java调用dll进行人脸识别（JNI的使用）

1.JNI部分

调用JNI基类（JNIBase.java）

 package lveyo.bcndyl.jni.opencv; public class JNIBase { public JNIBase(){} public JNIBase(String libraryName){ loadLibrary(libraryName); } private static void loadLibrary(String libraryName){ System.loadLibrary(libraryName); } }

实现基类(JNIOpencv.java)

package lveyo.bcndyl.jni.opencv; public class JNIOpencv extends JNIBase{ public JNIOpencv (String libraryName){ super(libraryName); } public JNIOpencv(){ System.loadLibrary("jniOpenCV"); } public native int[] detectFace(int minFaceWidth, int minFaceHeight, String cascade, String filename); }

 

2.编译文件

运行——>cmd——>进入java源文件夹

javac -d . JNIBase.java

javac -d . JNIOpencv.java

javah lveyo.bcndyl.jni.opencv.JNIOpencv //生成lveyo_bcndyl_jni_opencv_JNIOpencv.h头文件

 

二、visual C++6.0

3.建立工程（选择win32 Dynamic-Link Library）

tools---->Options---->Drectories(添加以下目录)

C:/jdk1.6/INCLUDE //根据具体jdk安装目录确定C:/jdk1.6/INCLUDE/WIN32

4.点击工程右键（先确定自己已经安装和配置好OpenCV，可以到OpenCV中文官网参考visualC++6.0配置安装opencv步骤）

选择“设置”，下拉选择“所有配置”，在右方区域点击连接，在对象/库模块下添加

cxcore.lib cv.lib ml.lib cvaux.lib highgui.lib cvcam.lib

5.新建头文件veyo_bcndyl_jni_opencv_JNIOpencv.h

添加在编译时生成的头文件代码：

#include <jni.h>#ifndef _Included_lveyo_bcndyl_jni_opencv_JNIOpencv#define _Included_lveyo_bcndyl_jni_opencv_JNIOpencv#ifdef __cplusplusextern "C" {#endifJNIEXPORT jintArray JNICALL Java_lveyo_bcndyl_jni_opencv_JNIOpencv_detectFace (JNIEnv *, jobject, jint, jint, jstring, jstring);#ifdef __cplusplus}#endif#endif

6.新建jniOpenCV.cpp程序文件（检测主程序）

//#include "stdafx.h" #include <jni.h> #include "lveyo_bcndyl_jni_opencv_JNIOpencv.h" #include "cv.h" #include "highgui.h" JNIEXPORT jintArray JNICALL Java_lveyo_bcndyl_jni_opencv_JNIOpencv_detectFace (JNIEnv *env, jobject obj, jint width, jint height, jstring cascade, jstring filename) { const char *str_cascade, *str_filename; str_cascade = env->GetStringUTFChars(cascade, false); str_filename = env->GetStringUTFChars(filename, false); jintArray faceArray; CvHaarClassifierCascade *cv_cascade = (CvHaarClassifierCascade*)cvLoad( str_cascade ); IplImage *image = cvLoadImage( str_filename, 1 ); if(image!=0){ CvMemStorage* storage = cvCreateMemStorage(0); CvSeq* faces; //double t = (double)cvGetTickCount(); /* use the fastest variant */ faces = cvHaarDetectObjects( image, cv_cascade, storage, 1.2, 2, CV_HAAR_DO_CANNY_PRUNING, cvSize(width, height) ); //t = (double)cvGetTickCount() - t; //printf( "detection time = %gms/n", t/((double)cvGetTickFrequency()*1000.) ); const int total = faces->total; faceArray = env-> NewIntArray(4*total); jint faceBuf[4]; for( int i = 0; i < total; i++ ) { CvRect face_rect = *(CvRect*)cvGetSeqElem( faces, i ); int pointX = face_rect.x; int pointY = face_rect.y; int faceWidth = face_rect.width; int faceHeight = face_rect.height; //printf("i %d, x %d, y %d, width %d, height %d/n", // i,pointX,pointY,faceWidth,faceHeight); faceBuf[0] = pointX; faceBuf[1] = pointY; faceBuf[2] = faceWidth; faceBuf[3] = faceHeight; env->SetIntArrayRegion(faceArray,i*4,4,faceBuf); } cvReleaseMemStorage( &storage ); cvReleaseImage( &image ); } cvReleaseHaarClassifierCascade( &cv_cascade ); env->ReleaseStringUTFChars(cascade, str_cascade); env->ReleaseStringUTFChars(filename, str_filename); return faceArray; }

然后进行编译、组建

7.进入c++源文件夹把DEBUG里生成的jniOpenCV.dll拖入D:/mylib里（具体原因参考我的另一篇文章《java调用dll（JNI的使用）》）同时把Opencv的C:/Program Files/OpenCV/bin目录下的所有dll文件拖入D:/mylib中(java程序调用需要)

 

8.D:/mylib中需要拖入haarcascade_frontalface_alt2.xml文件（OpenCV提供的人脸特征文件）

 

 

三、回到java部分

9.新建Test.java(进行程序测试)

注意先在D盘放入人物图片

package lveyo.bcndyl.jni.opencv; public class Test { public static void main(String[] args) { //初始化JNI调用类JNIOpencv JNIOpencv open = new JNIOpencv("jniOpenCV"); //要检测的图片文件 String filename = "d:/80010.jpg"; //OpenCv提供的人间的特征文件 String cascade = "d:/haarcascade_frontalface_alt2.xml"; //人脸检测，前两个参数为可检测的最小人脸的宽度和高度 //返回值为人脸在图中的坐标和宽高，{x, y, width, height} int[] faces = open.detectFace(40, 40, cascade, filename); if(faces != null && faces.length!=0){ //返回的人脸总数 System.out.println( "faces " + faces.length/4 ); //分别输出每个人脸的坐标信息 for (int temp : faces) { System.out.println(temp); } } } }

运行程序

如果嫌麻烦，这里有个下载地址：http://download.csdn.net/source/2834782