OpenCV 人脸检测自学(7)
目前需要提炼下ml部分的接口。目的是以后方便选择用哪种分类器。还是一头雾水啊。。。学到哪先记录到哪。
一。以CvSVM为例。下面是CvSVM类的定义：
class CV_EXPORTS_W CvSVM : public CvStatModel
{
public:
// SVM type
enum { C_SVC=100, NU_SVC=101, ONE_CLASS=102, EPS_SVR=103, NU_SVR=104 };
// SVM kernel type
enum { LINEAR=0, POLY=1, RBF=2, SIGMOID=3 };
// SVM params type
enum { C=0, GAMMA=1, P=2, NU=3, COEF=4, DEGREE=5 };
CV_WRAP CvSVM();
virtual ~CvSVM();
CvSVM( const CvMat* trainData, const CvMat* responses,
const CvMat* varIdx=0, const CvMat* sampleIdx=0,
CvSVMParams params=CvSVMParams() );
virtual bool train( const CvMat* trainData, const CvMat* responses,
const CvMat* varIdx=0, const CvMat* sampleIdx=0,//这两个参数好像不太用
CvSVMParams params=CvSVMParams() );
virtual float predict( const CvMat* sample, bool returnDFVal=false ) const;//
virtual float predict( const CvMat* samples, CV_OUT CvMat* results ) const;
CV_WRAP virtual int get_support_vector_count() const;
virtual const float* get_support_vector(int i) const;
virtual CvSVMParams get_params() const { return params; };
CV_WRAP virtual void clear();
static CvParamGrid get_default_grid( int param_id );
virtual void write( CvFileStorage* storage, const char* name ) const;
virtual void read( CvFileStorage* storage, CvFileNode* node );
CV_WRAP int get_var_count() const { return var_idx ? var_idx->cols : var_all; }
protected:
virtual bool set_params( const CvSVMParams& params );
virtual bool train1( int sample_count, int var_count, const float** samples,
const void* responses, double Cp, double Cn,
CvMemStorage* _storage, double* alpha, double& rho );
virtual bool do_train( int svm_type, int sample_count, int var_count, const float** samples,
const CvMat* responses, CvMemStorage* _storage, double* alpha );
virtual void create_kernel();
virtual void create_solver();
virtual float predict( const float* row_sample, int row_len, bool returnDFVal=false ) const;
virtual void write_params( CvFileStorage* fs ) const;
virtual void read_params( CvFileStorage* fs, CvFileNode* node );
CvSVMParams params;
CvMat* class_labels;
int var_all;
float** sv;
int sv_total;
CvMat* var_idx;
CvMat* class_weights;
CvSVMDecisionFunc* decision_func;
CvMemStorage* storage;
CvSVMSolver* solver;
CvSVMKernel* kernel;
};
SVM的接口基本上跟大部分分类器的差不多。train函数参数里一个是train_data,一个是response，最后一个是SVM对应的参数结构体。predict的参数就是一个1 x N的样本特征向量。下面是OpenCV提供的一个调用例子。
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/ml/ml.hpp>
using namespace cv;
int main()
{
// Data for visual representation
int width = 512, height = 512;
Mat image = Mat::zeros(height, width, CV_8UC3);
// Set up training data
float labels[4] = {1.0, -1.0, -1.0, -1.0};
Mat labelsMat(4, 1, CV_32FC1, labels);//对应于接口的_response
float trainingData[4][2] = { {501, 10}, {255, 10}, {501, 255}, {10, 501} };
Mat trainingDataMat(4, 2, CV_32FC1, trainingData);//对应于接口的_train_data
// Set up SVM's parameters
CvSVMParams params;
params.svm_type = CvSVM::C_SVC;
params.kernel_type = CvSVM::LINEAR;
params.term_crit = cvTermCriteria(CV_TERMCRIT_ITER, 100, 1e-6);
// Train the SVM
