人臉和性別識別（基於OpenCV）

描寫敘述

人臉識別包含四個步驟

1. 人臉檢測：定位人臉區域，僅僅關心是不是臉；
2. 人臉預處理：對人臉檢測出來的圖片進行調整優化。
3. 收集和學習人臉：收集要識別的人的預處理過的人臉，然后通過一些算法去學習怎樣識別；
4. 人臉識別：識別當前人臉與數據庫里的哪個人臉最類似。

人臉檢測

OpenCV集成了基於PCA LDA 和LBP的人臉檢測器。源文件自帶非常多各種訓練好的檢測器。下表是經常使用的XML文件

上面的XML文件能夠檢測正面人臉、眼睛或鼻子。檢測人臉我採用的是第一個或第二個Harr人臉檢測器。

識別率比較好。

第一步：載入Harr人臉檢測XML文件

try{
        faceCascade.load(faceCascadeFilename);
    }catch(cv::Exception& e){}

        if ( faceCascade.empty() ) {
        cerr << "ERROR: Could not load Face Detection cascade classifier [" << faceCascadeFilename << "]!" << endl;
        cerr << "Copy the file from your OpenCV data folder (eg: 'C:\\OpenCV\\data\\haarcascade_frontalface_alt2') into this WebcamFaceRec folder." << endl;
        exit(1);
    }
    cout << "Loaded the Face Detection cascade classifier [" << faceCascadeFilename << "]." << endl;

第二步：載入攝像頭，從視頻獲取圖像幀。

try{
        videoCapture.open(CameraID);
    }catch(cv::Exception& e){}

    if(!videoCapture.isOpened()){
        cerr << "ERROR: could not open Camera!" << endl;
        exit(1);
    }

videoCapture >> cameraFrame;

第三步：一幀圖像預處理

1、 灰度轉換：使用cvtColor()函數，將彩色圖像轉換為灰度圖像。台式機是3通道的BGR。移動設備則是4通道的BGRA格式

if(srcimg.channels() ==3 ){ cvtColor(srcimg,gray_img,CV_BGR2GRAY); }
    else if(srcimg.channels() ==4 ){ cvtColor(srcimg,gray_img,CV_BGRA2GRAY); }
    else { gray_img = srcimg; }

2、直方圖均衡化，在OpenCV函數中利用equalizeHist()函數運行直方圖均衡化，提升對照度和亮度。

equalizeHist(gray_img,equalized_Img);

第四步：檢測人臉

上面已經創建了級聯分類器並載入好XML文件。接着使用函數Classifier::detecMultiScale()函數來檢測人臉。這個函數的參數說明：
a、minFeatureSize: 該參數決定最小的人臉大小。通常能夠設為20*20或30*30像素。假設使用攝像機或移動設備檢測，則人臉一般非常接近攝像機，可把參數調大。80*80；
b、searchScaleFactor: 該參數決定有多少不同大小的人臉要搜索，通常設為1.1
c、minNeighbors： 該參數決定檢測器怎樣確定人臉已經被檢測到。通常設為3
d、flags: 該參數設定是否要查找全部的人臉或最大的人臉
（CASCADE_FIND_BIGGEST_OBJECT）

int flags = CASCADE_FIND_BIGGEST_OBJECT;
    //smallest object Size
    Size minFeatureSize = Size(20,20);
    // How detailed should the search be. Must be larger than 1.0.
    float searchScaleFactor = 1.1f;
    // How much the detections should be filtered out. This should depend on how bad false detections are to your system.
    // minNeighbors=2 means lots of good+bad detections, and minNeighbors=6 means only good detections are given but some are missed.
    int minNeighbors = 6;
    vector<Rect> faces;
    faceCascade.detectMultiScale(dectImg,faces,searchScaleFactor,
                                    minNeighbors,flags,minFeatureSize);
    //faceCascade.detectMultiScale(equalized_Img, faces);
    int i = 0;
    for(i = 0; i < faces.size(); i++){
        Rect face_id = faces[i];
        rectangle(orginalimg,face_id,Scalar(0,255,0),1);
    }

