├── CMakeLists.txt
├── README.md
├── bin
    ├── 1.jpg
    ├── ORBextractor
    └── orb_features.jpg
├── include
    └── ORBextractor.h
├── lib
    └── liborbextractor.so
├── param
    └── orb.yaml
├── screenshots
    ├── 1.jpg
    ├── 2.jpg
    ├── liu1.jpg
    └── liu2.jpg
└── src
    ├── ORBextractor
    ├── ORBextractor.cpp
    └── main.cpp


/CMakeLists.txt:
--------------------------------------------------------------------------------
 1 | cmake_minimum_required(VERSION 2.8)
 2 | project(orbextractor)
 3 | 
 4 | IF(NOT CMAKE_BUILD_TYPE)
 5 |   SET(CMAKE_BUILD_TYPE Release)
 6 | ENDIF()
 7 | 
 8 | MESSAGE("Build type: " ${CMAKE_BUILD_TYPE})
 9 | 
10 | set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS}  -Wall  -O3 -march=native ")
11 | set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall   -O3 -march=native")
12 | 
13 | # Check C++11 or C++0x support
14 | include(CheckCXXCompilerFlag)
15 | CHECK_CXX_COMPILER_FLAG("-std=c++11" COMPILER_SUPPORTS_CXX11)
16 | CHECK_CXX_COMPILER_FLAG("-std=c++0x" COMPILER_SUPPORTS_CXX0X)
17 | if(COMPILER_SUPPORTS_CXX11)
18 |    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
19 |    add_definitions(-DCOMPILEDWITHC11)
20 |    message(STATUS "Using flag -std=c++11.")
21 | elseif(COMPILER_SUPPORTS_CXX0X)
22 |    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++0x")
23 |    add_definitions(-DCOMPILEDWITHC0X)
24 |    message(STATUS "Using flag -std=c++0x.")
25 | else()
26 |    message(FATAL_ERROR "The compiler ${CMAKE_CXX_COMPILER} has no C++11 support. Please use a different C++ compiler.")
27 | endif()
28 | 
29 | #LIST(APPEND CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake_modules)
30 | 
31 | find_package(OpenCV 3.0 QUIET)
32 | if(NOT OpenCV_FOUND)
33 |    find_package(OpenCV 2.4.3 QUIET)
34 |    if(NOT OpenCV_FOUND)
35 |       message(FATAL_ERROR "OpenCV > 2.4.3 not found.")
36 |    endif()
37 | endif()
38 | 
39 | 
40 | include_directories(
41 | ${PROJECT_SOURCE_DIR}
42 | ${PROJECT_SOURCE_DIR}/include
43 | )
44 | 
45 | set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${PROJECT_SOURCE_DIR}/lib)
46 | 
47 | add_library(${PROJECT_NAME} SHARED
48 | src/ORBextractor.cpp
49 | )
50 | 
51 | target_link_libraries(${PROJECT_NAME}
52 | ${OpenCV_LIBS}
53 | )
54 | 
55 | # Build examples
56 | 
57 | set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${PROJECT_SOURCE_DIR}/bin)
58 | 
59 | add_executable(ORBextractor
60 | src/main.cpp)
61 | target_link_libraries(ORBextractor ${PROJECT_NAME})
62 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # orb-extractor
 2 | 
 3 | **29 Nov 2019**
 4 | 
 5 | I separate orb-feature extractor code from [orbslam2 project](https://github.com/raulmur/ORB_SLAM2).
 6 | 
 7 | # 1. Prerequisites
 8 | I have tested the library in **Ubuntu 18.04**
 9 |  
10 | ## C++11 or C++0x Compiler
11 | 
12 | ## OpenCV
13 | I use [OpenCV](http://opencv.org) to manipulate images and features.My version is 3.4.8.
14 | 
15 | # 2. Building this library and examples(main.cpp)
16 | 
17 | Clone the repository:
18 | ```
19 | git clone https://github.com/slaming/ORBExtractor
20 | ```
21 | ```
22 | cd ORBExtractor
23 | ```
24 | ```
25 | mkdir build
26 | ```
27 | ```
28 | cd build
29 | ```
30 | ```
31 | cmake ..
32 | ```
33 | ```
34 | build -j4
35 | ```
36 | 
37 | # 3.Examples(main.cpp)
38 | 
39 | ```
40 | cd bin
41 | ```
42 | ```
43 | ./ORBextractor ../pictures/liu2.jpg ../param/orb.yaml
44 | ```
45 | 
46 | ## Provide two results
47 | ### First
48 | <a href="https://github.com/slaming/ORBExtractor/blob/master/screenshots" target="_blank"><img src="https://github.com/slaming/ORBExtractor/blob/master/screenshots/liu1.jpg" 
49 | alt="ORBExtractor1" width="240" height="180" border="10" /></a>
50 | <a href="https://github.com/slaming/ORBExtractor/blob/master/screenshots" target="_blank"><img src="https://github.com/slaming/ORBExtractor/blob/master/screenshots/1.jpg" 
51 | alt="ORBExtractor1" width="240" height="180" border="10" /></a>
52 | ### Second
53 | <a href="https://github.com/slaming/ORBExtractor/blob/master/screenshots" target="_blank"><img src="https://github.com/slaming/ORBExtractor/blob/master/screenshots/liu2.jpg" 
54 | alt="ORBExtractor2" width="240" height="180" border="10" /></a>
55 | <a href="https://github.com/slaming/ORBExtractor/blob/master/screenshots" target="_blank"><img src="https://github.com/slaming/ORBExtractor/blob/master/screenshots/2.jpg" 
56 | alt="ORBExtractor2" width="240" height="180" border="10" /></a>
57 | 


--------------------------------------------------------------------------------
/bin/1.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/bin/1.jpg


--------------------------------------------------------------------------------
/bin/ORBextractor:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/bin/ORBextractor


--------------------------------------------------------------------------------
/bin/orb_features.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/bin/orb_features.jpg


--------------------------------------------------------------------------------
/include/ORBextractor.h:
--------------------------------------------------------------------------------
 1 | #ifndef ORBEXTRACTOR_H
 2 | #define ORBEXTRACTOR_H
 3 | 
 4 | #include <vector>
 5 | #include <list>
 6 | #include <opencv/cv.h>
 7 | 
 8 | 
 9 | namespace ORB_SLAM2
10 | {
11 | 
12 | class ExtractorNode
13 | {
14 | public:
15 |     ExtractorNode():bNoMore(false){}
16 | 
17 |     void DivideNode(ExtractorNode &n1, ExtractorNode &n2, ExtractorNode &n3, ExtractorNode &n4);
18 | 
19 |     std::vector<cv::KeyPoint> vKeys;
20 |     cv::Point2i UL, UR, BL, BR;
21 |     std::list<ExtractorNode>::iterator lit;
22 |     bool bNoMore;
23 | };
24 | 
25 | class ORBextractor
26 | {
27 | public:
28 |     
29 |     enum {HARRIS_SCORE=0, FAST_SCORE=1 };
30 | 
31 |     ORBextractor(int nfeatures, float scaleFactor, int nlevels,
32 |                  int iniThFAST, int minThFAST);
33 | 
34 |     ~ORBextractor(){}
35 | 
36 |     // Compute the ORB features and descriptors on an image.
37 |     // ORB are dispersed on the image using an octree.
38 |     // Mask is ignored in the current implementation.
39 |     void operator()( cv::InputArray image, cv::InputArray mask,
40 |       std::vector<cv::KeyPoint>& keypoints,
41 |       cv::OutputArray descriptors);
42 | 
43 |     int inline GetLevels(){
44 |         return nlevels;}
45 | 
46 |     float inline GetScaleFactor(){
47 |         return scaleFactor;}
48 | 
49 |     std::vector<float> inline GetScaleFactors(){
50 |         return mvScaleFactor;
51 |     }
52 | 
53 |     std::vector<float> inline GetInverseScaleFactors(){
54 |         return mvInvScaleFactor;
55 |     }
56 | 
57 |     std::vector<float> inline GetScaleSigmaSquares(){
58 |         return mvLevelSigma2;
59 |     }
60 | 
61 |     std::vector<float> inline GetInverseScaleSigmaSquares(){
62 |         return mvInvLevelSigma2;
63 |     }
64 | 
65 |     std::vector<cv::Mat> mvImagePyramid;
66 | 
67 | protected:
68 | 
69 |     void ComputePyramid(cv::Mat image);
70 |     void ComputeKeyPointsOctTree(std::vector<std::vector<cv::KeyPoint> >& allKeypoints);    
71 |     std::vector<cv::KeyPoint> DistributeOctTree(const std::vector<cv::KeyPoint>& vToDistributeKeys, const int &minX,
72 |                                            const int &maxX, const int &minY, const int &maxY, const int &nFeatures, const int &level);
73 | 
74 |     void ComputeKeyPointsOld(std::vector<std::vector<cv::KeyPoint> >& allKeypoints);
75 |     std::vector<cv::Point> pattern;
76 | 
77 |     int nfeatures;
78 |     double scaleFactor;
79 |     int nlevels;
80 |     int iniThFAST;
81 |     int minThFAST;
82 | 
83 |     std::vector<int> mnFeaturesPerLevel;
84 | 
85 |     std::vector<int> umax;
86 | 
87 |     std::vector<float> mvScaleFactor;
88 |     std::vector<float> mvInvScaleFactor;    
89 |     std::vector<float> mvLevelSigma2;
90 |     std::vector<float> mvInvLevelSigma2;
91 | };
92 | 
93 | } //namespace ORB_SLAM
94 | 
95 | #endif
96 | 
97 | 


--------------------------------------------------------------------------------
/lib/liborbextractor.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/lib/liborbextractor.so


