Eigen::Matrix3f Homography::calcHomography( const Eigen::Isometry3f &camA2plane, const Eigen::Isometry3f &camB2plane ) const
{
// calculate the plane normal (z-vector) from the point of camB
Eigen::Vector3f n = camB2plane.rotation().transpose() * -Eigen::Vector3f::UnitZ();
// calculate the distance from camB to plane
float d = camB2plane.inverse().translation().dot( n );
// get transform from camB to camA
Eigen::Isometry3f camBtoCamA = camA2plane.inverse() * camB2plane;
// get the homography
return calcHomography( camBtoCamA, n, d );
}
void KinectDatasetReader::MoveGroundTruthsToOrigin()
{
Eigen::Isometry3f origin = mGroundTruth[0].LocalTransform;
origin = origin.inverse();
for(size_t i = 0; i < mGroundTruth.size(); i++)
mGroundTruth[i].LocalTransform = origin * mGroundTruth[i].LocalTransform;
}
bool Utils::
composeViewMatrix(Eigen::Projective3f& oMatrix, const Eigen::Matrix3f& iCalib,
const Eigen::Isometry3f& iPose, const bool iIsOrthographic) {
Eigen::Projective3f calib = Eigen::Projective3f::Identity();
calib.matrix().col(2).swap(calib.matrix().col(3));
calib.matrix().topLeftCorner<2,3>() = iCalib.matrix().topRows<2>();
calib.matrix().bottomLeftCorner<1,3>() = iCalib.matrix().bottomRows<1>();
if (iIsOrthographic) calib.matrix().col(2).swap(calib.matrix().col(3));
oMatrix = calib*iPose.inverse();
return true;
}
int
main(int argc, char** argv){
string dirname;
string graphFilename;
float numThreads;
int ng_step;
int ng_minPoints;
int ng_imageRadius;
float ng_worldRadius;
float ng_maxCurvature;
float al_inlierDistance;
float al_inlierCurvatureRatio;
float al_inlierNormalAngle;
float al_inlierMaxChi2;
float al_scale;
float al_flatCurvatureThreshold;
float al_outerIterations;
float al_nonLinearIterations;
float al_linearIterations;
float al_minInliers;
float al_lambda;
float al_debug;
PointWithNormalStatistcsGenerator normalGenerator;
PointWithNormalAligner aligner;
g2o::CommandArgs arg;
arg.param("ng_step",ng_step,normalGenerator.step(),"compute a normal each x pixels") ;
arg.param("ng_minPoints",ng_minPoints,normalGenerator.minPoints(),"minimum number of points in a region to compute the normal");
arg.param("ng_imageRadius",ng_imageRadius,normalGenerator.imageRadius(), "radius of a ball in the works where to compute the normal for a pixel");
arg.param("ng_worldRadius",ng_worldRadius, normalGenerator.worldRadius(), "radius of a ball in the works where to compute the normal for a pixel");
arg.param("ng_maxCurvature",ng_maxCurvature, normalGenerator.maxCurvature(), "above this threshold the normal is not computed");
arg.param("al_inlierDistance", al_inlierDistance, aligner.inlierDistanceThreshold(), "max metric distance between two points to regard them as iniliers");
arg.param("al_inlierCurvatureRatio", al_inlierCurvatureRatio, aligner.inlierCurvatureRatioThreshold(), "max metric distance between two points to regard them as iniliers");
arg.param("al_inlierNormalAngle", al_inlierNormalAngle, aligner.inlierNormalAngularThreshold(), "max metric distance between two points to regard them as iniliers");
arg.param("al_inlierMaxChi2", al_inlierMaxChi2, aligner.inlierMaxChi2(), "max metric distance between two points to regard them as iniliers");
arg.param("al_minInliers", al_minInliers, aligner.minInliers(), "minimum numver of inliers to do the matching");
arg.param("al_scale", al_scale, aligner.scale(), "scale of the range image for the alignment");