人臉識別

為了識別人臉。須要收集足夠多的要識別的人的人臉圖像。

收集好之后，選擇適合人臉識別的機器學習算法。通過算法來學習收集的數據。從而訓練出一個模型並保存。下次進來一幀圖像，通過算法對模型里的參數進行匹配識別。人臉識別機器學習算法有非常多，如SVM（支持向量機），ANN（人工神經網絡）還有最經常使用的是基於特征臉的算法。OpenCV提供了CV::Algorithm類，類中有基於特征臉的（PCA 主成分分析）、Fisher臉（LDA 線性判別分析）和LPBH（局部二值模式直方圖）

使用里面的算法，第一步必須通過cv::Algorithm::creat< FaceRecognizer>創建一個FaceRecognizer對象。創建了FaceRecognizer對象之后。將收集的人臉數據和標簽傳遞給FaceRecognizer::train() 函數就可以進行訓練模型。

string facerecAlgorithm = "FaceRecognizer.Fisherfaces";
Ptr<FaceRecognizer> model;
// Use OpenCV's new FaceRecognizer in the "contrib" module:
model = Algorithm::create<FaceRecognizer>(facerecAlgorithm);
if (model.empty()) {
cerr << "ERROR: The FaceRecognizer [" << facerecAlgorithm;
 cerr << "] is not available in your version of OpenCV. ";
 cerr << "Please update to OpenCV v2.4.1 or newer." << endl;
 exit(1);
}

model->train(preprocessedFaces, faceLabels);

訓練好模型之后。通常是把模型保存下來，以免下次反復訓練。

直接載入模型就可以。下一步就是人臉識別。相同，opencv把識別算法集成在FaceRecognizer類中。簡單地調用FaceRecognizer::predict() 就能夠識別。

int identity = model->predict(preprocessedFace);

測試程序

/* * Copyright (c) 2011. Philipp Wagner <bytefish[at]gmx[dot]de>. * Released to public domain under terms of the BSD Simplified license. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * Neither the name of the organization nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * See <http://www.opensource.org/licenses/bsd-license> */

#include "opencv2/core/core.hpp"
#include "opencv2/contrib/contrib.hpp"
#include "opencv2/highgui/highgui.hpp"

#include <iostream>
#include <fstream>
#include <sstream>
#include <direct.h>

using namespace cv;
using namespace std;

//const char *faceCascadeFilename = "C:\\opencv\\sources\\data\\lbpcascades\\lbpcascade_frontalface.xml"; 
const char *faceCascadeFilename = "C:\\opencv\\sources\\data\\haarcascades\\haarcascade_frontalface_alt2.xml"; 
const char *eyeCascadeFilename1 = "C:\\opencv\\sources\\data\\haarcascades\\haarcascade_eye.xml";               // Basic eye detector for open eyes only.
const char *eyeCascadeFilename2 = "C:\\opencv\\sources\\data\\haarcascades\\haarcascade_eye_tree_eyeglasses.xml";
const char *face_lib            = "face_train_img//";

const int DESIRED_CAMERA_WIDTH = 640;
const int DESIRED_CAMERA_HEIGHT = 480;
const int Width = 92;
const int Height = 112;
int gender_width;
int gender_height;
int im_width;
int im_height;

string g_listname_t[]= 
{
    "Jack",
    "William",
    "huang",
    "Barton"
};


static Mat norm_0_255(InputArray _src) {
    Mat src = _src.getMat();
    // Create and return normalized image:
    Mat dst;
    switch(src.channels()) {
    case 1:
        cv::normalize(_src, dst, 0, 255, NORM_MINMAX, CV_8UC1);
        break;
    case 3:
        cv::normalize(_src, dst, 0, 255, NORM_MINMAX, CV_8UC3);
        break;
    default:
        src.copyTo(dst);
        break;
    }
    return dst;
}

