http://blog.csdn.net/zhazhiqiang/article/details/18664417
http://blog.csdn.net/hujingshuang/article/details/47337707/
http://blog.csdn.net/orsinozhu/article/details/40554211
http://blog.csdn.net/qq_14845119/article/details/52187774
http://blog.csdn.net/leifeng_soul/article/details/52608575
http://blog.csdn.net/zhazhiqiang/article/details/20723425
http://blog.csdn.net/alvine008/article/details/9097105
http://blog.csdn.net/love_linney/article/details/25192909
#ifndef MY_HOG_SVM_H
#define MY_HOG_SVM_H
#include <QObject>
#include "opencv2/core/core.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/highgui/highgui.hpp"
#include "opencv2/calib3d/calib3d.hpp"
#include "opencv2/objdetect/objdetect.hpp"
#include "opencv/cv.h"
#include <QDebug>
#include <QTime>
#include <QDateTime>
#include <QTimer>
#include <QtCore/qmath.h>
#include "opencv/ml.h"
#include <iostream>
#include <fstream>
#include <string>
#include <vector>
using namespace cv;
using namespace std;
class Mysvm : public CvSVM
{
public:
//获得SVM的决策函数中的alpha数组
double * get_alpha_vector()
{
return this->decision_func->alpha;
}
//获得SVM的决策函数中的rho参数,即偏移量
float get_rho()
{
return this->decision_func->rho;
}
};
class Mysvm;
class My_Hog_Svm : public QObject
{
Q_OBJECT
public:
explicit My_Hog_Svm(QObject *parent = 0);
private:
const int m_iImgHeight = 96;
const int m_iImgWidht =96;
const int m_iBlockSizeWidth = 32;
const int m_iCellSizeWidth =16;
const int m_iStrideSizeWidth =16;
private:
void MyTrain();
void Detection();
void GetFeatureVector();
void DrawBox();
};
#endif // MY_HOG_SVM_H
#include "my_hog_svm.h"
//#include "mysvm.h"
My_Hog_Svm::My_Hog_Svm(QObject *parent) : QObject(parent)
{
// //1:训练
// this->MyTrain();
// //2:检测
// this->Detection();
// //3:获得特征向量
// this->GetFeatureVector();
//4:画框
this->DrawBox();
}
void My_Hog_Svm::MyTrain()
{
vector<string> img_path; //样本路径
vector<int> img_catg; //标记正负样本
int nLine = 0; //样本总共的个数
string buf;
ifstream svm_data_true( "./TRAIN_HEAD/Pos.txt" ); //正样本路径
ifstream svm_data_false( "./TRAIN_HEAD/Neg.txt" ); //负样本路径
unsigned long n;
//获取样本的路径
while(svm_data_true) //正样本
{
if( getline( svm_data_true, buf ) )
{
nLine ++;
img_catg.push_back(1);
img_path.push_back( buf );
}
}
while(svm_data_false) //负样本
{
if(getline(svm_data_false, buf))
{
nLine ++;
img_catg.push_back(0);
img_path.push_back( buf );
}
}
svm_data_true.close();//关闭文件
svm_data_false.close();
Mat data_mat; //存放特征值的矩阵
Mat res_mat; //存放正负样本的标识
int nImgNum = nLine; //读入样本数量
//类型矩阵,存储每个样本的类型标志
res_mat = Mat::zeros( nImgNum, 1, CV_32FC1);
Mat src;
Mat trainImg = Mat::zeros(m_iImgHeight, m_iImgWidht, CV_8UC3);//需要分析的图片
//获取每一个文件的特征值矩阵
for( string::size_type i = 0; i != img_path.size(); i++ )
{
src = imread(img_path[i].c_str(), 0);
resize(src, trainImg, cv::Size(m_iImgWidht,m_iImgHeight), 0, 0, INTER_CUBIC); //调整训练的图片
HOGDescriptor *hog=new HOGDescriptor(cvSize(m_iImgWidht,m_iImgHeight) //参数:窗口大小,块的大小,块滑动增量,cell的大小,每个bin的特征值
,cvSize(m_iBlockSizeWidth,m_iBlockSizeWidth),cvSize(m_iCellSizeWidth,m_iCellSizeWidth),cvSize(m_iStrideSizeWidth,m_iStrideSizeWidth), 9); //具体意思见参考文章1,2
