multicalib.cpp 27 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
/*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, Baisheng Lai (laibaisheng@gmail.com), Zhejiang University,
// all rights reserved.
//
// 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*/

/**
 * This module was accepted as a GSoC 2015 project for OpenCV, authored by
 * Baisheng Lai, mentored by Bo Li.
 *
 * The omnidirectional camera in this module is denoted by the catadioptric
 * model. Please refer to Mei's paper for details of the camera model:
 *
 *      C. Mei and P. Rives, "Single view point omnidirectional camera
 *      calibration from planar grids", in ICRA 2007.
 *
 * The implementation of the calibration part is based on Li's calibration
 * toolbox:
 *
 *     B. Li, L. Heng, K. Kevin  and M. Pollefeys, "A Multiple-Camera System
 *     Calibration Toolbox Using A Feature Descriptor-Based Calibration
 *     Pattern", in IROS 2013.
 */

#include "precomp.hpp"
#include "multicalib.hpp"
#include "opencv2/core.hpp"
#include <string>
#include <vector>
#include <queue>
#include <iostream>
#include <CharucoBoardCornerFinder.h>
#include <CharucoGridCornerDetecor.h>
Matvey Safroshkin's avatar
Matvey Safroshkin committed
69
#include "opencv2/core/core_c.h"
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
namespace cv {
	namespace multicalib {

		MultiCameraCalibration::MultiCameraCalibration(int cameraType, int nCameras, const std::string& fileName,
			float patternWidth, float patternHeight, int verbose, int showExtration, int nMiniMatches, int flags, TermCriteria criteria,
			Ptr<FeatureDetector> detector, Ptr<DescriptorExtractor> descriptor,
			Ptr<DescriptorMatcher> matcher)
		{
			_camType = cameraType;
			_nCamera = nCameras;
			_flags = flags;
			_nMiniMatches = nMiniMatches;
			_filename = fileName;
			_patternWidth = patternWidth;
			_patternHeight = patternHeight;
			_criteria = criteria;
			_showExtraction = showExtration;
			_objectPointsForEachCamera.resize(_nCamera);
			_imagePointsForEachCamera.resize(_nCamera);
			_cameraMatrix.resize(_nCamera);
			_distortCoeffs.resize(_nCamera);
			_xi.resize(_nCamera);
			_omEachCamera.resize(_nCamera);
			_tEachCamera.resize(_nCamera);
			_detector = detector;
			_descriptor = descriptor;
			_matcher = matcher;
			_verbose = verbose;
			for (int i = 0; i < _nCamera; ++i)
			{
				_vertexList.push_back(vertex());
			}
		}

		double MultiCameraCalibration::run()
		{
			loadImages();
			initialize();
			double error = optimizeExtrinsics();
			return error;
		}

		std::vector<std::string> MultiCameraCalibration::readStringList()
		{
			std::vector<std::string> l;
			l.resize(0);
			FileStorage fs(_filename, FileStorage::READ);

			FileNode n = fs.getFirstTopLevelNode();

			FileNodeIterator it = n.begin(), it_end = n.end();
			for (; it != it_end; ++it)
				l.push_back((std::string)*it);

			return l;
		}

email's avatar
email committed
127
		void MultiCameraCalibration::loadImages( )
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
		{
			std::vector<std::string> file_list;
			file_list = readStringList();

			Ptr<FeatureDetector> detector  = new CharucoGridCornerDetector;


			randpattern::CHarucoBoardCornerFinder finder(_patternWidth, _patternHeight, _nMiniMatches, CV_32F, _verbose, this->_showExtraction, detector);
			Mat pattern = cv::imread(file_list[0]);
			finder.loadBoard(pattern);


			std::vector<std::vector<std::string> > filesEachCameraFull(_nCamera);
			std::vector<std::vector<int> > timestampFull(_nCamera);
			std::vector<std::vector<int> > timestampAvailable(_nCamera);

			for (int i = 1; i < (int)file_list.size(); ++i)
			{
				int cameraVertex, timestamp;
				std::string filename = file_list[i].substr(0, file_list[i].rfind('.'));
				size_t spritPosition1 = filename.rfind('/');
				size_t spritPosition2 = filename.rfind('\\');
				if (spritPosition1 != std::string::npos)
				{
					filename = filename.substr(spritPosition1 + 1, filename.size() - 1);
				}

				else if (spritPosition2 != std::string::npos)
				{
					filename = filename.substr(spritPosition2 + 1, filename.size() - 1);
				}
				sscanf(filename.c_str(), "%d-%d", &cameraVertex, &timestamp);
				filesEachCameraFull[cameraVertex].push_back(file_list[i]);
				timestampFull[cameraVertex].push_back(timestamp);
			}


