机器学习中SVM+HOG实现对饮料瓶的识别

首先是正样本的制作，这里我用opencv的拍照功能进行本地保存，接下来是拍照的代码。

void TakePhotoForPos()
{
	VideoCapture capture(0);
	Mat src;
	if (!capture.isOpened())
	{
		cout << "ERROR" << endl;
		system("pause");
	}
	int PressKeyTime = 0;
	while (true)
	{
		capture >> src;
		imshow("src", src);
		if (waitKey(30)==32)
		{
			ostringstream oss;
			oss << PressKeyTime <<endl;
			imwrite("D:\\traindatabase\\pos\\"+to_string(PressKeyTime)+".jpg", src);
			cout << oss.str() << endl;
			PressKeyTime++;
		}
		if (PressKeyTime>499)
		{
			break;
		}
	}
}

这里用空格键拍照，并保存到D:\traindatabase\pos目录下，一共500张。下面是声明。

#include<opencv.hpp>
#include<iostream>
#include<highgui/highgui.hpp>
#include<ml/ml.hpp>
#include<string>
#include<fstream>
#include<vector>

using namespace std;
using namespace cv;

利用cmd将图片名字写入txt文件中，这个与Linux命令相似，利用cd切换到当前目录，cd 文件目录。在输入 dir /s/b > path.txt。则将图片路径写入txt文件中。
当要进入其他盘时，可以win+R，打开运行窗口，再输入cmd，进去之后输入d：，这样就进入了D盘了，就可以对D盘中的文件进行读取了。
制作样本时将样本复制到word中制作标签。选中jpg换成jpg^p1，ctrl+h表示替换全局的这个值，加个小三角表示将转义字符，加p表示换行，在加入1表示后面追加1。这样就成功的添加了标签了。
路径文件为这种格式的：

D:\traindatabase\neg\1.jpg
0
D:\traindatabase\neg\10.jpg
0
D:\traindatabase\neg\100.jpg
0
D:\traindatabase\neg\1000.jpg
0
D:\traindatabase\neg\1001.jpg
0
D:\traindatabase\neg\1002.jpg
0
D:\traindatabase\pos\1003.jpg
1
D:\traindatabase\pos\1004.jpg
1

接下来我们通过使用HOG对样本进行分类，以及模型的训练。data_path保存的是样本及其标签的txt文件，save表示的是训练出的xml文件保存的路径。

void Train_SVMmodel(const string& data_path,const string& save_path)
{
	int ImgWidght = 64;
	int ImgHeight = 128;
	vector<string> img_path;
	vector<int> img_catg;
	int nLine = 0;
	string buf;
	ifstream svm_data(data_path);
	unsigned long n;
	while (svm_data)
	{
		if (getline(svm_data, buf))
		{
			nLine++;
			if (nLine % 2 == 0)
			{
				img_catg.push_back(atoi(buf.c_str()));	//atoi将字符串转化为整型，标志（0，1）
			}
			else
			{
				img_path.push_back(buf);	//图像路径
			}
		}
	}	//对文件进行解析
	svm_data.close();
	Mat data_mat, res_mat;
	int nImgNum = nLine / 2;
	data_mat = Mat::zeros(nImgNum, 3780, CV_32FC1);
	res_mat = Mat::zeros(nImgNum, 1, CV_32SC1);
	Mat src;
	Mat small;
	Mat trainImg = Mat(Size(ImgWidght, ImgHeight), 8, 3);
	for (string::size_type i = 0; i != img_path.size(); i++)
	{
		src = imread(img_path[i].c_str());
		if (src.empty())
		{
			cout << "can not load the image" << img_path[i] << endl;
			continue;
		}
		cout << "processing" << img_path[i].c_str() << endl;
		resize(src, small, Size(ImgWidght, ImgHeight), (0, 0), (0, 0), 1);
		//cvtColor(small, small, COLOR_RGB2GRAY);
		HOGDescriptor *hog = new HOGDescriptor(Size(ImgWidght, ImgHeight), Size(16, 16), Size(8, 8), Size(8, 8), 9);
		vector<float>descriptors;
		hog->compute(small, descriptors, Size(1, 1), Size(0, 0));
		cout << "HOG dimision is" << descriptors.size() << endl;
		n = 0;
		for (vector<float>::iterator iter = descriptors.begin(); iter != descriptors.end(); iter++)
		{
			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() << img_catg[i] << endl;
	}
	//cout << data_mat << endl;
	//cout << res_mat << endl;
	Ptr<ml::SVM>svm = ml::SVM::create();
	cout << "training" << endl;
	svm->setType(ml::SVM::C_SVC);
	svm->setKernel(ml::SVM::RBF);
	svm->setDegree(10.0);
	svm->setGamma(8.0);
	svm->setCoef0(1.0);
	svm->setC(10.0);
	svm->setNu(0.5);
	svm->setP(0);
	svm->setTermCriteria(TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 1000, 0.01));
	svm->train(data_mat, ml::ROW_SAMPLE, res_mat);
	cout << "End of training" << endl;
	svm->save(save_path);
}

