opencv 特征点提取算法 SIFT SURF ORB FAST LBP学习（二）

demo: http://download.youkuaiyun.com/detail/keen_zuxwang/9852587

MainActivity.java:

...
public int doFeatureDetect(int detect_type)
    {
        int length0, length1;
        Mat img_object = new Mat();
        Mat img_scene  = new Mat();

        Utils.bitmapToMat(bt1, img_object);
        Utils.bitmapToMat(bt2, img_scene);

        /*
                    FAST = 1,
                    STAR = 2,
                    SIFT = 3,
                    SURF = 4,
                    ORB = 5,
                    MSER = 6,
                    GFTT = 7,
                    HARRIS = 8,
                    SIMPLEBLOB = 9,
                    DENSE = 10,
                    BRISK = 11,
                    GRIDRETECTOR = 1000,
        */          
        //-- Step 1: Detect the keypoints using SURF Detector
        FeatureDetector detector;
        String str_detect;
        Log.i("Feature", "  detect_type = "+detect_type);
        switch(detect_type){
                case 1:
                     detector = FeatureDetector.create(FeatureDetector.FAST); // ok
                     str_detect = "FAST";
                     break;
                case 2:
                     detector = FeatureDetector.create(FeatureDetector.STAR); // ok
                     str_detect = "STAR";
                     break;
                case 3:
                     detector = FeatureDetector.create(FeatureDetector.SIFT);
                     str_detect = "SIFT";
                     break;
                case 4:
                     detector = FeatureDetector.create(FeatureDetector.SURF);
                     str_detect = "SURF";
                     break;
                case 5:
                     detector = FeatureDetector.create(FeatureDetector.ORB); // ok
                     str_detect = "ORB";
                     break;
                case 6:
                     detector = FeatureDetector.create(FeatureDetector.MSER);
                     str_detect = "MSER";
                     break;
                case 7:
                     detector = FeatureDetector.create(FeatureDetector.GFTT);
                     str_detect = "GFTT";
                     break;
                case 8:
                     detector = FeatureDetector.create(FeatureDetector.HARRIS); // ok
                     str_detect = "HARRIS";
                     break;
                case 9:
                     detector = FeatureDetector.create(FeatureDetector.SIMPLEBLOB);
                     str_detect = "SIMPLEBLOB";
                     break;
                case 10:
                     detector = FeatureDetector.create(FeatureDetector.DENSE);
                     str_detect = "DENSE";
                     break;
                case 11:
                     detector = FeatureDetector.create(FeatureDetector.BRISK);
                     str_detect = "BRISK";
                     break;
                case 12:
                     detector = FeatureDetector.create(FeatureDetector.GRIDRETECTOR);
                     str_detect = "GRIDRETECTOR";
                     break;
                default:
                     detector = FeatureDetector.create(FeatureDetector.FAST);
                     str_detect = "FAST";
                     break;
           }


           MatOfKeyPoint keypoints_object = new MatOfKeyPoint();
           MatOfKeyPoint keypoints_scene  = new MatOfKeyPoint();

           detector.detect(img_object, keypoints_object);
           detector.detect(img_scene, keypoints_scene);

           KeyPoint[] obj_item0 = keypoints_object.toArray();
           KeyPoint[] obj_item1 = keypoints_scene.toArray();

           length0 = obj_item0.length;
           length1 = obj_item1.length;

           Log.i("Feature", "  length0 = "+length0+", length1 = "+length1); 
           for( int i = 0; i < length0; i++ ) {
               //-- Get the keypoints
               Point pts = obj_item0[i].pt;
               Core.rectangle(img_object, 
                       new Point(pts.x-5, pts.y-5),
                       new Point(pts.x+5, pts.y+5),
                       new Scalar(0, 255, detect_type*20),  
                       2);
           }

           for( int i = 0; i < length1; i++ ) {
               //-- Get the keypoints 
               Point pts = obj_item1[i].pt;
               Core.rectangle(img_scene, 
                       new Point(pts.x-5, pts.y-5),
                       new Point(pts.x+5, pts.y+5),
                       new Scalar(0, 255, detect_type*20),  
                       2);
           }

           bt3 = Bitmap.createBitmap(img_object.cols(), img_object.rows(), Config.RGB_565);
           Utils.matToBitmap(img_object, bt3);
           iv0.setImageBitmap(bt3);

           bt4 = Bitmap.createBitmap(img_scene.cols(), img_scene.rows(), Config.RGB_565);
           Utils.matToBitmap(img_scene, bt4);
           iv1.setImageBitmap(bt4);

           Log.w("Feature","     FeatureDetector    \n"); 
           return 1;
    }
..

JNI:
orbhog.cpp

JNIEXPORT jlong JNICALL Java_com_example_orbhog_MainActivity_doHarris(JNIEnv *env, jclass clz, jlong imageGray)
{
    int win_size=15;//10;
    int r=3;
    int count=0;
    const int MAX_COUNT=500; // 500
    double quality=0.01;
    double min_distance= 15;//10;
    CvPoint2D32f *points[2]={0,0};

    points[1]=(CvPoint2D32f*)cvAlloc(MAX_COUNT*sizeof(points[0][0]));

    Mat pImg = Mat(*(Mat*)imageGray);
    IplImage  temp_src = pImg;
    IplImage* imageg = &temp_src;

    IplImage* grey=cvCreateImage(cvSize(imageg->width, imageg->height), IPL_DEPTH_8U, 1);
    cvCvtColor(imageg, grey, CV_BGR2GRAY); // 1、灰度图

    //goodFeaturesToTrack(pGray, features, maxCount, qLevel, minDist);

    //automatic initialization
    IplImage* eig=cvCreateImage(cvGetSize(grey),32,1);
    IplImage* temp=cvCreateImage(cvGetSize(grey),32,1);

    count=MAX_COUNT;
    cvGoodFeaturesToTrack(grey, eig, temp,
                    points[1],
                    &count,
                    quality,
                    min_distance,
                    0,
                    3,
                    0,
                    0.04);   //读取第一帧影像

    //能够将角点位置精确到亚像素级精度,提取易于跟踪的特征点，特征点精确描述
    cvFindCornerSubPix(grey, points[1], count,
                      cvSize(win_size,win_size),
                      cvSize(-1,-1), // cvSize(1,1)就表示成忽略掉相邻1个像素
                      cvTermCriteria(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS,20,0.03)); //迭代次数iteration最小精度epsilon

    for(int i=0;i<count;i++){
        cvRectangle(imageg, cvPoint(points[1][i].x-r,points[1][i].y-r),cvPoint(points[1][i].x+r, points[1][i].y+r), cvScalar(0,0,255), 2);
    }
    cvReleaseImage(&eig);
    cvReleaseImage(&temp);

    Mat *hist = new Mat(imageg);
    return (jlong) hist;
}

Harris:

Orb:

Fast:

Harris (cvGoodFeaturesToTrack):