mono_tum.cc系统构造函数——ORBSLAM2源码讲解（三）

---

前言

欢迎浏览我的SLAM专栏，一起加油淦穿SLAM！

---

一、mono_tum.cc*的源码及注释

本博客是以单目的形式来学习代码。
以下为ORB-SLAM2源码的Examples文件夹下的Monocular的mono_tum.cc的源码及注释：

/**
* This file is part of ORB-SLAM2.
*
* Copyright (C) 2014-2016 Raúl Mur-Artal  (University of Zaragoza)
* For more information see 
*
* ORB-SLAM2 is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* ORB-SLAM2 is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with ORB-SLAM2. If not, see .
*/


#include
#include
#include
#include

#include

#include
#include
using namespace std;


void LoadImages(const string &strFile, vector<string> &vstrImageFilenames,
                vector<double> &vTimestamps);

int main(int argc, char **argv)// 这个是主函数
{
    if(argc != 4)
    {
        cerr << endl << "Usage: ./mono_tum path_to_vocabulary path_to_settings path_to_sequence" << endl;
        return 1;
    }

    // Retrieve paths to images
    vector<string> vstrImageFilenames;
    vector<double> vTimestamps;
    string strFile = string(argv[3])+"/rgb.txt";
    LoadImages(strFile, vstrImageFilenames, vTimestamps);

    int nImages = vstrImageFilenames.size();

    // Create SLAM system. It initializes all system threads and gets ready to process frames.
    ORB_SLAM2::System SLAM(argv[1],argv[2],ORB_SLAM2::System::MONOCULAR,true);
// 此处重点讲解system函数，在这段代码下面第二章
    // Vector for tracking time statistics
    vector<float> vTimesTrack;
    vTimesTrack.resize(nImages);

    cout << endl << "-------" << endl;
    cout << "Start processing sequence ..." << endl;
    cout << "Images in the sequence: " << nImages << endl << endl;

    // Main loop
    cv::Mat im;
    for(int ni=0; ni<nImages; ni++)
    {
        // Read image from file
        im = cv::imread(string(argv[3])+"/"+vstrImageFilenames[ni],CV_LOAD_IMAGE_UNCHANGED);
        double tframe = vTimestamps[ni];

        if(im.empty())
        {
            cerr << endl << "Failed to load image at: "
                 << string(argv[3]) << "/" << vstrImageFilenames[ni] << endl;
            return 1;
        }

#ifdef COMPILEDWITHC11
        std::chrono::steady_clock::time_point t1 = std::chrono::steady_clock::now();
#else
        std::chrono::monotonic_clock::time_point t1 = std::chrono::monotonic_clock::now();
#endif

        // Pass the image to the SLAM system
        SLAM.TrackMonocular(im,tframe);

#ifdef COMPILEDWITHC11
        std::chrono::steady_clock::time_point t2 = std::chrono::steady_clock::now();
#else
        std::chrono::monotonic_clock::time_point t2 = std::chrono::monotonic_clock::now();
#endif

        double ttrack= std::chrono::duration_cast<std::chrono::duration<double> >(t2 - t1).count();

        vTimesTrack[ni]=ttrack;

        // Wait to load the next frame
        double T=0;
        if(ni<nImages-1)
            T = vTimestamps[ni+1]-tframe;
        else if(ni>0)
            T = tframe-vTimestamps[ni-1];

        if(ttrack<T)
            usleep((T-ttrack)*1e6);
    }

    // Stop all threads
    SLAM.Shutdown();

    // Tracking time statistics
    sort(vTimesTrack.begin(),vTimesTrack.end());
    float totaltime = 0;
    for(int ni=0; ni<nImages; ni++)
    {
        totaltime+=vTimesTrack[ni];
    }
    cout << "-------" << endl << endl;
    cout << "median tracking time: " << vTimesTrack[nImages/2] << endl;
    cout << "mean tracking time: " << totaltime/nImages << endl;

    // Save camera trajectory
    SLAM.SaveKeyFrameTrajectoryTUM("KeyFrameTrajectory.txt");

    return 0;
}

/**
 * @brief 导入图片
 * 
 * @param[in] strFile                   读入的文件名称
 * @param[in&out] vstrImageFilenames    彩色图片名称
 * @param[in&out] vTimestamps           记录时间戳
 */
void LoadImages(const string &strFile, vector<string> &vstrImageFilenames, vector<double> &vTimestamps)
{
    ifstream f;
    f.open(strFile.c_str());

