- 简介
- 准备库
- 注册账号
- 下载SDK
- 库结构
- 建项目
- 添加javacv依赖
- 添加虹软依赖
- 激活人脸库
- 提取特征
- 主程序
- 参考
现在人脸识别技术应用越来越广泛,对于个人来说,最大的障碍是模型训练太难,合理利用现成的库可以大大加速开发速度。虹软的库效果不错,并且免费试用,值得一试。本文就利用虹软的库做一个简单的人脸识别程序,仅仅是为了跑通。
准备库 注册账号到虹软视觉官网,点击右上角 开发者中心 ,进行账号注册和登录。
下载SDK进入 我的应用
点击右上角 添加SDK 按钮,根据自身需要选择对应的SDK,获得SDK_KEY并且下载SDK。本次下载的SDK是
建议先看下 开发说明文档,不看也没关系,反正几个函数,以后用到再翻也可以。
由于这个项目后继还有别的用途,所以建成springboot项目,并且添加了mysql,web,mybatis,log4j等,如果单纯为这个试验建项目,可以建个maven项目。
添加javacv依赖pom文件加上依赖:
org.bytedeco
javacv-platform
1.5.7
这个库很大,建议改阿里源进行下载。
添加虹软依赖在资源文件夹创建一个lib文件夹,将虹软的java依赖库复制进去
pom文件加上依赖
com.arcsoft.face arcsoft-sdk-face 3.0.0.0 system ${project.basedir}/src/main/resources/lib/arcsoft-sdk-face-3.0.0.0.jar
maven刷新一下
激活人脸库创建一个类来激活库
注意更改appId, sdkKey, lib路径
public class ActiveEng {
private static final Logger logger = Logger.getLogger(Javacv.class);
public static void main(String[] args) {
//从官网获取
String appId = "";//在账号下面找
String sdkKey = "";//在账号里面找
FaceEngine faceEngine = new FaceEngine("F:\arcsoft_lib");//sdk lib的位置
//激活引擎
int errorCode = faceEngine.activeOnline(appId, sdkKey);
if (errorCode != ErrorInfo.MOK.getValue() && errorCode != ErrorInfo.MERR_ASF_ALREADY_ACTIVATED.getValue()) {
logger.debug("引擎激活失败");
}
else {
logger.debug("引擎激活成功");
}
}
}
激活以后会在项目文件夹下生产一个ArcFace64.dat文件
在做1:N之前,要将人脸信息的特征提取出来,保存好,方便以后比对
在项目文件夹里面建两个文件夹:record和feature,把模板照片放到record里面,创建一个类提取特征。
public class SaveFaceFeature {
private static final Logger logger = Logger.getLogger(Javacv.class);
public static void main(String[] args) {
//识别库初始化
FaceEngine faceEngine = new FaceEngine("F:\arcsoft_lib");
//激活信息
ActiveFileInfo activeFileInfo=new ActiveFileInfo();
int errorCode = faceEngine.getActiveFileInfo(activeFileInfo);
if (errorCode != ErrorInfo.MOK.getValue() && errorCode != ErrorInfo.MERR_ASF_ALREADY_ACTIVATED.getValue()) {
logger.debug("获取激活文件信息失败,返回代码:" + errorCode);
} else {
logger.debug("获取激活文件信息成功,返回代码:" + errorCode);
}
//引擎配置
EngineConfiguration engineConfiguration = new EngineConfiguration();
engineConfiguration.setDetectMode(DetectMode.ASF_DETECT_MODE_IMAGE);//图像模式
engineConfiguration.setDetectFaceOrientPriority(DetectOrient.ASF_OP_0_ONLY); //逆时针0度
engineConfiguration.setDetectFaceMaxNum(1);
engineConfiguration.setDetectFaceScaleVal(32);
//功能配置
FunctionConfiguration functionConfiguration = new FunctionConfiguration();
//functionConfiguration.setSupportAge(true); //年龄检测
//functionConfiguration.setSupportFace3dAngle(true); //3D角度检测
functionConfiguration.setSupportFaceDetect(true); //人脸检测
functionConfiguration.setSupportFaceRecognition(true); //人脸识别
//functionConfiguration.setSupportGender(true); //性别检测
functionConfiguration.setSupportLiveness(true); //活体检测
//functionConfiguration.setSupportIRLiveness(true); //红外活体检测
engineConfiguration.setFunctionConfiguration(functionConfiguration);
//初始化引擎
errorCode = faceEngine.init(engineConfiguration);
if (errorCode != ErrorInfo.MOK.getValue()) {
logger.debug("初始化引擎失败" + errorCode);
}
//获取目录的路径
//使用相对路径
//换成绝对路径
//显示出来
