small_gicp/src/test/registration_test.cpp

304 lines
12 KiB
C++

#include <gtest/gtest.h>
#include <pcl/point_types.h>
#include <pcl/point_cloud.h>
#include <small_gicp/util/downsampling.hpp>
#include <small_gicp/util/normal_estimation_omp.hpp>
#include <small_gicp/pcl/pcl_point.hpp>
#include <small_gicp/pcl/pcl_point_traits.hpp>
#include <small_gicp/pcl/pcl_registration.hpp>
#include <small_gicp/pcl/pcl_registration_impl.hpp>
#include <small_gicp/points/point_cloud.hpp>
#include <small_gicp/factors/icp_factor.hpp>
#include <small_gicp/factors/gicp_factor.hpp>
#include <small_gicp/factors/plane_icp_factor.hpp>
#include <small_gicp/registration/reduction_omp.hpp>
#include <small_gicp/registration/reduction_tbb.hpp>
#include <small_gicp/registration/registration.hpp>
#include <small_gicp/benchmark/read_points.hpp>
#include <small_gicp/registration/registration_helper.hpp>
using namespace small_gicp;
class RegistrationTest : public testing::Test, public testing::WithParamInterface<std::tuple<const char*, const char*>> {
public:
void SetUp() override {
// Load points
const double downsampling_resolution = 0.3;
target = std::make_shared<PointCloud>(read_ply("data/target.ply"));
target = voxelgrid_sampling(*target, downsampling_resolution);
estimate_normals_covariances_omp(*target);
target_pcl = pcl::make_shared<pcl::PointCloud<pcl::PointNormalCovariance>>();
target_pcl->resize(target->size());
for (size_t i = 0; i < target->size(); i++) {
target_pcl->at(i).getVector4fMap() = target->point(i).cast<float>();
}
estimate_normals_covariances_omp(*target_pcl);
source = std::make_shared<PointCloud>(read_ply("data/source.ply"));
source = voxelgrid_sampling(*source, downsampling_resolution);
estimate_normals_covariances_omp(*source);
source_pcl = pcl::make_shared<pcl::PointCloud<pcl::PointNormalCovariance>>();
source_pcl->resize(source->size());
for (size_t i = 0; i < source->size(); i++) {
source_pcl->at(i).getVector4fMap() = source->point(i).cast<float>();
}
estimate_normals_covariances_omp(*source_pcl);
target_tree = std::make_shared<KdTree<PointCloud>>(target);
source_tree = std::make_shared<KdTree<PointCloud>>(source);
target_voxelmap = std::make_shared<GaussianVoxelMap>(1.0);
target_voxelmap->insert(*target);
source_voxelmap = std::make_shared<GaussianVoxelMap>(1.0);
source_voxelmap->insert(*source);
std::mt19937 mt;
std::uniform_real_distribution<> udist(-1.0, 1.0);
const int num_noise_poses = 4;
T_noise.resize(num_noise_poses);
for (int i = 0; i < num_noise_poses; i++) {
T_noise[i] = Eigen::Isometry3d::Identity();
if (i != 0) {
T_noise[i].translation() = Eigen::Vector3d(udist(mt), udist(mt), udist(mt)) * 0.5;
T_noise[i].linear() = Eigen::AngleAxisd(udist(mt) * 10.0 * M_PI / 180.0, Eigen::Vector3d(udist(mt), udist(mt), udist(mt)).normalized()).toRotationMatrix();
}
}
const int num_shifted_poses = 2;
T_source_shifted.resize(num_shifted_poses);
shifted.resize(num_shifted_poses);
for (int i = 0; i < num_shifted_poses; i++) {
T_source_shifted[i] = Eigen::Isometry3d::Identity();
T_source_shifted[i].translation() = Eigen::Vector3d(udist(mt), udist(mt), udist(mt)) * 2.0;
T_source_shifted[i].linear() = Eigen::AngleAxisd(udist(mt) * M_PI_2 + M_PI_2, Eigen::Vector3d(udist(mt), udist(mt), udist(mt)).normalized()).toRotationMatrix();
const Eigen::Isometry3d T_shifted_source = T_source_shifted[i].inverse();
shifted[i] = std::make_shared<PointCloud>(source->points);
std::ranges::transform(source->points, shifted[i]->points.begin(), [&](const auto& p) -> Eigen::Vector4d { return T_shifted_source * p; });
