利用html5、websocket和opencv实现人脸检测

最新推荐文章于 2024-06-04 13:52:27 发布

dengzhenxiang2730

最新推荐文章于 2024-06-04 13:52:27 发布

阅读量452

点赞数

文章标签：网络人工智能开发工具

原文链接：http://www.cnblogs.com/husthzy/p/5659672.html

版权

最近学习人脸识别相关的东西，在MFC下使用OpenCV做了一个简单的应用。训练需要较多的数据，windows应用程序终究还是不方便，于是想着做成CS模式：检测识别都放在服务器端，视频获取和显示都放在网页端。

在网上找了一些资料，实现了简单的人脸检测。人脸识别只要在这个框架上加点代码就行。主要参考了下面这篇文章：

http://www.open-open.com/home/space-361-do-blog-id-8960.html

jetty版本：jetty-9.2.17.v20160517

javacv版本：1.2

首先是html代码，主要实现：

获取视频并显示（html5, webrtc, javascript）；
通过websocket传输视频帧；
接收并显示服务器端返回的图像数据（包含人脸检测结果）

 1 <!doctype html>
 2 <html lang="zh-CN">
 3 <head>
 4     <meta charset="UTF-8">
 5     <title>FaceDetect</title>
 6 </head>
 7 <body>
 8 
 9 <div style="visibility:hidden; width:0; height:0;">
10     <canvas id="canvas" width="320"  height="240"></canvas>
11 </div>
12 
13 <div>
14     <video id="video" autoplay style="display: inline;"></video>
15     <img id="target" style="display:inline;"/>
16 </div>
17  
18 <script type="text/javascript">
19 
20     var ws = new WebSocket("ws://127.0.0.1:2014/");
21     ws.binaryType = "arraybuffer";
22 
23     ws.onopen = function() {
24         ws.send("I'm client");
25     };
26 
27     ws.onmessage = function (evt) { 
28         var bytes = new Uint8Array(evt.data);
29         var data = "";
30         var len = bytes.byteLength;
31         for (var i = 0; i < len; ++i) {
32             data += String.fromCharCode(bytes[i]);
33         }
34         var img = document.getElementById("target");
35         img.src = "data:image/png;base64,"+window.btoa(data);
36     };
37 
38     ws.onclose = function() {
39         alert("Closed");
40     };
41 
42     ws.onerror = function(err) {
43         alert("Error: " + err);
44     };
45     
46     var getUserMedia = (navigator.getUserMedia || navigator.webkitGetUserMedia || navigator.mozGetUserMedia || navigator.msGetUserMedia);
47  
48     var video = document.getElementById('video');
49     var canvas = document.getElementById('canvas');
50     var ctx = canvas.getContext('2d');
51     
52     getUserMedia.call(navigator, {
53         video: true,
54         audio: true
55     }, function(localMediaStream) {
56         video.src = window.URL.createObjectURL(localMediaStream);
57         video.onloadedmetadata = function(e) {
58             console.log("Label: " + localMediaStream.label);
59             console.log("AudioTracks" , localMediaStream.getAudioTracks());
60             console.log("VideoTracks" , localMediaStream.getVideoTracks());
61         };
62     }, function(e) {
63         console.log('Reeeejected!', e);
64     });
65 
66     function dataURItoBlob(dataURI) {
67         // convert base64/URLEncoded data component to raw binary data held in a string
68         var byteString;
69         if (dataURI.split(',')[0].indexOf('base64') >= 0)
70             byteString = atob(dataURI.split(',')[1]);
71         else
72             byteString = unescape(dataURI.split(',')[1]);
73 
74         // separate out the mime component
75         var mimeString = dataURI.split(',')[0].split(':')[1].split(';')[0];
76 
77         // write the bytes of the string to a typed array
78         var ia = new Uint8Array(byteString.length);
79         for (var i = 0; i < byteString.length; i++) {
80             ia[i] = byteString.charCodeAt(i);
81         }
82 
83         return new Blob([ia], {type:mimeString});
84     }
85 
86     timer = setInterval(
87             function () {
88                 ctx.drawImage(video, 0, 0, 320, 240);
89                 var data = canvas.toDataURL('image/jpeg', 1.0);
90                 newblob = dataURItoBlob(data);
91                 ws.send(newblob);
92             }, 250);
93 </script>
94 </body>
95 
96 </html>