--------------------------------------------------------------------------------
/param/orb.yaml:
--------------------------------------------------------------------------------
 1 | %YAML:1.0
 2 | 
 3 | #--------------------------------------------------------------------------------------------
 4 | # ORB Parameters
 5 | #--------------------------------------------------------------------------------------------
 6 | 
 7 | # ORB Extractor: Number of features per image
 8 | ORBextractor.nFeatures: 2000
 9 | 
10 | # ORB Extractor: Scale factor between levels in the scale pyramid 	
11 | ORBextractor.scaleFactor: 1.2
12 | 
13 | # ORB Extractor: Number of levels in the scale pyramid	
14 | ORBextractor.nLevels: 8
15 | 
16 | # ORB Extractor: Fast threshold
17 | # Image is divided in a grid. At each cell FAST are extracted imposing a minimum response.
18 | # Firstly we impose iniThFAST. If no corners are detected we impose a lower value minThFAST
19 | # You can lower these values if your images have low contrast			
20 | ORBextractor.iniThFAST: 20
21 | ORBextractor.minThFAST: 7
22 | 


--------------------------------------------------------------------------------
/screenshots/1.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/screenshots/1.jpg


--------------------------------------------------------------------------------
/screenshots/2.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/screenshots/2.jpg


--------------------------------------------------------------------------------
/screenshots/liu1.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/screenshots/liu1.jpg


--------------------------------------------------------------------------------
/screenshots/liu2.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/screenshots/liu2.jpg


--------------------------------------------------------------------------------
/src/ORBextractor:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/liuzhenboo/orb-extractor/538dbab86a60fd6632b4d222819fe2058332fcc2/src/ORBextractor


