【EmguCV】C#实现HOG与SVM的几个问题

关于SVM中的alpha、rho向量

由于EmguCV封装的更加彻底，在C#中并不能跟C++一样通过重载获得这两个中间变量

//继承自CvSVM的类，因为生成setSVMDetector()中用到的检测子参数时，需要用到训练好的SVM的decision_func参数，  
//但通过查看CvSVM源码可知decision_func参数是protected类型变量，无法直接访问到，只能继承之后通过函数访问  
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;  
    }  
};

见C++实例：训练SVM分类器进行HOG行人检测 http://blog.csdn.net/pb09013037/article/details/41256945）

---

为了获取这两个变量用于自定义HOG检测子，暂时想到的几种办法：

1、C#读取生成的XML文件

分类器训练好后一般需要进行保存，方便直接预测

SVM svm = new SVM();
bool trained = svm.Train(my_train.sampleFeatureMat, my_train.sampleLabelMat, null, null, p);
svm.Save(@"../HOG_SVM.xml");

这里给出我的C#提取SVM参数方式：

（只用于提取训练目标为1与-1两类的XML文件，如果类型大于2，则有多个rho与alpha数组，需要进一步组合）

using System;
using System.Text;
using System.Xml;
using System.IO;


namespace HOG_SVM
{
    class GetData
    {
        public double[] alpha;
        public double rho;

        XmlDocument doc;        
        StreamReader sr;
        int sv_count;
        string alpha_str;

        public GetData()
        {
            doc = new XmlDocument();
            doc.Load(Form1.LOAD_PATH);
            XmlNode nodes = doc.DocumentElement;
            get_rho(nodes);
            getAlpha_str(nodes);
            getSv_count(nodes);
            getAlpha(); 
        }

        public void get_rho(XmlNode nodes)
        {
            if (nodes.HasChildNodes)
            {
                foreach (XmlNode node in nodes.ChildNodes)
                {
                    if (nodes.Name == "rho")
                    {
                        rho = Double.Parse(nodes.InnerText);
                        return;
                    }
                    get_rho(node);
                }
            }
        }

        public void getAlpha_str(XmlNode nodes)
        {
            if (nodes.HasChildNodes)
            {
                foreach (XmlNode node in nodes.ChildNodes)
                {
                    if (nodes.Name == "alpha")
                    {
                        //sr = new StreamReader(new Stream(nodes.InnerText));
                        alpha_str = nodes.InnerText;
                        return;
                    }
                    getAlpha_str(node);
                }
            }
        }

        public void getSv_count(XmlNode nodes)
        {
            if (nodes.HasChildNodes)
            {
                foreach (XmlNode node in nodes.ChildNodes)
                {
                    if (nodes.Name == "sv_count")
                    {
                        sv_count = int.Parse(nodes.InnerText);
                        return;
                    }
                    getSv_count(node);
                }
            }
        }
        
        public void getAlpha()
        {
            byte[] array = Encoding.ASCII.GetBytes(alpha_str);
            MemoryStream stream = new MemoryStream(array);             //convert stream 2 string      
            sr = new StreamReader(stream);
            alpha = new double[sv_count];
            sr.ReadLine();
            int i = 0;
            while (true)
            {

                string tmp = sr.ReadLine();
                if (tmp == "")
                    continue;

                string[] tmp2 = tmp.Split(' ');
                foreach (string ele in tmp2)
                {
                    if (ele != "")
                    {
                        alpha[i] = double.Parse(ele);
                        i++;
                    }
                }

                if (i == sv_count)
                    break;
            }
        }        
    }
}

c#读取XML的方式比较多，还可以利用Linq操作xml，另外也可以参考以下链接：

c# 读取opencv 生成的svm训练好的xml分类器：http://blog.csdn.net/yeyang911/article/details/12905153

---

2、使用其他C#的SVM库

关于提取参数，自定义HOG Detector的问题，后来在网上搜到了这种方式

Training custom SVM to use with HOGDescriptor in OpenCV：

I was struggling with the same problem. Searching forums I have found, that the detector cannot be trained using CvSVM (I don't know the reason). I used LIBSVM for training the the detector. Here is the code to extract the detector for HOGDescriptor.setSVMDetector( w): For data details see LIBSVM documentation/header. I did all the training in C++, filling the LIBSVM training data from CV to LIBSVM; the code below extracts the detector vector needed for cv::HOGDescriptor. The w parameter is std::vector<float> w   

const double * const *sv_coef = model.sv_coef;
const svm_node * const *SV = model.SV;
int l = model.l;
model.label;

const svm_node* p_tmp = SV[0];
int len = 0;
while( p_tmp->index != -1 )
{
    len++;
    p_tmp++;
}
w.resize( len+1 );

for( int i=0; i<l; i++)
{
    double svcoef = sv_coef[0][i];
    const svm_node* p = SV[i];
    while( p->index != -1 )
    {
        w[p->index-1] += float(svcoef * p->value);
        p++;
    }
}
w[len] = float(-model.rho[0]);

来自: http://stackoverflow.com/questions/15339657/training-custom-svm-to-use-with-hogdescriptor-in-opencv

---

该回答提到的 LIBSVM 库就是比较好的替代手段，应该可以直接获取到这两个中间量，而不用再去解析XML。

可以去作者主页上下载LIBSVM库：http://www.csie.ntu.edu.tw/~cjlin/libsvm/#csharp

---

3、其他相关链接

• 前些天的【OpenCV】基于HOG与SVM的行人检测学习（原理小结）:

    http://www.cnblogs.com/KC-Mei/p/4534009.html

• training GPU HOGDescriptor for multi scale detection：

    http://answers.opencv.org/question/4351/training-gpu-hogdescriptor-for-multi-scale-detection/