estimate_normals_covariances_omp(*shifted[i]);
}
std::ifstream ifs("data/T_target_source.txt");
if (!ifs) {
std::cerr << "error: failed to open T_target_source.txt" << std::endl;
}
for (int i = 0; i < 4; i++) {
for (int j = 0; j < 4; j++) {
ifs >> T_target_source(i, j);
}
}
}
template <typename Registration>
void test_registration(Registration& registration) {
for (int i = 0; i < T_noise.size(); i++) {
auto result = registration.align(*target, *source, *target_tree, T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_source, result.T_target_source)) << "Forward transformation T_noise=" << i;
result = registration.align(*target_pcl, *source_pcl, *target_tree, T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_source, result.T_target_source)) << "Forward transformation (pcl) T_noise=" << i;
result = registration.align(*source, *target, *source_tree, T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_source.inverse(), result.T_target_source)) << "Inverse transformation T_noise=" << i;
result = registration.align(*source_pcl, *target_pcl, *source_tree, T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_source.inverse(), result.T_target_source)) << "Inverse transformation (pcl) T_noise=" << i;
for (int j = 0; j < T_source_shifted.size(); j++) {
const Eigen::Isometry3d T_target_shifted = T_target_source * T_source_shifted[j];
auto result = registration.align(*target, *shifted[j], *target_tree, T_source_shifted[j] * T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_shifted, result.T_target_source)) << "Forward transformation T_source_shifted=" << j << " T_noise=" << i;
}
}
}
template <typename Registration>
void test_registration_vgicp(Registration& registration) {
for (int i = 0; i < T_noise.size(); i++) {
auto result = registration.align(*target_voxelmap, *source, *target_voxelmap, T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_source, result.T_target_source)) << "Forward transformation T_noise=" << i;
result = registration.align(*source_voxelmap, *target, *source_voxelmap, T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_source.inverse(), result.T_target_source)) << "Inverse transformation T_noise=" << i;
for (int j = 0; j < T_source_shifted.size(); j++) {
const Eigen::Isometry3d T_target_shifted = T_target_source * T_source_shifted[j];
auto result = registration.align(*target_voxelmap, *shifted[j], *target_voxelmap, T_source_shifted[j] * T_noise[i]);
EXPECT_TRUE(compare_transformation(T_target_shifted, result.T_target_source)) << "Forward transformation T_source_shifted=" << j << " T_noise=" << i;
}
}
}
bool compare_transformation(const Eigen::Isometry3d& T1, const Eigen::Isometry3d& T2) {
const Eigen::Isometry3d e = T1.inverse() * T2;
const double error_rot = Eigen::AngleAxisd(e.linear()).angle();
const double error_trans = e.translation().norm();
const double rot_tol = 2.5 * M_PI / 180.0;
const double trans_tol = 0.2;
EXPECT_NEAR(error_rot, 0.0, rot_tol);
EXPECT_NEAR(error_trans, 0.0, trans_tol);
return error_rot < rot_tol && error_trans < trans_tol;
}
protected:
PointCloud::Ptr target; ///< Target points
PointCloud::Ptr source; ///< Source points
pcl::PointCloud<pcl::PointNormalCovariance>::Ptr target_pcl; ///< Target points (pcl)
pcl::PointCloud<pcl::PointNormalCovariance>::Ptr source_pcl; ///< Source points (pcl)
KdTree<PointCloud>::Ptr target_tree; ///< Nearest neighbor search for the target points
KdTree<PointCloud>::Ptr source_tree; ///< Nearest neighbor search for the source points
GaussianVoxelMap::Ptr target_voxelmap; ///< Gaussian voxel map for the target points