--------------------------------------------------------------------------------
/src/ORBextractor.cpp:
--------------------------------------------------------------------------------
   1 | #include <opencv2/core/core.hpp>
   2 | #include <opencv2/highgui/highgui.hpp>
   3 | #include <opencv2/features2d/features2d.hpp>
   4 | #include <opencv2/imgproc/imgproc.hpp>
   5 | #include <vector>
   6 | #include <iterator>
   7 | 
   8 | #include "ORBextractor.h"
   9 | #include <iostream>
  10 | 
  11 | 
  12 | using namespace cv;
  13 | using namespace std;
  14 | 
  15 | namespace ORB_SLAM2
  16 | {
  17 | 
  18 | const int PATCH_SIZE = 31;
  19 | const int HALF_PATCH_SIZE = 15;
  20 | const int EDGE_THRESHOLD = 19;
  21 | 
  22 | 
  23 | static float IC_Angle(const Mat& image, Point2f pt,  const vector<int> & u_max)
  24 | {
  25 |     int m_01 = 0, m_10 = 0;
  26 | 
  27 |     const uchar* center = &image.at<uchar> (cvRound(pt.y), cvRound(pt.x));
  28 | 
  29 |     // Treat the center line differently, v=0
  30 |     for (int u = -HALF_PATCH_SIZE; u <= HALF_PATCH_SIZE; ++u)
  31 |         m_10 += u * center[u];
  32 | 
  33 |     // Go line by line in the circuI853lar patch
  34 |     int step = (int)image.step1();
  35 |     for (int v = 1; v <= HALF_PATCH_SIZE; ++v)
  36 |     {
  37 |         // Proceed over the two lines
  38 |         int v_sum = 0;
  39 |         int d = u_max[v];
  40 |         for (int u = -d; u <= d; ++u)
  41 |         {
  42 |             int val_plus = center[u + v*step], val_minus = center[u - v*step];
  43 |             v_sum += (val_plus - val_minus);
  44 |             m_10 += u * (val_plus + val_minus);
  45 |         }
  46 |         m_01 += v * v_sum;
  47 |     }
  48 | 
  49 |     return fastAtan2((float)m_01, (float)m_10);
  50 | }
  51 | 
  52 | 
  53 | const float factorPI = (float)(CV_PI/180.f);
  54 | static void computeOrbDescriptor(const KeyPoint& kpt,
  55 |                                  const Mat& img, const Point* pattern,
  56 |                                  uchar* desc)
  57 | {
  58 |     float angle = (float)kpt.angle*factorPI;
  59 |     float a = (float)cos(angle), b = (float)sin(angle);
  60 | 
  61 |     const uchar* center = &img.at<uchar>(cvRound(kpt.pt.y), cvRound(kpt.pt.x));
  62 |     const int step = (int)img.step;
  63 | 
  64 |     #define GET_VALUE(idx) \
  65 |         center[cvRound(pattern[idx].x*b + pattern[idx].y*a)*step + \
  66 |                cvRound(pattern[idx].x*a - pattern[idx].y*b)]
  67 | 
  68 | 
  69 |     for (int i = 0; i < 32; ++i, pattern += 16)
  70 |     {
  71 |         int t0, t1, val;
  72 |         t0 = GET_VALUE(0); t1 = GET_VALUE(1);
  73 |         val = t0 < t1;
  74 |         t0 = GET_VALUE(2); t1 = GET_VALUE(3);
  75 |         val |= (t0 < t1) << 1;
  76 |         t0 = GET_VALUE(4); t1 = GET_VALUE(5);
  77 |         val |= (t0 < t1) << 2;
  78 |         t0 = GET_VALUE(6); t1 = GET_VALUE(7);
  79 |         val |= (t0 < t1) << 3;
  80 |         t0 = GET_VALUE(8); t1 = GET_VALUE(9);
  81 |         val |= (t0 < t1) << 4;
  82 |         t0 = GET_VALUE(10); t1 = GET_VALUE(11);
  83 |         val |= (t0 < t1) << 5;
  84 |         t0 = GET_VALUE(12); t1 = GET_VALUE(13);
  85 |         val |= (t0 < t1) << 6;
  86 |         t0 = GET_VALUE(14); t1 = GET_VALUE(15);
  87 |         val |= (t0 < t1) << 7;
  88 | 
  89 |         desc[i] = (uchar)val;
  90 |     }
  91 | 
  92 |     #undef GET_VALUE
  93 | }
  94 | 
  95 | 
  96 | static int bit_pattern_31_[256*4] =
  97 | {
  98 |     8,-3, 9,5/*mean (0), correlation (0)*/,
  99 |     4,2, 7,-12/*mean (1.12461e-05), correlation (0.0437584)*/,
 100 |     -11,9, -8,2/*mean (3.37382e-05), correlation (0.0617409)*/,
 101 |     7,-12, 12,-13/*mean (5.62303e-05), correlation (0.0636977)*/,
 102 |     2,-13, 2,12/*mean (0.000134953), correlation (0.085099)*/,
 103 |     1,-7, 1,6/*mean (0.000528565), correlation (0.0857175)*/,
 104 |     -2,-10, -2,-4/*mean (0.0188821), correlation (0.0985774)*/,
 105 |     -13,-13, -11,-8/*mean (0.0363135), correlation (0.0899616)*/,
 106 |     -13,-3, -12,-9/*mean (0.121806), correlation (0.099849)*/,
 107 |     10,4, 11,9/*mean (0.122065), correlation (0.093285)*/,
 108 |     -13,-8, -8,-9/*mean (0.162787), correlation (0.0942748)*/,
 109 |     -11,7, -9,12/*mean (0.21561), correlation (0.0974438)*/,
 110 |     7,7, 12,6/*mean (0.160583), correlation (0.130064)*/,
 111 |     -4,-5, -3,0/*mean (0.228171), correlation (0.132998)*/,
 112 |     -13,2, -12,-3/*mean (0.00997526), correlation (0.145926)*/,
 113 |     -9,0, -7,5/*mean (0.198234), correlation (0.143636)*/,
 114 |     12,-6, 12,-1/*mean (0.0676226), correlation (0.16689)*/,
 115 |     -3,6, -2,12/*mean (0.166847), correlation (0.171682)*/,
 116 |     -6,-13, -4,-8/*mean (0.101215), correlation (0.179716)*/,
 117 |     11,-13, 12,-8/*mean (0.200641), correlation (0.192279)*/,
 118 |     4,7, 5,1/*mean (0.205106), correlation (0.186848)*/,
 119 |     5,-3, 10,-3/*mean (0.234908), correlation (0.192319)*/,
 120 |     3,-7, 6,12/*mean (0.0709964), correlation (0.210872)*/,
 121 |     -8,-7, -6,-2/*mean (0.0939834), correlation (0.212589)*/,
 122 |     -2,11, -1,-10/*mean (0.127778), correlation (0.20866)*/,
 123 |     -13,12, -8,10/*mean (0.14783), correlation (0.206356)*/,
 124 |     -7,3, -5,-3/*mean (0.182141), correlation (0.198942)*/,
 125 |     -4,2, -3,7/*mean (0.188237), correlation (0.21384)*/,
 126 |     -10,-12, -6,11/*mean (0.14865), correlation (0.23571)*/,
 127 |     5,-12, 6,-7/*mean (0.222312), correlation (0.23324)*/,
 128 |     5,-6, 7,-1/*mean (0.229082), correlation (0.23389)*/,
 129 |     1,0, 4,-5/*mean (0.241577), correlation (0.215286)*/,
 130 |     9,11, 11,-13/*mean (0.00338507), correlation (0.251373)*/,
 131 |     4,7, 4,12/*mean (0.131005), correlation (0.257622)*/,
 132 |     2,-1, 4,4/*mean (0.152755), correlation (0.255205)*/,
 133 |     -4,-12, -2,7/*mean (0.182771), correlation (0.244867)*/,
 134 |     -8,-5, -7,-10/*mean (0.186898), correlation (0.23901)*/,
 135 |     4,11, 9,12/*mean (0.226226), correlation (0.258255)*/,
 136 |     0,-8, 1,-13/*mean (0.0897886), correlation (0.274827)*/,
 137 |     -13,-2, -8,2/*mean (0.148774), correlation (0.28065)*/,
 138 |     -3,-2, -2,3/*mean (0.153048), correlation (0.283063)*/,
 139 |     -6,9, -4,-9/*mean (0.169523), correlation (0.278248)*/,
 140 |     8,12, 10,7/*mean (0.225337), correlation (0.282851)*/,
 141 |     0,9, 1,3/*mean (0.226687), correlation (0.278734)*/,
 142 |     7,-5, 11,-10/*mean (0.00693882), correlation (0.305161)*/,
 143 |     -13,-6, -11,0/*mean (0.0227283), correlation (0.300181)*/,
 144 |     10,7, 12,1/*mean (0.125517), correlation (0.31089)*/,
 145 |     -6,-3, -6,12/*mean (0.131748), correlation (0.312779)*/,
 146 |     10,-9, 12,-4/*mean (0.144827), correlation (0.292797)*/,
 147 |     -13,8, -8,-12/*mean (0.149202), correlation (0.308918)*/,
 148 |     -13,0, -8,-4/*mean (0.160909), correlation (0.310013)*/,
 149 |     3,3, 7,8/*mean (0.177755), correlation (0.309394)*/,
 150 |     5,7, 10,-7/*mean (0.212337), correlation (0.310315)*/,
 151 |     -1,7, 1,-12/*mean (0.214429), correlation (0.311933)*/,
 152 |     3,-10, 5,6/*mean (0.235807), correlation (0.313104)*/,
 153 |     2,-4, 3,-10/*mean (0.00494827), correlation (0.344948)*/,
 154 |     -13,0, -13,5/*mean (0.0549145), correlation (0.344675)*/,
 155 |     -13,-7, -12,12/*mean (0.103385), correlation (0.342715)*/,
 156 |     -13,3, -11,8/*mean (0.134222), correlation (0.322922)*/,
 157 |     -7,12, -4,7/*mean (0.153284), correlation (0.337061)*/,
 158 |     6,-10, 12,8/*mean (0.154881), correlation (0.329257)*/,
 159 |     -9,-1, -7,-6/*mean (0.200967), correlation (0.33312)*/,
 160 |     -2,-5, 0,12/*mean (0.201518), correlation (0.340635)*/,
 161 |     -12,5, -7,5/*mean (0.207805), correlation (0.335631)*/,
 162 |     3,-10, 8,-13/*mean (0.224438), correlation (0.34504)*/,
 163 |     -7,-7, -4,5/*mean (0.239361), correlation (0.338053)*/,
 164 |     -3,-2, -1,-7/*mean (0.240744), correlation (0.344322)*/,
 165 |     2,9, 5,-11/*mean (0.242949), correlation (0.34145)*/,
 166 |     -11,-13, -5,-13/*mean (0.244028), correlation (0.336861)*/,
 167 |     -1,6, 0,-1/*mean (0.247571), correlation (0.343684)*/,
 168 |     5,-3, 5,2/*mean (0.000697256), correlation (0.357265)*/,
 169 |     -4,-13, -4,12/*mean (0.00213675), correlation (0.373827)*/,
 170 |     -9,-6, -9,6/*mean (0.0126856), correlation (0.373938)*/,
 171 |     -12,-10, -8,-4/*mean (0.0152497), correlation (0.364237)*/,
 172 |     10,2, 12,-3/*mean (0.0299933), correlation (0.345292)*/,
 173 |     7,12, 12,12/*mean (0.0307242), correlation (0.366299)*/,
 174 |     -7,-13, -6,5/*mean (0.0534975), correlation (0.368357)*/,
 175 |     -4,9, -3,4/*mean (0.099865), correlation (0.372276)*/,
 176 |     7,-1, 12,2/*mean (0.117083), correlation (0.364529)*/,
 177 |     -7,6, -5,1/*mean (0.126125), correlation (0.369606)*/,
 178 |     -13,11, -12,5/*mean (0.130364), correlation (0.358502)*/,
 179 |     -3,7, -2,-6/*mean (0.131691), correlation (0.375531)*/,
 180 |     7,-8, 12,-7/*mean (0.160166), correlation (0.379508)*/,
 181 |     -13,-7, -11,-12/*mean (0.167848), correlation (0.353343)*/,
 182 |     1,-3, 12,12/*mean (0.183378), correlation (0.371916)*/,
 183 |     2,-6, 3,0/*mean (0.228711), correlation (0.371761)*/,
 184 |     -4,3, -2,-13/*mean (0.247211), correlation (0.364063)*/,
 185 |     -1,-13, 1,9/*mean (0.249325), correlation (0.378139)*/,
 186 |     7,1, 8,-6/*mean (0.000652272), correlation (0.411682)*/,
 187 |     1,-1, 3,12/*mean (0.00248538), correlation (0.392988)*/,
 188 |     9,1, 12,6/*mean (0.0206815), correlation (0.386106)*/,
 189 |     -1,-9, -1,3/*mean (0.0364485), correlation (0.410752)*/,
 190 |     -13,-13, -10,5/*mean (0.0376068), correlation (0.398374)*/,
 191 |     7,7, 10,12/*mean (0.0424202), correlation (0.405663)*/,
 192 |     12,-5, 12,9/*mean (0.0942645), correlation (0.410422)*/,
