グレイコードパターンのデコード

目標

このチュートリアルでは、GrayCodePattern クラスを使って次のことを行う方法を学ぶ:

• 以前に取得したグレイコードパターンをデコードする。
• 視差マップを生成する。
• 点群 (pointcloud) を生成する。

コード

/*M///////////////////////////////////////////////////////////////////////////////////////
 //
 //  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
 //  copy or use the software.
 //
 //
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2015, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
 //   * Redistribution's of source code must retain the above copyright notice,
 //     this list of conditions and the following disclaimer.
 //
 //   * Redistribution's 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.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
 // (including, but not limited to, procurement of substitute goods or services;
 // loss of use, data, or profits; or business interruption) however caused
 // and on any theory of liability, whether in contract, strict liability,
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
 
#include <iostream>
#include <opencv2/core.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/calib3d.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/structured_light.hpp>
#include <opencv2/opencv_modules.hpp>
 
// (if you did not build the opencv_viz module, you will only see the disparity images)
#ifdef HAVE_OPENCV_VIZ
#include <opencv2/viz.hpp>
#endif
 
using namespace std;
using namespace cv;
 
static const char* keys =
{ "{@images_list | | Image list where the captured pattern images are saved}"
 "{@calib_param_path     | | Calibration_parameters            }"
 "{@proj_width      | | The projector width used to acquire the pattern          }"
 "{@proj_height     | | The projector height used to acquire the pattern}"
 "{@white_thresh     | | The white threshold height (optional)}"
 "{@black_thresh     | | The black threshold (optional)}" };
 
static void help()
{
  cout << "\nThis example shows how to use the \"Structured Light module\" to decode a previously acquired gray code pattern, generating a pointcloud"
 "\nCall:\n"
 "./example_structured_light_pointcloud <images_list> <calib_param_path> <proj_width> <proj_height> <white_thresh> <black_thresh>\n"
        << endl;
}
 
static bool readStringList( const string& filename, vector<string>& l )
{
  l.resize( 0 );
 FileStorage fs( filename, FileStorage::READ );
 if( !fs.isOpened() )
  {
    cerr << "failed to open " << filename << endl;
 return false;
  }
 FileNode n = fs.getFirstTopLevelNode();
 if( n.type() != FileNode::SEQ )
  {
    cerr << "cam 1 images are not a sequence! FAIL" << endl;
 return false;
  }
 
 FileNodeIterator it = n.begin(), it_end = n.end();
 for( ; it != it_end; ++it )
  {
    l.push_back( ( string ) *it );
  }
 
  n = fs["cam2"];
 if( n.type() != FileNode::SEQ )
  {
    cerr << "cam 2 images are not a sequence! FAIL" << endl;
 return false;
  }
 
  it = n.begin(), it_end = n.end();
 for( ; it != it_end; ++it )
  {
    l.push_back( ( string ) *it );
  }
 
 if( l.size() % 2 != 0 )
  {
    cout << "Error: the image list contains odd (non-even) number of elements\n";
 return false;
  }
 return true;
}
 
int main( int argc, char** argv )
{
 structured_light::GrayCodePattern::Params params;
 CommandLineParser parser( argc, argv, keys );
 String images_file = parser.get<String>( 0 );
 String calib_file = parser.get<String>( 1 );
 
 params.width = parser.get<int>( 2 );
 params.height = parser.get<int>( 3 );
 
 if( images_file.empty() || calib_file.empty() || params.width < 1 || params.height < 1 || argc < 5 || argc > 7 )
  {
    help();
 return -1;
  }
 
