OpenCV KNN 之 使用方法

http://blog.csdn.net/WL2002200/article/details/43149229

OpenCV 中KNN构造函数如下。

C++: CvKNearest::CvKNearest()
C++: CvKNearest::CvKNearest(const Mat& trainData, const Mat& responses, const Mat& sam-
pleIdx=Mat(), bool isRegression=false, int max_k=32 )

训练函数为：

C++: bool CvKNearest::train(
    const Mat& trainData, //训练数据
    const Mat& responses,//对应的响应值
    const Mat& sampleIdx=Mat(),//样本索引
    bool isRegression=false,//是否是回归，否则是分类问题
    int maxK=32, //最大K值
    bool updateBase=false//是否更新数据，是，则maxK需要小于原数据大小 )

查找函数：

C++: float CvKNearest::find_nearest(
const Mat& samples,//按行存储的测试数据
 int k, //K 值
Mat* results=0,//预测结果
const float** neighbors=0, //近邻指针向量
Mat* neighborResponses=0, //近邻值
Mat* dist=0 //距离矩阵) const

C++: float CvKNearest::find_nearest(
const Mat& samples,
int k,
Mat& results,
Mat& neighborResponses,
Mat& dists) const

还有一些其他辅助函数，无关紧要，略去了。

opencv 有KNN 的示例，改写成C++ 版本如下：

#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/ml/ml.hpp>

int main( )
{
    const int K = 10;
    int i, j, k, accuracy;
    float response;
    int train_sample_count = 100;
    cv::RNG rng_state(-1);
    cv::Mat trainData(train_sample_count,2,CV_32FC1);
    cv::Mat trainClasses(train_sample_count,1,CV_32FC1);
    cv::Mat img(cv::Size(500,500),CV_8UC3,cv::Scalar::all (0));
    float _sample[2];
    cv::Mat sample(1,2,CV_32FC1,_sample);

    cv::Mat trainData1, trainData2, trainClasses1, trainClasses2;

    // form the training samples
    trainData1 = trainData.rowRange (0,train_sample_count/2);
    rng_state.fill (trainData1,CV_RAND_NORMAL,cv::Scalar(200,200),cv::Scalar(50,50));

    trainData2 = trainData.rowRange (train_sample_count/2,train_sample_count);
    rng_state.fill (trainData2,CV_RAND_NORMAL,cv::Scalar(300,300),cv::Scalar(50,50));

    trainClasses1 = trainClasses.rowRange (0,train_sample_count/2);
    trainClasses1.setTo (1);

    trainClasses2 = trainClasses.rowRange (train_sample_count/2,train_sample_count);
    trainClasses2.setTo (2);

    // learn classifier
    CvKNearest knn( trainData, trainClasses, cv::Mat(), false, K );
    cv::Mat nearests( 1, K, CV_32FC1);

    for( i = 0; i < img.rows; i++ )
    {
        for( j = 0; j < img.cols; j++ )
        {
            sample.at<float>(0,0) = (float)j;
            sample.at<float>(0,1) = (float)i;

            // estimate the response and get the neighbors' labels
            response = knn.find_nearest(sample,K,0,0,&nearests,0);

            // compute the number of neighbors representing the majority
            for( k = 0, accuracy = 0; k < K; k++ )
            {
                if( nearests.at<float>(0,k) == response)
                    accuracy++;
            }
            // highlight the pixel depending on the accuracy (or confidence)
            img.at<cv::Vec3b>(i,j) = response == 1 ?
                        (accuracy > 5 ? cv::Vec3b(0,0,180) : cv::Vec3b(0,120,180)) :
                        (accuracy > 5 ? cv::Vec3b(0,180,0) : cv::Vec3b(0,120,120));
        }
    }

    // display the original training samples
    for( i = 0; i < train_sample_count/2; i++ )
    {
        cv::Point pt;
        pt.x = cvRound(trainData1.at<float>(i,0));
        pt.y = cvRound(trainData1.at<float>(i,1));
        cv::circle (img,pt,2,cv::Scalar(0,0,255),1,CV_FILLED);
        pt.x = cvRound(trainData2.at<float>(i,0));
        pt.y = cvRound(trainData2.at<float>(i,1));
        cv::circle (img,pt,2,cv::Scalar(0,255,0),1,CV_FILLED);
    }

    cv::namedWindow( "classifier result", 1 );
    cv::imshow( "classifier result", img );
    cv::waitKey(0);

    return 0;
}