因为这个样本是第一行为路径，第二行为标签，因此通过这样的方法可以将路径转化为数组，将对应的每个标签储存到数组中，实现对饮料瓶的二分类。接下来就是对SVM的参数的设置，这些参数的说明可以查看官方文档进行设置，最后将训练好的xml文件保存到当前文件夹，并通过此模型来判断物体是不是属于饮料瓶。
接下来是SVM的预测。filename是xml文件的路径。

	void Detect_Unit(const string& filename)
{
	//HOGDescriptor hog(Size(32, 64), Size(4, 4), Size(8, 8), Size(8, 8), 9);
	int DescriptorDim;
	Ptr<ml::SVM>svm = ml::SVM::load(filename);
	DescriptorDim = svm->getVarCount();
	Mat supportVector = svm->getSupportVectors();
	int supportVectorNum = supportVector.rows;
	cout << "支持向量个数为" << supportVectorNum << endl;
	vector<float>svm_alpha;
	vector<float>svm_svidx;
	float svm_rho;

	svm_rho = svm->getDecisionFunction(0, svm_alpha, svm_svidx);
	Mat alphaMat = Mat::zeros(1, supportVectorNum, CV_32FC1);
	Mat supportVectorMat = Mat::zeros(supportVectorNum, DescriptorDim, CV_32FC1);
	Mat resultMat = Mat::zeros(1, DescriptorDim, CV_32FC1);
	supportVectorMat = supportVector;
	//将alpha向量的数据复制到alphaMat中,返回SVM决策函数中的alpha向量
	for (int i = 0; i < supportVectorNum; i++)
	{
		alphaMat.at<float>(0, i) = svm_alpha[i];
	}
	//计算-（alphaMat*supportVectorMat），结果放在resultMat中。
	resultMat = -1 * alphaMat * supportVectorMat;
	vector<float>myDetector;
	//将resultMat中的数据复制到数组myDetector；
	for (int i = 0; i < DescriptorDim; 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_rho);
	HOGDescriptor myHOG;
	Size s1(16,16);
	Size s2(8, 8);
	myHOG.winSize = Size(64,128);
	myHOG.blockSize = s1;
	myHOG.blockStride = s2;
	myHOG.cellSize = s2;
	myHOG.nbins = 9;
	if (_access("D:\\svm\\svm\\HOGDetectorForOpenCv.txt",0)== -1)
	{
		ofstream fout("HOGDetectorForOpenCv.txt");
		for (unsigned int i = 0; i < myDetector.size(); i++)
		{
			fout << myDetector[i] << endl;
		}
	}
}

使用训练好的模型进行判断，可以预测结果。
下面对检测子进行保存并初始化：

/*加载检测子*/
/*如果有检测子的话跳过此步骤*/
void Load_Detect_Vector(const string& filename)
{
	/*当myDetector为空时加载*/
	ifstream finPos(filename);
	string buf;
	int nLine = 0;
	while (finPos)
	{
		if (getline(finPos, buf))
		{
			myDetector.push_back(atof(buf.c_str()));
			nLine++;
		}
	}
}