//把record文件里面的图片抽取特征,按身份证号存在feature文件夹里面
String recordPath = "./record"; // 模板目录路径
String featurePath = "./feature"; // 特征目录路径
File recordDocument = new File(recordPath);//获取路径
//模板路径的目录不存在的话直接跳出
if (!recordDocument.exists()) {
logger.debug("模板图像目录" + recordPath + "不存在");//不存在就输出
return;
}
File fa[] = recordDocument.listFiles();//用数组接收
//没有文件直接跳出
if(fa.length == 0){
logger.debug("目录里面没有图片");
}
List faceInfoList = new ArrayList<>();
ImageInfo imageInfo = new ImageInfo();
ImageInfoEx imageInfoEx = new ImageInfoEx();
FaceFeature feature = new FaceFeature();
//逐一提取特征并且保存
for (File record:fa) {
logger.debug("提取" + record.toString() + "的特征");
//imageInfoEx需要imageInfo
imageInfo = getRGBData(record);
//extractFaceFeature(byte[] data, int width, int height, ImageFormat imageFormat,
// FaceInfo faceInfo, FaceFeature feature)
//extractFaceFeature(ImageInfoEx imageInfoEx, FaceInfo faceInfo, FaceFeature feature)
//extractFaceFeature(byte[][] imageDataPlanes, int[] imageStrides, int width, int height,
// ImageFormat imageFormat, FaceInfo faceInfo, FaceFeature feature)
//三种方法可以提取特征
//我们选择第二种,需要ImageInfoEx和FaceInfo
//ImageInfoEx
imageInfoEx.setHeight(imageInfo.getHeight());
imageInfoEx.setWidth(imageInfo.getWidth());
imageInfoEx.setImageFormat(imageInfo.getImageFormat());
imageInfoEx.setImageDataPlanes(new byte[][]{imageInfo.getImageData()});
imageInfoEx.setImageStrides(new int[]{imageInfo.getWidth() * 3});
errorCode = faceEngine.detectFaces(imageInfoEx, DetectModel.ASF_DETECT_MODEL_RGB, faceInfoList);
//提取feature
errorCode = faceEngine.extractFaceFeature(imageInfoEx, faceInfoList.get(0), feature);
//保存特征
String featureFileName =featurePath + "/" +record.getName().substring(0, 18) + ".dat";
File featureFile = new File(featureFileName);
try {
OutputStream outputStream = new FileOutputStream(featureFile);
outputStream.write(feature.getFeatureData());
outputStream.flush();
outputStream.close();
} catch (FileNotFoundException e) {
logger.debug("创建输出流失败" + featureFileName);
e.printStackTrace();
} catch (IOException e) {
logger.debug("文件写入失败" + featureFileName);
e.printStackTrace();
}
logger.debug(featureFileName);
}
//引擎卸载
errorCode = faceEngine.unInit();
}
}
运行提取模板图像的特征,保存到feature文件夹备用
主程序创建一个实体类
public class RegistedFeatureInfo {
private String idNum;
private FaceFeature faceFeature;
public String getIdNum() {
return idNum;
}
public void setIdNum(String idNum) {
this.idNum = idNum;
}
public FaceFeature getFaceFeature() {
return faceFeature;
}
public void setFaceFeature(FaceFeature faceFeature) {
this.faceFeature = faceFeature;
}
@Override
public String toString() {
return "RegistedFeatureInfo{" +
"idNum='" + idNum + ''' +
", faceFeature=" + faceFeature +
'}';
}
}
主类
基本就是把特征读入list,javacv开窗抓frame,找人头提特征,逐个对比,过阈值显示出来。
javacv部分可以参考 eguid_1 大神的各个专栏,不要吝惜一百几十块钱。
public class Javacv {
private static final Logger logger = Logger.getLogger(Javacv.class);
public void openCam() throws FrameGrabber.Exception {
FaceEngine faceEngine = new FaceEngine("F:\arcsoft_lib");
