using System;
|
using System.Collections.Generic;
|
using System.ComponentModel;
|
using System.Data;
|
using System.Drawing;
|
using System.Linq;
|
using System.Text;
|
using System.Threading.Tasks;
|
using System.Windows.Forms;
|
using Emgu.CV;
|
using Emgu.CV.UI;
|
using Emgu.Util;
|
using Emgu.CV.Structure;
|
using Emgu.CV.OCR;
|
using Emgu.CV.ML;
|
using Emgu.CV.Features2D;
|
using System.IO;
|
using System.Xml;
|
using System.Threading;
|
|
namespace CNNClassifier
|
{
|
public partial class Form1 : Form
|
{
|
public Form1()
|
{
|
InitializeComponent();
|
}
|
|
private void Form1_Load(object sender, EventArgs e)
|
{
|
Control.CheckForIllegalCrossThreadCalls = false;
|
|
DirectoryInfo di_P = new DirectoryInfo("samples/pos");
|
DirectoryInfo di_N = new DirectoryInfo("samples/neg");
|
|
PSP.Text = di_P.FullName.ToString();
|
NSP.Text = di_N.FullName.ToString();
|
}
|
|
private void button1_Click(object sender, EventArgs e)
|
{
|
FolderBrowserDialog fDialog = new FolderBrowserDialog();
|
fDialog.Description = "please select positve floder";
|
|
if (fDialog.ShowDialog() == DialogResult.OK)
|
{
|
PSP.Text = fDialog.SelectedPath;
|
}
|
}
|
|
private void button2_Click(object sender, EventArgs e)
|
{
|
FolderBrowserDialog fDialog = new FolderBrowserDialog();
|
fDialog.Description = "please Negative positve floder";
|
|
if (fDialog.ShowDialog() == DialogResult.OK)
|
{
|
NSP.Text = fDialog.SelectedPath;
|
}
|
}
|
|
private void button3_Click(object sender, EventArgs e)
|
{
|
ThreadStart ts = new ThreadStart(trainSamples);
|
Thread t = new Thread(ts);
|
t.Start();
|
}
|
|
|
Matrix<float> HogFeatureData = new Matrix<float>(967, 900);//500样本+467负样本，HOG纬度=(36-12)/6 + 1) *(36-12)/6 + 1) * 4 *9
|
Matrix<int> featureClasses = new Matrix<int>(967, 1);//样本类别，1为正样本，-1为负样本
|
Emgu.CV.ML.SVM svm;
|
private void trainSamples()
|
{
|
DirectoryInfo diPos = new DirectoryInfo(PSP.Text);
|
DirectoryInfo diNeg = new DirectoryInfo(NSP.Text);//正副样本地址
|
int posNum = diPos.GetFiles().Length;
|
int negNum = diNeg.GetFiles().Length;
|
FileInfo[] posFiles = diPos.GetFiles();
|
FileInfo[] negFiles = diNeg.GetFiles();
|
|
#region 获取样本，并计算HOG特征
|
|
HOGDescriptor hog = new HOGDescriptor(new Size(36, 36),
|
new Size(12, 12),
|
new Size(6, 6),
|
new Size(6, 6),
|
9);// 定义HOG描述子
|
|
#region 正样本
|
for (int fNum = 0; fNum < posNum; fNum++)
|
{
|
string filePath = diPos.FullName + "\\" + posFiles[fNum].Name;
|
|
Image<Gray, byte> im = new Emgu.CV.Image<Gray, byte>(filePath);
|
|
float[] fArr = hog.Compute(im);//计算HOG特征
|
|
for (int i = 0; i < fArr.Length; i++)
|
{
|
HogFeatureData[fNum, i] = fArr[i];
|
}
|
|
featureClasses.Data[fNum, 0] = 1;
|
|
#region 进度
|
PFN.Text = diPos.GetFiles()[fNum].Name;
|
float tempF = (float.Parse((fNum + 1).ToString()) * 100 / float.Parse(posNum.ToString()));
|
PPB.Value = (int)tempF;
|
#endregion
|
}
|
|
#endregion
|
|
#region 负样本
|
for (int fNum = 0; fNum < diNeg.GetFiles().Length; fNum++)
|
{
|
string filePath = diNeg.FullName + "\\" + negFiles[fNum].Name;
|
Image<Gray, byte> im = new Emgu.CV.Image<Gray, byte>(filePath);
|
|
float[] fArr = hog.Compute(im);
|
|
for (int i = 0; i < fArr.Length; i++)
|
{
|
HogFeatureData[fNum + posNum, i] = fArr[i];
|
}
|
|
featureClasses.Data[fNum + posNum, 0] = -1;
|
|
#region 进度
|
NFN.Text = diNeg.GetFiles()[fNum].Name;
|
float tempF = (float.Parse((fNum + 1).ToString()) * 100 / float.Parse(negNum.ToString()));
|
NPB.Value = (int)tempF;
|
#endregion
|
}
|
#endregion
|
|
#endregion
|
|
#region 训练并保存训练结果
|
|
svm = new Emgu.CV.ML.SVM();