static void read_csv(const string& filename, vector<Mat>& images, vector<int>& labels, char separator = ';') {
    std::ifstream file(filename.c_str(), ifstream::in);
    if (!file) {
        string error_message = "No valid input file was given, please check the given filename.";
        CV_Error(CV_StsBadArg, error_message);
    }
    string line, path, classlabel;
    while (getline(file, line)) {
        stringstream liness(line);
        getline(liness, path, separator);
        getline(liness, classlabel);
        if(!path.empty() && !classlabel.empty()) {
            images.push_back(imread(path, 0));
            labels.push_back(atoi(classlabel.c_str()));
        }
    }
}

static void InitVideoCapture(VideoCapture &videoCapture, int CameraID)
{
    try{
        videoCapture.open(CameraID);
    }catch(cv::Exception& e){}

    if(!videoCapture.isOpened()){
        cerr << "ERROR: could not open Camera!" << endl;
        exit(1);
    }

    videoCapture.set(CV_CAP_PROP_FRAME_WIDTH, DESIRED_CAMERA_WIDTH);
    videoCapture.set(CV_CAP_PROP_FRAME_HEIGHT, DESIRED_CAMERA_HEIGHT);

    cout << "CameraID is :" << CameraID << endl;


}

static void InitDetectors(CascadeClassifier &faceCascade, CascadeClassifier &eyeCascade1, CascadeClassifier &eyeCascade2)
{
    try{
        faceCascade.load(faceCascadeFilename);
    }catch(cv::Exception& e){}

        if ( faceCascade.empty() ) {
        cerr << "ERROR: Could not load Face Detection cascade classifier [" << faceCascadeFilename << "]!" << endl;
        cerr << "Copy the file from your OpenCV data folder (eg: 'C:\\OpenCV\\data\\haarcascade_frontalface_alt2') into this WebcamFaceRec folder." << endl;
        exit(1);
    }
    cout << "Loaded the Face Detection cascade classifier [" << faceCascadeFilename << "]." << endl;

    // Load the Eye Detection cascade classifier xml file.
    try {   // Surround the OpenCV call by a try/catch block so we can give a useful error message!
        eyeCascade1.load(eyeCascadeFilename1);
    } catch (cv::Exception& e) {}
    if ( eyeCascade1.empty() ) {
        cerr << "ERROR: Could not load 1st Eye Detection cascade classifier [" << eyeCascadeFilename1 << "]!" << endl;
        cerr << "Copy the file from your OpenCV data folder (eg: 'C:\\OpenCV\\data\\haarcascades') into this WebcamFaceRec folder." << endl;
        exit(1);
    }
    cout << "Loaded the 1st Eye Detection cascade classifier [" << eyeCascadeFilename1 << "]." << endl;

    // Load the Eye Detection cascade classifier xml file.
    try {   // Surround the OpenCV call by a try/catch block so we can give a useful error message!
        eyeCascade2.load(eyeCascadeFilename2);
    } catch (cv::Exception& e) {}
    if ( eyeCascade2.empty() ) {
        cerr << "Could not load 2nd Eye Detection cascade classifier [" << eyeCascadeFilename2 << "]." << endl;
        // Dont exit if the 2nd eye detector did not load, because we have the 1st eye detector at least.
        //exit(1);
    }
    else
        cout << "Loaded the 2nd Eye Detection cascade classifier [" << eyeCascadeFilename2 << "]." << endl;


}

void readDataTraining(Ptr<FaceRecognizer> &model,vector<Mat> &images,vector<int> &labels,string &filePath )
{

    // These vectors hold the images and corresponding labels.