 193 |     6,3, 7,11/*mean (0.1074), correlation (0.413224)*/,
 194 |     5,-13, 6,10/*mean (0.109256), correlation (0.408646)*/,
 195 |     2,-12, 2,3/*mean (0.131691), correlation (0.416076)*/,
 196 |     3,8, 4,-6/*mean (0.165081), correlation (0.417569)*/,
 197 |     2,6, 12,-13/*mean (0.171874), correlation (0.408471)*/,
 198 |     9,-12, 10,3/*mean (0.175146), correlation (0.41296)*/,
 199 |     -8,4, -7,9/*mean (0.183682), correlation (0.402956)*/,
 200 |     -11,12, -4,-6/*mean (0.184672), correlation (0.416125)*/,
 201 |     1,12, 2,-8/*mean (0.191487), correlation (0.386696)*/,
 202 |     6,-9, 7,-4/*mean (0.192668), correlation (0.394771)*/,
 203 |     2,3, 3,-2/*mean (0.200157), correlation (0.408303)*/,
 204 |     6,3, 11,0/*mean (0.204588), correlation (0.411762)*/,
 205 |     3,-3, 8,-8/*mean (0.205904), correlation (0.416294)*/,
 206 |     7,8, 9,3/*mean (0.213237), correlation (0.409306)*/,
 207 |     -11,-5, -6,-4/*mean (0.243444), correlation (0.395069)*/,
 208 |     -10,11, -5,10/*mean (0.247672), correlation (0.413392)*/,
 209 |     -5,-8, -3,12/*mean (0.24774), correlation (0.411416)*/,
 210 |     -10,5, -9,0/*mean (0.00213675), correlation (0.454003)*/,
 211 |     8,-1, 12,-6/*mean (0.0293635), correlation (0.455368)*/,
 212 |     4,-6, 6,-11/*mean (0.0404971), correlation (0.457393)*/,
 213 |     -10,12, -8,7/*mean (0.0481107), correlation (0.448364)*/,
 214 |     4,-2, 6,7/*mean (0.050641), correlation (0.455019)*/,
 215 |     -2,0, -2,12/*mean (0.0525978), correlation (0.44338)*/,
 216 |     -5,-8, -5,2/*mean (0.0629667), correlation (0.457096)*/,
 217 |     7,-6, 10,12/*mean (0.0653846), correlation (0.445623)*/,
 218 |     -9,-13, -8,-8/*mean (0.0858749), correlation (0.449789)*/,
 219 |     -5,-13, -5,-2/*mean (0.122402), correlation (0.450201)*/,
 220 |     8,-8, 9,-13/*mean (0.125416), correlation (0.453224)*/,
 221 |     -9,-11, -9,0/*mean (0.130128), correlation (0.458724)*/,
 222 |     1,-8, 1,-2/*mean (0.132467), correlation (0.440133)*/,
 223 |     7,-4, 9,1/*mean (0.132692), correlation (0.454)*/,
 224 |     -2,1, -1,-4/*mean (0.135695), correlation (0.455739)*/,
 225 |     11,-6, 12,-11/*mean (0.142904), correlation (0.446114)*/,
 226 |     -12,-9, -6,4/*mean (0.146165), correlation (0.451473)*/,
 227 |     3,7, 7,12/*mean (0.147627), correlation (0.456643)*/,
 228 |     5,5, 10,8/*mean (0.152901), correlation (0.455036)*/,
 229 |     0,-4, 2,8/*mean (0.167083), correlation (0.459315)*/,
 230 |     -9,12, -5,-13/*mean (0.173234), correlation (0.454706)*/,
 231 |     0,7, 2,12/*mean (0.18312), correlation (0.433855)*/,
 232 |     -1,2, 1,7/*mean (0.185504), correlation (0.443838)*/,
 233 |     5,11, 7,-9/*mean (0.185706), correlation (0.451123)*/,
 234 |     3,5, 6,-8/*mean (0.188968), correlation (0.455808)*/,
 235 |     -13,-4, -8,9/*mean (0.191667), correlation (0.459128)*/,
 236 |     -5,9, -3,-3/*mean (0.193196), correlation (0.458364)*/,
 237 |     -4,-7, -3,-12/*mean (0.196536), correlation (0.455782)*/,
 238 |     6,5, 8,0/*mean (0.1972), correlation (0.450481)*/,
 239 |     -7,6, -6,12/*mean (0.199438), correlation (0.458156)*/,
 240 |     -13,6, -5,-2/*mean (0.211224), correlation (0.449548)*/,
 241 |     1,-10, 3,10/*mean (0.211718), correlation (0.440606)*/,
 242 |     4,1, 8,-4/*mean (0.213034), correlation (0.443177)*/,
 243 |     -2,-2, 2,-13/*mean (0.234334), correlation (0.455304)*/,
 244 |     2,-12, 12,12/*mean (0.235684), correlation (0.443436)*/,
 245 |     -2,-13, 0,-6/*mean (0.237674), correlation (0.452525)*/,
 246 |     4,1, 9,3/*mean (0.23962), correlation (0.444824)*/,
 247 |     -6,-10, -3,-5/*mean (0.248459), correlation (0.439621)*/,
 248 |     -3,-13, -1,1/*mean (0.249505), correlation (0.456666)*/,
 249 |     7,5, 12,-11/*mean (0.00119208), correlation (0.495466)*/,
 250 |     4,-2, 5,-7/*mean (0.00372245), correlation (0.484214)*/,
 251 |     -13,9, -9,-5/*mean (0.00741116), correlation (0.499854)*/,
 252 |     7,1, 8,6/*mean (0.0208952), correlation (0.499773)*/,
 253 |     7,-8, 7,6/*mean (0.0220085), correlation (0.501609)*/,
 254 |     -7,-4, -7,1/*mean (0.0233806), correlation (0.496568)*/,
 255 |     -8,11, -7,-8/*mean (0.0236505), correlation (0.489719)*/,
 256 |     -13,6, -12,-8/*mean (0.0268781), correlation (0.503487)*/,
 257 |     2,4, 3,9/*mean (0.0323324), correlation (0.501938)*/,
 258 |     10,-5, 12,3/*mean (0.0399235), correlation (0.494029)*/,
 259 |     -6,-5, -6,7/*mean (0.0420153), correlation (0.486579)*/,
 260 |     8,-3, 9,-8/*mean (0.0548021), correlation (0.484237)*/,
 261 |     2,-12, 2,8/*mean (0.0616622), correlation (0.496642)*/,
 262 |     -11,-2, -10,3/*mean (0.0627755), correlation (0.498563)*/,
 263 |     -12,-13, -7,-9/*mean (0.0829622), correlation (0.495491)*/,
 264 |     -11,0, -10,-5/*mean (0.0843342), correlation (0.487146)*/,
 265 |     5,-3, 11,8/*mean (0.0929937), correlation (0.502315)*/,
 266 |     -2,-13, -1,12/*mean (0.113327), correlation (0.48941)*/,
 267 |     -1,-8, 0,9/*mean (0.132119), correlation (0.467268)*/,
 268 |     -13,-11, -12,-5/*mean (0.136269), correlation (0.498771)*/,
 269 |     -10,-2, -10,11/*mean (0.142173), correlation (0.498714)*/,
 270 |     -3,9, -2,-13/*mean (0.144141), correlation (0.491973)*/,
 271 |     2,-3, 3,2/*mean (0.14892), correlation (0.500782)*/,
 272 |     -9,-13, -4,0/*mean (0.150371), correlation (0.498211)*/,
 273 |     -4,6, -3,-10/*mean (0.152159), correlation (0.495547)*/,
 274 |     -4,12, -2,-7/*mean (0.156152), correlation (0.496925)*/,
 275 |     -6,-11, -4,9/*mean (0.15749), correlation (0.499222)*/,
 276 |     6,-3, 6,11/*mean (0.159211), correlation (0.503821)*/,
 277 |     -13,11, -5,5/*mean (0.162427), correlation (0.501907)*/,
 278 |     11,11, 12,6/*mean (0.16652), correlation (0.497632)*/,
 279 |     7,-5, 12,-2/*mean (0.169141), correlation (0.484474)*/,
 280 |     -1,12, 0,7/*mean (0.169456), correlation (0.495339)*/,
 281 |     -4,-8, -3,-2/*mean (0.171457), correlation (0.487251)*/,
 282 |     -7,1, -6,7/*mean (0.175), correlation (0.500024)*/,
 283 |     -13,-12, -8,-13/*mean (0.175866), correlation (0.497523)*/,
 284 |     -7,-2, -6,-8/*mean (0.178273), correlation (0.501854)*/,
 285 |     -8,5, -6,-9/*mean (0.181107), correlation (0.494888)*/,
 286 |     -5,-1, -4,5/*mean (0.190227), correlation (0.482557)*/,
 287 |     -13,7, -8,10/*mean (0.196739), correlation (0.496503)*/,
 288 |     1,5, 5,-13/*mean (0.19973), correlation (0.499759)*/,
 289 |     1,0, 10,-13/*mean (0.204465), correlation (0.49873)*/,
 290 |     9,12, 10,-1/*mean (0.209334), correlation (0.49063)*/,
 291 |     5,-8, 10,-9/*mean (0.211134), correlation (0.503011)*/,
 292 |     -1,11, 1,-13/*mean (0.212), correlation (0.499414)*/,
 293 |     -9,-3, -6,2/*mean (0.212168), correlation (0.480739)*/,
 294 |     -1,-10, 1,12/*mean (0.212731), correlation (0.502523)*/,
 295 |     -13,1, -8,-10/*mean (0.21327), correlation (0.489786)*/,
 296 |     8,-11, 10,-6/*mean (0.214159), correlation (0.488246)*/,
 297 |     2,-13, 3,-6/*mean (0.216993), correlation (0.50287)*/,
 298 |     7,-13, 12,-9/*mean (0.223639), correlation (0.470502)*/,
 299 |     -10,-10, -5,-7/*mean (0.224089), correlation (0.500852)*/,
 300 |     -10,-8, -8,-13/*mean (0.228666), correlation (0.502629)*/,
 301 |     4,-6, 8,5/*mean (0.22906), correlation (0.498305)*/,
 302 |     3,12, 8,-13/*mean (0.233378), correlation (0.503825)*/,
 303 |     -4,2, -3,-3/*mean (0.234323), correlation (0.476692)*/,
 304 |     5,-13, 10,-12/*mean (0.236392), correlation (0.475462)*/,
 305 |     4,-13, 5,-1/*mean (0.236842), correlation (0.504132)*/,
 306 |     -9,9, -4,3/*mean (0.236977), correlation (0.497739)*/,
 307 |     0,3, 3,-9/*mean (0.24314), correlation (0.499398)*/,
 308 |     -12,1, -6,1/*mean (0.243297), correlation (0.489447)*/,
 309 |     3,2, 4,-8/*mean (0.00155196), correlation (0.553496)*/,
 310 |     -10,-10, -10,9/*mean (0.00239541), correlation (0.54297)*/,
 311 |     8,-13, 12,12/*mean (0.0034413), correlation (0.544361)*/,
 312 |     -8,-12, -6,-5/*mean (0.003565), correlation (0.551225)*/,
 313 |     2,2, 3,7/*mean (0.00835583), correlation (0.55285)*/,
 314 |     10,6, 11,-8/*mean (0.00885065), correlation (0.540913)*/,
 315 |     6,8, 8,-12/*mean (0.0101552), correlation (0.551085)*/,
 316 |     -7,10, -6,5/*mean (0.0102227), correlation (0.533635)*/,
 317 |     -3,-9, -3,9/*mean (0.0110211), correlation (0.543121)*/,
 318 |     -1,-13, -1,5/*mean (0.0113473), correlation (0.550173)*/,
 319 |     -3,-7, -3,4/*mean (0.0140913), correlation (0.554774)*/,
 320 |     -8,-2, -8,3/*mean (0.017049), correlation (0.55461)*/,
 321 |     4,2, 12,12/*mean (0.01778), correlation (0.546921)*/,
 322 |     2,-5, 3,11/*mean (0.0224022), correlation (0.549667)*/,
 323 |     6,-9, 11,-13/*mean (0.029161), correlation (0.546295)*/,
 324 |     3,-1, 7,12/*mean (0.0303081), correlation (0.548599)*/,
 325 |     11,-1, 12,4/*mean (0.0355151), correlation (0.523943)*/,
 326 |     -3,0, -3,6/*mean (0.0417904), correlation (0.543395)*/,
 327 |     4,-11, 4,12/*mean (0.0487292), correlation (0.542818)*/,
 328 |     2,-4, 2,1/*mean (0.0575124), correlation (0.554888)*/,
 329 |     -10,-6, -8,1/*mean (0.0594242), correlation (0.544026)*/,
 330 |     -13,7, -11,1/*mean (0.0597391), correlation (0.550524)*/,
 331 |     -13,12, -11,-13/*mean (0.0608974), correlation (0.55383)*/,
 332 |     6,0, 11,-13/*mean (0.065126), correlation (0.552006)*/,
 333 |     0,-1, 1,4/*mean (0.074224), correlation (0.546372)*/,
 334 |     -13,3, -9,-2/*mean (0.0808592), correlation (0.554875)*/,
 335 |     -9,8, -6,-3/*mean (0.0883378), correlation (0.551178)*/,
 336 |     -13,-6, -8,-2/*mean (0.0901035), correlation (0.548446)*/,
 337 |     5,-9, 8,10/*mean (0.0949843), correlation (0.554694)*/,
 338 |     2,7, 3,-9/*mean (0.0994152), correlation (0.550979)*/,
 339 |     -1,-6, -1,-1/*mean (0.10045), correlation (0.552714)*/,
 340 |     9,5, 11,-2/*mean (0.100686), correlation (0.552594)*/,