arg.param("al_flatCurvatureThreshold", al_flatCurvatureThreshold, aligner.flatCurvatureThreshold(), "curvature above which the patches are not considered planar");
arg.param("al_outerIterations", al_outerIterations, aligner.outerIterations(), "outer interations (incl. data association)");
arg.param("al_linearIterations", al_linearIterations, aligner.linearIterations(), "linear iterations for each outer one (uses R,t)");
arg.param("al_nonLinearIterations", al_nonLinearIterations, aligner.nonLinearIterations(), "nonlinear iterations for each outer one (uses q,t)");
arg.param("al_lambda", al_lambda, aligner.lambda(), "damping factor for the transformation update, the higher the smaller the step");
arg.param("al_debug", al_debug, aligner.debug(), "prints lots of stuff");
arg.param("numThreads", numThreads, 1, "numver of threads for openmp");
if (numThreads<1)
numThreads = 1;
std::vector<string> fileNames;
arg.paramLeftOver("dirname", dirname, "", "", true);
arg.paramLeftOver("graph-file", graphFilename, "out.g2o", "graph-file", true);
arg.parseArgs(argc, argv);
cerr << "dirname " << dirname << endl;
std::vector<string> filenames;
std::set<string> filenamesset = readDir(dirname);
for(set<string>::const_iterator it =filenamesset.begin(); it!=filenamesset.end(); it++) {
filenames.push_back(*it);
}
normalGenerator.setStep(ng_step);
normalGenerator.setMinPoints(ng_minPoints);
normalGenerator.setImageRadius(ng_imageRadius);
normalGenerator.setWorldRadius(ng_worldRadius);
normalGenerator.setMaxCurvature(ng_maxCurvature);
#ifdef _PWN_USE_OPENMP_
normalGenerator.setNumThreads(numThreads);
#endif //_PWN_USE_OPENMP_
aligner.setInlierDistanceThreshold(al_inlierDistance);
aligner.setInlierCurvatureRatioThreshold(al_inlierCurvatureRatio);
aligner.setInlierNormalAngularThreshold(al_inlierNormalAngle);
aligner.setInlierMaxChi2(al_inlierMaxChi2);
aligner.setMinInliers(al_minInliers);
aligner.setScale(al_scale);
aligner.setFlatCurvatureThreshold(al_flatCurvatureThreshold);
aligner.setOuterIterations(al_outerIterations);
aligner.setLinearIterations(al_linearIterations);
aligner.setNonLinearIterations(al_nonLinearIterations);
aligner.setLambda(al_lambda);
aligner.setDebug(al_debug);
#ifdef _PWN_USE_OPENMP_
aligner.setNumThreads(numThreads);
#endif //_PWN_USE_OPENMP_
Frame* referenceFrame= 0;
Eigen::Matrix3f cameraMatrix;
cameraMatrix <<
525.0f, 0.0f, 319.5f,
0.0f, 525.0f, 239.5f,
0.0f, 0.0f, 1.0f;
ostringstream os(graphFilename.c_str());
cerr << "there are " << filenames.size() << " files in the pool" << endl;
Isometry3f trajectory;
trajectory.setIdentity();
int previousIndex=-1;
int graphNum=0;
int nFrames = 0;
string baseFilename = graphFilename.substr( 0, graphFilename.find_last_of( '.' ) );
for (size_t i=0; i<filenames.size(); i++){
cerr << endl << endl << endl;
cerr << ">>>>>>>>>>>>>>>>>>>>>>>> PROCESSING " << filenames[i] << " <<<<<<<<<<<<<<<<<<<<" << endl;
Frame* currentFrame= new Frame();
if(!currentFrame->load(filenames[i])) {
cerr << "failure in loading image: " << filenames[i] << ", skipping" << endl;
delete currentFrame;
continue;
}
nFrames ++;
if (! referenceFrame ){
os << "PARAMS_CAMERACALIB 0 0 0 0 0 0 0 1 525 525 319.5 239.5"<< endl;
trajectory.setIdentity();
}
{
// write the vertex frame
Vector6f x = t2v(trajectory);
Vector3f t = x.head<3>();
Vector3f mq = x.tail<3>();
float w = mq.squaredNorm();
if (w>1){
mq.setZero();
w = 1.0f;
} else {
w = sqrt(1-w);
}
