Delphi-OpenCV/samples/LibDemo/cvExtractSURF/cv_ExtractSURF.dpr

// *****************************************************************
// Delphi-OpenCV Demo
// Copyright (C) 2013 Project Delphi-OpenCV
// ****************************************************************
// Contributor:
  // Laentir Valetov
// email:laex@bk.ru
// ****************************************************************
// You may retrieve the latest version of this file at the GitHub,
// located at git://github.com/Laex/Delphi-OpenCV.git
// ****************************************************************
// The contents of this file are used with permission, subject to
// the Mozilla Public License Version 1.1 (the "License"); you may
// not use this file except in compliance with the License. You may
// obtain a copy of the License at
// http://www.mozilla.org/MPL/MPL-1_1Final.html
//
// Software distributed under the License is distributed on an
// "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
// implied. See the License for the specific language governing
// rights and limitations under the License.
// ***************************************************************
// Original file:
// http://blog.vidikon.com/?p=213
// ***************************************************************

program cv_ExtractSURF;

{$APPTYPE CONSOLE}
{$POINTERMATH ON}
{$R *.res}

uses
  System.SysUtils,
  Winapi.Windows,
  ocv.highgui_c,
  ocv.core_c,
  ocv.core.types_c,
  ocv.imgproc_c,
  ocv.imgproc.types_c,
  ocv.compat,
  ocv.calib3d_c,
  ocv.nonfree,
  uResourcePaths;

// c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD>
// comparison of the two features
function compareSURFDescriptors(const d1: PSingle; const d2: PSingle; best: Double; length: Integer): Double;
var
  total_cost: Double;
  i: Integer;
  t0, t1, t2, t3: Double;
begin
  total_cost := 0;
  assert(length mod 4 = 0);
  i := 0;
  While i < length - 1 do
  begin
    t0 := d1[i] - d2[i];
    t1 := d1[i + 1] - d2[i + 1];
    t2 := d1[i + 2] - d2[i + 2];
    t3 := d1[i + 3] - d2[i + 3];
    total_cost := total_cost + t0 * t0 + t1 * t1 + t2 * t2 + t3 * t3;
    if (total_cost > best) then
      break;
    i := i + 4;
  end;
  result := total_cost;
end;

// c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<> <20>c<EFBFBD><63><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD>
// compares one feature of the scene with all the features
function naiveNearestNeighbor(const vec: PSingle; laplacian: Integer; const model_keypoints: pCvSeq;
  const model_descriptors: pCvSeq): Integer;
Var
  length, i, neighbor: Integer;
  d, dist1, dist2: Double;
  reader, kreader: TCvSeqReader;
  kp: pCvSURFPoint;
  mvec: PSingle;
begin
  length := model_descriptors.elem_size div sizeof(single);
  neighbor := -1;
  dist1 := 1E6;
  dist2 := 1E6;;
  // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63> c<><63><EFBFBD><EFBFBD>
  // The initial feature scenes
  cvStartReadSeq(model_keypoints, @kreader, 0);
  cvStartReadSeq(model_descriptors, @reader, 0);
  // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD> c<><63><EFBFBD><EFBFBD>
  // Iterating through all features of the scene
  for i := 0 to model_descriptors.total - 1 do
  begin
    kp := pCvSURFPoint(kreader.ptr);
    mvec := PSingle(reader.ptr);
    CV_NEXT_SEQ_ELEM(kreader.seq.elem_size, kreader);
    CV_NEXT_SEQ_ELEM(reader.seq.elem_size, reader);
    // <20><><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD> <20><><EFBFBD><EFBFBD>cc<63><63><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD>
    // To accelerate the first compared Laplacian features
    if (laplacian <> kp.laplacian) then
      continue;
    // c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD>
    // comparison of the features
    d := compareSURFDescriptors(vec, mvec, dist2, length);
    if (d < dist1) then
    begin
      // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD>
      // Found a better match features
      dist2 := dist1;
      dist1 := d;
      neighbor := i;
    end

    else if (d < dist2) then
      dist2 := d;
  end;
  if (dist1 < 0.6 * dist2) then
    Exit(neighbor);
  result := -1;
end;

