opencv3.2 计算HOG特征，并进行相似度比较

#include 
#include 
#include 
#include 
#include 
#include 
using namespace cv;
using namespace std;
using namespace Eigen;
int main() {
    Mat image1;
    image1 = imread("Image1.png",0);//图像大小640 * 480
    if (!image1.data) {
        printf("读取图像失败.\n");
        return -1;
    }
    cout << image1.size() << endl;
    
    /*
    //                                     滑动窗口大小  , block大小     ,block移动步长,   cell大小    ,bins个数
    HOGDescriptor *hog = new HOGDescriptor(cvSize(64, 48), cvSize(32, 32), cvSize(8, 8), cvSize(16, 16), 9);
    vector descriptors1;
    //           图像, 特征向量,     window移动步长, padding，用于填充图片以适应大小
    hog->compute(roi1, descriptors1, Size(2, 2), Size(0, 0));
        hog1.compute(image1, descriptors1);
        cout << "descriptors.size = " << descriptors1.size() << endl;//输出hog特征描绘子的维度
    */
    
    
    Mat roi1(image1, Rect(200,200,300,450));
    Mat roi2(image1, Rect(210, 210, 300, 450));
    Mat roi3(image1, Rect(0, 0, 300, 450));
    //                                     滑动窗口大小  , block大小     ,block移动步长,   cell大小    ,bins个数
    HOGDescriptor *hog = new HOGDescriptor(cvSize(64, 48), cvSize(16, 16), cvSize(8, 8), cvSize(16, 16), 9);
    vector descriptors1;
    vector descriptors2;
    vector descriptors3;
    
    hog->compute(roi1, descriptors1, Size(64, 28), Size(0, 0));
    
    hog->compute(roi2, descriptors2, Size(64, 28), Size(0, 0));
    hog->compute(roi3, descriptors3, Size(64, 28), Size(0, 0));
    
    VectorXf v1(descriptors1.size());
    VectorXf v2(descriptors1.size());
    VectorXf v3(descriptors1.size());
    for (int i = 0;i < descriptors1.size();i++) {
        v1[i] = descriptors1[i];
        v2[i] = descriptors2[i];
        v3[i] = descriptors3[i];
    }
    float norm1 = (v1-v2).transpose()*(v1-v2);
    float norm2 = (v2-v3).transpose()*(v2-v3);
    cout <