接下来是对结果的预测：

	void Detect_Final()
{
	/*VideoCapture detect(0);
	if (!detect.isOpened())
	{
		cout << "ERROR" << endl;
		system("pause");
	}*/
	//Mat frame;
	//Mat src;D:\traindatabase\pos\0.jpg
	Mat inter = imread("D:\\traindatabase\\pos\\4.jpg");
	//detect >> inter;
	if (inter.empty())
	{
		cout << "the picture is empty" << endl;
		system("pause");
	};
	//resize(inter, frame, Size(64, 128));
	//cvtColor(src, frame, COLOR_BGR2GRAY);
	vector<Rect>found, found_filtered;
	myHOG.detectMultiScale(inter, found, 0, Size(8, 8), Size(32, 32), 1.05, 2);
	//cout << "找到的矩形框个数为" << found.size() << endl;
	cout << found.size();
	//找出所有没有嵌套的矩形框，并放入found_filtered中，如果有嵌套，取最外面的一个
	for (unsigned int i = 0; i < found.size(); i++)
	{
		Rect r = found[i];
		unsigned int j = 0;
		for (; j < found.size(); j++)
			if (j != i && (r & found[j]) == r)
				break;
		if (j == found.size())
		{
			found_filtered.push_back(r);
		}
	}
	//画矩形框
	for (unsigned int i = 0; i < found_filtered.size(); i++)
	{
		Rect r = found_filtered[i];
		r.x += cvRound(r.width*0.1);
		r.width = cvRound(r.width*0.8);
		r.y += cvRound(r.height*0.07);
		r.height = cvRound(r.height*0.8);
		rectangle(inter, r.tl(), r.br(), Scalar(0, 255, 0), 3);
		putText(inter, "Drinks", Point2f(r.x + 5, r.y + 10), cv::FONT_HERSHEY_PLAIN, 0.4, Scalar(0, 255, 0), 1, 8, false);
	}
	imshow("output", inter);
	//resize(inter, inter, Size(4 * inter.cols, 4 * inter.rows));
	//imshow("frame", inter);
	//Mat output;
	//resize(src, output, Size(4 * src.cols, 4 * src.rows));
	//imshow("src", output);
	waitKey(0);
}

这就是SVM的简单的识别，这段代码仍然具有很多的错误，但是可以实现一些简单的识别。也存在着精度和速度上的问题，没有GPU的加速，而且HOG进行检测时需要算很多，特征向量有2000个，因此带来的计算问题还需要优化。特征向量矩阵和分类使用的xml文件如果有需要的可以私信我，因为文件比较大，在这就不方便发了。
接下来附上结果图片：

调用顺序为先训练xml文件再保存特征向量，再进行初始化并进行预测。
接下来是完整代码：

#include<opencv.hpp>
#include<iostream>
#include<highgui/highgui.hpp>
#include<ml/ml.hpp>
#include<string>
#include<fstream>
#include<vector>
#include<objdetect.hpp>
#include<stdlib.h>
#include<algorithm>
#include<io.h>

using namespace std;
using namespace cv;

vector<float>myDetector;	//得到自己的检测子
vector<string>FILENAME;	//正样本文件名称
HOGDescriptor myHOG;	//得到自己的HOG特征向量
string Forward = "D:\\traindatabase\\pos\\";
string Back = ".jpg";
int SumPicture = 500;	//正样本数量

