OpenCV机器学习库MLL

学习机器学习的时候，基本都是在用Matlab、Python写算法，做测试；

由于最近要用OpenCV写作业，兴起看了看OpenCV的机器学习模块(The Machine Learning Library——MLL)。

来看看MLL的主要构成：Statistical Model是个基类，下面的K-NN、SVM等都是其子类。

不太喜欢这个Statistical定语，Statistics在ML界横行的好多年，感觉温度已经降下来了。

来看下Statistical Model：

class CV_EXPORTS_W CvStatModel
{
public:
    CvStatModel();
    virtual ~CvStatModel();

    virtual void clear();

    CV_WRAP virtual void save( const char* filename, const char* name=0 ) const;
    CV_WRAP virtual void load( const char* filename, const char* name=0 );

    virtual void write( CvFileStorage* storage, const char* name ) const;
    virtual void read( CvFileStorage* storage, CvFileNode* node );
    virtual bool train(const Mat& train_data, const Mat& responses, Mat(), Mat(), CVParms params );
    virtual float predict(const Mat& sample, ...);

protected:
    const char* default_model_name;
};

void CvStatModel::clear() 清除内存重置模型状态；

 

void CvStatModel::save() /load() 保存/加载文件和模型；

void CvStatModel:read() /write() 读写文件和模型；

bool CvStatModel::train() 训练模型；

float CvStatModel::predict() 预测样本结果；

 

那么朴素贝叶斯、K-近邻、支持向量机、决策树等类都是继承CVStatModel；

使用这些方法的基本框架就是：

Method.train(train_data, responses, Mat(), Mat(), params);

Method.predict(sampleMat);

一个具体的例子<Support Vector Machines for Non-Linearly Separable Data>

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

#define NTRAINING_SAMPLES 100   // Number of training samples per class
#define FRAC_LINEAR_SEP  0.9f    // Fraction of samples which compose the linear separable part

using namespace cv;
using namespace std;

void help()
{
    cout<< "\n--------------------------------------------------------------------------" << endl
        << "This program shows Support Vector Machines for Non-Linearly Separable Data. " << endl
        << "Usage:"                                                              << endl
        << "./non_linear_svms" << endl
        << "--------------------------------------------------------------------------"  << endl
        << endl;
}

int main()
{
    help();

    // Data for visual representation
    const int WIDTH = 512, HEIGHT = 512;
    Mat I = Mat::zeros(HEIGHT, WIDTH, CV_8UC3);

    //--------------------- 1. Set up training data randomly ---------------------------------------
    Mat trainData(2*NTRAINING_SAMPLES, 2, CV_32FC1);
    Mat labels  (2*NTRAINING_SAMPLES, 1, CV_32FC1);
    
    RNG rng(100); // Random value generation class

    // Set up the linearly separable part of the training data
    int nLinearSamples = (int) (FRAC_LINEAR_SEP * NTRAINING_SAMPLES);

    // Generate random points for the class 1
    Mat trainClass = trainData.rowRange(0, nLinearSamples);
    // The x coordinate of the points is in [0, 0.4)
    Mat c = trainClass.colRange(0, 1);
    rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(0.4 * WIDTH));
    // The y coordinate of the points is in [0, 1)
    c = trainClass.colRange(1,2);
    rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));

    // Generate random points for the class 2
    trainClass = trainData.rowRange(2*NTRAINING_SAMPLES-nLinearSamples, 2*NTRAINING_SAMPLES);
    // The x coordinate of the points is in [0.6, 1]
    c = trainClass.colRange(0 , 1); 
    rng.fill(c, RNG::UNIFORM, Scalar(0.6*WIDTH), Scalar(WIDTH));
    // The y coordinate of the points is in [0, 1)
    c = trainClass.colRange(1,2);
    rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));

    //------------------ Set up the non-linearly separable part of the training data ---------------

    // Generate random points for the classes 1 and 2
    trainClass = trainData.rowRange(  nLinearSamples, 2*NTRAINING_SAMPLES-nLinearSamples);
    // The x coordinate of the points is in [0.4, 0.6)
    c = trainClass.colRange(0,1);
    rng.fill(c, RNG::UNIFORM, Scalar(0.4*WIDTH), Scalar(0.6*WIDTH)); 
    // The y coordinate of the points is in [0, 1)
    c = trainClass.colRange(1,2);
    rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));
    
    //------------------------- Set up the labels for the classes ---------------------------------
    labels.rowRange(                0,  NTRAINING_SAMPLES).setTo(1);  // Class 1
    labels.rowRange(NTRAINING_SAMPLES, 2*NTRAINING_SAMPLES).setTo(2);  // Class 2

    //------------------------ 2. Set up the support vector machines parameters --------------------
    CvSVMParams params;
    params.svm_type    = SVM::C_SVC;
    params.C     = 0.1;
    params.kernel_type = SVM::LINEAR;
    params.term_crit  = TermCriteria(CV_TERMCRIT_ITER, (int)1e7, 1e-6);

    //------------------------ 3. Train the svm ----------------------------------------------------
    cout << "Starting training process" << endl;
    CvSVM svm;
    svm.train(trainData, labels, Mat(), Mat(), params);
    cout << "Finished training process" << endl;
    
    //------------------------ 4. Show the decision regions ----------------------------------------
 Vec3b green(0,100,0), blue (100,0,0);
    for (int i = 0; i < I.rows; ++i)
        for (int j = 0; j < I.cols; ++j)
        {
            Mat sampleMat = (Mat_<float>(1,2) << i, j);
            float response = svm.predict(sampleMat);

            if      (response == 1)    I.at<Vec3b>(j, i)  = green;
            else if (response == 2)    I.at<Vec3b>(j, i)  = blue;
        }

    //----------------------- 5. Show the training data --------------------------------------------
    int thick = -1;
    int lineType = 8;
    float px, py;
    // Class 1
    for (int i = 0; i < NTRAINING_SAMPLES; ++i)
    {
        px = trainData.at<float>(i,0);
        py = trainData.at<float>(i,1);
        circle(I, Point( (int) px,  (int) py ), 3, Scalar(0, 255, 0), thick, lineType);
    }
    // Class 2
    for (int i = NTRAINING_SAMPLES; i <2*NTRAINING_SAMPLES; ++i)
    {
        px = trainData.at<float>(i,0);
        py = trainData.at<float>(i,1);
        circle(I, Point( (int) px, (int) py ), 3, Scalar(255, 0, 0), thick, lineType);
    }

    //------------------------- 6. Show support vectors --------------------------------------------
    thick = 2;
    lineType  = 8;
    int x    = svm.get_support_vector_count();

    for (int i = 0; i < x; ++i)
    {
        const float* v = svm.get_support_vector(i);
        circle( I,  Point( (int) v[0], (int) v[1]), 6, Scalar(128, 128, 128), thick, lineType);
    }

    imwrite("result.png", I);                    // save the Image
    imshow("SVM for Non-Linear Training Data", I); // show it to the user
    waitKey(0);
}

结果：