vector<float>descriptors;//结果数组,特征值的个数
hog->compute(trainImg, descriptors, Size(m_iStrideSizeWidth,m_iStrideSizeWidth), Size(0,0)); //调用计算函数开始计算,滑动块增量
if (i==0)
{
//初始化存放所有图片特征值的容器
data_mat = Mat::zeros( nImgNum, descriptors.size(), CV_32FC1 ); //根据输入图片大小进行分配空间
}
n=0;
for(vector<float>::iterator iter=descriptors.begin();iter!=descriptors.end();iter++)
{
//为容器赋值
data_mat.at<float>(i,n) = *iter;
n++;
}
res_mat.at<float>(i, 0) = img_catg[i];
cout<<" end processing "<<img_path[i].c_str()<<" label:"<<img_catg[i]<<" HOG dims: "<<descriptors.size()<<endl;
}
//svm训练
Mysvm* svm = new Mysvm();
//训练SVM分类器
//迭代终止条件,当迭代满1000次或误差小于FLT_EPSILON时停止迭代
CvTermCriteria criteria = cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 1000, FLT_EPSILON);
//SVM参数:SVM类型为C_SVC;线性核函数;松弛因子C=0.01
CvSVMParams param(CvSVM::C_SVC, CvSVM::LINEAR, 0, 1, 0, 0.01, 0, 0, 0, criteria);
//☆☆☆☆☆☆☆☆☆(5) SVM学习 ☆☆☆☆☆☆☆☆☆☆☆☆
svm->train( data_mat, res_mat, Mat(), Mat(), param );
//☆☆利用训练数据和确定的学习参数,进行SVM学习☆☆☆☆
svm->save( "./TRAIN_HEAD/SVM_DATA.xml" );
qDebug() << "Finish!";
}
void My_Hog_Svm::Detection()
{
Mysvm* svm = new Mysvm();
svm->load("./TRAIN_HEAD/SVM_DATA.xml");
Mat trainImg = Mat::zeros(m_iImgHeight, m_iImgWidht, CV_8UC3);//需要分析的图片
string buf;
//检测样本
vector<string> img_tst_path;
ifstream img_tstNeg( "./TRAIN_HEAD/testNeg.txt" );
ifstream img_tstPos( "./TRAIN_HEAD/testPos.txt" );
while( img_tstNeg )
{
if( getline( img_tstNeg, buf ) )
{
img_tst_path.push_back( buf );
}
}
img_tstNeg.close();
while( img_tstPos )
{
if( getline( img_tstPos, buf ) )
{
img_tst_path.push_back( buf );
}
}
img_tstPos.close();
Mat test;
char line[512];
ofstream predict_txt( "./TRAIN_HEAD/SVM_PREDICT.txt" );
for( string::size_type j = 0; j != img_tst_path.size(); j++ )
{
test = imread( img_tst_path[j].c_str(), 0);//读入图像
resize(test, trainImg, cv::Size(m_iImgWidht,m_iImgHeight), 0, 0, INTER_CUBIC);//要搞成同样的大小才可以检测到
HOGDescriptor *hog=new HOGDescriptor(cvSize(m_iImgWidht,m_iImgHeight)
,cvSize(m_iBlockSizeWidth,m_iBlockSizeWidth),cvSize(m_iCellSizeWidth,m_iCellSizeWidth),cvSize(m_iStrideSizeWidth,m_iStrideSizeWidth), 9);
vector<float>descriptors;//结果数组
hog->compute(trainImg, descriptors,Size(m_iStrideSizeWidth,m_iStrideSizeWidth), Size(0,0)); //调用计算函数开始计算
cout<<"The Detection Result:"<<endl;
Mat SVMtrainMat = Mat::zeros(1,descriptors.size(),CV_32FC1);
int n=0;
for(vector<float>::iterator iter=descriptors.begin();iter!=descriptors.end();iter++)
{
SVMtrainMat.at<float>(0,n) = *iter;
n++;
}
int ret = svm->predict(SVMtrainMat);
std::sprintf( line, "%s %d\r\n", img_tst_path[j].c_str(), ret );
printf("%s %d\r\n", img_tst_path[j].c_str(), ret);
predict_txt<<line;
}
predict_txt.close();
cout << "Finish" <<endl;
return ;
}
/*************************************************************************************************
线性SVM训练完成后得到的XML文件里面,有一个数组,叫做support vector,还有一个数组,叫做alpha,有一个浮点数,叫做rho;
将alpha矩阵同support vector相乘,注意,alpha*supportVector,将得到一个列向量。之后,再该列向量的最后添加一个元素rho。
如此,变得到了一个分类器,利用该分类器,直接替换opencv中行人检测默认的那个分类器(cv::HOGDescriptor::setSVMDetector()),