 341 |     11,-3, 12,-8/*mean (0.101091), correlation (0.532394)*/,
 342 |     3,0, 3,5/*mean (0.101147), correlation (0.525576)*/,
 343 |     -1,4, 0,10/*mean (0.105263), correlation (0.531498)*/,
 344 |     3,-6, 4,5/*mean (0.110785), correlation (0.540491)*/,
 345 |     -13,0, -10,5/*mean (0.112798), correlation (0.536582)*/,
 346 |     5,8, 12,11/*mean (0.114181), correlation (0.555793)*/,
 347 |     8,9, 9,-6/*mean (0.117431), correlation (0.553763)*/,
 348 |     7,-4, 8,-12/*mean (0.118522), correlation (0.553452)*/,
 349 |     -10,4, -10,9/*mean (0.12094), correlation (0.554785)*/,
 350 |     7,3, 12,4/*mean (0.122582), correlation (0.555825)*/,
 351 |     9,-7, 10,-2/*mean (0.124978), correlation (0.549846)*/,
 352 |     7,0, 12,-2/*mean (0.127002), correlation (0.537452)*/,
 353 |     -1,-6, 0,-11/*mean (0.127148), correlation (0.547401)*/
 354 | };
 355 | 
 356 | ORBextractor::ORBextractor(int _nfeatures, float _scaleFactor, int _nlevels,
 357 |          int _iniThFAST, int _minThFAST):
 358 |     nfeatures(_nfeatures), scaleFactor(_scaleFactor), nlevels(_nlevels),
 359 |     iniThFAST(_iniThFAST), minThFAST(_minThFAST)
 360 | {
 361 |     mvScaleFactor.resize(nlevels);
 362 |     mvLevelSigma2.resize(nlevels);
 363 |     mvScaleFactor[0]=1.0f;
 364 |     mvLevelSigma2[0]=1.0f;
 365 |     for(int i=1; i<nlevels; i++)
 366 |     {
 367 |         mvScaleFactor[i]=mvScaleFactor[i-1]*scaleFactor;
 368 |         mvLevelSigma2[i]=mvScaleFactor[i]*mvScaleFactor[i];
 369 |     }
 370 | 
 371 |     mvInvScaleFactor.resize(nlevels);
 372 |     mvInvLevelSigma2.resize(nlevels);
 373 |     for(int i=0; i<nlevels; i++)
 374 |     {
 375 |         mvInvScaleFactor[i]=1.0f/mvScaleFactor[i];
 376 |         mvInvLevelSigma2[i]=1.0f/mvLevelSigma2[i];
 377 |     }
 378 | 
 379 |     mvImagePyramid.resize(nlevels);
 380 | 
 381 |     mnFeaturesPerLevel.resize(nlevels);
 382 |     float factor = 1.0f / scaleFactor;
 383 |     float nDesiredFeaturesPerScale = nfeatures*(1 - factor)/(1 - (float)pow((double)factor, (double)nlevels));
 384 | 
 385 |     int sumFeatures = 0;
 386 |     for( int level = 0; level < nlevels-1; level++ )
 387 |     {
 388 |         mnFeaturesPerLevel[level] = cvRound(nDesiredFeaturesPerScale);
 389 |         sumFeatures += mnFeaturesPerLevel[level];
 390 |         nDesiredFeaturesPerScale *= factor;
 391 |     }
 392 |     mnFeaturesPerLevel[nlevels-1] = std::max(nfeatures - sumFeatures, 0);
 393 | 
 394 |     const int npoints = 512;
 395 |     const Point* pattern0 = (const Point*)bit_pattern_31_;
 396 |     std::copy(pattern0, pattern0 + npoints, std::back_inserter(pattern));
 397 | 
 398 |     //This is for orientation
 399 |     // pre-compute the end of a row in a circular patch
 400 |     umax.resize(HALF_PATCH_SIZE + 1);
 401 | 
 402 |     int v, v0, vmax = cvFloor(HALF_PATCH_SIZE * sqrt(2.f) / 2 + 1);
 403 |     int vmin = cvCeil(HALF_PATCH_SIZE * sqrt(2.f) / 2);
 404 |     const double hp2 = HALF_PATCH_SIZE*HALF_PATCH_SIZE;
 405 |     for (v = 0; v <= vmax; ++v)
 406 |         umax[v] = cvRound(sqrt(hp2 - v * v));
 407 | 
 408 |     // Make sure we are symmetric
 409 |     for (v = HALF_PATCH_SIZE, v0 = 0; v >= vmin; --v)
 410 |     {
 411 |         while (umax[v0] == umax[v0 + 1])
 412 |             ++v0;
 413 |         umax[v] = v0;
 414 |         ++v0;
 415 |     }
 416 | }
 417 | 
 418 | static void computeOrientation(const Mat& image, vector<KeyPoint>& keypoints, const vector<int>& umax)
 419 | {
 420 |     for (vector<KeyPoint>::iterator keypoint = keypoints.begin(),
 421 |          keypointEnd = keypoints.end(); keypoint != keypointEnd; ++keypoint)
 422 |     {
 423 |         keypoint->angle = IC_Angle(image, keypoint->pt, umax);
 424 |     }
 425 | }
 426 | 
 427 | void ExtractorNode::DivideNode(ExtractorNode &n1, ExtractorNode &n2, ExtractorNode &n3, ExtractorNode &n4)
 428 | {
 429 |     const int halfX = ceil(static_cast<float>(UR.x-UL.x)/2);
 430 |     const int halfY = ceil(static_cast<float>(BR.y-UL.y)/2);
 431 | 
 432 |     //Define boundaries of childs
 433 |     n1.UL = UL;
 434 |     n1.UR = cv::Point2i(UL.x+halfX,UL.y);
 435 |     n1.BL = cv::Point2i(UL.x,UL.y+halfY);
 436 |     n1.BR = cv::Point2i(UL.x+halfX,UL.y+halfY);
 437 |     n1.vKeys.reserve(vKeys.size());
 438 | 
 439 |     n2.UL = n1.UR;
 440 |     n2.UR = UR;
 441 |     n2.BL = n1.BR;
 442 |     n2.BR = cv::Point2i(UR.x,UL.y+halfY);
 443 |     n2.vKeys.reserve(vKeys.size());
 444 | 
 445 |     n3.UL = n1.BL;
 446 |     n3.UR = n1.BR;
 447 |     n3.BL = BL;
 448 |     n3.BR = cv::Point2i(n1.BR.x,BL.y);
 449 |     n3.vKeys.reserve(vKeys.size());
 450 | 
 451 |     n4.UL = n3.UR;
 452 |     n4.UR = n2.BR;
 453 |     n4.BL = n3.BR;
 454 |     n4.BR = BR;
 455 |     n4.vKeys.reserve(vKeys.size());
 456 | 
 457 |     //Associate points to childs
 458 |     for(size_t i=0;i<vKeys.size();i++)
 459 |     {
 460 |         const cv::KeyPoint &kp = vKeys[i];
 461 |         if(kp.pt.x<n1.UR.x)
 462 |         {
 463 |             if(kp.pt.y<n1.BR.y)
 464 |                 n1.vKeys.push_back(kp);
 465 |             else
 466 |                 n3.vKeys.push_back(kp);
 467 |         }
 468 |         else if(kp.pt.y<n1.BR.y)
 469 |             n2.vKeys.push_back(kp);
 470 |         else
 471 |             n4.vKeys.push_back(kp);
 472 |     }
 473 | 
 474 |     if(n1.vKeys.size()==1)
 475 |         n1.bNoMore = true;
 476 |     if(n2.vKeys.size()==1)
 477 |         n2.bNoMore = true;
 478 |     if(n3.vKeys.size()==1)
 479 |         n3.bNoMore = true;
 480 |     if(n4.vKeys.size()==1)
 481 |         n4.bNoMore = true;
 482 | 
 483 | }
 484 | 
 485 | vector<cv::KeyPoint> ORBextractor::DistributeOctTree(const vector<cv::KeyPoint>& vToDistributeKeys, const int &minX,
 486 |                                        const int &maxX, const int &minY, const int &maxY, const int &N, const int &level)
 487 | {
 488 |     // Compute how many initial nodes   
 489 |     const int nIni = round(static_cast<float>(maxX-minX)/(maxY-minY));
 490 | 
 491 |     const float hX = static_cast<float>(maxX-minX)/nIni;
 492 | 
 493 |     list<ExtractorNode> lNodes;
 494 | 
 495 |     vector<ExtractorNode*> vpIniNodes;
 496 |     vpIniNodes.resize(nIni);
 497 | 
 498 |     for(int i=0; i<nIni; i++)
 499 |     {
 500 |         ExtractorNode ni;
 501 |         ni.UL = cv::Point2i(hX*static_cast<float>(i),0);
 502 |         ni.UR = cv::Point2i(hX*static_cast<float>(i+1),0);
 503 |         ni.BL = cv::Point2i(ni.UL.x,maxY-minY);
 504 |         ni.BR = cv::Point2i(ni.UR.x,maxY-minY);
 505 |         ni.vKeys.reserve(vToDistributeKeys.size());
 506 | 
 507 |         lNodes.push_back(ni);
 508 |         vpIniNodes[i] = &lNodes.back();
 509 |     }
 510 | 
 511 |     //Associate points to childs
 512 |     for(size_t i=0;i<vToDistributeKeys.size();i++)
 513 |     {
 514 |         const cv::KeyPoint &kp = vToDistributeKeys[i];
 515 |         vpIniNodes[kp.pt.x/hX]->vKeys.push_back(kp);
 516 |     }
 517 | 
 518 |     list<ExtractorNode>::iterator lit = lNodes.begin();
 519 | 
 520 |     while(lit!=lNodes.end())
 521 |     {
 522 |         if(lit->vKeys.size()==1)
 523 |         {
 524 |             lit->bNoMore=true;
 525 |             lit++;
 526 |         }
 527 |         else if(lit->vKeys.empty())
 528 |             lit = lNodes.erase(lit);
 529 |         else
 530 |             lit++;
 531 |     }
 532 | 
 533 |     bool bFinish = false;
 534 | 
 535 |     int iteration = 0;
 536 | 
 537 |     vector<pair<int,ExtractorNode*> > vSizeAndPointerToNode;
 538 |     vSizeAndPointerToNode.reserve(lNodes.size()*4);
 539 | 
 540 |     // 根据兴趣点分布,利用N叉树方法对图像进行划分区域
 541 |     while(!bFinish)
 542 |     {
 543 |         iteration++;
 544 | 
 545 |         int prevSize = lNodes.size();
 546 | 
 547 |         lit = lNodes.begin();
 548 | 
 549 |         int nToExpand = 0;
 550 | 
 551 |         vSizeAndPointerToNode.clear();
 552 | 
 553 |         // 将目前的子区域经行划分
 554 |         while(lit!=lNodes.end())
 555 |         {
 556 |             if(lit->bNoMore)
 557 |             {
 558 |                 // If node only contains one point do not subdivide and continue
 559 |                 lit++;
 560 |                 continue;
 561 |             }
 562 |             else
 563 |             {
 564 |                 // If more than one point, subdivide
 565 |                 ExtractorNode n1,n2,n3,n4;
 566 |                 lit->DivideNode(n1,n2,n3,n4); // 再细分成四个子区域
 567 | 
 568 |                 // Add childs if they contain points
 569 |                 if(n1.vKeys.size()>0)
 570 |                 {
 571 |                     lNodes.push_front(n1);                    
 572 |                     if(n1.vKeys.size()>1)
 573 |                     {
 574 |                         nToExpand++;
 575 |                         vSizeAndPointerToNode.push_back(make_pair(n1.vKeys.size(),&lNodes.front()));
 576 |                         lNodes.front().lit = lNodes.begin();
 577 |                     }
 578 |                 }
 579 |                 if(n2.vKeys.size()>0)
 580 |                 {
 581 |                     lNodes.push_front(n2);
 582 |                     if(n2.vKeys.size()>1)
 583 |                     {
 584 |                         nToExpand++;
 585 |                         vSizeAndPointerToNode.push_back(make_pair(n2.vKeys.size(),&lNodes.front()));
 586 |                         lNodes.front().lit = lNodes.begin();
 587 |                     }
 588 |                 }
 589 |                 if(n3.vKeys.size()>0)
 590 |                 {
 591 |                     lNodes.push_front(n3);
 592 |                     if(n3.vKeys.size()>1)
 593 |                     {
 594 |                         nToExpand++;
 595 |                         vSizeAndPointerToNode.push_back(make_pair(n3.vKeys.size(),&lNodes.front()));
 596 |                         lNodes.front().lit = lNodes.begin();
 597 |                     }
 598 |                 }