			// calibrate each camera individually
			for (int camera = 0; camera < _nCamera; ++camera)
			{
				Mat image, cameraMatrix, distortCoeffs;

				// find image and object points
				for (int imgIdx = 0; imgIdx < (int)filesEachCameraFull[camera].size(); ++imgIdx)
				{
					image = imread(filesEachCameraFull[camera][imgIdx], IMREAD_GRAYSCALE);
					if (!image.empty() && _verbose)
					{
						std::cout << "open image " << filesEachCameraFull[camera][imgIdx] << " successfully" << std::endl;
					}
					else if (image.empty() && _verbose)
					{
						std::cout << "open image" << filesEachCameraFull[camera][imgIdx] << " failed" << std::endl;
					}
					std::vector<Mat> imgObj = finder.computeObjectImagePointsForSingle(image);
					if ((int)imgObj[0].total() > _nMiniMatches)
					{
						_imagePointsForEachCamera[camera].push_back(imgObj[0]);
						_objectPointsForEachCamera[camera].push_back(imgObj[1]);
						timestampAvailable[camera].push_back(timestampFull[camera][imgIdx]);
					}
					else if ((int)imgObj[0].total() <= _nMiniMatches && _verbose)
					{
						std::cout << "image " << filesEachCameraFull[camera][imgIdx] << " has too few matched points " << std::endl;
					}
				}

				// calibrate
				Mat idx;
				double rms = 0.0;
				if (_camType == PINHOLE)
				{
					rms = cv::calibrateCamera(_objectPointsForEachCamera[camera], _imagePointsForEachCamera[camera],
						image.size(), _cameraMatrix[camera], _distortCoeffs[camera], _omEachCamera[camera],
email's avatar
email committed
202
						_tEachCamera[camera], _flags);
Matvey Safroshkin's avatar
Matvey Safroshkin committed
203

email's avatar
email committed
204
					
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
					idx = Mat(1, (int)_omEachCamera[camera].size(), CV_32S);
					for (int i = 0; i < (int)idx.total(); ++i)
					{
						idx.at<int>(i) = i;
					}
				}
				//else if (_camType == FISHEYE)
				//{
				//    rms = cv::fisheye::calibrate(_objectPointsForEachCamera[camera], _imagePointsForEachCamera[camera],
				//        image.size(), _cameraMatrix[camera], _distortCoeffs[camera], _omEachCamera[camera],
				//        _tEachCamera[camera], _flags);
				//    idx = Mat(1, (int)_omEachCamera[camera].size(), CV_32S);
				//    for (int i = 0; i < (int)idx.total(); ++i)
				//    {
				//        idx.at<int>(i) = i;
				//    }
				//}
				else if (_camType == OMNIDIRECTIONAL)
				{
					rms = cv::omnidir::calibrate(_objectPointsForEachCamera[camera], _imagePointsForEachCamera[camera],
						image.size(), _cameraMatrix[camera], _xi[camera], _distortCoeffs[camera], _omEachCamera[camera],
						_tEachCamera[camera], _flags, TermCriteria(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, 300, 1e-7),
						idx);
				}
				_cameraMatrix[camera].convertTo(_cameraMatrix[camera], CV_32F);
				_distortCoeffs[camera].convertTo(_distortCoeffs[camera], CV_32F);
				_xi[camera].convertTo(_xi[camera], CV_32F);
				//else
				//{
				//    CV_Error_(CV_StsOutOfRange, "Unknown camera type, use PINHOLE or OMNIDIRECTIONAL");
				//}

				for (int i = 0; i < (int)_omEachCamera[camera].size(); ++i)
				{
					int cameraVertex, timestamp, photoVertex;
					cameraVertex = camera;
					timestamp = timestampAvailable[camera][idx.at<int>(i)];

					photoVertex = this->getPhotoVertex(timestamp);

					if (_omEachCamera[camera][i].type() != CV_32F)
					{
						_omEachCamera[camera][i].convertTo(_omEachCamera[camera][i], CV_32F);
					}
					if (_tEachCamera[camera][i].type() != CV_32F)
					{
						_tEachCamera[camera][i].convertTo(_tEachCamera[camera][i], CV_32F);
					}