    // skip first three lines
    // 前三行是注释，跳过
    string s0;
    getline(f,s0);
    getline(f,s0);
    getline(f,s0);

    while(!f.eof())
    {
        string s;
        getline(f,s);
        if(!s.empty())
        {
            stringstream ss;
            ss << s;
            double t;
            string sRGB;
            ss >> t;
            vTimestamps.push_back(t);
            ss >> sRGB;
            vstrImageFilenames.push_back(sRGB);
        }
    }
}

二、System函数

1.system.h

control+system进入include文件夹下的system.h文件
代码注释如下：

public:

    // Initialize the SLAM system. It launches the Local Mapping, Loop Closing and Viewer threads.
    //构造函数，用来初始化整个系统。
    System(const string &strVocFile,            //指定ORB字典文件的路径
           const string &strSettingsFile,       //指定配置文件的路径
           const eSensor sensor,                //指定所使用的传感器类型
           const bool bUseViewer = true);       //指定是否使用可视化界面 TODO 

2.system.cc

control+system进入src文件夹下的system.cc文件
代码注释如下：

//系统的构造函数，将会启动其他的线程
System::System(const string &strVocFile,					//词典文件路径
			   const string &strSettingsFile,				//配置文件路径
			   const eSensor sensor,						//传感器类型
               const bool bUseViewer):						//是否使用可视化界面
					 mSensor(sensor), 							//初始化传感器类型
					 mpViewer(static_cast<Viewer*>(NULL)),		//空。。。对象指针？  TODO 
					 mbReset(false),							//无复位标志
					 mbActivateLocalizationMode(false),			//没有这个模式转换标志
        			 mbDeactivateLocalizationMode(false)		//没有这个模式转换标志
{
    // Output welcome message
    cout << endl <<
    "ORB-SLAM2 Copyright (C) 2014-2016 Raul Mur-Artal, University of Zaragoza." << endl <<
    "This program comes with ABSOLUTELY NO WARRANTY;" << endl  <<
    "This is free software, and you are welcome to redistribute it" << endl <<
    "under certain conditions. See LICENSE.txt." << endl << endl;

    // 输出当前传感器类型
    cout << "Input sensor was set to: ";

    if(mSensor==MONOCULAR)
        cout << "Monocular" << endl;
    else if(mSensor==STEREO)
        cout << "Stereo" << endl;
    else if(mSensor==RGBD)
        cout << "RGB-D" << endl;

    //Check settings file
    cv::FileStorage fsSettings(strSettingsFile.c_str(), 	//将配置文件名转换成为字符串
    						   cv::FileStorage::READ);		//只读
    //如果打开失败，就输出调试信息
    if(!fsSettings.isOpened())
    {
       cerr << "Failed to open settings file at: " << strSettingsFile << endl;
       //然后退出
       exit(-1);
    }
        //Load ORB Vocabulary
    cout << endl << "Loading ORB Vocabulary. This could take a while..." << endl;

    //建立一个新的ORB字典
    mpVocabulary = new ORBVocabulary();
    //获取字典加载状态
    bool bVocLoad = mpVocabulary->loadFromTextFile(strVocFile);
    //如果加载失败，就输出调试信息
    if(!bVocLoad)
    {
        cerr << "Wrong path to vocabulary. " << endl;
        cerr << "Falied to open at: " << strVocFile << endl;
        //然后退出
        exit(-1);
    }
    //否则则说明加载成功
    cout << "Vocabulary loaded!" << endl << endl;

    //Create KeyFrame Database
    mpKeyFrameDatabase = new KeyFrameDatabase(*mpVocabulary);

    //Create the Map
    mpMap = new Map();

    //Create Drawers. These are used by the Viewer
    //这里的帧绘制器和地图绘制器将会被可视化的Viewer所使用
    mpFrameDrawer = new FrameDrawer(mpMap);
    mpMapDrawer = new MapDrawer(mpMap, strSettingsFile);

    //在本主进程中初始化追踪线程
    //Initialize the Tracking thread
    //(it will live in the main thread of execution, the one that called this constructor)
    mpTracker = new Tracking(this,						//现在还不是很明白为什么这里还需要一个this指针  TODO  
    						 mpVocabulary,				//字典
    						 mpFrameDrawer, 			//帧绘制器
    						 mpMapDrawer,				//地图绘制器
                             mpMap, 					//地图
                             mpKeyFrameDatabase, 		//关键帧地图
                             strSettingsFile, 			//设置文件路径
                             mSensor);					//传感器类型iomanip

