test.opencv;import org.opencv.core.*;
import org.opencv.highgui.HighGui;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import org.opencv.objdetect.CascadeClassifier;
import org.opencv.videoio.VideoCapture;
import org.opencv.videoio.VideoWriter;
import org.opencv.videoio.Videoio;import java.util.Arrays;/*** Opencv 图片人脸识别、实时摄像头人脸识别、视频文件人脸识别*/
public class FaceVideo {// 初始化人脸探测器static CascadeClassifier faceDetector;static int i = 0;static {// 判断系统String os = System.getProperty("os.name");// 加载动态库if (os != null && os.toLowerCase().startsWith("windows")) {// Windows操作系统// todo windows 系统部署加载 .dll 文件 - 路径跟据自己存放位置更改System.load("D:\\opencv\\opencv\\build\\java\\x64\\opencv_java470.dll");
// ClassLoader.getSystemResource("dlls/opencv_java470.dll");} else if (os != null && os.toLowerCase().startsWith("linux")) {// Linux操作系统// todo Linux 服务器部署加载 .so 文件 - 路径跟据自己存放位置更改System.load("/opt/face/libopencv_java440.so");}// 引入 特征分类器配置 文件:haarcascade_frontalface_alt.xml 文件路径String property = "D:\\opencv\\opencv\\build\\etc\\haarcascades\\haarcascade_frontalface_alt.xml";System.out.println(property);faceDetector = new CascadeClassifier(property);}private static final String PATH_PREFIX = "C:\\Users\\dev\\Desktop\\";public static void main(String[] args) {// 1- 从摄像头实时人脸识别,识别成功保存图片到本地
// getVideoFromCamera();// 2- 从本地视频文件中识别人脸
// getVideoFromFile();// 3- 本地图片人脸识别,识别成功并保存人脸图片到本地
// face("5-1.jpg");// 4- 比对本地2张图的人脸相似度 (越接近1越相似)double compareHist = compare_image(PATH_PREFIX + "5-1.jpg", PATH_PREFIX + "6-1.jpg");System.out.println(compareHist);if (compareHist > 0.72) {System.out.println("人脸匹配");} else {System.out.println("人脸不匹配");}}/*** OpenCV-4.7.0 从摄像头实时读取*/public static void getVideoFromCamera() {//1 如果要从摄像头获取视频 则要在 VideoCapture 的构造方法写 0VideoCapture capture = new VideoCapture(0);Mat video = new Mat();int index = 0;if (capture.isOpened()) {while (i < 3) {// 匹配成功3次退出capture.read(video);HighGui.imshow("实时人脸识别", getFace(video));index = HighGui.waitKey(100);if (index == 27) {capture.release();break;}}} else {System.out.println("摄像头未开启");}try {capture.release();Thread.sleep(1000);System.exit(0);} catch (InterruptedException e) {e.printStackTrace();}return;}/*** OpenCV-4.7.0 从视频文件中读取*/public static void getVideoFromFile() {VideoCapture capture = new VideoCapture();capture.open(PATH_PREFIX + "yimi.mp4");//1 读取视频文件的路径if (!capture.isOpened()) {System.out.println("读取视频文件失败!");return;}Mat video = new Mat();int index = 0;while (capture.isOpened()) {capture.read(video);//2 视频文件的视频写入 Mat video 中HighGui.imshow("本地视频识别人脸", getFace(video));//3 显示图像index = HighGui.waitKey(100);//4 获取键盘输入if (index == 27) {//5 如果是 Esc 则退出capture.release();return;}}}/*** OpenCV-4.7.0 人脸识别** @param image 待处理Mat图片(视频中的某一帧)* @return 处理后的图片*/public static Mat getFace(Mat image) {// 1 读取OpenCV自带的人脸识别特征XML文件(faceDetector)
// CascadeClassifier facebook = new CascadeClassifier("D:\\Sofeware\\opencv\\sources\\data\\haarcascades\\haarcascade_frontalface_alt.xml");// 2 特征匹配类MatOfRect face = new MatOfRect();// 3 特征匹配faceDetector.detectMultiScale(image, face);Rect[] rects = face.toArray();System.out.println("匹配到 " + rects.length + " 个人脸");if (rects != null && rects.length >= 1) {// 4 为每张识别到的人脸画一个圈for (int i = 0; i < rects.length; i++) {Imgproc.rectangle(image, new Point(rects[i].x, rects[i].y), new Point(rects[i].x + rects[i].width, rects[i].y + rects[i].height), new Scalar(0, 255, 0));Imgproc.putText(image, "Human", new Point(rects[i].x, rects[i].y), Imgproc.FONT_HERSHEY_SCRIPT_SIMPLEX, 1.0, new Scalar(0, 255, 0), 1, Imgproc.LINE_AA, false);//Mat dst=image.clone();//Imgproc.resize(image, image, new Size(300,300));}i++;if (i == 3) {// 获取匹配成功第10次的照片Imgcodecs.imwrite(PATH_PREFIX + "face.png", image);}}return image;}/*** OpenCV-4.7.0 图片人脸识别*/public static void face(String filename) {// 1 读取OpenCV自带的人脸识别特征XML文件// OpenCV 图像识别库一般位于 opencv\sources\data 下面