|
svm.Type = SVM.SvmType.CSvc;
|
svm.SetKernel(SVM.SvmKernelType.Linear);//线性
|
svm.C =1;
|
svm.TermCriteria = new MCvTermCriteria(1000, 0.001);//1000次或者收敛达到0.001就跳出
|
|
|
svm.Train(HogFeatureData, Emgu.CV.ML.MlEnum.DataLayoutType.RowSample, featureClasses);
|
svm.Save("HogFeatures.xml");
|
|
#endregion
|
|
|
MessageBox.Show("训练完毕，训练数据保存在：HogFeatures.xml！");
|
}
|
|
private void HogSvmPredict()
|
{
|
#region HOG检测
|
|
#region 生成自定义detector向量
|
XmlDocument xml = new XmlDocument();
|
xml.Load(trainedFile.Text);
|
get_rho(xml.DocumentElement);
|
get_Alpha(xml.DocumentElement);
|
getSv_count(xml.DocumentElement);
|
getAlpha();
|
|
int supportVectors = svm.GetSupportVectors().Height;
|
int DescriptorDim = svm.GetSupportVectors().Width;
|
|
//int supportVectors = 1;
|
//int DescriptorDim = 900;
|
|
|
svmMat = new Matrix<float>(supportVectors, DescriptorDim);
|
get_supportVector(xml.DocumentElement);
|
|
|
Matrix<float> alphaMat = new Matrix<float>(1, supportVectors);
|
for (int i = 0; i < alphaArr.Length; i++)
|
{
|
alphaMat[0, i] = alphaArr[i];
|
}
|
|
|
Matrix<float> resultMat = new Matrix<float>(1, DescriptorDim);
|
|
resultMat = -1 * alphaMat * svmMat;
|
|
|
float[] mydetector = new float[DescriptorDim + 1];
|
for (int i = 0; i < DescriptorDim; i++)
|
{
|
mydetector[i] = resultMat[0, i];
|
}
|
mydetector[DescriptorDim] = rhoValue;
|
|
#endregion
|
|
Rectangle[] regions;
|
Mat image = new Mat("111.jpg", Emgu.CV.CvEnum.LoadImageType.Color);
|
|
HOGDescriptor hog = new HOGDescriptor(new Size(36, 36), new Size(12, 12), new Size(6, 6), new Size(6, 6), 9);// 定义HOG描述子
|
hog.SetSVMDetector(mydetector);
|
MCvObjectDetection[] results = hog.DetectMultiScale(image);
|
regions = new Rectangle[results.Length];
|
for (int i = 0; i < results.Length; i++)
|
{
|
regions[i] = results[i].Rect;
|
}
|
|
//using (HOGDescriptor hog = new HOGDescriptor())
|
//{
|
// hog.SetSVMDetector(mydetector);
|
// //var b = HOGDescriptor.GetDefaultPeopleDetector();
|
// //des.SetSVMDetector(HOGDescriptor.GetDefaultPeopleDetector());
|
|
// MCvObjectDetection[] results = hog.DetectMultiScale(image);
|
// regions = new Rectangle[results.Length];
|
// for (int i = 0; i < results.Length; i++)
|
// {
|
// regions[i] = results[i].Rect;
|
// }
|
//}
|
|
using (Graphics g = Graphics.FromImage(image.Bitmap))
|
{
|
foreach (Rectangle rc in regions)
|
{
|
g.DrawRectangle(new Pen(Color.White, 2), rc);//给识别出的行人画矩形框
|
}
|
}
|
GC.Collect();
|
imageBox1.Image = image;
|
#endregion
|
}
|
|
#region 常用方法
|
int sv_count = 0;
|
string alphaValue = "";
|
float rhoValue = 0;
|
Matrix<float> svmMat;
|
#region 获取xml文件中的getSv_count
|
|
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);
|
}
|
}
|
}
|
#endregion
|
|
#region 获取xml文件中的alpha值 alphaArr
|
|
public void get_Alpha(XmlNode nodes)
|
{
|
if (nodes.HasChildNodes)
|
{
|
foreach (XmlNode node in nodes.ChildNodes)
|
{
|
if (nodes.Name == "alpha")
|
{
|
alphaValue = nodes.InnerText;
|
return;
|
}
|
get_Alpha(node);
|
}
|
}
|
}
|
|
public float[] alphaArr;
|
public void getAlpha()
|
{
|
byte[] array = Encoding.ASCII.GetBytes(alphaValue);
|
MemoryStream stream = new MemoryStream(array);
|
StreamReader sr = new StreamReader(stream);
|
alphaArr = new float[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 != "")
|
{
|
alphaArr[i] = float.Parse(ele);
|
i++;
|
}
|
}
|
if (i == sv_count)
|
break;
|
}
|
}
|
#endregion
|
|
#region 获取xml文件中的rho值 rhoValue
|
|
public void get_rho(XmlNode nodes)
|
{
|
if (nodes.HasChildNodes)
|
{
|
foreach (XmlNode node in nodes.ChildNodes)