/*训练自己的SVM模型，并以此得到检测子*/
void Train_SVMmodel(const string& data_path,const string& save_path)
{
	int ImgWidght = 64;
	int ImgHeight = 128;
	vector<string> img_path;
	vector<int> img_catg;
	int nLine = 0;
	string buf;
	ifstream svm_data(data_path);
	unsigned long n;
	while (svm_data)
	{
		if (getline(svm_data, buf))
		{
			nLine++;
			if (nLine % 2 == 0)
			{
				img_catg.push_back(atoi(buf.c_str()));	//atoi将字符串转化为整型，标志（0，1）
			}
			else
			{
				img_path.push_back(buf);	//图像路径
			}
		}
	}
	svm_data.close();
	Mat data_mat, res_mat;
	int nImgNum = nLine / 2;
	data_mat = Mat::zeros(nImgNum, 3780, CV_32FC1);
	res_mat = Mat::zeros(nImgNum, 1, CV_32SC1);
	Mat src;
	Mat small;
	Mat trainImg = Mat(Size(ImgWidght, ImgHeight), 8, 3);
	for (string::size_type i = 0; i != img_path.size(); i++)
	{
		src = imread(img_path[i].c_str());
		if (src.empty())
		{
			cout << "can not load the image" << img_path[i] << endl;
			continue;
		}
		cout << "processing" << img_path[i].c_str() << endl;
		resize(src, small, Size(ImgWidght, ImgHeight), (0, 0), (0, 0), 1);
		//cvtColor(small, small, COLOR_RGB2GRAY);
		HOGDescriptor *hog = new HOGDescriptor(Size(ImgWidght, ImgHeight), Size(16, 16), Size(8, 8), Size(8, 8), 9);
		vector<float>descriptors;
		hog->compute(small, descriptors, Size(1, 1), Size(0, 0));
		cout << "HOG dimision is" << descriptors.size() << endl;
		n = 0;
		for (vector<float>::iterator iter = descriptors.begin(); iter != descriptors.end(); iter++)
		{
			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() << img_catg[i] << endl;
	}
	//cout << data_mat << endl;
	//cout << res_mat << endl;
	Ptr<ml::SVM>svm = ml::SVM::create();
	cout << "training" << endl;
	svm->setType(ml::SVM::C_SVC);
	svm->setKernel(ml::SVM::LINEAR);
	svm->setDegree(10.0);
	svm->setGamma(8.0);
	svm->setCoef0(1.0);
	svm->setC(10.0);
	svm->setNu(0.5);
	svm->setP(0);
	svm->setTermCriteria(TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 3000, 1e-6));
	svm->train(data_mat, ml::ROW_SAMPLE, res_mat);
	cout << "End of training" << endl;
	svm->save(save_path);
}

/*得到自己的检测子*/
void Detect_Unit(const string& filename)
{
	//HOGDescriptor hog(Size(32, 64), Size(4, 4), Size(8, 8), Size(8, 8), 9);
	int DescriptorDim;
	Ptr<ml::SVM>svm = ml::SVM::load(filename);
	DescriptorDim = svm->getVarCount();
	Mat supportVector = svm->getSupportVectors();
	int supportVectorNum = supportVector.rows;
	cout << "支持向量个数为" << supportVectorNum << endl;
	vector<float>svm_alpha;
	vector<float>svm_svidx;
	float svm_rho;

	svm_rho = svm->getDecisionFunction(0, svm_alpha, svm_svidx);
	Mat alphaMat = Mat::zeros(1, supportVectorNum, CV_32FC1);
	Mat supportVectorMat = Mat::zeros(supportVectorNum, DescriptorDim, CV_32FC1);
	Mat resultMat = Mat::zeros(1, DescriptorDim, CV_32FC1);
	supportVectorMat = supportVector;
	//将alpha向量的数据复制到alphaMat中,返回SVM决策函数中的alpha向量
	for (int i = 0; i < supportVectorNum; i++)
	{
		alphaMat.at<float>(0, i) = svm_alpha[i];
	}
	//计算-（alphaMat*supportVectorMat），结果放在resultMat中。
	resultMat = -1 * alphaMat * supportVectorMat;
	vector<float>myDetector;
	//将resultMat中的数据复制到数组myDetector；
	for (int i = 0; i < DescriptorDim; 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_rho);
	HOGDescriptor myHOG;
	Size s1(16,16);
	Size s2(8, 8);
	myHOG.winSize = Size(64,128);
	myHOG.blockSize = s1;
	myHOG.blockStride = s2;
	myHOG.cellSize = s2;
	myHOG.nbins = 9;
	if (_access("D:\\svm\\svm\\HOGDetectorForOpenCv.txt",0)== -1)
	{
		ofstream fout("HOGDetectorForOpenCv.txt");
		for (unsigned int i = 0; i < myDetector.size(); i++)
		{
			fout << myDetector[i] << endl;
		}
	}
}

/*加载检测子*/
/*如果有检测子的话跳过此步骤*/
void Load_Detect_Vector()
{
	/*当myDetector为空时加载*/
	ifstream finPos("D;\\svm\\svm\\HOGDetectorForOpenCv.txt");
	string buf;
	int nLine = 0;
	while (finPos)
	{
		if (getline(finPos, buf))
		{
			myDetector.push_back(atof(buf.c_str()));
			nLine++;
		}
	}
}