					Mat transform = Mat::eye(4, 4, CV_32F);
					Mat R, T;
					Rodrigues(_omEachCamera[camera][i], R);
					T = (_tEachCamera[camera][i]).reshape(1, 3);
					R.copyTo(transform.rowRange(0, 3).colRange(0, 3));
					T.copyTo(transform.rowRange(0, 3).col(3));

					this->_edgeList.push_back(edge(cameraVertex, photoVertex, idx.at<int>(i), transform));
				}
				std::cout << "initialized for camera " << camera << " rms = " << rms << std::endl;
				std::cout << "initialized camera matrix for camera " << camera << " is" << std::endl;
				std::cout << _cameraMatrix[camera] << std::endl;
				std::cout << "xi for camera " << camera << " is " << _xi[camera] << std::endl;
			}

		}

		int MultiCameraCalibration::getPhotoVertex(int timestamp)
		{
			int photoVertex = INVALID;

			// find in existing photo vertex
			for (int i = 0; i < (int)_vertexList.size(); ++i)
			{
				if (_vertexList[i].timestamp == timestamp)
				{
					photoVertex = i;
					break;
				}
			}

			// add a new photo vertex
			if (photoVertex == INVALID)
			{
				_vertexList.push_back(vertex(Mat::eye(4, 4, CV_32F), timestamp));
				photoVertex = (int)_vertexList.size() - 1;
			}

			return photoVertex;
		}

		void MultiCameraCalibration::initialize()
		{
			int nVertices = (int)_vertexList.size();
			int nEdges = (int)_edgeList.size();

			// build graph
			Mat G = Mat::zeros(nVertices, nVertices, CV_32S);
			for (int edgeIdx = 0; edgeIdx < nEdges; ++edgeIdx)
			{
				G.at<int>(this->_edgeList[edgeIdx].cameraVertex, this->_edgeList[edgeIdx].photoVertex) = edgeIdx + 1;
			}
			G = G + G.t();

			// traverse the graph
			std::vector<int> pre, order;
			graphTraverse(G, 0, order, pre);

			for (int i = 0; i < _nCamera; ++i)
			{
				if (pre[i] == INVALID)
				{
					std::cout << "camera" << i << "is not connected" << std::endl;
				}
			}

			for (int i = 1; i < (int)order.size(); ++i)
			{
				int vertexIdx = order[i];
				Mat prePose = this->_vertexList[pre[vertexIdx]].pose;
				int edgeIdx = G.at<int>(vertexIdx, pre[vertexIdx]) - 1;
				Mat transform = this->_edgeList[edgeIdx].transform;

				if (vertexIdx < _nCamera)
				{
					this->_vertexList[vertexIdx].pose = transform * prePose.inv();
					this->_vertexList[vertexIdx].pose.convertTo(this->_vertexList[vertexIdx].pose, CV_32F);
					if (_verbose)
					{
						std::cout << "initial pose for camera " << vertexIdx << " is " << std::endl;
						std::cout << this->_vertexList[vertexIdx].pose << std::endl;
					}
				}
				else
				{
					this->_vertexList[vertexIdx].pose = prePose.inv() * transform;
					this->_vertexList[vertexIdx].pose.convertTo(this->_vertexList[vertexIdx].pose, CV_32F);
				}
			}
		}

		double MultiCameraCalibration::optimizeExtrinsics()
		{
			// get om, t vector
			int nVertex = (int)this->_vertexList.size();