ActiveFileInfo activeFileInfo=new ActiveFileInfo();
AtomicInteger errorCode = new AtomicInteger(faceEngine.getActiveFileInfo(activeFileInfo));
if (errorCode.get() != ErrorInfo.MOK.getValue() && errorCode.get() != ErrorInfo.MERR_ASF_ALREADY_ACTIVATED.getValue()) {
logger.debug("获取激活文件信息失败,返回代码:" + errorCode);
} else {
logger.debug("获取激活文件信息成功,返回代码:" + errorCode);
}
//引擎配置
EngineConfiguration engineConfiguration = new EngineConfiguration();
engineConfiguration.setDetectMode(DetectMode.ASF_DETECT_MODE_IMAGE);//图像模式
engineConfiguration.setDetectFaceOrientPriority(DetectOrient.ASF_OP_0_ONLY); //逆时针0度
engineConfiguration.setDetectFaceMaxNum(1);
engineConfiguration.setDetectFaceScaleVal(32);
//功能配置
FunctionConfiguration functionConfiguration = new FunctionConfiguration();
//functionConfiguration.setSupportAge(true); //年龄检测
//functionConfiguration.setSupportFace3dAngle(true); //3D角度检测
functionConfiguration.setSupportFaceDetect(true); //人脸检测
functionConfiguration.setSupportFaceRecognition(true); //人脸识别
//functionConfiguration.setSupportGender(true); //性别检测
functionConfiguration.setSupportLiveness(true); //活体检测
//functionConfiguration.setSupportIRLiveness(true); //红外活体检测
engineConfiguration.setFunctionConfiguration(functionConfiguration);
//初始化引擎
errorCode.set(faceEngine.init(engineConfiguration));
if (errorCode.get() != ErrorInfo.MOK.getValue()) {
logger.debug("初始化引擎失败" + errorCode);
}
OpenCVFrameGrabber grabber = new OpenCVFrameGrabber(0);//新建opencv抓取器,一般的电脑和移动端设备中摄像头默认序号是0,不排除其他情况
grabber.start();//开始获取摄像头数据
CanvasFrame canvas = new CanvasFrame("摄像头预览");//新建一个预览窗口
canvas.setDefaultCloseOperation(JFrame.DO_NOTHING_ON_CLOSE);
//grabber抓的是fram
//faceEngine.detectFaces 要用到 imageInfo
//imageInfo要通过getRGBData获得
//getRGBData(byte[] bytes) getRGBData(InputStream input)
//byte[] 或者 InputStream 格式的数据都可以用
//grabber抓的是frame
//如何将frame转化为 byte[]或者 InputStream
//目前只能frame转BufferedImage转 byte[]
Frame frame = new Frame();
List faceInfoList = new ArrayList();
ImageInfo imageInfo = new ImageInfo();
ByteArrayOutputStream bStream = new ByteArrayOutputStream();
int countTimes = 0;
OpenCVFrameConverter converter = new OpenCVFrameConverter.ToMat();
boolean faceDetected = false;
Point pointA = new Point();
Point pointB = new Point();
//创建feature的list
List rfiList = new ArrayList<>();
//读取feature文件夹里面的数据,写入list里面
//文件夹位置
String featurePath = "./feature"; // 特征目录路径
File featureDocument = new File(featurePath);//获取路径
//模板路径的目录不存在的话直接跳出
if (!featureDocument.exists()) {
logger.debug("人员特征目录" + featurePath + "不存在");//不存在就输出
return;
}
File fa[] = featureDocument.listFiles();//用数组接收
//没有文件直接跳出
if(fa.length == 0){
logger.debug("没有特征数据");
}
//读取
//is
for (File feature:fa) {
try {
//创建输入流
InputStream inputStream = new FileInputStream(feature.getAbsoluteFile());
//创建特征
RegistedFeatureInfo registedFeatureInfo = new RegistedFeatureInfo();
FaceFeature faceFeature = new FaceFeature();
//创建缓存
byte[] bytes = new byte[faceFeature.FEATURE_SIZE];
