OpenCV人脸识别 C++

学习记录

记录人脸识别过程

环境

1. OpenCV 4.3.0
2. 树莓派4B(在Windows系统下也可以)

功能

默认状态下为人脸检测，可以进行人脸训练和识别

代码

 /******************************************************* 
  > File Name: main.cpp
  > Author: admin
  > Mail: [email protected]
  > Created Time: Fri 24 Apr 2020 09:46:00 CST
  > Modified Time:2020年07月01日 星期三 15时13分11秒
  > Note: No
 *******************************************************/

#include 
#include 
#include 
#include 
#include 
#include 

using namespace std;

// OpenCV includes
#include "opencv2/core.hpp"
#include "opencv2/imgproc.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/ml.hpp"
#include "opencv2/objdetect/objdetect.hpp"
#include "opencv2/face.hpp"

using namespace cv;
using namespace cv::ml;

cv::Ptr<cv::face::FaceRecognizer> recognizer;
CascadeClassifier faceCascade;
char g_moduleFlag = 0;
vector<string> g_ids;

Mat preProcessImage(Mat input);
Mat extractFace(Mat input, vector<Rect> *faces);
bool detection(void);
void init(void);
int videoProcess(void);
Mat testvideo(Mat img);
int trainOnce(Mat img);
int updateLabels(void);


//输出灰度图像，直方图均衡，长宽缩小0.5倍
Mat preProcessImage(Mat input)
{
	//cout<<"-->>function preprocessImage"<
    Mat result;
    
    resize(input, input, Size(), 0.5, 0.5, INTER_AREA);
    cvtColor(input, result, COLOR_BGR2GRAY);
    // Equalize the histogram
    equalizeHist(result, result);
    return result;
}

//仅支持单脸
//未检测到脸时返回纯黑
//返回灰度脸图，缩小了1.1倍
Mat extractFace(Mat input, vector<Rect> *faces)
{
	//cout<<"-->>function extractFace"<
    Mat result = preProcessImage(input);        
	
	Mat zero= Mat::zeros(result.rows, result.cols, CV_8UC1);
    
    faceCascade.detectMultiScale(result, *faces, 	\
								1.1, 2, 0|CV_HAL_CMP_GE, Size(100, 100), Size(200, 200) );
     
    if((*faces).size() == 0)
        return zero;
    
    Mat grayc = result.clone();
    grayc = grayc((*faces)[0]);//

    return grayc;
}

//检测人脸
Mat detection(Mat img)
{   
	//cout<<"-->>function detection"<
    vector<Rect> faces;

    Mat grayt = extractFace(img, &faces);

    Scalar he= sum(grayt);
    if(he[0] == 0)
	{
		resize(img, img, Size(), 0.5, 0.5, INTER_AREA);
        return img;
    }
	
	resize(img, img, Size(), 0.5, 0.5, INTER_AREA);
    rectangle(img, faces[0], Scalar(255, 255, 255), 2);

    return img;
}


int updateLabels(void)
{
	cout<<"-->>function updateLabels"<<endl;

	int inputlabel;
	
	cout<<"请输入标签号："<<endl;	
	cin>>inputlabel;
	cin.ignore();
	//g_ids.push_back(to_string(inputlabel));
	
	cout<<"-->>function updateLabels end"<<endl;
	return inputlabel;
}

int trainOnce(Mat img)
{
    cout<<"-->>function trainOnce"<<endl;
    
    std::vector<cv::Mat> referenceImages;
    std::vector<int> labels; 
    // vectors of reference image and their labels
       
    vector<Rect> faces;
    Mat frame;
	int inputlabel;
     
	frame = extractFace(img, &faces);
	Scalar he= sum(frame);
	if(he[0] == 0)
	{
		cout <<" No face"<<endl;
		return -1;
	}
	