			Mat extrinParam(1, (nVertex - 1) * 6, CV_32F);
			int offset = 0;
			// the pose of the vertex[0] is eye
			for (int i = 1; i < nVertex; ++i)
			{
				Mat rvec, tvec;
				cv::Rodrigues(this->_vertexList[i].pose.rowRange(0, 3).colRange(0, 3), rvec);
				this->_vertexList[i].pose.rowRange(0, 3).col(3).copyTo(tvec);

				rvec.reshape(1, 1).copyTo(extrinParam.colRange(offset, offset + 3));
				tvec.reshape(1, 1).copyTo(extrinParam.colRange(offset + 3, offset + 6));
				offset += 6;
			}
			//double error_pre = computeProjectError(extrinParam);
			// optimization
			const double alpha_smooth = 0.01;
			double change = 1;
			for (int iter = 0; ; ++iter)
			{
				if ((_criteria.type == 1 && iter >= _criteria.maxCount) ||
					(_criteria.type == 2 && change <= _criteria.epsilon) ||
					(_criteria.type == 3 && (change <= _criteria.epsilon || iter >= _criteria.maxCount)))
					break;
				double alpha_smooth2 = 1 - std::pow(1 - alpha_smooth, (double)iter + 1.0);
				Mat JTJ_inv, JTError;
				this->computeJacobianExtrinsic(extrinParam, JTJ_inv, JTError);
				Mat G = alpha_smooth2*JTJ_inv * JTError;
				if (G.depth() == CV_64F)
				{
					G.convertTo(G, CV_32F);
				}

				extrinParam = extrinParam + G.reshape(1, 1);

				change = norm(G) / norm(extrinParam);
				//double error = computeProjectError(extrinParam);
			}

			double error = computeProjectError(extrinParam);

			std::vector<Vec3f> rvecVertex, tvecVertex;
			vector2parameters(extrinParam, rvecVertex, tvecVertex);
			for (int verIdx = 1; verIdx < (int)_vertexList.size(); ++verIdx)
			{
				Mat R;
				Mat pose = Mat::eye(4, 4, CV_32F);
				Rodrigues(rvecVertex[verIdx - 1], R);
				R.copyTo(pose.colRange(0, 3).rowRange(0, 3));
				Mat(tvecVertex[verIdx - 1]).reshape(1, 3).copyTo(pose.rowRange(0, 3).col(3));
				_vertexList[verIdx].pose = pose;
				if (_verbose && verIdx < _nCamera)
				{
					std::cout << "final camera pose of camera " << verIdx << " is" << std::endl;
					std::cout << pose << std::endl;
				}
			}
			return error;
		}

		void MultiCameraCalibration::computeJacobianExtrinsic(const Mat& extrinsicParams, Mat& JTJ_inv, Mat& JTE)
		{
			int nParam = (int)extrinsicParams.total();
			int nEdge = (int)_edgeList.size();
			std::vector<int> pointsLocation(nEdge + 1, 0);

			for (int edgeIdx = 0; edgeIdx < nEdge; ++edgeIdx)
			{
				int nPoints = (int)_objectPointsForEachCamera[_edgeList[edgeIdx].cameraVertex][_edgeList[edgeIdx].photoIndex].total();
				pointsLocation[edgeIdx + 1] = pointsLocation[edgeIdx] + nPoints * 2;
			}

			JTJ_inv = Mat(nParam, nParam, CV_64F);
			JTE = Mat(nParam, 1, CV_64F);

			Mat J = Mat::zeros(pointsLocation[nEdge], nParam, CV_64F);
			Mat E = Mat::zeros(pointsLocation[nEdge], 1, CV_64F);

			for (int edgeIdx = 0; edgeIdx < nEdge; ++edgeIdx)
			{
				int photoVertex = _edgeList[edgeIdx].photoVertex;
				int photoIndex = _edgeList[edgeIdx].photoIndex;
				int cameraVertex = _edgeList[edgeIdx].cameraVertex;

				Mat objectPoints = _objectPointsForEachCamera[cameraVertex][photoIndex];
				Mat imagePoints = _imagePointsForEachCamera[cameraVertex][photoIndex];

				Mat rvecTran, tvecTran;
				Mat R = _edgeList[edgeIdx].transform.rowRange(0, 3).colRange(0, 3);
				tvecTran = _edgeList[edgeIdx].transform.rowRange(0, 3).col(3);
				cv::Rodrigues(R, rvecTran);

				Mat rvecPhoto = extrinsicParams.colRange((photoVertex - 1) * 6, (photoVertex - 1) * 6 + 3);
				Mat tvecPhoto = extrinsicParams.colRange((photoVertex - 1) * 6 + 3, (photoVertex - 1) * 6 + 6);