facedetect.html

然后是服务器端代码（jetty, websocket, javacv），主要实现：

接收客户传送的视频帧数据；
使用JavaCV实现人脸检测；
在原始图像上绘制检测结果，将新图像返回给客户

 1 package com.husthzy.face;
 2 
 3 import org.eclipse.jetty.server.Server;
 4 
 5 public class WebsocketServer extends Thread {
 6     @Override
 7     public void run() {
 8         super.run();
 9 
10         try {
11             Server server = new Server(2014);
12             server.setHandler(new FaceDetectionHandler());
13             server.setStopTimeout(0);
14             server.start();
15             server.join();
16         } catch (Exception e) {
17             e.printStackTrace();
18         }
19     }
20 
21     public static void main(String[] args) {
22         WebsocketServer mWebSocketServer = new WebsocketServer();
23         mWebSocketServer.start();
24     }
25 }

WebsocketServer.java

  1 package com.husthzy.face;
  2 
  3 import static org.bytedeco.javacpp.opencv_core.CV_8UC1;
  4 import static org.bytedeco.javacpp.opencv_imgcodecs.cvDecodeImage;
  5 import static org.bytedeco.javacpp.opencv_imgproc.COLOR_BGRA2GRAY;
  6 import static org.bytedeco.javacpp.opencv_imgproc.cvtColor;
  7 import static org.bytedeco.javacpp.opencv_imgproc.equalizeHist;
  8 import static org.bytedeco.javacpp.opencv_imgproc.rectangle;
  9 
 10 import java.awt.image.BufferedImage;
 11 import java.io.ByteArrayOutputStream;
 12 import java.io.IOException;
 13 import java.nio.ByteBuffer;
 14 import java.util.ArrayList;
 15 
 16 import javax.imageio.ImageIO;
 17 
 18 import org.bytedeco.javacpp.BytePointer;
 19 import org.bytedeco.javacpp.opencv_core;
 20 import org.bytedeco.javacpp.opencv_core.IplImage;
 21 import org.bytedeco.javacpp.opencv_core.Mat;
 22 import org.bytedeco.javacpp.opencv_core.Rect;
 23 import org.bytedeco.javacpp.opencv_core.RectVector;
 24 import org.bytedeco.javacpp.opencv_core.Scalar;
 25 import org.bytedeco.javacpp.opencv_objdetect.CascadeClassifier;
 26 import org.bytedeco.javacv.Frame;
 27 import org.bytedeco.javacv.Java2DFrameConverter;
 28 import org.bytedeco.javacv.OpenCVFrameConverter.ToMat;
 29 import org.eclipse.jetty.websocket.api.Session;
 30 import org.eclipse.jetty.websocket.api.annotations.OnWebSocketClose;
 31 import org.eclipse.jetty.websocket.api.annotations.OnWebSocketConnect;
 32 import org.eclipse.jetty.websocket.api.annotations.OnWebSocketError;
 33 import org.eclipse.jetty.websocket.api.annotations.OnWebSocketMessage;
 34 import org.eclipse.jetty.websocket.api.annotations.WebSocket;
 35 import org.eclipse.jetty.websocket.server.WebSocketHandler;
 36 import org.eclipse.jetty.websocket.servlet.WebSocketServletFactory;
 37 
 38 @WebSocket
 39 public class FaceDetectionHandler extends WebSocketHandler {
 40 
 41     private static final String CASCADE_FILE = "haarcascade_frontalface_alt.xml";
 42 
 43     private Session mSession;
 44     private static ArrayList<FaceDetectionHandler> sessions = new ArrayList<FaceDetectionHandler>();
 45 
 46     private CascadeClassifier face_cascade = new CascadeClassifier(CASCADE_FILE);
 47 
 48     public static ArrayList<FaceDetectionHandler> getAllSessions() {
 49         return sessions;
 50     }
 51 
 52     @Override
 53     public void configure(WebSocketServletFactory factory) {
 54         factory.register(FaceDetectionHandler.class);
 55         factory.getPolicy().setMaxBinaryMessageSize(1024 * 512);
 56     }
 57 
 58     @OnWebSocketClose
 59     public void onClose(int statusCode, String reason) {
 60         sessions.remove(this);
 61         System.out.println(
 62                 "Close: statusCode = " + statusCode + ", reason = " + reason + ", sessions = " + sessions.size());
 63     }
 64 
 65     @OnWebSocketError
 66     public void onError(Throwable t) {
 67         System.out.println("Error: " + t.getMessage());
 68     }
 69 
 70     @OnWebSocketConnect
 71     public void onConnect(Session session) {
 72         mSession = session;
 73         sessions.add(this);
 74 
 75         System.out.println("Connect: " + session.getRemoteAddress().getAddress());