下一章，重点讲上面这个Tracking函数，这是整个orb-slam2最重要的地方，包括ORBextractor函数提取特征点，下面第三章重点讲解。

---

三、Tracking

1.Tracking.cc

control+Tracking进入src文件夹下的Tracking.cc文件
代码注释如下：

///构造函数
Tracking::Tracking(
    System *pSys,                       //系统实例
    ORBVocabulary* pVoc,                //BOW字典
    FrameDrawer *pFrameDrawer,          //帧绘制器
    MapDrawer *pMapDrawer,              //地图点绘制器
    Map *pMap,                          //地图句柄
    KeyFrameDatabase* pKFDB,            //关键帧产生的词袋数据库
    const string &strSettingPath,       //配置文件路径
    const int sensor):                  //传感器类型
        mState(NO_IMAGES_YET),                              //当前系统还没有准备好
        mSensor(sensor),                                
        mbOnlyTracking(false),                              //处于SLAM模式
        mbVO(false),                                        //当处于纯跟踪模式的时候，这个变量表示了当前跟踪状态的好坏
        mpORBVocabulary(pVoc),          
        mpKeyFrameDB(pKFDB), 
        mpInitializer(static_cast<Initializer*>(NULL)),     //暂时给地图初始化器设置为空指针
        mpSystem(pSys), 
        mpViewer(NULL),                                     //注意可视化的查看器是可选的，因为ORB-SLAM2最后是被编译成为一个库，所以对方人拿过来用的时候也应该有权力说我不要可视化界面（何况可视化界面也要占用不少的CPU资源）
        mpFrameDrawer(pFrameDrawer),
        mpMapDrawer(pMapDrawer), 
        mpMap(pMap), 
        mnLastRelocFrameId(0)                               //恢复为0,没有进行这个过程的时候的默认值
{
    // Load camera parameters from settings file
    // Step 1 从配置文件中加载相机参数
    cv::FileStorage fSettings(strSettingPath, cv::FileStorage::READ);
    float fx = fSettings["Camera.fx"];
    float fy = fSettings["Camera.fy"];
    float cx = fSettings["Camera.cx"];
    float cy = fSettings["Camera.cy"];

    //     |fx  0   cx|
    // K = |0   fy  cy|
    //     |0   0   1 |
    //构造相机内参矩阵
    cv::Mat K = cv::Mat::eye(3,3,CV_32F);
    K.at<float>(0,0) = fx;
    K.at<float>(1,1) = fy;
    K.at<float>(0,2) = cx;
    K.at<float>(1,2) = cy;
    K.copyTo(mK);

    // 图像矫正系数
    // [k1 k2 p1 p2 k3]
    cv::Mat DistCoef(4,1,CV_32F);
    DistCoef.at<float>(0) = fSettings["Camera.k1"];
    DistCoef.at<float>(1) = fSettings["Camera.k2"];
    DistCoef.at<float>(2) = fSettings["Camera.p1"];
    DistCoef.at<float>(3) = fSettings["Camera.p2"];
    const float k3 = fSettings["Camera.k3"];
    //有些相机的畸变系数中会没有k3项
    if(k3!=0)
    {
        DistCoef.resize(5);
        DistCoef.at<float>(4) = k3;
    }
    DistCoef.copyTo(mDistCoef);

    // 双目摄像头baseline * fx 50
    mbf = fSettings["Camera.bf"];

    float fps = fSettings["Camera.fps"];
    if(fps==0)
        fps=30;

    // Max/Min Frames to insert keyframes and to check relocalisation
    mMinFrames = 0;
    mMaxFrames = fps;

    //输出
    cout << endl << "Camera Parameters: " << endl;
    cout << "- fx: " << fx << endl;
    cout << "- fy: " << fy << endl;
    cout << "- cx: " << cx << endl;
    cout << "- cy: " << cy << endl;
    cout << "- k1: " << DistCoef.at<float>(0) << endl;
    cout << "- k2: " << DistCoef.at<float>(1) << endl;
    if(DistCoef.rows==5)
        cout << "- k3: " << DistCoef.at<float>(4) << endl;
    cout << "- p1: " << DistCoef.at<float>(2) << endl;
    cout << "- p2: " << DistCoef.at<float>(3) << endl;
    cout << "- fps: " << fps << endl;

    // 1:RGB 0:BGR
    int nRGB = fSettings["Camera.RGB"];
    mbRGB = nRGB;

    if(mbRGB)
        cout << "- color order: RGB (ignored if grayscale)" << endl;
    else
        cout << "- color order: BGR (ignored if grayscale)" << endl;