 // Set up GraycodePattern with params
 Ptr<structured_light::GrayCodePattern> graycode = structured_light::GrayCodePattern::create( params );
 size_t white_thresh = 0;
 size_t black_thresh = 0;
 
 if( argc == 7 )
  {
 // If passed, setting the white and black threshold, otherwise using default values
    white_thresh = parser.get<unsigned>( 4 );
    black_thresh = parser.get<unsigned>( 5 );
 
    graycode->setWhiteThreshold( white_thresh );
    graycode->setBlackThreshold( black_thresh );
  }
 
  vector<string> imagelist;
 bool ok = readStringList( images_file, imagelist );
 if( !ok || imagelist.empty() )
  {
    cout << "can not open " << images_file << " or the string list is empty" << endl;
    help();
 return -1;
  }
 
 FileStorage fs( calib_file, FileStorage::READ );
 if( !fs.isOpened() )
  {
    cout << "Failed to open Calibration Data File." << endl;
    help();
 return -1;
  }
 
 // Loading calibration parameters
 Mat cam1intrinsics, cam1distCoeffs, cam2intrinsics, cam2distCoeffs, R, T;
  fs["cam1_intrinsics"] >> cam1intrinsics;
  fs["cam2_intrinsics"] >> cam2intrinsics;
  fs["cam1_distorsion"] >> cam1distCoeffs;
  fs["cam2_distorsion"] >> cam2distCoeffs;
  fs["R"] >> R;
  fs["T"] >> T;
 
  cout << "cam1intrinsics" << endl << cam1intrinsics << endl;
  cout << "cam1distCoeffs" << endl << cam1distCoeffs << endl;
  cout << "cam2intrinsics" << endl << cam2intrinsics << endl;
  cout << "cam2distCoeffs" << endl << cam2distCoeffs << endl;
  cout << "T" << endl << T << endl << "R" << endl << R << endl;
 
 if( (!R.data) || (!T.data) || (!cam1intrinsics.data) || (!cam2intrinsics.data) || (!cam1distCoeffs.data) || (!cam2distCoeffs.data) )
  {
    cout << "Failed to load cameras calibration parameters" << endl;
    help();
 return -1;
  }
 
 size_t numberOfPatternImages = graycode->getNumberOfPatternImages();
  vector<vector<Mat> > captured_pattern;
  captured_pattern.resize( 2 );
  captured_pattern[0].resize( numberOfPatternImages );
  captured_pattern[1].resize( numberOfPatternImages );
 
 Mat color = imread( imagelist[numberOfPatternImages], IMREAD_COLOR );
 Size imagesSize = color.size();
 
 // Stereo rectify
  cout << "Rectifying images..." << endl;
 Mat R1, R2, P1, P2, Q;
 Rect validRoi[2];
 stereoRectify( cam1intrinsics, cam1distCoeffs, cam2intrinsics, cam2distCoeffs, imagesSize, R, T, R1, R2, P1, P2, Q, 0,
                -1, imagesSize, &validRoi[0], &validRoi[1] );
 
 Mat map1x, map1y, map2x, map2y;
 initUndistortRectifyMap( cam1intrinsics, cam1distCoeffs, R1, P1, imagesSize, CV_32FC1, map1x, map1y );
 initUndistortRectifyMap( cam2intrinsics, cam2distCoeffs, R2, P2, imagesSize, CV_32FC1, map2x, map2y );
 
 // Loading pattern images
 for( size_t i = 0; i < numberOfPatternImages; i++ )
  {
    captured_pattern[0][i] = imread( imagelist[i], IMREAD_GRAYSCALE );
    captured_pattern[1][i] = imread( imagelist[i + numberOfPatternImages + 2], IMREAD_GRAYSCALE );
 
 if( (!captured_pattern[0][i].data) || (!captured_pattern[1][i].data) )
    {
      cout << "Empty images" << endl;
      help();
 return -1;
    }
 
 remap( captured_pattern[1][i], captured_pattern[1][i], map1x, map1y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 remap( captured_pattern[0][i], captured_pattern[0][i], map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 
  }
  cout << "done" << endl;
 
  vector<Mat> blackImages;
  vector<Mat> whiteImages;
 
  blackImages.resize( 2 );
  whiteImages.resize( 2 );
 