    // Read in the data. This can fail if no valid
    // input filename is given.
    try {
        read_csv(filePath, images, labels);
    } catch (cv::Exception& e) {
        cerr << "Error opening file \"" << filePath << "\". Reason: " << e.msg << endl;
        // nothing more we can do
        exit(1);
    }
    // Quit if there are not enough images for this demo.
    if(images.size() <= 1) {
        string error_message = "This demo needs at least 2 images to work. Please add more images to your data set!";
        CV_Error(CV_StsError, error_message);
    }


/* Mat testSample = images[images.size() - 1]; int testLabel = labels[labels.size() - 1]; images.pop_back(); labels.pop_back();*/


    model->train(images, labels);
    //int predictedLabel = model->predict(testSample);
    //
    // To get the confidence of a prediction call the model with:
    //
    // int predictedLabel = -1;
    // double confidence = 0.0;
    // model->predict(testSample, predictedLabel, confidence);
    //
    /*string result_message = format("Predicted class = %d / Actual class = %d.", predictedLabel, testLabel); cout << result_message << endl;*/
}


void preprocessing(Mat &srcimg, Mat &dstimg)
{
    Mat gray_img;
    if(srcimg.channels() ==3 ){
         cvtColor(srcimg,gray_img,CV_BGR2GRAY);
     }
    else if(srcimg.channels() ==4 ){
         cvtColor(srcimg,gray_img,CV_BGRA2GRAY);
     }
    else {
         gray_img = srcimg;
     }

     /*Mat equalized_Img; equalizeHist(gray_img,equalized_Img);*/
    dstimg = gray_img;
}


void faceDectRecog(CascadeClassifier &faceCascade,Ptr<FaceRecognizer> &model,Ptr<FaceRecognizer> &gender_model,Ptr<FaceRecognizer> &fishermodel,Mat &orginalimg,Mat &dectImg)
{
    static int num = 0;
    int predictedLabel = 0;
    int gender_predict = 0;
    // Only search for just 1 object (the biggest in the image).
    int flags = CASCADE_FIND_BIGGEST_OBJECT;
    //smallest object Size
    Size minFeatureSize = Size(20,20);
    // How detailed should the search be. Must be larger than 1.0.
    float searchScaleFactor = 1.1f;
    // How much the detections should be filtered out. This should depend on how bad false detections are to your system.
    // minNeighbors=2 means lots of good+bad detections, and minNeighbors=6 means only good detections are given but some are missed.
    int minNeighbors = 8;
    vector<Rect> faces;
    faceCascade.detectMultiScale(dectImg,faces,searchScaleFactor,
                                    minNeighbors,flags,minFeatureSize);
    //faceCascade.detectMultiScale(equalized_Img, faces);

/* pca + Lda Mat eigenvalues = gender_model->getMat("eigenvalues");//提取model中的特征值，該特征值默認由大到小排列 Mat W = gender_model->getMat("eigenvectors");//提取model中的特征向量，特征向量的排列方式與特征值排列順序一一相應 int xth = 121;//打算保留前121個特征向量，代碼中沒有體現原因，但選擇121是經過斟酌的，首先，在我的實驗中，"前121個特征值之和/全部特征值總和>0.97"；其次，121=11^2，能夠將結果表示成一個11*11的2維圖像方陣，交給fisherface去計算。 //vector<Mat> reduceDemensionimages;//降維后的圖像矩陣 Mat evs = Mat(W, Range::all(), Range(0, xth));//選擇前xth個特征向量，其余舍棄 Mat mean = gender_model->getMat("mean"); */


    int i = 0;
    for(i = 0; i < faces.size(); i++){
        Rect face_id = faces[i];

        Mat face = dectImg(face_id);