    // Load ORB parameters

    // Step 2 加载ORB特征点有关的参数,并新建特征点提取器

    // 每一帧提取的特征点数 1000
    int nFeatures = fSettings["ORBextractor.nFeatures"];
    // 图像建立金字塔时的变化尺度 1.2
    float fScaleFactor = fSettings["ORBextractor.scaleFactor"];
    // 尺度金字塔的层数 8
    int nLevels = fSettings["ORBextractor.nLevels"];
    // 提取fast特征点的默认阈值 20
    int fIniThFAST = fSettings["ORBextractor.iniThFAST"];
    // 如果默认阈值提取不出足够fast特征点，则使用最小阈值 8
    int fMinThFAST = fSettings["ORBextractor.minThFAST"];

    // tracking过程都会用到mpORBextractorLeft作为特征点提取器，重点是这里提取ORB特征
    mpORBextractorLeft = new ORBextractor(
        nFeatures,      //参数的含义还是看上面的注释吧
        fScaleFactor,
        nLevels,
        fIniThFAST,
        fMinThFAST);

    // 如果是双目，tracking过程中还会用用到mpORBextractorRight作为右目特征点提取器
    if(sensor==System::STEREO)
        mpORBextractorRight = new ORBextractor(nFeatures,fScaleFactor,nLevels,fIniThFAST,fMinThFAST);

    // 在单目初始化的时候，会用mpIniORBextractor来作为特征点提取器
    if(sensor==System::MONOCULAR)
        mpIniORBextractor = new ORBextractor(2*nFeatures,fScaleFactor,nLevels,fIniThFAST,fMinThFAST);

    cout << endl  << "ORB Extractor Parameters: " << endl;
    cout << "- Number of Features: " << nFeatures << endl;
    cout << "- Scale Levels: " << nLevels << endl;
    cout << "- Scale Factor: " << fScaleFactor << endl;
    cout << "- Initial Fast Threshold: " << fIniThFAST << endl;
    cout << "- Minimum Fast Threshold: " << fMinThFAST << endl;

    if(sensor==System::STEREO || sensor==System::RGBD)
    {
        // 判断一个3D点远/近的阈值 mbf * 35 / fx
        //ThDepth其实就是表示基线长度的多少倍
        mThDepth = mbf*(float)fSettings["ThDepth"]/fx;
        cout << endl << "Depth Threshold (Close/Far Points): " << mThDepth << endl;
    }

    if(sensor==System::RGBD)
    {
        // 深度相机disparity转化为depth时的因子
        mDepthMapFactor = fSettings["DepthMapFactor"];
        if(fabs(mDepthMapFactor)<1e-5)
            mDepthMapFactor=1;
        else
            mDepthMapFactor = 1.0f/mDepthMapFactor;
    }

2.ORBextractor.cc

control+Tracking进入src文件夹下的ORBextractor.cc文件
代码注释如下：

//特征点提取器的构造函数

ORBextractor::ORBextractor(int _nfeatures,		//指定要提取的特征点数目
						   float _scaleFactor,	//指定图像金字塔的缩放系数
						   int _nlevels,		//指定图像金字塔的层数
						   int _iniThFAST,		//指定初始的FAST特征点提取参数，可以提取出最明显的角点
						   int _minThFAST):		//如果初始阈值没有检测到角点，降低到这个阈值提取出弱一点的角点
    nfeatures(_nfeatures), scaleFactor(_scaleFactor), nlevels(_nlevels),
    iniThFAST(_iniThFAST), minThFAST(_minThFAST)//设置这些参数
{
	//存储每层图像缩放系数的vector调整为符合图层数目的大小
    mvScaleFactor.resize(nlevels);  
	//存储这个sigma^2，其实就是每层图像相对初始图像缩放因子的平方
    mvLevelSigma2.resize(nlevels);
	//对于初始图像，这两个参数都是1
    mvScaleFactor[0]=1.0f;
    mvLevelSigma2[0]=1.0f;
	//然后逐层计算图像金字塔中图像相当于初始图像的缩放系数 
    for(int i=1; i<nlevels; i++)  
    {
		//其实就是这样累乘计算得出来的
        mvScaleFactor[i]=mvScaleFactor[i-1]*scaleFactor;
		//原来这里的sigma^2就是每层图像相对于初始图像缩放因子的平方
        mvLevelSigma2[i]=mvScaleFactor[i]*mvScaleFactor[i];
    }