// <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD>
// Function searches for matching pairs
procedure findPairs(const objectKeypoints: pCvSeq; const objectDescriptors: pCvSeq; const imageKeypoints: pCvSeq;
  const imageDescriptors: pCvSeq; Var ptpairs: TArray<Integer>);
var
  i: Integer;
  reader, kreader: TCvSeqReader;
  kp: pCvSURFPoint;
  descriptor: PSingle;
  nearest_neighbor: Integer;
begin
  // <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>cc<63><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // Sets the initial features of object recognition
  cvStartReadSeq(objectKeypoints, @kreader);
  cvStartReadSeq(objectDescriptors, @reader);
  SetLength(ptpairs, 0);
  // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // Iterating through all features of the object
  for i := 0 to objectDescriptors.total - 1 do
  begin
    kp := pCvSURFPoint(kreader.ptr);
    descriptor := PSingle(reader.ptr);
    CV_NEXT_SEQ_ELEM(kreader.seq.elem_size, kreader);
    CV_NEXT_SEQ_ELEM(reader.seq.elem_size, reader);
    // c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63> c<> <20>c<EFBFBD><63><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD> <20><> c<><63><EFBFBD><EFBFBD>
    // comparison of the current features with all the features of the scene
    nearest_neighbor := naiveNearestNeighbor(descriptor, kp.laplacian, imageKeypoints, imageDescriptors);
    if (nearest_neighbor >= 0) then
    begin
      // <20><><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD>
      // Match the features found
      SetLength(ptpairs, length(ptpairs) + 2);
      ptpairs[High(ptpairs) - 1] := i;
      ptpairs[High(ptpairs)] := nearest_neighbor;
    end;
  end;
end;

// <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
// Finding rough position of the object
function locatePlanarObject(const objectKeypoints: pCvSeq; const objectDescriptors: pCvSeq; const imageKeypoints: pCvSeq;
  const imageDescriptors: pCvSeq; const src_corners: TArray<TCvPoint>; dst_corners: TArray<TCvPoint>): Integer;

var
  h: array [0 .. 8] of Double;
  _h: TCvMat;
  ptpairs: TArray<Integer>;
  pt1, pt2: TArray<TCvPoint2D32f>;
  _pt1, _pt2: TCvMat;
  i, n: Integer;
  x, y, _Z, _X, _Y: Double;

begin
  _h := cvMat(3, 3, CV_64F, @h);
  // <20><><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD> <20><> <20><><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>, <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD>
  // We are looking for a pair of features on each image that correspond to each other
  findPairs(objectKeypoints, objectDescriptors, imageKeypoints, imageDescriptors, ptpairs);
  n := length(ptpairs) div 2;
  // <20>c<EFBFBD><63> <20><><EFBFBD> <20><><EFBFBD><EFBFBD>, <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // If found little pair, then have to go - object not found
  if (n < 4) then
    Exit(0);
  // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  SetLength(pt1, n);
  SetLength(pt2, n);
  // c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // read the coordinates of the "singular" points
  for i := 0 to n - 1 do
  begin
    pt1[i] := pCvSURFPoint(cvGetSeqElem(objectKeypoints, ptpairs[i * 2])).pt;
    pt2[i] := pCvSURFPoint(cvGetSeqElem(imageKeypoints, ptpairs[i * 2 + 1])).pt;
  end;

  // <20><> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // Using computed vectors - creating a matrix
  _pt1 := cvMat(1, n, CV_32FC2, @pt1[0]);
  _pt2 := cvMat(1, n, CV_32FC2, @pt2[0]);
  // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20> c <20><><EFBFBD>, <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD>
  // Find the transformation between the original image and the fact that looking
  if (cvFindHomography(@_pt1, @_pt2, @_h, CV_RANSAC, 5) = 0) then
    Exit(0);
  // <20><> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> (<28> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> _h) <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>, <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
  // Using the values transformation (in the matrix _h) find
  // the coordinates of a quadrilateral, indicative of the object
  for i := 0 to 3 do
  begin
    x := src_corners[i].x;
    y := src_corners[i].y;
    _Z := 1. / (h[6] * x + h[7] * y + h[8]);
    _X := (h[0] * x + h[1] * y + h[2]) * _Z;
    _Y := (h[3] * x + h[4] * y + h[5]) * _Z;
    dst_corners[i] := cvPoint(cvRound(_X), cvRound(_Y));
  end;
  Exit(1);
end;