 599 |                 if(n4.vKeys.size()>0)
 600 |                 {
 601 |                     lNodes.push_front(n4);
 602 |                     if(n4.vKeys.size()>1)
 603 |                     {
 604 |                         nToExpand++;
 605 |                         vSizeAndPointerToNode.push_back(make_pair(n4.vKeys.size(),&lNodes.front()));
 606 |                         lNodes.front().lit = lNodes.begin();
 607 |                     }
 608 |                 }
 609 | 
 610 |                 lit=lNodes.erase(lit);
 611 |                 continue;
 612 |             }
 613 |         }       
 614 | 
 615 |         // Finish if there are more nodes than required features
 616 |         // or all nodes contain just one point
 617 |         if((int)lNodes.size()>=N || (int)lNodes.size()==prevSize)
 618 |         {
 619 |             bFinish = true;
 620 |         }
 621 |         // 当再划分之后所有的Node数大于要求数目时
 622 |         else if(((int)lNodes.size()+nToExpand*3)>N)
 623 |         {
 624 | 
 625 |             while(!bFinish)
 626 |             {
 627 | 
 628 |                 prevSize = lNodes.size();
 629 | 
 630 |                 vector<pair<int,ExtractorNode*> > vPrevSizeAndPointerToNode = vSizeAndPointerToNode;
 631 |                 vSizeAndPointerToNode.clear();
 632 | 
 633 |                 // 对需要划分的部分进行排序, 即对兴趣点数较多的区域进行划分
 634 |                 sort(vPrevSizeAndPointerToNode.begin(),vPrevSizeAndPointerToNode.end());
 635 |                 for(int j=vPrevSizeAndPointerToNode.size()-1;j>=0;j--)
 636 |                 {
 637 |                     ExtractorNode n1,n2,n3,n4;
 638 |                     vPrevSizeAndPointerToNode[j].second->DivideNode(n1,n2,n3,n4);
 639 | 
 640 |                     // Add childs if they contain points
 641 |                     if(n1.vKeys.size()>0)
 642 |                     {
 643 |                         lNodes.push_front(n1);
 644 |                         if(n1.vKeys.size()>1)
 645 |                         {
 646 |                             vSizeAndPointerToNode.push_back(make_pair(n1.vKeys.size(),&lNodes.front()));
 647 |                             lNodes.front().lit = lNodes.begin();
 648 |                         }
 649 |                     }
 650 |                     if(n2.vKeys.size()>0)
 651 |                     {
 652 |                         lNodes.push_front(n2);
 653 |                         if(n2.vKeys.size()>1)
 654 |                         {
 655 |                             vSizeAndPointerToNode.push_back(make_pair(n2.vKeys.size(),&lNodes.front()));
 656 |                             lNodes.front().lit = lNodes.begin();
 657 |                         }
 658 |                     }
 659 |                     if(n3.vKeys.size()>0)
 660 |                     {
 661 |                         lNodes.push_front(n3);
 662 |                         if(n3.vKeys.size()>1)
 663 |                         {
 664 |                             vSizeAndPointerToNode.push_back(make_pair(n3.vKeys.size(),&lNodes.front()));
 665 |                             lNodes.front().lit = lNodes.begin();
 666 |                         }
 667 |                     }
 668 |                     if(n4.vKeys.size()>0)
 669 |                     {
 670 |                         lNodes.push_front(n4);
 671 |                         if(n4.vKeys.size()>1)
 672 |                         {
 673 |                             vSizeAndPointerToNode.push_back(make_pair(n4.vKeys.size(),&lNodes.front()));
 674 |                             lNodes.front().lit = lNodes.begin();
 675 |                         }
 676 |                     }
 677 | 
 678 |                     lNodes.erase(vPrevSizeAndPointerToNode[j].second->lit);
 679 | 
 680 |                     if((int)lNodes.size()>=N)
 681 |                         break;
 682 |                 }
 683 | 
 684 |                 if((int)lNodes.size()>=N || (int)lNodes.size()==prevSize)
 685 |                     bFinish = true;
 686 | 
 687 |             }
 688 |         }
 689 |     }
 690 | 
 691 |     // Retain the best point in each node
 692 |     // 保留每个区域响应值最大的一个兴趣点
 693 |     vector<cv::KeyPoint> vResultKeys;
 694 |     vResultKeys.reserve(nfeatures);
 695 |     for(list<ExtractorNode>::iterator lit=lNodes.begin(); lit!=lNodes.end(); lit++)
 696 |     {
 697 |         vector<cv::KeyPoint> &vNodeKeys = lit->vKeys;
 698 |         cv::KeyPoint* pKP = &vNodeKeys[0];
 699 |         float maxResponse = pKP->response;
 700 | 
 701 |         for(size_t k=1;k<vNodeKeys.size();k++)
 702 |         {
 703 |             if(vNodeKeys[k].response>maxResponse)
 704 |             {
 705 |                 pKP = &vNodeKeys[k];
 706 |                 maxResponse = vNodeKeys[k].response;
 707 |             }
 708 |         }
 709 | 
 710 |         vResultKeys.push_back(*pKP);
 711 |     }
 712 | 
 713 |     return vResultKeys;
 714 | }
 715 | 
 716 | void ORBextractor::ComputeKeyPointsOctTree(vector<vector<KeyPoint> >& allKeypoints)
 717 | {
 718 |     allKeypoints.resize(nlevels);
 719 | 
 720 |     const float W = 30;
 721 | 
 722 |     // 对每一层图像做处理
 723 |     for (int level = 0; level < nlevels; ++level)
 724 |     {
 725 |         const int minBorderX = EDGE_THRESHOLD-3;
 726 |         const int minBorderY = minBorderX;
 727 |         const int maxBorderX = mvImagePyramid[level].cols-EDGE_THRESHOLD+3;
 728 |         const int maxBorderY = mvImagePyramid[level].rows-EDGE_THRESHOLD+3;
 729 | 
 730 |         vector<cv::KeyPoint> vToDistributeKeys;
 731 |         vToDistributeKeys.reserve(nfeatures*10);
 732 | 
 733 |         const float width = (maxBorderX-minBorderX);
 734 |         const float height = (maxBorderY-minBorderY);
 735 | 
 736 |         const int nCols = width/W;
 737 |         const int nRows = height/W;
 738 |         const int wCell = ceil(width/nCols);
 739 |         const int hCell = ceil(height/nRows);
 740 | 
 741 |         for(int i=0; i<nRows; i++)
 742 |         {
 743 |             const float iniY =minBorderY+i*hCell;
 744 |             float maxY = iniY+hCell+6;
 745 | 
 746 |             if(iniY>=maxBorderY-3)
 747 |                 continue;
 748 |             if(maxY>maxBorderY)
 749 |                 maxY = maxBorderY;
 750 | 
 751 |             for(int j=0; j<nCols; j++)
 752 |             {
 753 |                 const float iniX =minBorderX+j*wCell;
 754 |                 float maxX = iniX+wCell+6;
 755 |                 if(iniX>=maxBorderX-6)
 756 |                     continue;
 757 |                 if(maxX>maxBorderX)
 758 |                     maxX = maxBorderX;
 759 | 
 760 |                 // FAST提取兴趣点, 自适应阈值
 761 |                 vector<cv::KeyPoint> vKeysCell;
 762 |                 FAST(mvImagePyramid[level].rowRange(iniY,maxY).colRange(iniX,maxX),
 763 |                      vKeysCell,iniThFAST,true);
 764 | 
 765 |                 if(vKeysCell.empty())
 766 |                 {
 767 |                     FAST(mvImagePyramid[level].rowRange(iniY,maxY).colRange(iniX,maxX),
 768 |                          vKeysCell,minThFAST,true);
 769 |                 }
 770 | 
 771 |                 if(!vKeysCell.empty())
 772 |                 {
 773 |                     for(vector<cv::KeyPoint>::iterator vit=vKeysCell.begin(); vit!=vKeysCell.end();vit++)
 774 |                     {
 775 |                         (*vit).pt.x+=j*wCell;
 776 |                         (*vit).pt.y+=i*hCell;
 777 |                         vToDistributeKeys.push_back(*vit);
 778 |                     }
 779 |                 }
 780 | 
 781 |             }
 782 |         }
 783 | 
 784 |         vector<KeyPoint> & keypoints = allKeypoints[level];
 785 |         keypoints.reserve(nfeatures);
 786 | 
 787 |         // 根据mnFeaturesPerLevel,即该层的兴趣点数,对特征点进行剔除
 788 |         keypoints = DistributeOctTree(vToDistributeKeys, minBorderX, maxBorderX,
 789 |                                       minBorderY, maxBorderY,mnFeaturesPerLevel[level], level);
 790 | 
 791 |         const int scaledPatchSize = PATCH_SIZE*mvScaleFactor[level];
 792 | 
 793 |         // Add border to coordinates and scale information
 794 |         const int nkps = keypoints.size();
 795 |         for(int i=0; i<nkps ; i++)
 796 |         {
 797 |             keypoints[i].pt.x+=minBorderX;
 798 |             keypoints[i].pt.y+=minBorderY;
 799 |             keypoints[i].octave=level;
 800 |             keypoints[i].size = scaledPatchSize;
 801 |         }
 802 |     }
 803 | 
 804 |     // compute orientations
 805 |     for (int level = 0; level < nlevels; ++level)
 806 |         computeOrientation(mvImagePyramid[level], allKeypoints[level], umax);
 807 | }
 808 | 
 809 | void ORBextractor::ComputeKeyPointsOld(std::vector<std::vector<KeyPoint> > &allKeypoints)
 810 | {
 811 |     allKeypoints.resize(nlevels);
 812 | 
 813 |     float imageRatio = (float)mvImagePyramid[0].cols/mvImagePyramid[0].rows;
 814 | 
 815 |     for (int level = 0; level < nlevels; ++level)
 816 |     {
 817 |         const int nDesiredFeatures = mnFeaturesPerLevel[level];
 818 | 
 819 |         const int levelCols = sqrt((float)nDesiredFeatures/(5*imageRatio));
 820 |         const int levelRows = imageRatio*levelCols;
 821 | 
 822 |         const int minBorderX = EDGE_THRESHOLD;
 823 |         const int minBorderY = minBorderX;
 824 |         const int maxBorderX = mvImagePyramid[level].cols-EDGE_THRESHOLD;
 825 |         const int maxBorderY = mvImagePyramid[level].rows-EDGE_THRESHOLD;
 826 | 
 827 |         const int W = maxBorderX - minBorderX;
 828 |         const int H = maxBorderY - minBorderY;
 829 |         const int cellW = ceil((float)W/levelCols);
 830 |         const int cellH = ceil((float)H/levelRows);
 831 | 
 832 |         const int nCells = levelRows*levelCols;
 833 |         const int nfeaturesCell = ceil((float)nDesiredFeatures/nCells);
 834 | 
 835 |         vector<vector<vector<KeyPoint> > > cellKeyPoints(levelRows, vector<vector<KeyPoint> >(levelCols));