// CascadeClassifier facebook=new CascadeClassifier("D:\\Sofeware\\opencv\\sources\\data\\haarcascades\\haarcascade_frontalface_alt.xml");// 2 读取测试图片String imgPath = PATH_PREFIX + filename;Mat image = Imgcodecs.imread(imgPath);if (image.empty()) {System.out.println("image 内容不存在!");return;}// 3 特征匹配MatOfRect face = new MatOfRect();faceDetector.detectMultiScale(image, face);// 4 匹配 Rect 矩阵 数组Rect[] rects = face.toArray();System.out.println("匹配到 " + rects.length + " 个人脸");// 5 为每张识别到的人脸画一个圈int i = 1;for (Rect rect : face.toArray()) {Imgproc.rectangle(image, new Point(rect.x, rect.y), new Point(rect.x + rect.width, rect.y + rect.height),new Scalar(0, 255, 0), 3);imageCut(imgPath, PATH_PREFIX + i + ".jpg", rect.x, rect.y, rect.width, rect.height);// 进行图片裁剪i++;}// 6 展示图片HighGui.imshow("人脸识别", image);HighGui.waitKey(0);}/*** 裁剪人脸** @param imagePath* @param outFile* @param posX* @param posY* @param width* @param height*/public static void imageCut(String imagePath, String outFile, int posX, int posY, int width, int height) {// 原始图像Mat image = Imgcodecs.imread(imagePath);// 截取的区域:参数,坐标X,坐标Y,截图宽度,截图长度Rect rect = new Rect(posX, posY, width, height);// 两句效果一样Mat sub = image.submat(rect); // Mat sub = new Mat(image, rect);Mat mat = new Mat();Size size = new Size(width, height);Imgproc.resize(sub, mat, size);// 将人脸进行截图并保存Imgcodecs.imwrite(outFile, mat);System.out.println(String.format("图片裁切成功,裁切后图片文件为: %s", outFile));}/*** 人脸比对** @param img_1* @param img_2* @return*/public static double compare_image(String img_1, String img_2) {Mat mat_1 = conv_Mat(img_1);Mat mat_2 = conv_Mat(img_2);Mat hist_1 = new Mat();Mat hist_2 = new Mat();//颜色范围MatOfFloat ranges = new MatOfFloat(0f, 256f);//直方图大小, 越大匹配越精确 (越慢)MatOfInt histSize = new MatOfInt(1000);Imgproc.calcHist(Arrays.asList(mat_1), new MatOfInt(0), new Mat(), hist_1, histSize, ranges);Imgproc.calcHist(Arrays.asList(mat_2), new MatOfInt(0), new Mat(), hist_2, histSize, ranges);// CORREL 相关系数double res = Imgproc.compareHist(hist_1, hist_2, Imgproc.CV_COMP_CORREL);return res;}/*** 灰度化人脸** @param img* @return*/public static Mat conv_Mat(String img) {Mat image0 = Imgcodecs.imread(img);Mat image1 = new Mat();// 灰度化Imgproc.cvtColor(image0, image1, Imgproc.COLOR_BGR2GRAY);// 探测人脸MatOfRect faceDetections = new MatOfRect();faceDetector.detectMultiScale(image1, faceDetections);// rect中人脸图片的范围for (Rect rect : faceDetections.toArray()) {Mat face = new Mat(image1, rect);return face;}return null;}/*** OpenCV-4.7.0 将摄像头拍摄的视频写入本地*/public static void writeVideo() {//1 如果要从摄像头获取视频 则要在 VideoCapture 的构造方法写 0VideoCapture capture = new VideoCapture(0);Mat video = new Mat();int index = 0;Size size = new Size(capture.get(Videoio.CAP_PROP_FRAME_WIDTH), capture.get(Videoio.CAP_PROP_FRAME_HEIGHT));VideoWriter writer = new VideoWriter("D:/a.mp4", VideoWriter.fourcc('D', 'I', 'V', 'X'), 15.0, size, true);while (capture.isOpened()) {capture.read(video);//2 将摄像头的视频写入 Mat video 中writer.write(video);HighGui.imshow("像头获取视频", video);//3 显示图像index = HighGui.waitKey(100);//4 获取键盘输入if (index == 27) {//5 如果是 Esc 则退出capture.release();writer.release();return;}}}}