Article Directory
foreword
This is the second article in this series. The main function of this article is to detect and recognize faces. The algorithm used in the face detection part is the cascade classifier, and the LBPH algorithm used in the face recognition part. The recognition accuracy can be compared to the face recognition system used in ordinary schools. Without further ado, let’s go to the code
1. Form design
Two, the code part
using Emgu.CV;
using Emgu.CV.CvEnum;
using Emgu.CV.Face;
using Emgu.CV.Structure;
using Emgu.CV.Util;
using System;
using System.Drawing;
using System.IO;
using System.Text;
using System.Windows.Forms;
namespace facerecognize
{
public partial class Form1 : Form
{
//在方法外面定义各个需要的变量和类，这样所有的方法都可调用相应的变量
Capture _capture;
//实例化人脸检测级联分类器
CascadeClassifier face_detect = new CascadeClassifier(@"haarcascade\haarcascade_frontalface_alt2.xml");//haarcascade_frontalface_alt2
Mat face_image = new Mat();
VectorOfInt image_lable = new VectorOfInt();
VectorOfMat image_data = new VectorOfMat();
string[] name;
//实例化局部二值模式人脸识别器
FaceRecognizer face_recognize = new LBPHFaceRecognizer();
bool after_train = false;
bool catch_face = false;
ASCIIEncoding ascii = new ASCIIEncoding();
public Form1()
{
InitializeComponent();
}
//定义摄像头捕捉到图片时调用的方法
void frame(object sender,EventArgs e)
{
Mat scr = _capture.QueryFrame();
//裁剪图片，提高电脑运行速度
CvInvoke.Resize(scr, scr, new Size(320, 240));
//检测人脸
Rectangle[] recs = face_detect.DetectMultiScale(scr);
foreach(Rectangle rec in recs)
{
CvInvoke.Rectangle(scr, rec, new MCvScalar(0, 0, 255));
//将人脸区域图片显示在imagebox2中，整张图片显示在imagebox1中
imageBox2.Image = new Mat(scr, rec);
imageBox1.Image = scr;
}
if(catch_face)
{
if(labeltextBox.Text!=null)
{
//这里输入的只能是英文字符，若是中文，则会出现乱码，笔者一开始以为是编码问题，后面尝试好几种解码方法，都没能解决这个问题。
string labeltext = labeltextBox.Text;
face_image = new Image<Gray, byte>(new Bitmap(imageBox2.Image.Bitmap)).Resize(120, 120, Inter.Area).Mat;
//将图片以image_data-文本框中输入的名字-系统时间.jpg格式保存，方便后面将文本框中名字取出
face_image.Save("Image_data" + "\\" + labeltext + "_" + DateTime.Now.Year + DateTime.Now.Month + DateTime.Now.Day + ".jpg");
MessageBox.Show("保存成功");
catch_face = false;
}
}
if(after_train)
{
foreach(Rectangle rec in recs)
{
//局部变量text存放人脸名字
string text = "";
Mat scr_image = new Mat(scr, rec);
//将人脸图片缩减，并且转为mat格式
scr_image = scr_image.ToImage<Gray, byte>().Resize(120, 120, Inter.Area).Mat;
//对人脸进行识别
FaceRecognizer.PredictionResult result = face_recognize.Predict(scr_image);
//name中存的是上面文本框存的各个人脸对应的名字，与label也是一一对应的
text = name[result.Label];
if(result.Distance>3000)
{
text = "Miss";
}
CvInvoke.PutText(scr, text, rec.Location, FontFace.HersheyComplex, 1, new MCvScalar(0, 0, 255));
imageBox1.Image = scr;
}
}
}
private void imageBox1_Click(object sender, EventArgs e)
{
}
//打开摄像头
private void opencap_Click(object sender, EventArgs e)
{
_capture = new Capture(0);
Application.Idle += frame;
}
//保存图片，并将catch_face重置为true
private void savebtn_Click(object sender, EventArgs e)
{
catch_face = true;
}
//开始训练图片
private void trainbtn_Click(object sender, EventArgs e)
{
//将iamge_data中获得所有文件存在path中，后面对其进行切片处理，并且将取得的人脸名字存在name中
string[] path = Directory.GetFiles("Image_data");
Mat[] data = new Mat[path.Length];
int[] label = new int[path.Length];
name = new string[path.Length];
for(int i =0;i<path.Length;i++)
{
Image<Gray, byte> image = new Image<Gray, byte>(path[i]);
data[i] = image.Mat;
label[i] = i;
//返回最后出现\\的索引
int a = path[i].LastIndexOf('\\');
//返回最后出现_的索引
int b = path[i].LastIndexOf('_');
//从a+1处开始截取字符，截取b-a-1个，这就是人脸名字在的区域
name[i] = path[i].Substring(a+1, b - a - 1);
}
image_data.Push(data);
image_lable.Push(label);
//训练图片
face_recognize.Train(image_data, image_lable);
face_recognize.Save("face\\data.txt");
MessageBox.Show("训练成功");
after_train = true;
}
}
}
3. Results
Four. Summary
The recognition accuracy is still very good, but the part worth improving is that you cannot enter Chinese in the text box, otherwise there will be garbled characters in the file name. The author has also changed several encoding methods, but still has not solved this problem.
I hope to make progress together with all the friends who love cv. Readers who know the solution can leave a message in the comment area.
Note: Please note the source for reprinting, originality is not easy.