 836 | 
 837 |         vector<vector<int> > nToRetain(levelRows,vector<int>(levelCols,0));
 838 |         vector<vector<int> > nTotal(levelRows,vector<int>(levelCols,0));
 839 |         vector<vector<bool> > bNoMore(levelRows,vector<bool>(levelCols,false));
 840 |         vector<int> iniXCol(levelCols);
 841 |         vector<int> iniYRow(levelRows);
 842 |         int nNoMore = 0;
 843 |         int nToDistribute = 0;
 844 | 
 845 | 
 846 |         float hY = cellH + 6;
 847 | 
 848 |         for(int i=0; i<levelRows; i++)
 849 |         {
 850 |             const float iniY = minBorderY + i*cellH - 3;
 851 |             iniYRow[i] = iniY;
 852 | 
 853 |             if(i == levelRows-1)
 854 |             {
 855 |                 hY = maxBorderY+3-iniY;
 856 |                 if(hY<=0)
 857 |                     continue;
 858 |             }
 859 | 
 860 |             float hX = cellW + 6;
 861 | 
 862 |             for(int j=0; j<levelCols; j++)
 863 |             {
 864 |                 float iniX;
 865 | 
 866 |                 if(i==0)
 867 |                 {
 868 |                     iniX = minBorderX + j*cellW - 3;
 869 |                     iniXCol[j] = iniX;
 870 |                 }
 871 |                 else
 872 |                 {
 873 |                     iniX = iniXCol[j];
 874 |                 }
 875 | 
 876 | 
 877 |                 if(j == levelCols-1)
 878 |                 {
 879 |                     hX = maxBorderX+3-iniX;
 880 |                     if(hX<=0)
 881 |                         continue;
 882 |                 }
 883 | 
 884 | 
 885 |                 Mat cellImage = mvImagePyramid[level].rowRange(iniY,iniY+hY).colRange(iniX,iniX+hX);
 886 | 
 887 |                 cellKeyPoints[i][j].reserve(nfeaturesCell*5);
 888 | 
 889 |                 FAST(cellImage,cellKeyPoints[i][j],iniThFAST,true);
 890 | 
 891 |                 if(cellKeyPoints[i][j].size()<=3)
 892 |                 {
 893 |                     cellKeyPoints[i][j].clear();
 894 | 
 895 |                     FAST(cellImage,cellKeyPoints[i][j],minThFAST,true);
 896 |                 }
 897 | 
 898 | 
 899 |                 const int nKeys = cellKeyPoints[i][j].size();
 900 |                 nTotal[i][j] = nKeys;
 901 | 
 902 |                 if(nKeys>nfeaturesCell)
 903 |                 {
 904 |                     nToRetain[i][j] = nfeaturesCell;
 905 |                     bNoMore[i][j] = false;
 906 |                 }
 907 |                 else
 908 |                 {
 909 |                     nToRetain[i][j] = nKeys;
 910 |                     nToDistribute += nfeaturesCell-nKeys;
 911 |                     bNoMore[i][j] = true;
 912 |                     nNoMore++;
 913 |                 }
 914 | 
 915 |             }
 916 |         }
 917 | 
 918 | 
 919 |         // Retain by score
 920 | 
 921 |         while(nToDistribute>0 && nNoMore<nCells)
 922 |         {
 923 |             int nNewFeaturesCell = nfeaturesCell + ceil((float)nToDistribute/(nCells-nNoMore));
 924 |             nToDistribute = 0;
 925 | 
 926 |             for(int i=0; i<levelRows; i++)
 927 |             {
 928 |                 for(int j=0; j<levelCols; j++)
 929 |                 {
 930 |                     if(!bNoMore[i][j])
 931 |                     {
 932 |                         if(nTotal[i][j]>nNewFeaturesCell)
 933 |                         {
 934 |                             nToRetain[i][j] = nNewFeaturesCell;
 935 |                             bNoMore[i][j] = false;
 936 |                         }
 937 |                         else
 938 |                         {
 939 |                             nToRetain[i][j] = nTotal[i][j];
 940 |                             nToDistribute += nNewFeaturesCell-nTotal[i][j];
 941 |                             bNoMore[i][j] = true;
 942 |                             nNoMore++;
 943 |                         }
 944 |                     }
 945 |                 }
 946 |             }
 947 |         }
 948 | 
 949 |         vector<KeyPoint> & keypoints = allKeypoints[level];
 950 |         keypoints.reserve(nDesiredFeatures*2);
 951 | 
 952 |         const int scaledPatchSize = PATCH_SIZE*mvScaleFactor[level];
 953 | 
 954 |         // Retain by score and transform coordinates
 955 |         for(int i=0; i<levelRows; i++)
 956 |         {
 957 |             for(int j=0; j<levelCols; j++)
 958 |             {
 959 |                 vector<KeyPoint> &keysCell = cellKeyPoints[i][j];
 960 |                 KeyPointsFilter::retainBest(keysCell,nToRetain[i][j]);
 961 |                 if((int)keysCell.size()>nToRetain[i][j])
 962 |                     keysCell.resize(nToRetain[i][j]);
 963 | 
 964 | 
 965 |                 for(size_t k=0, kend=keysCell.size(); k<kend; k++)
 966 |                 {
 967 |                     keysCell[k].pt.x+=iniXCol[j];
 968 |                     keysCell[k].pt.y+=iniYRow[i];
 969 |                     keysCell[k].octave=level;
 970 |                     keysCell[k].size = scaledPatchSize;
 971 |                     keypoints.push_back(keysCell[k]);
 972 |                 }
 973 |             }
 974 |         }
 975 | 
 976 |         if((int)keypoints.size()>nDesiredFeatures)
 977 |         {
 978 |             KeyPointsFilter::retainBest(keypoints,nDesiredFeatures);
 979 |             keypoints.resize(nDesiredFeatures);
 980 |         }
 981 |     }
 982 | 
 983 |     // and compute orientations
 984 |     for (int level = 0; level < nlevels; ++level)
 985 |         computeOrientation(mvImagePyramid[level], allKeypoints[level], umax);
 986 | }
 987 | 
 988 | static void computeDescriptors(const Mat& image, vector<KeyPoint>& keypoints, Mat& descriptors,
 989 |                                const vector<Point>& pattern)
 990 | {
 991 |     descriptors = Mat::zeros((int)keypoints.size(), 32, CV_8UC1);
 992 | 
 993 |     for (size_t i = 0; i < keypoints.size(); i++)
 994 |         computeOrbDescriptor(keypoints[i], image, &pattern[0], descriptors.ptr((int)i));
 995 | }
 996 | 
 997 | void ORBextractor::operator()( InputArray _image, InputArray _mask, vector<KeyPoint>& _keypoints,
 998 |                       OutputArray _descriptors)
 999 | { 
1000 |     if(_image.empty())
1001 |         return;
1002 | 
1003 |     Mat image = _image.getMat();
1004 |     assert(image.type() == CV_8UC1 );
1005 | 
1006 |     // Pre-compute the scale pyramid
1007 |     // 构建图像金字塔
1008 |     ComputePyramid(image);
1009 | 
1010 |     // 计算每层图像的兴趣点
1011 |     vector < vector<KeyPoint> > allKeypoints; // vector<vector<KeyPoint>>
1012 |     ComputeKeyPointsOctTree(allKeypoints);
1013 |     //ComputeKeyPointsOld(allKeypoints);
1014 | 
1015 |     Mat descriptors;
1016 | 
1017 |     int nkeypoints = 0;
1018 |     for (int level = 0; level < nlevels; ++level)
1019 |         nkeypoints += (int)allKeypoints[level].size();
1020 |     if( nkeypoints == 0 )
1021 |         _descriptors.release();
1022 |     else
1023 |     {
1024 |         _descriptors.create(nkeypoints, 32, CV_8U);
1025 |         descriptors = _descriptors.getMat();
1026 |     }
1027 | 
1028 |     _keypoints.clear();
1029 |     _keypoints.reserve(nkeypoints);
1030 | 
1031 |     int offset = 0;
1032 |     for (int level = 0; level < nlevels; ++level)
1033 |     {
1034 |         vector<KeyPoint>& keypoints = allKeypoints[level];
1035 |         int nkeypointsLevel = (int)keypoints.size();
1036 | 
1037 |         if(nkeypointsLevel==0)
1038 |             continue;
1039 | 
1040 |         // preprocess the resized image 对图像进行高斯模糊
1041 |         Mat workingMat = mvImagePyramid[level].clone();
1042 |         GaussianBlur(workingMat, workingMat, Size(7, 7), 2, 2, BORDER_REFLECT_101);
1043 | 
1044 |         // Compute the descriptors 计算描述子
1045 |         Mat desc = descriptors.rowRange(offset, offset + nkeypointsLevel);
1046 |         computeDescriptors(workingMat, keypoints, desc, pattern);
1047 | 
1048 |         offset += nkeypointsLevel;
1049 | 
1050 |         // Scale keypoint coordinates
1051 |         if (level != 0)
1052 |         {
1053 |             float scale = mvScaleFactor[level]; //getScale(level, firstLevel, scaleFactor);
1054 |             for (vector<KeyPoint>::iterator keypoint = keypoints.begin(),
1055 |                  keypointEnd = keypoints.end(); keypoint != keypointEnd; ++keypoint)
1056 |                 keypoint->pt *= scale;
1057 |         }
1058 |         // And add the keypoints to the output
1059 |         _keypoints.insert(_keypoints.end(), keypoints.begin(), keypoints.end());
1060 |     }
1061 | }
1062 | 
1063 | /**
1064 |  * 构建图像金字塔
1065 |  * @param image 输入图像
1066 |  */
1067 | void ORBextractor::ComputePyramid(cv::Mat image)
1068 | {
1069 |     for (int level = 0; level < nlevels; ++level)
1070 |     {
1071 |         float scale = mvInvScaleFactor[level];
1072 |         Size sz(cvRound((float)image.cols*scale), cvRound((float)image.rows*scale));
1073 |         Size wholeSize(sz.width + EDGE_THRESHOLD*2, sz.height + EDGE_THRESHOLD*2);
1074 |         Mat temp(wholeSize, image.type()), masktemp;
1075 |         mvImagePyramid[level] = temp(Rect(EDGE_THRESHOLD, EDGE_THRESHOLD, sz.width, sz.height));
1076 | 
1077 |         // Compute the resized image
1078 |         if( level != 0 )
1079 |         {
1080 |             resize(mvImagePyramid[level-1], mvImagePyramid[level], sz, 0, 0, cv::INTER_LINEAR);
1081 | 
1082 |             copyMakeBorder(mvImagePyramid[level], temp, EDGE_THRESHOLD, EDGE_THRESHOLD, EDGE_THRESHOLD, EDGE_THRESHOLD,
1083 |                            BORDER_REFLECT_101+BORDER_ISOLATED);            
1084 |         }
1085 |         else
1086 |         {
1087 |             copyMakeBorder(image, temp, EDGE_THRESHOLD, EDGE_THRESHOLD, EDGE_THRESHOLD, EDGE_THRESHOLD,
1088 |                            BORDER_REFLECT_101);            
1089 |         }
1090 |     }
1091 | 
1092 | }
1093 | 
1094 | } //namespace ORB_SLAM
1095 | 