|
{
|
if (nodes.Name == "rho")
|
{
|
rhoValue = float.Parse(nodes.InnerText);
|
return;
|
}
|
get_rho(node);
|
}
|
}
|
}
|
#endregion
|
|
#region 从xml中读取已训练的数据
|
|
public void get_supportVector(XmlNode nodes)
|
{
|
if (nodes.HasChildNodes)
|
{
|
foreach (XmlNode node in nodes.ChildNodes)
|
{
|
if (nodes.Name == "support_vectors")
|
{
|
for (int i = 0; i < node.ChildNodes.Count; i++)
|
{
|
string strValue = node.ChildNodes[i].InnerText;
|
string[] tempStrArr = strValue.Replace("\r\n", "")
|
.Replace(" ", " ").Replace(" ", " ").Replace(" ", " ")
|
.Replace(" ", " ").Replace(" ", " ").Replace(" ", " ")
|
.Replace(" ", " ").Split(' ');
|
int z = 0;
|
for (int j = 0; j < tempStrArr.Length; j++)
|
{
|
if (tempStrArr[j] != "")
|
{
|
svmMat[i, z] = float.Parse(tempStrArr[j]);
|
z++;
|
}
|
}
|
}
|
return;
|
}
|
get_supportVector(node);
|
}
|
}
|
}
|
#endregion
|
#endregion
|
|
private void button7_Click(object sender, EventArgs e)
|
{
|
OpenFileDialog fileDialog = new OpenFileDialog();
|
fileDialog.Title = "选择文件";
|
fileDialog.Filter = "XML文件 (*.xml)|*.xml";
|
fileDialog.FilterIndex = 1;
|
fileDialog.RestoreDirectory = true;
|
|
if (fileDialog.ShowDialog() == DialogResult.OK)
|
{
|
trainedFile.Text = fileDialog.FileName;
|
}
|
}
|
|
|
private void button8_Click(object sender, EventArgs e)
|
{
|
#region 读取已经训练好的XML
|
svm = new Emgu.CV.ML.SVM();
|
FileStorage fsr = new FileStorage("HogFeatures.xml", FileStorage.Mode.Read);
|
svm.Read(fsr.GetFirstTopLevelNode());
|
#endregion
|
|
#region 生成自定义detector向量 mydetector
|
XmlDocument xml = new XmlDocument();
|
xml.Load(trainedFile.Text);
|
get_rho(xml.DocumentElement);
|
get_Alpha(xml.DocumentElement);
|
getSv_count(xml.DocumentElement);
|
getAlpha();
|
|
int supportVectors = svm.GetSupportVectors().Height;
|
int DescriptorDim = svm.GetSupportVectors().Width;
|
|
svmMat = new Matrix<float>(supportVectors, DescriptorDim);
|
get_supportVector(xml.DocumentElement);
|
|
|
Matrix<float> alphaMat = new Matrix<float>(1, supportVectors);
|
for (int i = 0; i < alphaArr.Length; i++)
|
{
|
alphaMat[0, i] = alphaArr[i];
|
}
|
|
|
Matrix<float> resultMat = new Matrix<float>(1, DescriptorDim);
|
|
resultMat = -1 * alphaMat * svmMat;
|
|
|
float[] mydetector = new float[DescriptorDim + 1];
|
for (int i = 0; i < DescriptorDim; i++)
|
{
|
mydetector[i] = resultMat[0, i];
|
}
|
mydetector[DescriptorDim] = rhoValue;
|
|
#endregion
|
|
|
myRect[] regions;
|
Mat image = new Mat("333.jpg", Emgu.CV.CvEnum.LoadImageType.Color);
|
|
using (HOGDescriptor hog = new HOGDescriptor(new Size(36, 36), new Size(12, 12), new Size(6, 6), new Size(6, 6), 9))
|
{
|
hog.SetSVMDetector(mydetector);
|
|
MCvObjectDetection[] results = hog.DetectMultiScale(image);
|
regions = new myRect[results.Length];
|
for (int i = 0; i < results.Length; i++)
|
{
|
regions[i] = new myRect();
|
regions[i].rct = results[i].Rect;
|
regions[i].score = results[i].Score;
|
}
|
}
|
|
using (Graphics g = Graphics.FromImage(image.Bitmap))
|
{
|
foreach (myRect rc in regions)
|
{
|
g.DrawRectangle(new Pen(Color.White, 2), rc.rct);//给识别出的行人画矩形框
|
#region 得到匹配姓名，并画出
|
|
Font font = new Font("宋体", 16, GraphicsUnit.Pixel);
|
SolidBrush fontLine = new SolidBrush(Color.Yellow);
|
float xPos = rc.rct.X;
|
float yPos = rc.rct.Y - 21;
|
g.DrawString(Math.Round(rc.score, 2).ToString(), font, fontLine, xPos, yPos);
|
#endregion
|
}
|
}
|
GC.Collect();
|
imageBox1.Image = image;
|
|
GC.Collect();
|
}
|
|
public class myRect
|
{
|
public Rectangle rct;
|
public float score;
|
}
|
}
|
}
|