CvSVM SVM;
SVM.train(trainingDataMat, labelsMat, Mat(), Mat(), params);
Vec3b green(0,255,0), blue (255,0,0);
// Show the decision regions given by the SVM
for (int i = 0; i < image.rows; ++i)
for (int j = 0; j < image.cols; ++j)
{
Mat sampleMat = (Mat_<float>(1,2) << i,j);
float response = SVM.predict(sampleMat);
if (response == 1)
image.at<Vec3b>(j, i) = green;
else if (response == -1)
image.at<Vec3b>(j, i) = blue;
}
// Show the training data
int thickness = -1;
int lineType = 8;
circle( image, Point(501, 10), 5, Scalar( 0, 0, 0), thickness, lineType);
circle( image, Point(255, 10), 5, Scalar(255, 255, 255), thickness, lineType);
circle( image, Point(501, 255), 5, Scalar(255, 255, 255), thickness, lineType);
circle( image, Point( 10, 501), 5, Scalar(255, 255, 255), thickness, lineType);
// Show support vectors
thickness = 2;
lineType = 8;
int c = SVM.get_support_vector_count();
for (int i = 0; i < c; ++i)
{
const float* v = SVM.get_support_vector(i);
circle( image, Point( (int) v[0], (int) v[1]), 6, Scalar(128, 128, 128), thickness, lineType);
}
imwrite("result.png", image); // save the image
imshow("SVM Simple Example", image); // show it to the user
waitKey(0);
}
二。以cascadeclassifier为例说下它如何跟ml.hpp的关系。这个关系有点复杂，不像SVM那么标准了。
在traincascade\boost.cpp中
bool CvCascadeBoost::train( const CvFeatureEvaluator* _featureEvaluator,//包含了sum，tilted，特征的位置等信息
int _numSamples,
int _precalcValBufSize, int _precalcIdxBufSize,
const CvCascadeBoostParams& _params )
这个是训练一个强分类器的接口，里面调用训练一个弱分类器的接口是：ml\ml.hpp
bool
CvBoostTree::train( CvDTreeTrainData* _train_data,
const CvMat* _subsample_idx, CvBoost* _ensemble )
可是从ml.hpp文件中可以看到大部分从cvStatModel里面继承来的分类器的训练函数的结构应该是：
virtual bool train( const CvMat* train_data, [int tflag,] ..., const CvMat* responses, ...,
[const CvMat* var_idx,] ..., [const CvMat* sample_idx,] ...
[const CvMat* var_type,] ..., [const CvMat* missing_mask,] <misc_training_alg_params> ... )=0;
用括号括起来的是可选的参数，但是train_data的意思是一行是一个样本的所有特征（好像这么一行都叫特征向量。。。），行数是样本的数目。responses是响应值的矩阵，应该是一个n x 1的矩阵。
而在我们的例子里这两个参数都跑到CvDTreeTrainData* _train_data这里面去了。
1.
featureEvaluator->init( (CvFeatureParams*)featureParams, numPos + numNeg, cascadeParams.winSize );
在CvCascadeBoost初始化_featureEvaluator的时候就已经根据选择的特征类型，正样本的大小把所有样本的积分图空间申请了。还有就是在这个初始化的时候也把对应的responses申请了。
2.
bool CvCascadeClassifier::updateTrainingSet( double& acceptanceRatio)//featureEvaluator->setImage( img, isPositive ? 1 : 0, i );
在这把积分图和response都计算出来。
3. TrainCascade\boost.cpp
data = new CvCascadeBoostTrainData( _featureEvaluator, _numSamples,
_precalcValBufSize, _precalcIdxBufSize, _params );
在这里计算所有的样本的特征值。这样上面1.2.步骤中的featureEvaluator的信息也都在data中了，所以data直接送到CvBoostTree::train的接口中去了：
CvCascadeBoostTree* tree = new CvCascadeBoostTree;
if( !tree->train( data, subsample_mask, this ) )//应该是训练一个弱分类器tree
{
delete tree;
break;
}
cvSeqPush( weak, &tree );//把弱分类器添加到强分类器里面
所以我们要封装train的时候需要把从父类vfr_machine_learning_package的接口的_train_data和response给处理下成CvDTreeTrainData的data，然后才能调用if( !tree->train( data, subsample_mask, this ) )