os << "VERTEX_SE3:QUAT " << i << " " << t.transpose() << " " << mq.transpose() << " " << w << endl;
os << "DEPTH_IMAGE_DATA 0 " << filenames[i] << " 0 0 " << endl;
}
currentFrame->computeStats(normalGenerator,cameraMatrix);
if (referenceFrame) {
referenceFrame->setAligner(aligner, true);
currentFrame->setAligner(aligner, false);
Matrix6f omega;
Vector6f mean;
float tratio;
float rratio;
aligner.setImageSize(currentFrame->depthImage.rows(), currentFrame->depthImage.cols());
Eigen::Isometry3f X;
X.setIdentity();
double ostart = get_time();
float error;
int result = aligner.align(error, X, mean, omega, tratio, rratio);
cerr << "inliers=" << result << " error/inliers: " << error/result << endl;
cerr << "localTransform : " << endl;
cerr << X.inverse().matrix() << endl;
trajectory=trajectory*X;
cerr << "globaltransform: " << endl;
cerr << trajectory.matrix() << endl;
double oend = get_time();
cerr << "alignment took: " << oend-ostart << " sec." << endl;
cerr << "aligner scaled image size: " << aligner.scaledImageRows() << " " << aligner.scaledImageCols() << endl;
if(rratio < 100 && tratio < 100) {
// write the edge frame
Vector6f x = mean;
Vector3f t = x.head<3>();
Vector3f mq = x.tail<3>();
float w = mq.squaredNorm();
if (w>1){
mq.setZero();
w = 1.0f;
} else {
w = sqrt(1-w);
}
os << "EDGE_SE3:QUAT " << previousIndex << " " << i << " ";
os << t.transpose() << " " << mq.transpose() << " " << w << " ";
for (int r=0; r<6; r++){
for (int c=r; c<6; c++){
os << omega(r,c) << " ";
}
}
os << endl;
} else {
if (nFrames >10) {
char buf[1024];
sprintf(buf, "%s-%03d.g2o", baseFilename.c_str(), graphNum);
ofstream gs(buf);
gs << os.str();
gs.close();
}
os.str("");
os.clear();
if (referenceFrame)
delete referenceFrame;
referenceFrame = 0;
graphNum ++;
nFrames = 0;
}
}
previousIndex = i;
if (referenceFrame)
delete referenceFrame;
referenceFrame = currentFrame;
}
char buf[1024];
sprintf(buf, "%s-%03d.g2o", baseFilename.c_str(), graphNum);
cerr << "saving final frames, n: " << nFrames << " in file [" << buf << "]" << endl;
cerr << "filesize:" << os.str().length() << endl;
ofstream gs(buf);
gs << os.str();
cout << os.str();
gs.close();
}
void ICPOdometry::getIncrementalTransformation(Eigen::Vector3f & trans, Eigen::Matrix<float, 3, 3, Eigen::RowMajor> & rot, int threads, int blocks)
{
iterations[0] = 10;
iterations[1] = 5;
iterations[2] = 4;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> Rprev = rot;
Eigen::Vector3f tprev = trans;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> Rcurr = Rprev;
Eigen::Vector3f tcurr = tprev;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> Rprev_inv = Rprev.inverse();
Mat33 & device_Rprev_inv = device_cast<Mat33>(Rprev_inv);
float3& device_tprev = device_cast<float3>(tprev);
cv::Mat resultRt = cv::Mat::eye(4, 4, CV_64FC1);
for(int i = NUM_PYRS - 1; i >= 0; i--)
{
for(int j = 0; j < iterations[i]; j++)
{
Eigen::Matrix<float, 6, 6, Eigen::RowMajor> A_icp;
Eigen::Matrix<float, 6, 1> b_icp;
Mat33& device_Rcurr = device_cast<Mat33> (Rcurr);
float3& device_tcurr = device_cast<float3>(tcurr);
DeviceArray2D<float>& vmap_curr = vmaps_curr_[i];
DeviceArray2D<float>& nmap_curr = nmaps_curr_[i];
DeviceArray2D<float>& vmap_g_prev = vmaps_g_prev_[i];
DeviceArray2D<float>& nmap_g_prev = nmaps_g_prev_[i];
float residual[2];
icpStep(device_Rcurr,
device_tcurr,
vmap_curr,
nmap_curr,
device_Rprev_inv,
device_tprev,
intr(i),
vmap_g_prev,
nmap_g_prev,
distThres_,
angleThres_,