				Mat rvecCamera, tvecCamera;
				if (cameraVertex > 0)
				{
					rvecCamera = extrinsicParams.colRange((cameraVertex - 1) * 6, (cameraVertex - 1) * 6 + 3);
					tvecCamera = extrinsicParams.colRange((cameraVertex - 1) * 6 + 3, (cameraVertex - 1) * 6 + 6);
				}
				else
				{
					rvecCamera = Mat::zeros(3, 1, CV_32F);
					tvecCamera = Mat::zeros(3, 1, CV_32F);
				}

				Mat jacobianPhoto, jacobianCamera, error;
				computePhotoCameraJacobian(rvecPhoto, tvecPhoto, rvecCamera, tvecCamera, rvecTran, tvecTran,
					objectPoints, imagePoints, this->_cameraMatrix[cameraVertex], this->_distortCoeffs[cameraVertex],
					this->_xi[cameraVertex], jacobianPhoto, jacobianCamera, error);
				if (cameraVertex > 0)
				{
					jacobianCamera.copyTo(J.rowRange(pointsLocation[edgeIdx], pointsLocation[edgeIdx + 1]).
						colRange((cameraVertex - 1) * 6, cameraVertex * 6));
				}
				jacobianPhoto.copyTo(J.rowRange(pointsLocation[edgeIdx], pointsLocation[edgeIdx + 1]).
					colRange((photoVertex - 1) * 6, photoVertex * 6));
				error.copyTo(E.rowRange(pointsLocation[edgeIdx], pointsLocation[edgeIdx + 1]));
			}
			//std::cout << J.t() * J << std::endl;
			JTJ_inv = (J.t() * J + 1e-10).inv();
			JTE = J.t() * E;

		}
		void MultiCameraCalibration::computePhotoCameraJacobian(const Mat& rvecPhoto, const Mat& tvecPhoto, const Mat& rvecCamera,
			const Mat& tvecCamera, Mat& rvecTran, Mat& tvecTran, const Mat& objectPoints, const Mat& imagePoints, const Mat& K,
			const Mat& distort, const Mat& xi, Mat& jacobianPhoto, Mat& jacobianCamera, Mat& E)
		{
			Mat drvecTran_drecvPhoto, drvecTran_dtvecPhoto,
				drvecTran_drvecCamera, drvecTran_dtvecCamera,
				dtvecTran_drvecPhoto, dtvecTran_dtvecPhoto,
				dtvecTran_drvecCamera, dtvecTran_dtvecCamera;

			MultiCameraCalibration::compose_motion(rvecPhoto, tvecPhoto, rvecCamera, tvecCamera, rvecTran, tvecTran,
				drvecTran_drecvPhoto, drvecTran_dtvecPhoto, drvecTran_drvecCamera, drvecTran_dtvecCamera,
				dtvecTran_drvecPhoto, dtvecTran_dtvecPhoto, dtvecTran_drvecCamera, dtvecTran_dtvecCamera);

			if (rvecTran.depth() == CV_64F)
			{
				rvecTran.convertTo(rvecTran, CV_32F);
			}
			if (tvecTran.depth() == CV_64F)
			{
				tvecTran.convertTo(tvecTran, CV_32F);
			}
			float xif = 0.0f;
			if (_camType == OMNIDIRECTIONAL)
			{
				xif = xi.at<float>(0);
			}

			Mat imagePoints2, jacobian, dx_drvecCamera, dx_dtvecCamera, dx_drvecPhoto, dx_dtvecPhoto;
			if (_camType == PINHOLE)
			{
				cv::projectPoints(objectPoints, rvecTran, tvecTran, K, distort, imagePoints2, jacobian);
			}
			//else if (_camType == FISHEYE)
			//{
			//    cv::fisheye::projectPoints(objectPoints, imagePoints2, rvecTran, tvecTran, K, distort, 0, jacobian);
			//}
			else if (_camType == OMNIDIRECTIONAL)
			{
				cv::omnidir::projectPoints(objectPoints, imagePoints2, rvecTran, tvecTran, K, xif, distort, jacobian);
			}
			if (objectPoints.depth() == CV_32F)
			{
				Mat(imagePoints - imagePoints2).convertTo(E, CV_64FC2);
			}
			else
			{
				E = imagePoints - imagePoints2;
			}
			E = E.reshape(1, (int)imagePoints.total() * 2);