Var
  object_filename, scene_filename: AnsiString;
  storage: pCvMemStorage;
  colors: array [0 .. 8] of TCvScalar = (
    (
      val:
      (
        0,
        0,
        255,
        0
      )
    ),
    (
      val:
      (
        0,
        128,
        255,
        0
      )
    ),
    (
      val:
      (
        0,
        255,
        255,
        0
      )
    ),
    (
      val:
      (
        0,
        255,
        0,
        0
      )
    ),
    (
      val:
      (
        255,
        128,
        0,
        0
      )
    ),
    (
      val:
      (
        255,
        255,
        0,
        0
      )
    ),
    (
      val:
      (
        255,
        0,
        0,
        0
      )
    ),
    (
      val:
      (
        255,
        0,
        255,
        0
      )
    ),
    (
      val:
      (
        255,
        255,
        255,
        0
      )
    )
  );

  _object, image, object_color: pIplImage;
  objectKeypoints, objectDescriptors, imageKeypoints, imageDescriptors: pCvSeq;
  i: Integer;
  params: TCvSURFParams;
  tt: Double;
  src_corners, dst_corners: TArray<TCvPoint>;
  correspond: pIplImage;
  r1, r2: TCvPoint;
  ptpairs: TArray<Integer>;
  _r1, _r2: pCvSURFPoint;
  r: pCvSURFPoint;
  center: TCvPoint;
  radius: Integer;

begin
  try
    initModule_nonfree;
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // initialization parameters
    object_filename := iif(ParamCount = 2, ParamStr(1), cResourceMedia + 'box.png');
    scene_filename := iif(ParamCount = 2, ParamStr(2), cResourceMedia + 'box_in_scene.png');
    storage := cvCreateMemStorage(0);
    cvNamedWindow('Object', 1);
    cvNamedWindow('Object Correspond', 1);
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // Loading Images
    _object := cvLoadImage(pcvChar(@object_filename[1]), CV_LOAD_IMAGE_GRAYSCALE);
    image := cvLoadImage(pcvChar(@scene_filename[1]), CV_LOAD_IMAGE_GRAYSCALE);