sumData,
outData,
A_icp.data(),
b_icp.data(),
&residual[0],
threads,
blocks);
lastICPError = sqrt(residual[0]) / residual[1];
lastICPCount = residual[1];
Eigen::Matrix<double, 6, 1> result;
Eigen::Matrix<double, 6, 6, Eigen::RowMajor> dA_icp = A_icp.cast<double>();
Eigen::Matrix<double, 6, 1> db_icp = b_icp.cast<double>();
lastA = dA_icp;
lastb = db_icp;
result = lastA.ldlt().solve(lastb);
Eigen::Isometry3f incOdom;
OdometryProvider::computeProjectiveMatrix(resultRt, result, incOdom);
Eigen::Isometry3f currentT;
currentT.setIdentity();
currentT.rotate(Rprev);
currentT.translation() = tprev;
currentT = currentT * incOdom.inverse();
tcurr = currentT.translation();
Rcurr = currentT.rotation();
}
}
trans = tcurr;
rot = Rcurr;
}
void Aligner::align() {
struct timeval tvStart, tvEnd;
gettimeofday(&tvStart,0);
if (! _projector || !_linearizer || !_correspondenceGenerator){
cerr << "I do nothing since you did not set all required algorithms" << endl;
return;
}
// the current points are seen from the frame of the sensor
_projector->setTransform(_sensorOffset);
_projector->project(_correspondenceGenerator->currentIndexImage(),
_correspondenceGenerator->currentDepthImage(),
_currentScene->points());
_T = _initialGuess;
for(int i = 0; i < _outerIterations; i++) {
//cout << "********************* Iteration " << i << " *********************" << endl;
/************************************************************************
* Correspondence Computation *
************************************************************************/
//cout << "Computing correspondences...";
// compute the indices of the current scene from the point of view of the sensor
_projector->setTransform(_T*_sensorOffset);
_projector->project(_correspondenceGenerator->referenceIndexImage(),
_correspondenceGenerator->referenceDepthImage(),
_referenceScene->points());
// Correspondences computation.
_correspondenceGenerator->compute(*_referenceScene, *_currentScene, _T.inverse());
//cout << " done." << endl;
//cout << "# inliers found: " << _correspondenceGenerator->numCorrespondences() << endl;
/************************************************************************
* Alignment *
************************************************************************/
Eigen::Isometry3f invT = _T.inverse();
for (int k = 0; k < _innerIterations; k++) {
invT.matrix().block<1, 4>(3, 0) << 0, 0, 0, 1;
Matrix6f H;
Vector6f b;
_linearizer->setT(invT);
_linearizer->update();
H = _linearizer->H() + Matrix6f::Identity() * 10.0f;
b = _linearizer->b();
// add the priors
for (size_t i=0; i<_priors.size(); i++){
const SE3Prior* prior = _priors[i];
Vector6f priorError = prior->error(invT);
Matrix6f priorJacobian = prior->jacobian(invT);
Matrix6f priorInformationRemapped = prior->errorInformation(invT);
Matrix6f Hp = priorJacobian.transpose()*priorInformationRemapped*priorJacobian;
Vector6f bp = priorJacobian.transpose()*priorInformationRemapped*priorError;
H += Hp;
b += bp;
}
Vector6f dx = H.ldlt().solve(-b);
Eigen::Isometry3f dT = v2t(dx);
invT = dT * invT;
}
_T = invT.inverse();
_T.matrix().block<1, 4>(3, 0) << 0, 0, 0, 1;
}
gettimeofday(&tvEnd, 0);
double tStart = tvStart.tv_sec*1000+tvStart.tv_usec*0.001;
double tEnd = tvEnd.tv_sec*1000+tvEnd.tv_usec*0.001;
_totalTime = tEnd - tStart;
_error = _linearizer->error();
_inliers = _linearizer->inliers();
}
void PointWithNormalStatistcsGenerator::computeNormalsAndSVD(PointWithNormalVector& points, PointWithNormalSVDVector& svds, const Eigen::MatrixXi& indices,