			dx_drvecCamera = jacobian.colRange(0, 3) * drvecTran_drvecCamera + jacobian.colRange(3, 6) * dtvecTran_drvecCamera;
			dx_dtvecCamera = jacobian.colRange(0, 3) * drvecTran_dtvecCamera + jacobian.colRange(3, 6) * dtvecTran_dtvecCamera;
			dx_drvecPhoto = jacobian.colRange(0, 3) * drvecTran_drecvPhoto + jacobian.colRange(3, 6) * dtvecTran_drvecPhoto;
			dx_dtvecPhoto = jacobian.colRange(0, 3) * drvecTran_dtvecPhoto + jacobian.colRange(3, 6) * dtvecTran_dtvecPhoto;

			jacobianCamera = cv::Mat(dx_drvecCamera.rows, 6, CV_64F);
			jacobianPhoto = cv::Mat(dx_drvecPhoto.rows, 6, CV_64F);

			dx_drvecCamera.copyTo(jacobianCamera.colRange(0, 3));
			dx_dtvecCamera.copyTo(jacobianCamera.colRange(3, 6));
			dx_drvecPhoto.copyTo(jacobianPhoto.colRange(0, 3));
			dx_dtvecPhoto.copyTo(jacobianPhoto.colRange(3, 6));

		}
		void MultiCameraCalibration::graphTraverse(const Mat& G, int begin, std::vector<int>& order, std::vector<int>& pre)
		{
			CV_Assert(!G.empty() && G.rows == G.cols);
			int nVertex = G.rows;
			order.resize(0);
			pre.resize(nVertex, INVALID);
			pre[begin] = -1;
			std::vector<bool> visited(nVertex, false);
			std::queue<int> q;
			visited[begin] = true;
			q.push(begin);
			order.push_back(begin);

			while (!q.empty())
			{
				int v = q.front();
				q.pop();
				Mat idx;
				// use my findNonZero maybe
				findRowNonZero(G.row(v), idx);
				for (int i = 0; i < (int)idx.total(); ++i)
				{
					int neighbor = idx.at<int>(i);
					if (!visited[neighbor])
					{
						visited[neighbor] = true;
						q.push(neighbor);
						order.push_back(neighbor);
						pre[neighbor] = v;
					}
				}
			}
		}

		void MultiCameraCalibration::findRowNonZero(const Mat& row, Mat& idx)
		{
			CV_Assert(!row.empty() && row.rows == 1 && row.channels() == 1);
			Mat _row;
			std::vector<int> _idx;
			row.convertTo(_row, CV_32F);
			for (int i = 0; i < (int)row.total(); ++i)
			{
				if (_row.at<float>(i) != 0)
				{
					_idx.push_back(i);
				}
			}
			idx.release();
			idx.create(1, (int)_idx.size(), CV_32S);
			for (int i = 0; i < (int)_idx.size(); ++i)
			{
				idx.at<int>(i) = _idx[i];
			}
		}

		double MultiCameraCalibration::computeProjectError(Mat& parameters)
		{
			int nVertex = (int)_vertexList.size();
			CV_Assert((int)parameters.total() == (nVertex - 1) * 6 && parameters.depth() == CV_32F);
			int nEdge = (int)_edgeList.size();

			// recompute the transform between photos and cameras

			std::vector<edge> edgeList = this->_edgeList;
			std::vector<Vec3f> rvecVertex, tvecVertex;
			vector2parameters(parameters, rvecVertex, tvecVertex);

			float totalError = 0;
			int totalNPoints = 0;
			for (int edgeIdx = 0; edgeIdx < nEdge; ++edgeIdx)
			{
				Mat RPhoto, RCamera, TPhoto, TCamera, transform;
				int cameraVertex = edgeList[edgeIdx].cameraVertex;
				int photoVertex = edgeList[edgeIdx].photoVertex;
				int PhotoIndex = edgeList[edgeIdx].photoIndex;
				TPhoto = Mat(tvecVertex[photoVertex - 1]).reshape(1, 3);