    //接下来的两个向量保存上面的参数的倒数
    mvInvScaleFactor.resize(nlevels);
    mvInvLevelSigma2.resize(nlevels);
    for(int i=0; i<nlevels; i++)
    {
        mvInvScaleFactor[i]=1.0f/mvScaleFactor[i];
        mvInvLevelSigma2[i]=1.0f/mvLevelSigma2[i];
    }

    //调整图像金字塔vector以使得其符合设定的图像层数
    mvImagePyramid.resize(nlevels);

	//每层需要提取出来的特征点个数，这个向量也要根据图像金字塔设定的层数进行调整
    mnFeaturesPerLevel.resize(nlevels);
	
	//图片降采样缩放系数的倒数
    float factor = 1.0f / scaleFactor;
	//第0层图像应该分配的特征点数量
    float nDesiredFeaturesPerScale = nfeatures*(1 - factor)/(1 - (float)pow((double)factor, (double)nlevels));

	//用于在特征点个数分配的，特征点的累计计数清空
    int sumFeatures = 0;
	//开始逐层计算要分配的特征点个数，顶层图像除外（看循环后面）
    for( int level = 0; level < nlevels-1; level++ )
    {
		//分配 cvRound : 返回个参数最接近的整数值
        mnFeaturesPerLevel[level] = cvRound(nDesiredFeaturesPerScale);
		//累计
        sumFeatures += mnFeaturesPerLevel[level];
		//乘系数
        nDesiredFeaturesPerScale *= factor;
    }
    //由于前面的特征点个数取整操作，可能会导致剩余一些特征点个数没有被分配，所以这里就将这个余出来的特征点分配到最高的图层中
    mnFeaturesPerLevel[nlevels-1] = std::max(nfeatures - sumFeatures, 0);

	//成员变量pattern的长度，也就是点的个数，这里的512表示512个点（上面的数组中是存储的坐标所以是256*2*2）
    const int npoints = 512;
	//获取用于计算BRIEF描述子的随机采样点点集头指针
	//注意到pattern0数据类型为Points*,bit_pattern_31_是int[]型，所以这里需要进行强制类型转换
    const Point* pattern0 = (const Point*)bit_pattern_31_;	
	//使用std::back_inserter的目的是可以快覆盖掉这个容器pattern之前的数据
	//其实这里的操作就是，将在全局变量区域的、int格式的随机采样点以cv::point格式复制到当前类对象中的成员变量中
    std::copy(pattern0, pattern0 + npoints, std::back_inserter(pattern));

    //This is for orientation
	//下面的内容是和特征点的旋转计算有关的
    // pre-compute the end of a row in a circular patch
	//预先计算圆形patch中行的结束位置
	//+1中的1表示那个圆的中间行
    umax.resize(HALF_PATCH_SIZE + 1);
	
	//cvFloor返回不大于参数的最大整数值，cvCeil返回不小于参数的最小整数值，cvRound则是四舍五入
    int v,		//循环辅助变量
		v0,		//辅助变量
		vmax = cvFloor(HALF_PATCH_SIZE * sqrt(2.f) / 2 + 1);	//计算圆的最大行号，+1应该是把中间行也给考虑进去了
				//NOTICE 注意这里的最大行号指的是计算的时候的最大行号，此行的和圆的角点在45°圆心角的一边上，之所以这样选择
				//是因为圆周上的对称特性
				
	//这里的二分之根2就是对应那个45°圆心角
    
    int vmin = cvCeil(HALF_PATCH_SIZE * sqrt(2.f) / 2);
	//半径的平方
    const double hp2 = HALF_PATCH_SIZE*HALF_PATCH_SIZE;

	//利用圆的方程计算每行像素的u坐标边界（max）
    for (v = 0; v <= vmax; ++v)
        umax[v] = cvRound(sqrt(hp2 - v * v));		//结果都是大于0的结果，表示x坐标在这一行的边界

    // Make sure we are symmetric
	//这里其实是使用了对称的方式计算上四分之一的圆周上的umax，目的也是为了保持严格的对称（如果按照常规的想法做，由于cvRound就会很容易出现不对称的情况，
	//同时这些随机采样的特征点集也不能够满足旋转之后的采样不变性了）
	for (v = HALF_PATCH_SIZE, v0 = 0; v >= vmin; --v)
    {
        while (umax[v0] == umax[v0 + 1])
            ++v0;
        umax[v] = v0;
        ++v0;
    }
}