const Eigen::Matrix3f& cameraMatrix, const Eigen::Isometry3f& transform){
_integralImage.compute(indices,points);
int q=0;
int outerStep = _numThreads * _step;
PixelMapper pixelMapper;
pixelMapper.setCameraMatrix(cameraMatrix);
pixelMapper.setTransform(transform);
Eigen::Isometry3f inverseTransform = transform.inverse();
#pragma omp parallel
{
#ifdef _PWN_USE_OPENMP_
int threadNum = omp_get_thread_num();
#else // _PWN_USE_OPENMP_
int threadNum = 0;
Eigen::SelfAdjointEigenSolver<Eigen::Matrix3f> eigenSolver;
#endif // _PWN_USE_OPENMP_
for (int c=threadNum; c<indices.cols(); c+=outerStep) {
#ifdef _PWN_USE_OPENMP_
Eigen::SelfAdjointEigenSolver<Eigen::Matrix3f> eigenSolver;
#endif // _PWN_USE_OPENMP_
for (int r=0; r<indices.rows(); r+=_step, q++){
int index = indices(r,c);
//cerr << "index(" << r <<"," << c << ")=" << index << endl;
if (index<0)
continue;
// determine the region
PointWithNormal& point = points[index];
PointWithNormalSVD& originalSVD = svds[index];
PointWithNormalSVD svd;
Eigen::Vector3f normal = point.normal();
Eigen::Vector3f coord = pixelMapper.projectPoint(point.point()+Eigen::Vector3f(_worldRadius, _worldRadius, 0));
svd._z=point(2);
coord.head<2>()*=(1./coord(2));
int dx = abs(c - coord[0]);
int dy = abs(r - coord[1]);
if (dx>_imageRadius)
dx = _imageRadius;
if (dy>_imageRadius)
dy = _imageRadius;
PointAccumulator acc = _integralImage.getRegion(c-dx, c+dx, r-dy, r+dy);
svd._mean=point.point();
if (acc.n()>_minPoints){
Eigen::Vector3f mean = acc.mean();
svd._mean = mean;
svd._n = acc.n();
Eigen::Matrix3f cov = acc.covariance();
eigenSolver.compute(cov);
svd._U=eigenSolver.eigenvectors();
svd._singularValues=eigenSolver.eigenvalues();
if (svd._singularValues(0) <0)
svd._singularValues(0) = 0;
/*
cerr << "\t svd.singularValues():" << svd.singularValues() << endl;
cerr << "\t svd.U():" << endl << svd.U() << endl;
//cerr << "\t svd.curvature():" << svd.curvature() << endl;
*/
normal = eigenSolver.eigenvectors().col(0).normalized();
if (normal.dot(inverseTransform * mean) > 0.0f)
normal =-normal;
svd.updateCurvature();
//cerr << "n(" << index << ") c:" << svd.curvature() << endl << point.tail<3>() << endl;
if (svd.curvature()>_maxCurvature){
//cerr << "region: " << c-dx << " " << c+dx << " " << r-dx << " " << r+dx << " points: " << acc.n() << endl;
normal.setZero();
}
} else {
normal.setZero();
svd = PointWithNormalSVD();
}
if (svd.n() > originalSVD.n()){
originalSVD = svd;
point.setNormal(normal);
}
}
}
}
}
void RGBDOdometry::getIncrementalTransformation(Eigen::Vector3f & trans,
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> & rot,
const bool & rgbOnly,
const float & icpWeight,
const bool & pyramid,
const bool & fastOdom,
const bool & so3)
{
bool icp = !rgbOnly && icpWeight > 0;
bool rgb = rgbOnly || icpWeight < 100;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> Rprev = rot;
Eigen::Vector3f tprev = trans;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> Rcurr = Rprev;
Eigen::Vector3f tcurr = tprev;
if(rgb)
{
for(int i = 0; i < NUM_PYRS; i++)
{
computeDerivativeImages(nextImage[i], nextdIdx[i], nextdIdy[i]);
}
}
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> resultR = Eigen::Matrix<double, 3, 3, Eigen::RowMajor>::Identity();
if(so3)
{
int pyramidLevel = 2;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> R_lr = Eigen::Matrix<float, 3, 3, Eigen::RowMajor>::Identity();