/*最终的检测*/
void Detect_Final()
{
	/*VideoCapture detect(0);
	if (!detect.isOpened())
	{
		cout << "ERROR" << endl;
		system("pause");
	}*/
	//Mat frame;
	//Mat src;D:\traindatabase\pos\0.jpg
	HOGDescriptor myHOG(Size(64,128),Size(16,16),Size(8,8),Size(8,8),9);	//得到自己的HOG特征向量
	myHOG.setSVMDetector(myDetector);
	Mat inter = imread("D:\\traindatabase\\pos\\4.jpg");
	//detect >> inter;
	if (inter.empty())
	{
		cout << "the picture is empty" << endl;
		system("pause");
	};
	//resize(inter, frame, Size(64, 128));
	//cvtColor(src, frame, COLOR_BGR2GRAY);
	vector<Rect>found, found_filtered;
	myHOG.detectMultiScale(inter, found, 0, Size(8, 8), Size(32, 32), 1.05, 2);
	//cout << "找到的矩形框个数为" << found.size() << endl;
	cout << found.size();
	//找出所有没有嵌套的矩形框，并放入found_filtered中，如果有嵌套，取最外面的一个
	for (unsigned int i = 0; i < found.size(); i++)
	{
		Rect r = found[i];
		unsigned int j = 0;
		for (; j < found.size(); j++)
			if (j != i && (r & found[j]) == r)
				break;
		if (j == found.size())
		{
			found_filtered.push_back(r);
		}
	}
	//画矩形框
	for (unsigned int i = 0; i < found_filtered.size(); i++)
	{
		Rect r = found_filtered[i];
		r.x += cvRound(r.width*0.1);
		r.width = cvRound(r.width*0.8);
		r.y += cvRound(r.height*0.07);
		r.height = cvRound(r.height*0.8);
		rectangle(inter, r.tl(), r.br(), Scalar(0, 255, 0), 3);
		putText(inter, "Drinks", Point2f(r.x + 5, r.y + 10), cv::FONT_HERSHEY_PLAIN, 0.4, Scalar(0, 255, 0), 1, 8, false);
	}
	imshow("output", inter);
	//resize(inter, inter, Size(4 * inter.cols, 4 * inter.rows));
	//imshow("frame", inter);
	//Mat output;
	//resize(src, output, Size(4 * src.cols, 4 * src.rows));
	//imshow("src", output);
	waitKey(0);
}

/*利用opencv来对正样本进行拍照*/
void GenerateFileName()
{
	for (size_t i = 0; i < SumPicture; i++)
	{
		ostringstream filename;
		/*stringstream ss;
		string res;
		ss << i;
		ss >> res;*/
		filename << i << Back << endl;
		FILENAME.push_back(filename.str());
	}
}

/*利用opencv拍照得到正样本*/
void TakePhotoForPos()
{
	VideoCapture capture(0);
	Mat src;
	if (!capture.isOpened())
	{
		cout << "ERROR" << endl;
		system("pause");
	}
	int PressKeyTime = 0;
	while (true)
	{
		capture >> src;
		imshow("src", src);
		if (waitKey(30)==32)
		{
			ostringstream oss;
			oss << PressKeyTime <<endl;
			imwrite("D:\\traindatabase\\pos\\"+to_string(PressKeyTime)+".jpg", src);
			cout << oss.str() << endl;
			PressKeyTime++;
		}
		if (PressKeyTime>499)
		{
			break;
		}
	}
}

int main(int argc, char** argv)
{
	/*训练SVM+HOG*/
	//TakePhotoForPos();
	//Train_SVMmodel("D:\\traindatabase\\path.txt", "SVM_HOG.xml");
	//Detect_Unit("SVM_HOG.xml");
	Load_Detect_Vector();
	Detect_Final();
	return 0;
}

整体效果还是不太好，后续会继续补充。完整工程在我的GitHub上https://github.com/MEET123-123/Use-the-SVM-HOG-to-detect-the-buttle欢迎大家交流