 // Loading images (all white + all black) needed for shadows computation
 cvtColor( color, whiteImages[0], COLOR_RGB2GRAY );
 
  whiteImages[1] = imread( imagelist[2 * numberOfPatternImages + 2], IMREAD_GRAYSCALE );
  blackImages[0] = imread( imagelist[numberOfPatternImages + 1], IMREAD_GRAYSCALE );
  blackImages[1] = imread( imagelist[2 * numberOfPatternImages + 2 + 1], IMREAD_GRAYSCALE );
 
 remap( color, color, map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 
 remap( whiteImages[0], whiteImages[0], map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 remap( whiteImages[1], whiteImages[1], map1x, map1y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 
 remap( blackImages[0], blackImages[0], map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 remap( blackImages[1], blackImages[1], map1x, map1y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
 
  cout << endl << "Decoding pattern ..." << endl;
 Mat disparityMap;
 bool decoded = graycode->decode( captured_pattern, disparityMap, blackImages, whiteImages,
                                  structured_light::DECODE_3D_UNDERWORLD );
 if( decoded )
  {
    cout << endl << "pattern decoded" << endl;
 
 // To better visualize the result, apply a colormap to the computed disparity
 double min;
 double max;
 minMaxIdx(disparityMap, &min, &max);
 Mat cm_disp, scaledDisparityMap;
    cout << "disp min " << min << endl << "disp max " << max << endl;
 convertScaleAbs( disparityMap, scaledDisparityMap, 255 / ( max - min ) );
 applyColorMap( scaledDisparityMap, cm_disp, COLORMAP_JET );
 
 // Show the result
 resize( cm_disp, cm_disp, Size( 640, 480 ), 0, 0, INTER_LINEAR_EXACT );
 imshow( "cm disparity m", cm_disp );
 
 // Compute the point cloud
 Mat pointcloud;
    disparityMap.convertTo( disparityMap, CV_32FC1 );
 reprojectImageTo3D( disparityMap, pointcloud, Q, true, -1 );
 
 // Compute a mask to remove background
 Mat dst, thresholded_disp;
 threshold( scaledDisparityMap, thresholded_disp, 0, 255, THRESH_OTSU + THRESH_BINARY );
 resize( thresholded_disp, dst, Size( 640, 480 ), 0, 0, INTER_LINEAR_EXACT );
 imshow( "threshold disp otsu", dst );
 
#ifdef HAVE_OPENCV_VIZ
 // Apply the mask to the point cloud
 Mat pointcloud_tresh, color_tresh;
    pointcloud.copyTo( pointcloud_tresh, thresholded_disp );
    color.copyTo( color_tresh, thresholded_disp );
 
 // Show the point cloud on viz
 viz::Viz3d myWindow( "Point cloud with color" );
    myWindow.setBackgroundMeshLab();
    myWindow.showWidget( "coosys", viz::WCoordinateSystem() );
    myWindow.showWidget( "pointcloud", viz::WCloud( pointcloud_tresh, color_tresh ) );
    myWindow.showWidget( "text2d", viz::WText( "Point cloud", Point(20, 20), 20, viz::Color::green() ) );
    myWindow.spin();
#endif // HAVE_OPENCV_VIZ
 
  }
 
 waitKey();
 return 0;
}

解説

まず初めに、必要な引数をプログラムに渡す必要がある。1 つ目は、以前に取得したパターン画像の名前リストで、以下のように構成された .yaml ファイルに格納されている:

%YAML:1.0
cam1:
   - "/data/pattern_cam1_im1.png"
   - "/data/pattern_cam1_im2.png"
           ..............
   - "/data/pattern_cam1_im42.png"
   - "/data/pattern_cam1_im43.png"
   - "/data/pattern_cam1_im44.png"
cam2:
   - "/data/pattern_cam2_im1.png"
   - "/data/pattern_cam2_im2.png"
           ..............
   - "/data/pattern_cam2_im42.png"
   - "/data/pattern_cam2_im43.png"
   - "/data/pattern_cam2_im44.png"

