【原】训练自己的haar-like特征分类器并识别物体（3）

在前两篇文章中，我介绍了《训练自己的haar-like特征分类器并识别物体》的前三个步骤：

1.准备训练样本图片，包括正例及反例样本

2.生成样本描述文件

3.训练样本

4.目标识别

==============

本文将着重说明最后一个阶段——目标识别，也即利用前面训练出来的分类器文件(.xml文件)对图片中的物体进行识别，并在图中框出在该物体。由于逻辑比较简单，这里直接上代码：

int _tmain(int argc, _TCHAR* argv[])
{
	char *cascade_name = CASCADE_HEAD_MY; //上文最终生成的xml文件命名为"CASCADE_HEAD_MY.xml"
	cascade = (CvHaarClassifierCascade*)cvLoad( cascade_name, 0, 0, 0 ); //加载xml文件

	if( !cascade ) 
	{ 
		fprintf( stderr, "ERROR: Could not load classifier cascade
" );
		system("pause");
		return -1;
	}
	storage = cvCreateMemStorage(0); 
	cvNamedWindow( "face", 1 ); 

	const char* filename = "(12).bmp"; 
	IplImage* image = cvLoadImage( filename, 1 );

	if( image ) 
	{ 
		detect_and_draw( image ); //函数见下方
		cvWaitKey(0); 
		cvReleaseImage( &image );   
	}
	cvDestroyWindow("result");
	return 0;
}

void detect_and_draw(IplImage* img ) 
{ 
    double scale=1.2; 
    static CvScalar colors[] = { 
        {{0,0,255}},{{0,128,255}},{{0,255,255}},{{0,255,0}}, 
        {{255,128,0}},{{255,255,0}},{{255,0,0}},{{255,0,255}} 
    };//Just some pretty colors to draw with

    //Image Preparation 
    // 
    IplImage* gray = cvCreateImage(cvSize(img->width,img->height),8,1); 
    IplImage* small_img=cvCreateImage(cvSize(cvRound(img->width/scale),cvRound(img->height/scale)),8,1); 
    cvCvtColor(img,gray, CV_BGR2GRAY); 
    cvResize(gray, small_img, CV_INTER_LINEAR);

    cvEqualizeHist(small_img,small_img); //直方图均衡

    //Detect objects if any 
    // 
    cvClearMemStorage(storage); 
    double t = (double)cvGetTickCount(); 
    CvSeq* objects = cvHaarDetectObjects(small_img, 
        cascade, 
        storage, 
        1.1, 
        2, 
        0/*CV_HAAR_DO_CANNY_PRUNING*/, 
        cvSize(30,30));

    t = (double)cvGetTickCount() - t; 
    printf( "detection time = %gms
", t/((double)cvGetTickFrequency()*1000.) );

    //Loop through found objects and draw boxes around them 
    for(int i=0;i<(objects? objects->total:0);++i) 
    { 
        CvRect* r=(CvRect*)cvGetSeqElem(objects,i); 
        cvRectangle(img, cvPoint(r->x*scale,r->y*scale), cvPoint((r->x+r->width)*scale,(r->y+r->height)*scale), colors[i%8]); 
    } 
    for( int i = 0; i < (objects? objects->total : 0); i++ ) 
    { 
        CvRect* r = (CvRect*)cvGetSeqElem( objects, i ); 
        CvPoint center; 
        int radius; 
        center.x = cvRound((r->x + r->width*0.5)*scale); 
        center.y = cvRound((r->y + r->height*0.5)*scale); 
        radius = cvRound((r->width + r->height)*0.25*scale); 
        cvCircle( img, center, radius, colors[i%8], 3, 8, 0 ); 
    }

    cvShowImage( "result", img ); 
    cvReleaseImage(&gray); 
    cvReleaseImage(&small_img); 
}

===================================

其实上面的代码可以运用于大部分模式识别问题，无论是自己生成的xml文件还是opencv自带的xml文件。在opencv的工程目录opencvdata文件夹下有大量的xml文件，这些都是opencv开源项目中的程序员们自己训练出来的。然而，效果一般不会合你预期，所以才有了本系列文章。天下没有免费的午餐，想要获得更高的查准率与查全率，不付出点努力是不行的！