--------------------------------------------------------------------------------
/src/main.cpp:
--------------------------------------------------------------------------------
  1 | #include "ORBextractor.h"
  2 | #include <fstream>
  3 | #include <iostream>
  4 | #include <opencv2/features2d.hpp>
  5 | #include <opencv2/imgproc.hpp>
  6 | #include <opencv2/highgui/highgui.hpp>
  7 | using namespace cv;
  8 | using namespace std;
  9 | using namespace ORB_SLAM2;
 10 | void GrabImage(cv::Mat &mImGray, int mbRGB)
 11 | {
 12 |     if(mImGray.channels()==3)
 13 |     {
 14 |         if(mbRGB)
 15 |         {
 16 |             cvtColor(mImGray,mImGray,CV_RGB2GRAY);
 17 |         }
 18 |         else
 19 |         {
 20 |             cvtColor(mImGray,mImGray,CV_BGR2GRAY);
 21 |         }
 22 |     }
 23 |     else if(mImGray.channels()==4)
 24 |     {
 25 |         if(mbRGB)
 26 |         {
 27 |             cvtColor(mImGray,mImGray,CV_RGBA2GRAY);
 28 |         }
 29 |         else
 30 |         {
 31 |             cvtColor(mImGray,mImGray,CV_BGRA2GRAY);
 32 |         }
 33 |     }
 34 | 
 35 | }
 36 | int main(int argc, char **argv)
 37 | {
 38 |     if(argc != 3)
 39 |     {
 40 |         cerr << endl << "Usage: ./ORBextractor path_to_a_picture named liu.png path to orb.yaml" << endl;
 41 |         return 1;
 42 |     }
 43 |     cout << "liu";
 44 |     string strSettingPath = argv[2];
 45 | 
 46 |     // 载入参数文件
 47 |     cv::FileStorage fSettings(strSettingPath, cv::FileStorage::READ);
 48 | 
 49 |     // 1:RGB 0:BGR
 50 |     int nRGB = fSettings["Camera.RGB"];
 51 | 
 52 |     if(nRGB)
 53 |         cout << "- color order: RGB (ignored if grayscale)" << endl;
 54 |     else
 55 |         cout << "- color order: BGR (ignored if grayscale)" << endl;
 56 | 
 57 |     // 每一帧提取的特征点数 1000
 58 |     int nFeatures = fSettings["ORBextractor.nFeatures"];
 59 |     // 图像建立金字塔时的变化尺度 1.2
 60 |     float fScaleFactor = fSettings["ORBextractor.scaleFactor"];
 61 |     // 尺度金字塔的层数 8
 62 |     int nLevels = fSettings["ORBextractor.nLevels"];
 63 |     // 提取fast特征点的默认阈值 20
 64 |     int fIniThFAST = fSettings["ORBextractor.iniThFAST"];
 65 |     // 如果默认阈值提取不出足够fast特征点，则使用最小阈值 8
 66 |     int fMinThFAST = fSettings["ORBextractor.minThFAST"];
 67 |     cout << endl  << "ORB Extractor Parameters: " << endl;
 68 |     cout << "- Number of Features: " << nFeatures << endl;
 69 |     cout << "- Scale Levels: " << nLevels << endl;
 70 |     cout << "- Scale Factor: " << fScaleFactor << endl;
 71 |     cout << "- Initial Fast Threshold: " << fIniThFAST << endl;
 72 |     cout << "- Minimum Fast Threshold: " << fMinThFAST << endl;
 73 |     
 74 | //******************************************************************************************************************************************
 75 |     // 初始化mpORBextractorLeft作为特征点提取器  
 76 |     ORBextractor* mpORBextractorLeft = new ORBextractor(nFeatures,fScaleFactor,nLevels,fIniThFAST,fMinThFAST);
 77 |    
 78 |     // 载入原图
 79 |     cv::Mat imLeft;   
 80 |     imLeft = cv::imread(argv[1],CV_LOAD_IMAGE_UNCHANGED);
 81 |     if( imLeft.empty() )                      
 82 |     {
 83 |         cout <<  "Could not open or find the image" << std::endl ;
 84 |         return -1;
 85 |     }
 86 |     // 转为灰度图
 87 |     cv::Mat &mImGray = imLeft;
 88 |     GrabImage(mImGray,nRGB);
 89 |  
 90 |     // 当前帧图像中提取的特征点集合
 91 |     std::vector<cv::KeyPoint> mvKeys;
 92 |     // 特征点对应的描述子
 93 |     cv::Mat mDescriptors;
 94 |     (*mpORBextractorLeft)(mImGray,cv::Mat(),mvKeys,mDescriptors);
 95 |     
 96 |     // diaplay
 97 |     Mat out1;
 98 |     //drawKeypoints(imLeft, mvKeys, out1, Scalar::all(-1), DrawMatchesFlags::DEFAULT); 
 99 |     drawKeypoints(imLeft, mvKeys, out1, Scalar::all(-1), DrawMatchesFlags::DRAW_RICH_KEYPOINTS); 
100 |     namedWindow("orb_features", WINDOW_NORMAL);
101 |     imshow("orb_features", out1);
102 |     waitKey(0);
103 |     imwrite("orb_features.jpg", out1);
104 |     return 0;
105 | 
106 | }    


--------------------------------------------------------------------------------