例えば、このチュートリアルで使用したデータセットは、解像度 1280x800 のプロジェクタを使用して取得されたものであり、42 枚のパターン画像 (番号 1 から 42 まで) + 1 枚の白 (番号 43) と 1 枚の黒 (番号 44) が、2 台のカメラの両方で撮影された。

次に、別の .yml ファイルに格納されたカメラのキャリブレーション引数を、パターンの投影に使用したプロジェクタの幅と高さ、および、省略可能だが白と黒のしきい値とともに、チュートリアルプログラムに渡す必要がある。

このようにして、GrayCodePattern クラスの引数を、パターン取得時に使用したプロジェクタの幅と高さで設定でき、GrayCodePattern オブジェクトへのポインタを作成できる:

structured_light::GrayCodePattern::Params params;
   ....
params.width = parser.get<int>( 2 );
params.height = parser.get<int>( 3 );
   ....
// Set up GraycodePattern with params
Ptr<structured_light::GrayCodePattern> graycode = structured_light::GrayCodePattern::create( params );

白と黒のしきい値が引数として渡された場合 (これらのしきい値はデコードされるピクセル数に影響する)、その値を設定できる。そうでなければ、アルゴリズムはデフォルト値を使用する。

size_t white_thresh = 0;
size_t black_thresh = 0;
if( argc == 7 )
{
 // If passed, setting the white and black threshold, otherwise using default values
  white_thresh = parser.get<size_t>( 4 );
  black_thresh = parser.get<size_t>( 5 );
  graycode->setWhiteThreshold( white_thresh );
  graycode->setBlackThreshold( black_thresh );
}

この時点で、GrayCodePattern クラスの decode メソッドを使用するには、取得したパターン画像を Mat のベクトルのベクトルに格納しなければならない。外側のベクトルはカメラが 2 台あるためサイズが 2 である。1 つ目のベクトルは左カメラから撮影したパターン画像を格納し、2 つ目は右カメラで取得したものを格納する。パターン画像の数は当然両カメラで同じであり、getNumberOfPatternImages() メソッドを使用して取得できる:

size_t numberOfPatternImages = graycode->getNumberOfPatternImages();
vector<vector<Mat> > captured_pattern;
captured_pattern.resize( 2 );
captured_pattern[0].resize( numberOfPatternImages );
captured_pattern[1].resize( numberOfPatternImages );
 
.....
 
for( size_t i = 0; i < numberOfPatternImages; i++ )
{
  captured_pattern[0][i] = imread( imagelist[i], IMREAD_GRAYSCALE );
  captured_pattern[1][i] = imread( imagelist[i + numberOfPatternImages + 2], IMREAD_GRAYSCALE );
   ......
}

白黒画像については、Mat の 2 つの異なるベクトルに格納しなければならない:

vector<Mat> blackImages;
vector<Mat> whiteImages;
blackImages.resize( 2 );
whiteImages.resize( 2 );
// Loading images (all white + all black) needed for shadows computation
cvtColor( color, whiteImages[0], COLOR_RGB2GRAY );
whiteImages[1] = imread( imagelist[2 * numberOfPatternImages + 2], IMREAD_GRAYSCALE );
blackImages[0] = imread( imagelist[numberOfPatternImages + 1], IMREAD_GRAYSCALE );
blackImages[1] = imread( imagelist[2 * numberOfPatternImages + 2 + 1], IMREAD_GRAYSCALE );

強調しておくべき重要な点として、パターン画像、白、黒のすべての画像は、グレースケール画像として読み込み、decode メソッドに渡す前にレクティファイ (平行化) しておく必要がある:

// Stereo rectify
cout << "Rectifying images..." << endl;
Mat R1, R2, P1, P2, Q;
Rect validRoi[2];
stereoRectify( cam1intrinsics, cam1distCoeffs, cam2intrinsics, cam2distCoeffs, imagesSize, R, T, R1, R2, P1, P2, Q, 0,
              -1, imagesSize, &validRoi[0], &validRoi[1] );
Mat map1x, map1y, map2x, map2y;
initUndistortRectifyMap( cam1intrinsics, cam1distCoeffs, R1, P1, imagesSize, CV_32FC1, map1x, map1y );
initUndistortRectifyMap( cam2intrinsics, cam2distCoeffs, R2, P2, imagesSize, CV_32FC1, map2x, map2y );
       ........
for( size_t i = 0; i < numberOfPatternImages; i++ )
{
       ........
  remap( captured_pattern[1][i], captured_pattern[1][i], map1x, map1y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
  remap( captured_pattern[0][i], captured_pattern[0][i], map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
}
       ........
remap( color, color, map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
remap( whiteImages[0], whiteImages[0], map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
remap( whiteImages[1], whiteImages[1], map1x, map1y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
remap( blackImages[0], blackImages[0], map2x, map2y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );
remap( blackImages[1], blackImages[1], map1x, map1y, INTER_NEAREST, BORDER_CONSTANT, Scalar() );

このようにして、decode メソッドを呼び出してパターンをデコードし、1 台目のカメラ (左) で計算された対応する視差マップを生成できる:

Mat disparityMap;
bool decoded = graycode->decode(captured_pattern, disparityMap, blackImages, whiteImages,
                                structured_light::DECODE_3D_UNDERWORLD);

結果をより見やすくするために、計算された視差にカラーマップを適用する:

double min;
double max;
minMaxIdx(disparityMap, &min, &max);
Mat cm_disp, scaledDisparityMap;
cout << "disp min " << min << endl << "disp max " << max << endl;
convertScaleAbs( disparityMap, scaledDisparityMap, 255 / ( max - min ) );
applyColorMap( scaledDisparityMap, cm_disp, COLORMAP_JET );
// Show the result
resize( cm_disp, cm_disp, Size( 640, 480 ) );
imshow( "cm disparity m", cm_disp )

この時点で、reprojectImageTo3D メソッドを使用して点群を生成できる。計算された視差を CV_32FC1 の Mat に変換するように注意すること (decode メソッドは CV_64FC1 の視差マップを計算する):

Mat pointcloud;
disparityMap.convertTo( disparityMap, CV_32FC1 );
reprojectImageTo3D( disparityMap, pointcloud, Q, true, -1 );

次に、不要な背景を除去するためのマスクを計算する:

Mat dst, thresholded_disp;
threshold( scaledDisparityMap, thresholded_disp, 0, 255, THRESH_OTSU + THRESH_BINARY );
resize( thresholded_disp, dst, Size( 640, 480 ) );
imshow( "threshold disp otsu", dst );

cam1 の白画像は、再構築された点群に物体の色をマッピングするために、以前にカラー画像としても読み込まれていた:

Mat color = imread( imagelist[numberOfPatternImages], IMREAD_COLOR );

こうして、背景除去マスクが点群とカラー画像に適用される:

Mat pointcloud_tresh, color_tresh;
pointcloud.copyTo(pointcloud_tresh, thresholded_disp);
color.copyTo(color_tresh, thresholded_disp);

最後に、スキャンされた物体の計算された点群を viz 上で可視化できる:

viz::Viz3d myWindow( "Point cloud with color");
myWindow.setBackgroundMeshLab();
myWindow.showWidget( "coosys", viz::WCoordinateSystem());
myWindow.showWidget( "pointcloud", viz::WCloud( pointcloud_tresh, color_tresh ) );
myWindow.showWidget( "text2d", viz::WText( "Point cloud", Point(20, 20), 20, viz::Color::green() ) );
myWindow.spin();