就可以利用你的训练样本训练出来的分类器进行行人检测了。
***************************************************************************************************/
void My_Hog_Svm::GetFeatureVector()
{
Mysvm* svm = new Mysvm();
svm->load("./TRAIN_HEAD/SVM_DATA.xml"); //获取特征值的个数
int l_iFeatureNum = svm->get_var_count();//特征向量的维数,即HOG描述子的维数
int supportVectorNum = svm->get_support_vector_count();//支持向量的个数
qDebug()<<"支持向量个数:"<<supportVectorNum;
Mat alphaMat = Mat::zeros(1, supportVectorNum, CV_32FC1);//alpha向量,长度等于支持向量个数
Mat supportVectorMat = Mat::zeros(supportVectorNum, l_iFeatureNum, CV_32FC1);//支持向量矩阵
Mat resultMat = Mat::zeros(1, l_iFeatureNum, CV_32FC1);//alpha向量乘以支持向量矩阵的结果
//将支持向量的数据复制到supportVectorMat矩阵中
for(int i=0; i<supportVectorNum; i++)
{
const float * pSVData = svm->get_support_vector(i);//返回第i个支持向量的数据指针
for(int j=0; j<l_iFeatureNum; j++)
{
supportVectorMat.at<float>(i,j) = pSVData[j];
}
}
//将alpha向量的数据复制到alphaMat中
double * pAlphaData = svm->get_alpha_vector();//返回SVM的决策函数中的alpha向量
for(int i=0; i<supportVectorNum; i++)
{
alphaMat.at<float>(0,i) = pAlphaData[i];
}
//gemm(alphaMat, supportVectorMat, -1, 0, 1, resultMat);//不知道为什么加负号?
resultMat = -1 * alphaMat * supportVectorMat;
//得到最终的setSVMDetector(const vector<float>& detector)参数中可用的检测子
vector<float> myDetector;
//将resultMat中的数据复制到数组myDetector中
for(int i=0; i<l_iFeatureNum; i++)
{
myDetector.push_back(resultMat.at<float>(0,i));
}
//最后添加偏移量rho,得到检测子
// myDetector.push_back(svm->get_rho());
qDebug()<<"检测子维数:"<<myDetector.size() +1;
//保存检测子参数到文件
FILE* fp = fopen("./TRAIN_HEAD/hogSVMDetector-peopleFlow.txt","wb");
if( NULL == fp )
{
return ;
}
for(int i=0; i<myDetector.size(); i++)
{
fprintf(fp, "%f \n", myDetector[i]);
}
fprintf(fp, "%f", svm->get_rho());
fclose(fp);
qDebug() << "Finish!";
return;
}
void My_Hog_Svm::DrawBox()
{
vector<Rect> found;
Mat img = imread("./11.jpg");
vector<float> myDetector;
ifstream fileIn("./TRAIN_HEAD/hogSVMDetector-peopleFlow.txt", ios::in);
float val = 0.0f;
while(!fileIn.eof())
{
fileIn>>val;
myDetector.push_back(val);
}
fileIn.close();
HOGDescriptor defaultHog(cvSize(m_iImgWidht,m_iImgHeight) //参数:窗口大小,块的大小,块滑动增量,cell的大小,每个bin的特征值
,cvSize(m_iBlockSizeWidth,m_iBlockSizeWidth),cvSize(m_iCellSizeWidth,m_iCellSizeWidth),cvSize(m_iStrideSizeWidth,m_iStrideSizeWidth), 9);
defaultHog.setSVMDetector(myDetector);
//进行检测
defaultHog.detectMultiScale(img, found);
//画长方形,框出行人
for(int i = 0; i < found.size(); i++)
{
Rect r = found[i];
rectangle(img, r.tl(), r.br(), Scalar(255, 255, 255), 3);
}
namedWindow("检测行人", CV_WINDOW_AUTOSIZE);
imshow("检测行人", img);
waitKey(0);
// Mat l_pImageEle;
// namedWindow("Video");
// VideoCapture capture("./a.avi");
// while(1) //循环每一帧
// {
// static int l_iNum = 0;
// if(!capture.read(l_pImageEle))
// {
// return;
// }
// if(l_iNum%34 ==0)
// {
// //进行检测
// defaultHog.detectMultiScale(l_pImageEle, found);
// //画长方形,框出行人
// for(int i = 0; i < found.size(); i++)
// {
// Rect r = found[i];
// rectangle(l_pImageEle, r.tl(), r.br(), Scalar(255, 255, 255), 3);
// }
// }
// imshow("Video", l_pImageEle);
// cvWaitKey(34);
// }
}