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> K = Eigen::Matrix<double, 3, 3, Eigen::RowMajor>::Zero();
K(0, 0) = intr(pyramidLevel).fx;
K(1, 1) = intr(pyramidLevel).fy;
K(0, 2) = intr(pyramidLevel).cx;
K(1, 2) = intr(pyramidLevel).cy;
K(2, 2) = 1;
float lastError = std::numeric_limits<float>::max() / 2;
float lastCount = std::numeric_limits<float>::max() / 2;
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> lastResultR = Eigen::Matrix<double, 3, 3, Eigen::RowMajor>::Identity();
for(int i = 0; i < 10; i++)
{
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> jtj;
Eigen::Matrix<float, 3, 1> jtr;
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> homography = K * resultR * K.inverse();
mat33 imageBasis;
memcpy(&imageBasis.data[0], homography.cast<float>().eval().data(), sizeof(mat33));
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> K_inv = K.inverse();
mat33 kinv;
memcpy(&kinv.data[0], K_inv.cast<float>().eval().data(), sizeof(mat33));
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> K_R_lr = K * resultR;
mat33 krlr;
memcpy(&krlr.data[0], K_R_lr.cast<float>().eval().data(), sizeof(mat33));
float residual[2];
TICK("so3Step");
so3Step(lastNextImage[pyramidLevel],
nextImage[pyramidLevel],
imageBasis,
kinv,
krlr,
sumDataSO3,
outDataSO3,
jtj.data(),
jtr.data(),
&residual[0],
GPUConfig::getInstance().so3StepThreads,
GPUConfig::getInstance().so3StepBlocks);
TOCK("so3Step");
lastSO3Error = sqrt(residual[0]) / residual[1];
lastSO3Count = residual[1];
//Converged
if(lastSO3Error < lastError && lastCount == lastSO3Count)
{
break;
}
else if(lastSO3Error > lastError + 0.001) //Diverging
{
lastSO3Error = lastError;
lastSO3Count = lastCount;
resultR = lastResultR;
break;
}
lastError = lastSO3Error;
lastCount = lastSO3Count;
lastResultR = resultR;
Eigen::Vector3f delta = jtj.ldlt().solve(jtr);
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> rotUpdate = OdometryProvider::rodrigues(delta.cast<double>());
R_lr = rotUpdate.cast<float>() * R_lr;
for(int x = 0; x < 3; x++)
{
for(int y = 0; y < 3; y++)
{
resultR(x, y) = R_lr(x, y);
}
}
}
}
iterations[0] = fastOdom ? 3 : 10;
iterations[1] = pyramid ? 5 : 0;
iterations[2] = pyramid ? 4 : 0;
Eigen::Matrix<float, 3, 3, Eigen::RowMajor> Rprev_inv = Rprev.inverse();
mat33 device_Rprev_inv = Rprev_inv;
float3 device_tprev = *reinterpret_cast<float3*>(tprev.data());
Eigen::Matrix<double, 4, 4, Eigen::RowMajor> resultRt = Eigen::Matrix<double, 4, 4, Eigen::RowMajor>::Identity();
if(so3)
{
for(int x = 0; x < 3; x++)
{
for(int y = 0; y < 3; y++)
{
resultRt(x, y) = resultR(x, y);
}
}
}
for(int i = NUM_PYRS - 1; i >= 0; i--)
{
if(rgb)
{
projectToPointCloud(lastDepth[i], pointClouds[i], intr, i);
}
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> K = Eigen::Matrix<double, 3, 3, Eigen::RowMajor>::Zero();
K(0, 0) = intr(i).fx;
K(1, 1) = intr(i).fy;
K(0, 2) = intr(i).cx;
K(1, 2) = intr(i).cy;
K(2, 2) = 1;
lastRGBError = std::numeric_limits<float>::max();
for(int j = 0; j < iterations[i]; j++)
{
Eigen::Matrix<double, 4, 4, Eigen::RowMajor> Rt = resultRt.inverse();
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> R = Rt.topLeftCorner(3, 3);
Eigen::Matrix<double, 3, 3, Eigen::RowMajor> KRK_inv = K * R * K.inverse();
mat33 krkInv;
memcpy(&krkInv.data[0], KRK_inv.cast<float>().eval().data(), sizeof(mat33));
Eigen::Vector3d Kt = Rt.topRightCorner(3, 1);
Kt = K * Kt;
