OpenCV学习4-----K-Nearest Neighbors（KNN）demo

最近用到KNN方法，学习一下OpenCV给出的demo。

demo大意是随机生成两团二维空间中的点，然后在500*500的二维空间平面上，计算每一个点属于哪一个类，然后用红色和绿色显示出来每一个点

如下是一系demo里用到的相关函数。

运行效果：

红色背景应该是表示每一个像素的类别标签和红色的点的标签相同。同理，绿色背景表示绿色的像素与绿色的点是同一个类的。

demo.cpp:

#include "ml.h"
#include "highgui.h"

int main( int argc, char** argv )
{
    const int K = ;
    int i, j, k, accuracy;
    float response;
    int train_sample_count = ;
    CvRNG rng_state = cvRNG(-);
    CvMat* trainData = cvCreateMat( train_sample_count, , CV_32FC1 );
    CvMat* trainClasses = cvCreateMat( train_sample_count, , CV_32FC1 );
    IplImage* img = cvCreateImage( cvSize( ,  ), ,  );
    float _sample[];
    CvMat sample = cvMat( , , CV_32FC1, _sample );
    cvZero( img );

    CvMat trainData1, trainData2, trainClasses1, trainClasses2;

    // form the training samples
    cvGetRows( trainData, &trainData1, , train_sample_count/ );
    cvRandArr( &rng_state, &trainData1, CV_RAND_NORMAL, cvScalar(,), cvScalar(,) );

    cvGetRows( trainData, &trainData2, train_sample_count/, train_sample_count );
    cvRandArr( &rng_state, &trainData2, CV_RAND_NORMAL, cvScalar(,), cvScalar(,) );

    cvGetRows( trainClasses, &trainClasses1, , train_sample_count/ );
    cvSet( &trainClasses1, cvScalar() );

    cvGetRows( trainClasses, &trainClasses2, train_sample_count/, train_sample_count );
    cvSet( &trainClasses2, cvScalar() );

    // learn classifier
    CvKNearest knn( trainData, trainClasses, , false, K );
    CvMat* nearests = cvCreateMat( , K, CV_32FC1);

    for( i = ; i < img->height; i++ )
    {
        for( j = ; j < img->width; j++ )
        {
            sample.data.fl[] = (float)j;
            sample.data.fl[] = (float)i;

            // estimate the response and get the neighbors' labels
            response = knn.find_nearest(&sample,K,,,nearests,);

            // compute the number of neighbors representing the majority
            for( k = , accuracy = ; k < K; k++ )
            {
                if( nearests->data.fl[k] == response)
                    accuracy++;
            }
            // highlight the pixel depending on the accuracy (or confidence)
            cvSet2D( img, i, j, response ==  ?
                (accuracy >  ? CV_RGB(,,) : CV_RGB(,,)) :
                (accuracy >  ? CV_RGB(,,) : CV_RGB(,,)) );
        }
    }

    // display the original training samples
    for( i = ; i < train_sample_count/; i++ )
    {
        CvPoint pt;
        pt.x = cvRound(trainData1.data.fl[i*]);
        pt.y = cvRound(trainData1.data.fl[i*+]);
        cvCircle( img, pt, , CV_RGB(,,), CV_FILLED );
        pt.x = cvRound(trainData2.data.fl[i*]);
        pt.y = cvRound(trainData2.data.fl[i*+]);
        cvCircle( img, pt, , CV_RGB(,,), CV_FILLED );
    }

    cvNamedWindow( "classifier result",  );
    cvShowImage( "classifier result", img );
    cvWaitKey();

    cvReleaseMat( &trainClasses );
    cvReleaseMat( &trainData );
    return ;
}

参考：

https://docs.opencv.org/2.4/modules/ml/doc/k_nearest_neighbors.html

https://docs.opencv.org/2.4/modules/core/doc/old_basic_structures.html?highlight=cvset#void cvSet2D(CvArr* arr, int idx0, int idx1, CvScalar value)