//读数据进缓存
inputStream.read(bytes);
//写数据进特征
faceFeature.setFeatureData(bytes);
registedFeatureInfo.setFaceFeature(faceFeature);
registedFeatureInfo.setIdNum(feature.getName().substring(0, 18));
//添加到list里面
rfiList.add(registedFeatureInfo);
//关闭is
inputStream.close();
logger.debug("现有" + rfiList.size() + "人在库");
} catch (FileNotFoundException e) {
e.printStackTrace();
} catch (IOException e) {
e.printStackTrace();
}
}
//窗口是否关闭
while(canvas.isDisplayable()){
frame = grabber.grab();
countTimes++;
if(countTimes % 10 == 0){
BufferedImage bufferedImage = Java2DFrameUtils.toBufferedImage(frame);
try {
ImageIO.write(bufferedImage, "jpg", bStream);
} catch (IOException e) {
e.printStackTrace();
}
imageInfo = getRGBData(bStream.toByteArray());
bStream.reset();
errorCode.set(faceEngine.detectFaces(imageInfo.getImageData(),
imageInfo.getWidth(),
imageInfo.getHeight(),
imageInfo.getImageFormat(),
faceInfoList));
if(faceInfoList.size() > 0){
//logger.debug(faceInfoList.get(0).getRect());
//更新矩形的信息
pointA.x(faceInfoList.get(0).getRect().getLeft());
pointA.y(faceInfoList.get(0).getRect().getTop());
pointB.x(faceInfoList.get(0).getRect().getRight());
pointB.y(faceInfoList.get(0).getRect().getBottom());
if(pointB.x()-pointA.x() > 150){
faceDetected = true;
//提取这个人的特征
//特征提取2
FaceFeature faceFeature2 = new FaceFeature();
faceEngine.extractFaceFeature(imageInfo.getImageData(), imageInfo.getWidth(),
imageInfo.getHeight(), imageInfo.getImageFormat(), faceInfoList.get(0), faceFeature2);
//看看这个人是谁
FaceSimilar faceSimilar = new FaceSimilar();
Float fSimilar = 0.0F;
String idNumR = "";
for (RegistedFeatureInfo person:rfiList) {
//对比
faceEngine.compareFaceFeature(faceFeature2, person.getFaceFeature(), faceSimilar);
//如果faceSimilar大于存储值,更新
if(faceSimilar.getScore() > fSimilar){
fSimilar = faceSimilar.getScore();
idNumR = person.getIdNum();
}
}
if(fSimilar > 0.8F){
logger.debug("人员对比最高值为" + fSimilar.toString() + ";id为" + idNumR);
}
}else{
faceDetected = false;
}
} else {
faceDetected = false;
}
}
//如果有头像
if(faceDetected){
//将frame转换成mat
Mat mat = Java2DFrameUtils.toMat(frame);
//加上方框
//rectangle(grabbedImage, new Point(x, y), new Point(x + w, y + h), Scalar.RED, 1, CV_AA, 0);
rectangle(mat, pointA, pointB, Scalar.GREEN, 5, CV_AA, 0);
//mat转换成frame
frame = converter.convert(mat);
}
canvas.showImage(frame);
try {
Thread.sleep(10);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
grabber.close();//停止抓取
}
}
springboot缺省使用headless模式,要禁用,否则会报awt异常
public class RlsbjApplication {
public static void main(String[] args) {
//springboot缺省用headless模式启动,要禁用这种模式,否则会报错
SpringApplicationBuilder builder = new SpringApplicationBuilder(RlsbjApplication.class);
builder.headless(false).run(args);
//SpringApplication.run(RlsbjApplication.class, args);
Javacv javacv = new Javacv();
try {
javacv.openCam();
} catch (FrameGrabber.Exception e) {
e.printStackTrace();
}
}
}
运行效果
参考1:https://blog.csdn.net/eguid_1/article/details/51659578