float3 kt = {(float)Kt(0), (float)Kt(1), (float)Kt(2)};
int sigma = 0;
int rgbSize = 0;
if(rgb)
{
TICK("computeRgbResidual");
computeRgbResidual(pow(minimumGradientMagnitudes[i], 2.0) / pow(sobelScale, 2.0),
nextdIdx[i],
nextdIdy[i],
lastDepth[i],
nextDepth[i],
lastImage[i],
nextImage[i],
corresImg[i],
sumResidualRGB,
maxDepthDeltaRGB,
kt,
krkInv,
sigma,
rgbSize,
GPUConfig::getInstance().rgbResThreads,
GPUConfig::getInstance().rgbResBlocks);
TOCK("computeRgbResidual");
}
float sigmaVal = std::sqrt((float)sigma / rgbSize == 0 ? 1 : rgbSize);
float rgbError = std::sqrt(sigma) / (rgbSize == 0 ? 1 : rgbSize);
if(rgbOnly && rgbError > lastRGBError)
{
break;
}
lastRGBError = rgbError;
lastRGBCount = rgbSize;
if(rgbOnly)
{
sigmaVal = -1; //Signals the internal optimisation to weight evenly
}
Eigen::Matrix<float, 6, 6, Eigen::RowMajor> A_icp;
Eigen::Matrix<float, 6, 1> b_icp;
mat33 device_Rcurr = Rcurr;
float3 device_tcurr = *reinterpret_cast<float3*>(tcurr.data());
DeviceArray2D<float>& vmap_curr = vmaps_curr_[i];
DeviceArray2D<float>& nmap_curr = nmaps_curr_[i];
DeviceArray2D<float>& vmap_g_prev = vmaps_g_prev_[i];
DeviceArray2D<float>& nmap_g_prev = nmaps_g_prev_[i];
float residual[2];
if(icp)
{
TICK("icpStep");
icpStep(device_Rcurr,
device_tcurr,
vmap_curr,
nmap_curr,
device_Rprev_inv,
device_tprev,
intr(i),
vmap_g_prev,
nmap_g_prev,
distThres_,
angleThres_,
sumDataSE3,
outDataSE3,
A_icp.data(),
b_icp.data(),
&residual[0],
GPUConfig::getInstance().icpStepThreads,
GPUConfig::getInstance().icpStepBlocks);
TOCK("icpStep");
}
lastICPError = sqrt(residual[0]) / residual[1];
lastICPCount = residual[1];
Eigen::Matrix<float, 6, 6, Eigen::RowMajor> A_rgbd;
Eigen::Matrix<float, 6, 1> b_rgbd;
if(rgb)
{
TICK("rgbStep");
rgbStep(corresImg[i],
sigmaVal,
pointClouds[i],
intr(i).fx,
intr(i).fy,
nextdIdx[i],
nextdIdy[i],
sobelScale,
sumDataSE3,
outDataSE3,
A_rgbd.data(),
b_rgbd.data(),
GPUConfig::getInstance().rgbStepThreads,
GPUConfig::getInstance().rgbStepBlocks);
TOCK("rgbStep");
}
Eigen::Matrix<double, 6, 1> result;
Eigen::Matrix<double, 6, 6, Eigen::RowMajor> dA_rgbd = A_rgbd.cast<double>();
Eigen::Matrix<double, 6, 6, Eigen::RowMajor> dA_icp = A_icp.cast<double>();
Eigen::Matrix<double, 6, 1> db_rgbd = b_rgbd.cast<double>();
Eigen::Matrix<double, 6, 1> db_icp = b_icp.cast<double>();
if(icp && rgb)
{
double w = icpWeight;
lastA = dA_rgbd + w * w * dA_icp;
lastb = db_rgbd + w * db_icp;
result = lastA.ldlt().solve(lastb);
}
else if(icp)
{
lastA = dA_icp;
lastb = db_icp;
result = lastA.ldlt().solve(lastb);
}
else if(rgb)
{
lastA = dA_rgbd;
lastb = db_rgbd;
result = lastA.ldlt().solve(lastb);
}
else
{
assert(false && "Control shouldn't reach here");
}
Eigen::Isometry3f rgbOdom;
OdometryProvider::computeUpdateSE3(resultRt, result, rgbOdom);
Eigen::Isometry3f currentT;
currentT.setIdentity();
currentT.rotate(Rprev);
currentT.translation() = tprev;
currentT = currentT * rgbOdom.inverse();
tcurr = currentT.translation();
Rcurr = currentT.rotation();
}
}
if(rgb && (tcurr - tprev).norm() > 0.3)
{
Rcurr = Rprev;
tcurr = tprev;
}
if(so3)
{
for(int i = 0; i < NUM_PYRS; i++)
{
std::swap(lastNextImage[i], nextImage[i]);
}
}
trans = tcurr;
rot = Rcurr;
}
void Aligner::align() {
assert(_projector && "Aligner: missing _projector");
assert(_linearizer && "Aligner: missing _linearizer");
assert(_correspondenceFinder && "Aligner: missing _correspondenceFinder");
assert(_referenceCloud && "Aligner: missing _referenceCloud");