	
	
	inputlabel = updateLabels();	
	referenceImages.push_back(frame);
	labels.push_back(inputlabel);
	
	if(g_moduleFlag == 0)//是否有模型
	{
		recognizer->train(referenceImages, labels);
		cout <<"g_moduleFlag == 0!"<<endl;
		g_moduleFlag = 1;
	}	
	else
		recognizer->update(referenceImages, labels);
	
    cout <<"trained!"<<endl;
	
    recognizer->write("../trainer.yml");
    
    return 0;
}

//摄像头测试
Mat testvideo(Mat img)
{   
	//cout<<"-->>function testvideo"<
    int predictedLabel = -1;
    double confidence = 0.0;
    
    vector<Rect> faces;
	Mat zero= Mat::zeros(img.rows, img.cols, CV_8UC1);
	
	if(g_moduleFlag == 0)
	{
		cout<<"未训练，无法测试"<<endl;
		return zero;
	}
	
    Mat grayt = extractFace(img, &faces);
	
    Scalar he= sum(grayt);
    if(he[0] == 0)
	{
		//cout<<"图片中无人脸"<
		resize(img, img, Size(), 0.5, 0.5, INTER_AREA);
        return img;
    }
	
	resize(img, img, Size(), 0.5, 0.5, INTER_AREA);
    rectangle(img, faces[0], Scalar(255, 255, 255), 2);
    
    // predict the label of this image
    recognizer->predict(grayt,     // face image 
                        predictedLabel, // predicted label of this image 
                        confidence);    // confidence of the prediction

    string name;

    if(confidence<80)
	{
        name = to_string(predictedLabel);             
    }
    else{
        name = g_ids[0];
        confidence = 0;
    }
    
    stringstream ss;
    ss << (int)confidence;
    putText(img, ss.str().c_str(), Point2d(faces[0].x, faces[0].y+30),	\
            FONT_HERSHEY_SIMPLEX, 1, Scalar(0,0,255), 2, LINE_AA);
    
    putText(img, name, Point2d(faces[0].x, faces[0].y+faces[0].height), 	\
            FONT_HERSHEY_SIMPLEX, 1, Scalar(0,0,0), 2, LINE_AA);
 
    return img;
}

int videoProcess(void)
{
    cout<<"-->>function videoProcess"<<endl;
    
    Mat camera; 
    vector<Rect> faces;
	char ch = 48;
	char sh = ch;
	int count = 0;
    
    VideoCapture cap(0);
        
    if( !cap.isOpened() )
    {   
        cout<< "No video!"<<endl; 
        return -1;   
    }   
    
	cout<<"请在图片中选择摄像头模式："<<endl;
	cout<<"	0 显示视频"<<endl;
	cout<<"	1 训练人脸"<<endl;
	cout<<"	2 持续测试人脸"<<endl;
	
	
    while(true)
    {
        cap >> camera;
		flip(camera,camera,1);
		
		if(ch == 48){//0	普通加载摄像头
			camera = detection(camera);
			//resize(camera, camera, Size(), 0.5, 0.5, INTER_AREA);
			imshow("Camera", camera);
			sh = ch;
			ch = waitKey(1);
		}
		else if( ch == 48+1){//1	训练人脸
			trainOnce(camera);
			resize(camera, camera, Size(), 0.5, 0.5, INTER_AREA);
			imshow("Camera", camera);
			ch = 48+2;
		}
		else if(ch == 48+2){//2		持续测试人脸
			camera = testvideo(camera);
			
			Scalar he= sum(camera);
			if(he[0] == 0)	//人脸未训练
				ch = 48;
			else{
				imshow("Camera", camera);
				sh = ch;
				ch = waitKey(1);
			}
		}
		else if(ch == 27)//esc
			break;
		else
			ch = sh;			
    }
    cap.release();
    
    return 0;
}

void init(void)
{
	cout<<"-->>function init"<<endl;
	string faceCascadeName = "../haar/haarcascade_frontalface_alt.xml";
    if( !faceCascade.load(faceCascadeName) )
    {
        cerr << "Error loading cascade file. Exiting!" << endl;
        exit(0);
    }
	
	g_ids.push_back("Unknown");
	recognizer = cv::face::LBPHFaceRecognizer::create(  1, 8, 8, 8, 200. );
}

int main(int argc, char* argv[])
{
	cout<<"-->>function main"<<endl;
	Mat camera;
	
	init();
	videoProcess();	
	
	destroyAllWindows();
	cout<<"End of program"<<endl;
    return 0;
}