				//edgeList[edgeIdx].transform = Mat::ones(4, 4, CV_32F);
				transform = Mat::eye(4, 4, CV_32F);
				cv::Rodrigues(rvecVertex[photoVertex - 1], RPhoto);
				if (cameraVertex == 0)
				{
					RPhoto.copyTo(transform.rowRange(0, 3).colRange(0, 3));
					TPhoto.copyTo(transform.rowRange(0, 3).col(3));
				}
				else
				{
					TCamera = Mat(tvecVertex[cameraVertex - 1]).reshape(1, 3);
					cv::Rodrigues(rvecVertex[cameraVertex - 1], RCamera);
					Mat(RCamera*RPhoto).copyTo(transform.rowRange(0, 3).colRange(0, 3));
					Mat(RCamera * TPhoto + TCamera).copyTo(transform.rowRange(0, 3).col(3));
				}

				transform.copyTo(edgeList[edgeIdx].transform);
				Mat rvec, tvec;
				cv::Rodrigues(transform.rowRange(0, 3).colRange(0, 3), rvec);
				transform.rowRange(0, 3).col(3).copyTo(tvec);

				Mat objectPoints, imagePoints, proImagePoints;
				objectPoints = this->_objectPointsForEachCamera[cameraVertex][PhotoIndex];
				imagePoints = this->_imagePointsForEachCamera[cameraVertex][PhotoIndex];
email's avatar
email committed
639
640
641
642
				



643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791

				if (this->_camType == PINHOLE)
				{
					cv::projectPoints(objectPoints, rvec, tvec, _cameraMatrix[cameraVertex], _distortCoeffs[cameraVertex],
						proImagePoints);
				}
				//else if (this->_camType == FISHEYE)
				//{
				//    cv::fisheye::projectPoints(objectPoints, proImagePoints, rvec, tvec, _cameraMatrix[cameraVertex],
				//        _distortCoeffs[cameraVertex]);
				//}
				else if (this->_camType == OMNIDIRECTIONAL)
				{
					float xi = _xi[cameraVertex].at<float>(0);

					cv::omnidir::projectPoints(objectPoints, proImagePoints, rvec, tvec, _cameraMatrix[cameraVertex],
						xi, _distortCoeffs[cameraVertex]);
				}
				Mat error = imagePoints - proImagePoints;
				Vec2f* ptr_err = error.ptr<Vec2f>();
				for (int i = 0; i < (int)error.total(); ++i)
				{
					totalError += sqrt(ptr_err[i][0] * ptr_err[i][0] + ptr_err[i][1] * ptr_err[i][1]);
				}
				totalNPoints += (int)error.total();
			}
			double meanReProjError = totalError / totalNPoints;
			_error = meanReProjError;
			return meanReProjError;
		}

		void MultiCameraCalibration::compose_motion(InputArray _om1, InputArray _T1, InputArray _om2, InputArray _T2, Mat& om3, Mat& T3, Mat& dom3dom1,
			Mat& dom3dT1, Mat& dom3dom2, Mat& dom3dT2, Mat& dT3dom1, Mat& dT3dT1, Mat& dT3dom2, Mat& dT3dT2)
		{
			Mat om1, om2, T1, T2;
			_om1.getMat().convertTo(om1, CV_64F);
			_om2.getMat().convertTo(om2, CV_64F);
			_T1.getMat().reshape(1, 3).convertTo(T1, CV_64F);
			_T2.getMat().reshape(1, 3).convertTo(T2, CV_64F);
			/*Mat om2 = _om2.getMat();
			Mat T1 = _T1.getMat().reshape(1, 3);
			Mat T2 = _T2.getMat().reshape(1, 3);*/

			//% Rotations:
			Mat R1, R2, R3, dR1dom1(9, 3, CV_64FC1), dR2dom2;
			cv::Rodrigues(om1, R1, dR1dom1);
			cv::Rodrigues(om2, R2, dR2dom2);
			/*JRodriguesMatlab(dR1dom1, dR1dom1);
			JRodriguesMatlab(dR2dom2, dR2dom2);*/
			dR1dom1 = dR1dom1.t();
			dR2dom2 = dR2dom2.t();