assert(_currentCloud && "Aligner: missing _currentCloud");
struct timeval tvStart, tvEnd;
gettimeofday(&tvStart, 0);
// The current points are seen from the frame of the sensor
_projector->setTransform(_currentSensorOffset);
_projector->project(_correspondenceFinder->currentIndexImage(),
_correspondenceFinder->currentDepthImage(),
_currentCloud->points());
_T = _initialGuess;
for(int i = 0; i < _outerIterations; i++) {
/************************************************************************
* Correspondence Computation *
************************************************************************/
// Compute the indices of the current scene from the point of view of the sensor
_T.matrix().row(3) << 0.0f, 0.0f, 0.0f, 1.0f;
_projector->setTransform(_T * _referenceSensorOffset);
_projector->project(_correspondenceFinder->referenceIndexImage(),
_correspondenceFinder->referenceDepthImage(),
_referenceCloud->points());
// Correspondences computation.
_correspondenceFinder->compute(*_referenceCloud, *_currentCloud, _T.inverse());
/************************************************************************
* Alignment *
************************************************************************/
Eigen::Isometry3f invT = _T.inverse();
for(int k = 0; k < _innerIterations; k++) {
invT.matrix().block<1, 4>(3, 0) << 0.0f, 0.0f, 0.0f, 1.0f;
Matrix6f H;
Vector6f b;
_linearizer->setT(invT);
_linearizer->update();
H = _linearizer->H() + Matrix6f::Identity();
b = _linearizer->b();
H += Matrix6f::Identity() * 1000.0f;
// Add the priors
for(size_t j = 0; j < _priors.size(); j++) {
const SE3Prior *prior = _priors[j];
Vector6f priorError = prior->error(invT);
Matrix6f priorJacobian = prior->jacobian(invT);
Matrix6f priorInformationRemapped = prior->errorInformation(invT);
Matrix6f Hp = priorJacobian.transpose() * priorInformationRemapped * priorJacobian;
Vector6f bp = priorJacobian.transpose() * priorInformationRemapped * priorError;
H += Hp;
b += bp;
}
Vector6f dx = H.ldlt().solve(-b);
Eigen::Isometry3f dT = v2t(dx);
invT = dT * invT;
}
_T = invT.inverse();
_T = v2t(t2v(_T));
_T.matrix().block<1, 4>(3, 0) << 0.0f, 0.0f, 0.0f, 1.0f;
}
gettimeofday(&tvEnd, 0);
double tStart = tvStart.tv_sec * 1000.0f + tvStart.tv_usec * 0.001f;
double tEnd = tvEnd.tv_sec * 1000.0f + tvEnd.tv_usec * 0.001f;
_totalTime = tEnd - tStart;
_error = _linearizer->error();
_inliers = _linearizer->inliers();
_computeStatistics(_mean, _omega, _translationalEigenRatio, _rotationalEigenRatio);
if (_rotationalEigenRatio > _rotationalMinEigenRatio ||
_translationalEigenRatio > _translationalMinEigenRatio) {
if (_debug) {
cerr << endl;
cerr << "************** WARNING SOLUTION MIGHT BE INVALID (eigenratio failure) **************" << endl;
cerr << "tr: " << _translationalEigenRatio << " rr: " << _rotationalEigenRatio << endl;
cerr << "************************************************************************************" << endl;
}
}
else {
if (_debug) {
cerr << "************** I FOUND SOLUTION VALID SOLUTION (eigenratio ok) *******************" << endl;
cerr << "tr: " << _translationalEigenRatio << " rr: " << _rotationalEigenRatio << endl;
cerr << "************************************************************************************" << endl;
}
}
if (_debug) {
cout << "Solution statistics in (t, mq): " << endl;
cout << "mean: " << _mean.transpose() << endl;
cout << "Omega: " << endl;
cout << _omega << endl;
}
}