    if (not Assigned(_object)) or (not Assigned(image)) then
    begin
      WriteLn(Format('Can not load %s and/or %s', [object_filename, scene_filename]));
      WriteLn('Usage: find_obj [<object_filename> <scene_filename>]');
      Halt;
    end;
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD>
    // Translation grayscale
    object_color := cvCreateImage(cvGetSize(_object), 8, 3);
    cvCvtColor(_object, object_color, CV_GRAY2BGR);
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> CvSURFParams c <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20> 128 <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // Initialization of the structure CvSURFParams c size descriptors 128 items
    params := CvSURFParams(500, 1);
    // <20><>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD>
    // note the time
    tt := cvGetTickCount();
    // <20><><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>cc<63><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // We are looking for particular object recognition
    cvExtractSURF(_object, nil, @objectKeypoints, @objectDescriptors, storage, params);
    WriteLn(Format('Object Descriptors: %d', [objectDescriptors.total]));
    // <20><><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63> c<><63><EFBFBD><EFBFBD>
    // We are looking for particular scenes
    cvExtractSURF(image, nil, @imageKeypoints, @imageDescriptors, storage, params);
    WriteLn(Format('Image Descriptors: %d', [imageDescriptors.total]));
    // c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> (<28> <20><><EFBFBD><EFBFBD> 167 <20><><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD>)
    // how long it took (I 167 milliseconds)
    tt := cvGetTickCount() - tt;
    WriteLn(Format('Extraction time = %gms', [tt / (cvGetTickFrequency() * 1000)]));
    // <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>, <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63>
    // Set the image borders, within which features will be compared
    SetLength(src_corners, 4);
    src_corners[0] := cvPoint(0, 0);
    src_corners[1] := cvPoint(_object.width, 0);
    src_corners[2] := cvPoint(_object.width, _object.height);
    src_corners[3] := cvPoint(0, _object.height);
    SetLength(dst_corners, 4);
    // c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> (<28> <20><><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD> c<><63><EFBFBD><EFBFBD> <20> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>)
    // <20><><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20> <20><><EFBFBD> <20><><EFBFBD><EFBFBD>
    // creation of an additional image (it will be a scene and object)
    // Run the example and you will understand what I mean
    correspond := cvCreateImage(cvSize(image.width, _object.height + image.height), 8, 1);
    cvSetImageROI(correspond, cvRect(0, 0, _object.width, _object.height));
    cvCopy(_object, correspond);
    cvSetImageROI(correspond, cvRect(0, _object.height, correspond.width, correspond.height));
    cvCopy(image, correspond);
    cvResetImageROI(correspond);
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>, <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // Call the function that retrieves the object on the screen
    if (locatePlanarObject(objectKeypoints, objectDescriptors, imageKeypoints, imageDescriptors, src_corners, dst_corners) <> 0)
    then
    begin
      // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
      // Draw out the desired quadrangle
      for i := 0 to 3 do
      begin
        r1 := dst_corners[i mod 4];
        r2 := dst_corners[(i + 1) mod 4];
        cvLine(correspond, cvPoint(r1.x, r1.y + _object.height), cvPoint(r2.x, r2.y + _object.height), colors[8]);
      end;
    end;
    // <20>c<EFBFBD><63> <20> <20><><EFBFBD><EFBFBD> <20><>c<EFBFBD><63> <20><><EFBFBD><EFBFBD>c<EFBFBD><63> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><> <20><><EFBFBD><EFBFBD><EFBFBD>, <20><> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD> <20><>, <20><><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><> <20><>c<EFBFBD><63><EFBFBD><EFBFBD> 23.3.
    // c<><63><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD>c<EFBFBD> <20>c<EFBFBD> c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD> <20> <20><><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // again finds all the matching pairs of features in both pictures
    findPairs(objectKeypoints, objectDescriptors, imageKeypoints, imageDescriptors, ptpairs);
    // <20><><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD> <20><> <20><>c<EFBFBD><63><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
    // Between pairs of features in the figure are held straight
    i := 0;
    While i < length(ptpairs) do
    begin
      _r1 := pCvSURFPoint(cvGetSeqElem(objectKeypoints, ptpairs[i]));
      _r2 := pCvSURFPoint(cvGetSeqElem(imageKeypoints, ptpairs[i + 1]));
      cvLine(correspond, cvPointFrom32f(_r1.pt), cvPoint(cvRound(_r2.pt.x), cvRound(_r2.pt.y + _object.height)), colors[8]);
      i := i + 2;
    end;
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD> <20><>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20><> <20><>c<EFBFBD><63><EFBFBD><EFBFBD> 23.4.
    cvShowImage('Object Correspond', correspond);
    // <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <20>c<EFBFBD><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD> (<28><>c. 23.5)
    // Highlight features of circles
    for i := 0 to objectKeypoints.total - 1 do
    begin
      r := pCvSURFPoint(cvGetSeqElem(objectKeypoints, i));
      center.x := cvRound(r.pt.x);
      center.y := cvRound(r.pt.y);
      radius := cvRound(r.size * 1.2 / 9 * 2);
      cvCircle(object_color, center, radius, colors[0], 1, 8, 0);
    end;
    cvShowImage('Object', object_color);
    cvWaitKey(0);
    cvDestroyAllWindows;
  except
    on E: Exception do
      WriteLn(E.ClassName, ': ', E.Message);
  end;

end.