 76     }
 77 
 78     @OnWebSocketMessage
 79     public void onMessage(String message) {
 80         System.out.println("Message: " + message);
 81     }
 82 
 83     @OnWebSocketMessage
 84     public void onBinaryMessage(byte data[], int offset, int length) {
 85         System.out.println("Binary Message len:" + length);
 86         if (length > 10000) {
 87             try {
 88                 byte[] sdata = process(data);
 89                 ByteBuffer byteBuffer = ByteBuffer.wrap(sdata);
 90                 mSession.getRemote().sendBytes(byteBuffer);
 91                 byteBuffer.clear();
 92             } catch (IOException e) {
 93                 e.printStackTrace();
 94             }
 95         }
 96     }
 97 
 98     public byte[] process(byte data[]) {
 99         IplImage originalImage = cvDecodeImage(opencv_core.cvMat(1, data.length, CV_8UC1, new BytePointer(data)));
100 
101         Mat videoMat = new Mat(originalImage);
102         Mat videoMatGray = new Mat();
103         // Convert the current frame to grayscale:
104         cvtColor(videoMat, videoMatGray, COLOR_BGRA2GRAY);
105         equalizeHist(videoMatGray, videoMatGray);
106 
107         // Point p = new Point();
108         RectVector faces = new RectVector();
109         face_cascade.detectMultiScale(videoMatGray, faces);
110         for (int i = 0; i < faces.size(); i++) {
111             Rect face_i = faces.get(i);
112 
113             //Mat face = new Mat(videoMatGray, face_i);
114             // If fisher face recognizer is used, the face need to be
115             // resized.
116             // resize(face, face_resized, new Size(im_width, im_height),
117             // 1.0, 1.0, INTER_CUBIC);
118 
119             // Now perform the prediction, see how easy that is:
120             // int prediction = lbphFaceRecognizer.predict(face);
121 
122             // And finally write all we've found out to the original image!
123             // First of all draw a green rectangle around the detected face:
124             rectangle(videoMat, face_i, new Scalar(0, 255, 0, 1));
125 
126             System.out.println("face pos: x:" + face_i.x() + " y:" + face_i.y());
127 
128             // Create the text we will annotate the box with:
129             //String box_text = "Prediction = " + prediction;
130             // Calculate the position for annotated text (make sure we don't
131             // put illegal values in there):
132             //int pos_x = Math.max(face_i.tl().x() - 10, 0);
133             //int pos_y = Math.max(face_i.tl().y() - 10, 0);
134             // And now put it into the image:
135             //putText(videoMat, box_text, new Point(pos_x, pos_y), FONT_HERSHEY_PLAIN, 1.0, new Scalar(0, 255, 0, 2.0));
136         }
137 
138         // JavaCVUtil.imShow(videoMat, "test");
139 
140         return getMatByteBuffer(videoMat);
141     }
142 
143     private byte[] getMatByteBuffer(Mat m) {
144         byte[] result = null;
145         try {
146             ToMat convert = new ToMat();
147             Frame frame = convert.convert(m);
148             Java2DFrameConverter java2dFrameConverter = new Java2DFrameConverter();
149             BufferedImage bufferedImage = java2dFrameConverter.convert(frame);
150             ByteArrayOutputStream out = new ByteArrayOutputStream();
151             ImageIO.write(bufferedImage, "png", out);
152             result = out.toByteArray();
153             out.close();
154         } catch (IOException exception) {
155             exception.printStackTrace();
156         }
157         return result;
158     }
159 }

FaceDetectionhandler.java

转载于:https://www.cnblogs.com/husthzy/p/5659672.html

dengzhenxiang2730

关注

0
点赞
踩
6

收藏

觉得还不错? 一键收藏
0
评论
利用html5、websocket和opencv实现人脸检测

最近学习人脸识别相关的东西，在MFC下使用OpenCV做了一个简单的应用。训练需要较多的数据，windows应用程序终究还是不方便，于是想着做成CS模式：检测识别都放在服务器端，视频获取和显示都放在网页端。在网上找了一些资料，实现了简单的人脸检测。人脸识别只要在这个框架上加点代码就行。主要参考了下面这篇文章：http://www.open-open.com/home/spa...
复制链接

扫一扫