        Mat face_resized;
        Mat gender_resized;
        cv::resize(face, face_resized, Size(im_width, im_height), 1.0, 1.0, INTER_CUBIC);
        cv::resize(face, gender_resized, Size(gender_width, gender_height), 1.0, 1.0, INTER_CUBIC);


        rectangle(orginalimg,face_id,Scalar(0,255,0),1);

        predictedLabel = model->predict(face_resized);

        string result_message;
        /*result_message = format("Predicted = %d ", predictedLabel); cout << result_message << endl;*/
/* PCA +LDA Mat projection = subspaceProject(evs, mean, gender_resized.reshape(1,1));//做子空間投影 //reduceDemensionimages.push_back(projection.reshape(1,sqrt(xth*1.0))); Mat reduceDemensionimages = projection.reshape(1,sqrt(xth*1.0)); gender_predict = fishermodel->predict(reduceDemensionimages);*/

        string box_text;
        box_text = format( "Prediction = " );


        if ( predictedLabel >= 0 && predictedLabel <=3 )
        {
            box_text.append( g_listname_t[predictedLabel] );
        }
        else box_text.append( "Unknown" );

        gender_predict = gender_model->predict(face_resized);
        if(gender_predict == 0)
        {
            result_message = format("Predicted: female");
            box_text.append( " female" );
        }

        else if (gender_predict == 1)
        {
            result_message = format("Predicted: male");
            box_text.append( " male" );
        }

        else result_message = format("Predicted: Unknow");
        cout << result_message << endl;
        // Calculate the position for annotated text (make sure we don't
        // put illegal values in there):
        int pos_x = std::max(face_id.tl().x - 10, 0);
        int pos_y = std::max(face_id.tl().y - 10, 0);
        // And now put it into the image:
         putText(orginalimg, box_text, Point(pos_x, pos_y), FONT_HERSHEY_PLAIN, 1.0, CV_RGB(0,255,0), 2.0);

    }
}


void faceDect(CascadeClassifier &faceCascade,Mat &orginalimg,Mat &dectImg)
{

    // Only search for just 1 object (the biggest in the image).
    int flags = CASCADE_FIND_BIGGEST_OBJECT;
    //smallest object Size
    Size minFeatureSize = Size(20,20);
    // How detailed should the search be. Must be larger than 1.0.
    float searchScaleFactor = 1.1f;
    // How much the detections should be filtered out. This should depend on how bad false detections are to your system.
    // minNeighbors=2 means lots of good+bad detections, and minNeighbors=6 means only good detections are given but some are missed.
    int minNeighbors = 6;
    vector<Rect> faces;
    faceCascade.detectMultiScale(dectImg,faces,searchScaleFactor,
                                    minNeighbors,flags,minFeatureSize);
    //faceCascade.detectMultiScale(equalized_Img, faces);
    int i = 0;
    for(i = 0; i < faces.size(); i++){
        Rect face_id = faces[i];
        rectangle(orginalimg,face_id,Scalar(0,255,0),1);
    }
}

void CaptureFace(CascadeClassifier &faceCascade,Ptr<FaceRecognizer> &model,Mat &orginalimg,Mat &dectImg)
{
    static int num = 0;
    // Only search for just 1 object (the biggest in the image).
    int flags = CASCADE_FIND_BIGGEST_OBJECT;
    //smallest object Size
    Size minFeatureSize = Size(20,20);
    // How detailed should the search be. Must be larger than 1.0.
    float searchScaleFactor = 1.1f;
    // How much the detections should be filtered out. This should depend on how bad false detections are to your system.
    // minNeighbors=2 means lots of good+bad detections, and minNeighbors=6 means only good detections are given but some are missed.
    int minNeighbors = 6;
    vector<Rect> faces;
    faceCascade.detectMultiScale(dectImg,faces,searchScaleFactor,
                                    minNeighbors,flags,minFeatureSize);
    //faceCascade.detectMultiScale(equalized_Img, faces);
    int i = 0;
    for(i = 0; i < faces.size(); i++){
        Rect face_id = faces[i];

        Mat face = dectImg(face_id);

