opencv3.10 +LDA线性判别分析举例

1. LDA类的声明文件在core.hpp中，可以通过包含opencv2/opencv.hpp来包含。

/**
   @brief Linear Discriminant Analysis
   @todo document this class
 */
class CV_EXPORTS LDA
{
public:
    /** @brief constructor
    Initializes a LDA with num_components (default 0).
    */
    explicit LDA(int num_components = 0);

    /** Initializes and performs a Discriminant Analysis with Fisher's
     Optimization Criterion on given data in src and corresponding labels
     in labels. If 0 (or less) number of components are given, they are
     automatically determined for given data in computation.
    */
    LDA(InputArrayOfArrays src, InputArray labels, int num_components = 0);

    /** Serializes this object to a given filename.
      */
    void save(const String& filename) const;

    /** Deserializes this object from a given filename.
      */
    void load(const String& filename);

    /** Serializes this object to a given cv::FileStorage.
      */
    void save(FileStorage& fs) const;

    /** Deserializes this object from a given cv::FileStorage.
      */
    void load(const FileStorage& node);

    /** destructor
      */
    ~LDA();

    /** Compute the discriminants for data in src (row aligned) and labels.
      */
    void compute(InputArrayOfArrays src, InputArray labels);

    /** Projects samples into the LDA subspace.
        src may be one or more row aligned samples.
      */
    Mat project(InputArray src);

    /** Reconstructs projections from the LDA subspace.
        src may be one or more row aligned projections.
      */
    Mat reconstruct(InputArray src);

    /** Returns the eigenvectors of this LDA.
      */
    Mat eigenvectors() const { return _eigenvectors; }

    /** Returns the eigenvalues of this LDA.
      */
    Mat eigenvalues() const { return _eigenvalues; }

    static Mat subspaceProject(InputArray W, InputArray mean, InputArray src);
    static Mat subspaceReconstruct(InputArray W, InputArray mean, InputArray src);

protected:
    bool _dataAsRow; // unused, but needed for 3.0 ABI compatibility.
    int _num_components;
    Mat _eigenvectors;
    Mat _eigenvalues;
    void lda(InputArrayOfArrays src, InputArray labels);
};

2. 使用举例

#include "opencv2/highgui/highgui.hpp"
#include "opencv2/features2d.hpp"
#include "opencv2/imgproc/imgproc.hpp"
//Opencv 3.X 系列的LDA类在core.hpp中，而opencv.hpp中有了对core.hpp的包含
#include <opencv2/opencv.hpp>
#include <iostream>

using namespace std;
using namespace cv;
int main(void)
{
	//样本数据
	double sampledata[7][2]={{2.95,6.63},{2.53,7.79},{3.57,5.65},{3.16,5.47},{2.58,4.46},{2.16,6.22},{3.27,3.52}};
	Mat dataMat=Mat(7,2,CV_64FC1,sampledata);
	//样本标签
	int dataLable[7]={0,0,0,0,1,1,1};
	Mat labelMat=Mat(7,1,CV_32SC1,dataLable);
	//LDA计算-两种方法
	/*
	第一种直接使用带参数的构造函数LDA (InputArrayOfArrays src, InputArray labels, int num_components=0)，实际上内部也是调用了成员函数void  compute (InputArrayOfArrays src, InputArray labels)来计算特征值特特征向量
	第二种方法就是构造函数LDA (int num_components=0)，然后调用成员函数compute函数
	*/
	LDA lda=LDA();
	lda.compute(dataMat,labelMat);
	//输出特征向量-所求的子空间
	Mat eivector=lda.eigenvectors();
	cout<<"rows:"<<eivector.rows<<endl;
	cout<<"cols:"<<eivector.cols<<endl;
	cout<<"The eigenvector is:"<<endl;
	for(int i=0;i<eivector.rows;i++)
	{
		for(int j=0;j<eivector.cols;j++)
		{
			cout<<eivector.ptr<double>(i)[j]<<" ";
		}
		cout<<endl;
	}
	//输出特征值
	cout<<"----------------------------"<<endl;
	Mat eigenValue=lda.eigenvalues();
	cout<<"rows:"<<eigenValue.rows<<endl;
	cout<<"cols:"<<eigenValue.cols<<endl;
	for(int i=0;i<eigenValue.rows;i++)
	{
		for(int j=0;j<eigenValue.cols;j++)
		{
			cout<<eigenValue.ptr<double>(i)[j]<<" ";
		}
		cout<<endl;
	}
	//投影-将样本投影到新的空间
	cout<<"----------------------------"<<endl;
	Mat dataSub =lda.project(dataMat);
	cout<<"rows:"<<dataSub.rows<<endl;
	cout<<"cols:"<<dataSub.cols<<endl;
	for(int i=0;i<dataSub.rows;i++)
	{
		for(int j=0;j<dataSub.cols;j++)
		{
			cout<<dataSub.ptr<double>(i)[j]<<" ";
		}
		cout<<endl;
	}


	system("pause");
	return 0;
}