匹配模板imge（缩放）到主/较大图像 [英] Matching template imge(scaled) to Main/larger image

问题描述

我想在主图片中找到/检查subImage / template image，并想知道它的坐标，

I want to find/check subImage/template image in main image and want to know its coordinates,

我使用下面的链接提供的代码来实现它，

I have used code given at following link to implement it,

在iOS中检查图像中的子图像是否存在

如果模板图像的大小与匹配的大小完全相同，

It is working fine, if the size of the template image is exactly same as size of the matching part of larger image.

但是如果子图像缩小或放大比较大图像的匹配部分时，结果不正确。

But it is not giving result properly if the subimage is scaled down or scaled up than matching part of larger image.

推荐答案

使用OpenCV特性检测。它比模板匹配更准确。

Use OpenCV Feature Detection. it is more accurate than template matching..

请尝试使用此代码..

Please try with this code..

-(void)featureDetection:(UIImage*)largerImage withImage:(UIImage*)subImage
{
    cv::Mat tempMat1 = [largerImage CVMat];
    cv::Mat tempMat2 = [subImage CVMat];

    cv::cvtColor(tempMat1, tempMat1, CV_RGB2GRAY);
    cv::cvtColor(tempMat2, tempMat2, CV_RGB2GRAY);

    if( !tempMat1.data || !tempMat2.data ) {
        return;
    }

    //-- Step 1: Detect the keypoints using SURF Detector
    int minHessian = 25;

    cv::SurfFeatureDetector detector( minHessian ); // More Accurate bt take more time..
    //cv::FastFeatureDetector detector( minHessian ); //Less Accurate bt take less time..

    std::vector<cv::KeyPoint> keypoints_1, keypoints_2;

    detector.detect( tempMat1, keypoints_1 );
    detector.detect( tempMat2, keypoints_2 );

    //-- Step 2: Calculate descriptors (feature vectors)
    cv::SurfDescriptorExtractor extractor;

    cv::Mat descriptors_1, descriptors_2;

    extractor.compute( tempMat1, keypoints_1, descriptors_1 );
    extractor.compute( tempMat2, keypoints_2, descriptors_2 );

    std::vector<cv::Point2f> obj_corners(4);

    //Get the corners from the object
    obj_corners[0] = (cvPoint(0,0));
    obj_corners[1] = (cvPoint(tempMat2.cols,0));
    obj_corners[2] = (cvPoint(tempMat2.cols,tempMat2.rows));
    obj_corners[3] = (cvPoint(0, tempMat2.rows));

    //-- Step 3: Matching descriptor vectors with a brute force matcher
    //cv::BruteForceMatcher < cv::L2<float> > matcher;
    cv::FlannBasedMatcher matcher;
    //std::vector< cv::DMatch > matches;
    std::vector<cv::vector<cv::DMatch > > matches;

    std::vector<cv::DMatch > good_matches;
    std::vector<cv::Point2f> obj;
    std::vector<cv::Point2f> scene;
    std::vector<cv::Point2f> scene_corners(4);
    cv::Mat H;

    matcher.knnMatch( descriptors_2, descriptors_1, matches,2);

    for(int i = 0; i < cv::min(tempMat1.rows-1,(int) matches.size()); i++)  {

        if((matches[i][0].distance < 0.6*(matches[i][1].distance)) && ((int) matches[i].size()<=2 && (int) matches[i].size()>0))  {
            good_matches.push_back(matches[i][0]);
        }
    }
    cv::Mat img_matches;
    drawMatches( tempMat2, keypoints_2, tempMat1, keypoints_1, good_matches, img_matches );

    NSLog(@"good matches %lu",good_matches.size());

    if (good_matches.size() >= 4)  {

         for( int i = 0; i < good_matches.size(); i++ ) {
             //Get the keypoints from the good matches
             obj.push_back( keypoints_2[ good_matches[i].queryIdx ].pt );
             scene.push_back( keypoints_1[ good_matches[i].trainIdx ].pt );
         }

         H = findHomography( obj, scene, CV_RANSAC );

         perspectiveTransform( obj_corners, scene_corners, H);

         NSLog(@"%f %f",scene_corners[0].x,scene_corners[0].y);
         NSLog(@"%f %f",scene_corners[1].x,scene_corners[1].y);
         NSLog(@"%f %f",scene_corners[2].x,scene_corners[2].y);
         NSLog(@"%f %f",scene_corners[3].x,scene_corners[3].y);


         //Draw lines between the corners (the mapped object in the scene image )
         line( tempMat1, scene_corners[0], scene_corners[1], cvScalar(0, 255, 0), 4 );

         line( tempMat1, scene_corners[1], scene_corners[2], cvScalar( 0, 255, 0), 4 );

         line( tempMat1, scene_corners[2], scene_corners[3], cvScalar( 0, 255, 0), 4 );

         line( tempMat1, scene_corners[3], scene_corners[0], cvScalar( 0, 255, 0), 4 );
     }

     // View matching..

     UIImage *resultimage = [UIImage imageWithCVMat:img_matches];
     UIImageView *imageview = [[UIImageView alloc] initWithImage:resultimage];
     imageview.frame = CGRectMake(0, 0, 320, 240);
     [self.view addSubview:imageview];

     // View Result

     UIImage *resultimage2 = [UIImage imageWithCVMat:tempMat1];
     UIImageView *imageview2 = [[UIImageView alloc] initWithImage:resultimage2];
     imageview2.frame = CGRectMake(0, 240, 320, 240);
     [self.view addSubview:imageview2];
}

这篇关于匹配模板imge（缩放）到主/较大图像的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！