        Mat face_resized;
        cv::resize(face, face_resized, Size(im_width, im_height), 1.0, 1.0, INTER_CUBIC);

        char c[4];
        itoa(num,c,10);
        string s = face_lib + (string)c + ".png";
        imwrite(s,face_resized);
        cout << "Capture the" << num << "face" << endl;
        cout << s << ";" << face_id << endl;
        num ++;

        rectangle(orginalimg,face_id,Scalar(0,255,0),1);
    }
}



int main(int argc, const char *argv[]) 
{
    int mode;
    int i;

    // Get the path to your CSV.
    string fn_csv = string("at.txt");
    string gender_csv = string("gender.txt");
    string temp_csv = string("test.txt");
    CascadeClassifier faceCascade;
    CascadeClassifier eyeCascade1;
    CascadeClassifier eyeCascade2;
    VideoCapture videoCapture;
    Ptr<FaceRecognizer> model;
    Ptr<FaceRecognizer> temp_model;
    Ptr<FaceRecognizer> gender_model;
    int CameraID = 0;

    vector<Mat> images;
    vector<int> labels;

    vector<Mat> temp_images;
    vector<int> temp_labels;

    vector<Mat> gender_images;
    vector<int> gender_labels;
    cout << "Compiled with OpenCV version " << CV_VERSION << endl << endl;

    InitDetectors(faceCascade,eyeCascade1,eyeCascade2);

    InitVideoCapture(videoCapture,CameraID);

    printf("\n");
    printf("FaceDec and Recognition V0.1\n");
    printf("Usage: mode 0 : FaceDect; 1: train your own face; 2: Recognition \n");
    printf("please input mode\n");
    scanf("%d",&mode);

    //model = createEigenFaceRecognizer();
    temp_model = createEigenFaceRecognizer();
    model =createEigenFaceRecognizer();
    gender_model =createEigenFaceRecognizer();

    Ptr<FaceRecognizer> fishermodel = createFisherFaceRecognizer();  


    //gender_model = createEigenFaceRecognizer();

    if(mode == 3)
    {
        readDataTraining(model,images,labels,fn_csv);
        readDataTraining(gender_model,gender_images,gender_labels,gender_csv);
        gender_width = gender_images[0].cols;
        gender_height = gender_images[0].rows;
        im_width = images[0].cols;
        im_height = images[0].rows;

        model->save("Face_recog.yml");
        gender_model->save("gender_recog.yml");
    }
    //readDataTraining(temp_model,temp_images,gender_labels,temp_csv);






    model->load("Face_recog.yml");

    gender_width = Width;
    gender_height = Height;
    im_width = Width;
    im_height = Height;

    gender_model->load("eigenface_gender.yml");//保存訓練結果，供檢測時使用 
    fishermodel->load("fisher.yml");

    printf("gender_width :%d gender_height :%d im_width: %d im_height:%d\n",gender_width,gender_height,im_width,im_height);


    int num = 0;
    Mat cameraFrame;

    if(mode == 4)
    {
        read_csv(temp_csv, temp_images, temp_labels);
        for(i = 0; i < temp_images.size(); i ++)
        {
            Mat temp_img;
            preprocessing(temp_images[i],temp_img);
            CaptureFace(faceCascade,temp_model,temp_images[i],temp_img);
        }


    }
    for(;;){
        videoCapture >> cameraFrame;
        if( cameraFrame.empty()){
            cerr << "Error : could not grap next frame " << endl;
        }
        Mat processFrame = cameraFrame.clone();
        Mat preprocess_img;

        preprocessing(processFrame,preprocess_img);

        switch(mode){
        case 0:
            faceDect(faceCascade,processFrame,preprocess_img);
            break;
        case 1:
            CaptureFace(faceCascade,model,processFrame,preprocess_img);
        case 2:
            faceDectRecog(faceCascade,model,gender_model,fishermodel,processFrame,preprocess_img);
        default:
            break;
        }

        imshow("face_recognizer",processFrame);

        char key = (char) waitKey(300);
        if(key == 27)
            break;



    }
    return 0;
}