			R3 = R2 * R1;
			Mat dR3dR2, dR3dR1;
			//dAB(R2, R1, dR3dR2, dR3dR1);
			matMulDeriv(R2, R1, dR3dR2, dR3dR1);
			Mat dom3dR3;
			cv::Rodrigues(R3, om3, dom3dR3);
			//JRodriguesMatlab(dom3dR3, dom3dR3);
			dom3dR3 = dom3dR3.t();

			dom3dom1 = dom3dR3 * dR3dR1 * dR1dom1;
			dom3dom2 = dom3dR3 * dR3dR2 * dR2dom2;
			dom3dT1 = Mat::zeros(3, 3, CV_64FC1);
			dom3dT2 = Mat::zeros(3, 3, CV_64FC1);

			//% Translations:
			Mat T3t = R2 * T1;
			Mat dT3tdR2, dT3tdT1;
			//dAB(R2, T1, dT3tdR2, dT3tdT1);
			matMulDeriv(R2, T1, dT3tdR2, dT3tdT1);

			Mat dT3tdom2 = dT3tdR2 * dR2dom2;
			T3 = T3t + T2;
			dT3dT1 = dT3tdT1;
			dT3dT2 = Mat::eye(3, 3, CV_64FC1);
			dT3dom2 = dT3tdom2;
			dT3dom1 = Mat::zeros(3, 3, CV_64FC1);
		}

		void MultiCameraCalibration::vector2parameters(const Mat& parameters, std::vector<Vec3f>& rvecVertex, std::vector<Vec3f>& tvecVertexs)
		{
			int nVertex = (int)_vertexList.size();
			CV_Assert((int)parameters.channels() == 1 && (int)parameters.total() == 6 * (nVertex - 1));
			CV_Assert(parameters.depth() == CV_32F);
			parameters.reshape(1, 1);

			rvecVertex.reserve(0);
			tvecVertexs.resize(0);

			for (int i = 0; i < nVertex - 1; ++i)
			{
				rvecVertex.push_back(Vec3f(parameters.colRange(i * 6, i * 6 + 3)));
				tvecVertexs.push_back(Vec3f(parameters.colRange(i * 6 + 3, i * 6 + 6)));
			}
		}

		void MultiCameraCalibration::parameters2vector(const std::vector<Vec3f>& rvecVertex, const std::vector<Vec3f>& tvecVertex, Mat& parameters)
		{
			CV_Assert(rvecVertex.size() == tvecVertex.size());
			int nVertex = (int)rvecVertex.size();
			// the pose of the first camera is known
			parameters.create(1, 6 * (nVertex - 1), CV_32F);

			for (int i = 0; i < nVertex - 1; ++i)
			{
				Mat(rvecVertex[i]).reshape(1, 1).copyTo(parameters.colRange(i * 6, i * 6 + 3));
				Mat(tvecVertex[i]).reshape(1, 1).copyTo(parameters.colRange(i * 6 + 3, i * 6 + 6));
			}
		}

		void MultiCameraCalibration::writeParameters(const std::string& filename)
		{
			FileStorage fs(filename, FileStorage::WRITE);

			fs << "nCameras" << _nCamera;

			for (int camIdx = 0; camIdx < _nCamera; ++camIdx)
			{
				std::stringstream tmpStr;
				tmpStr << camIdx;
				std::string cameraMatrix = "camera_matrix_" + tmpStr.str();
				std::string cameraPose = "camera_pose_" + tmpStr.str();
				std::string cameraDistortion = "camera_distortion_" + tmpStr.str();
				std::string cameraXi = "xi_" + tmpStr.str();

				fs << cameraMatrix << _cameraMatrix[camIdx];
				fs << cameraDistortion << _distortCoeffs[camIdx];
				if (_camType == OMNIDIRECTIONAL)
				{
					fs << cameraXi << _xi[camIdx].at<float>(0);
				}

				fs << cameraPose << _vertexList[camIdx].pose;
			}

			fs << "meanReprojectError" << _error;

			for (int photoIdx = _nCamera; photoIdx < (int)_vertexList.size(); ++photoIdx)
			{
				std::stringstream tmpStr;
				tmpStr << _vertexList[photoIdx].timestamp;
				std::string photoTimestamp = "pose_timestamp_" + tmpStr.str();

				fs << photoTimestamp << _vertexList[photoIdx].pose;
			}
		}
	}
} // namespace multicalib, cv