GaussianVoxelMap::Ptr source_voxelmap; ///< Gaussian voxel map for the target points
Eigen::Isometry3d T_target_source; ///< Ground truth transformation
std::vector<Eigen::Isometry3d> T_noise;
std::vector<Eigen::Isometry3d> T_source_shifted;
std::vector<PointCloud::Ptr> shifted;
};
// Load check
TEST_F(RegistrationTest, LoadCheck) {
EXPECT_FALSE(target->empty());
EXPECT_FALSE(source->empty());
EXPECT_FALSE(target_pcl->empty());
EXPECT_FALSE(source_pcl->empty());
}
// PCL interface test
TEST_F(RegistrationTest, PCLInterfaceTest) {
RegistrationPCL<pcl::PointNormalCovariance, pcl::PointNormalCovariance> registration;
registration.setRegistrationType("GICP");
registration.setMaxCorrespondenceDistance(1.0);
// Forward align
registration.setInputTarget(target_pcl);
registration.setInputSource(source_pcl);
pcl::PointCloud<pcl::PointNormalCovariance> aligned;
registration.align(aligned);
EXPECT_EQ(aligned.size(), source_pcl->size());
EXPECT_TRUE(compare_transformation(T_target_source, Eigen::Isometry3d(registration.getFinalTransformation().cast<double>())));
// Swap and backward align
registration.swapSourceAndTarget();
registration.align(aligned);
EXPECT_EQ(aligned.size(), target_pcl->size());
EXPECT_TRUE(compare_transformation(T_target_source.inverse(), Eigen::Isometry3d(registration.getFinalTransformation().cast<double>())));
// Clear and forward align
registration.clearTarget();
registration.clearSource();
registration.setInputTarget(target_pcl);
registration.setInputSource(source_pcl);
registration.align(aligned);
EXPECT_EQ(aligned.size(), source_pcl->size());
EXPECT_TRUE(compare_transformation(T_target_source, Eigen::Isometry3d(registration.getFinalTransformation().cast<double>())));
registration.setRegistrationType("VGICP");
registration.setVoxelResolution(1.0);
// Forward align
registration.setInputTarget(target_pcl);
registration.setInputSource(source_pcl);
registration.align(aligned);
EXPECT_EQ(aligned.size(), source_pcl->size());
EXPECT_TRUE(compare_transformation(T_target_source, Eigen::Isometry3d(registration.getFinalTransformation().cast<double>())));
// Swap and backward align
registration.swapSourceAndTarget();
registration.align(aligned);
EXPECT_EQ(aligned.size(), target_pcl->size());
EXPECT_TRUE(compare_transformation(T_target_source.inverse(), Eigen::Isometry3d(registration.getFinalTransformation().cast<double>())));
// Clear and forward align
registration.clearTarget();
registration.clearSource();
registration.setInputTarget(target_pcl);
registration.setInputSource(source_pcl);
registration.align(aligned);
EXPECT_EQ(aligned.size(), source_pcl->size());
EXPECT_TRUE(compare_transformation(T_target_source, Eigen::Isometry3d(registration.getFinalTransformation().cast<double>())));
}
INSTANTIATE_TEST_SUITE_P(
RegistrationTest,
RegistrationTest,
testing::Combine(testing::Values("ICP", "PLANE_ICP", "GICP", "VGICP"), testing::Values("SERIAL", "TBB", "OMP")),
[](const auto& info) {
std::stringstream sst;
sst << std::get<0>(info.param) << "_" << std::get<1>(info.param);
return sst.str();
});
TEST_P(RegistrationTest, RegistrationTest) {
const std::string factor = std::get<0>(GetParam());
const std::string reduction = std::get<1>(GetParam());
if (factor == "ICP") {
if (reduction == "SERIAL") {
Registration<ICPFactor, SerialReduction> reg;
test_registration(reg);
} else if (reduction == "TBB") {
Registration<ICPFactor, ParallelReductionTBB> reg;
test_registration(reg);
} else if (reduction == "OMP") {
Registration<ICPFactor, ParallelReductionOMP> reg;
test_registration(reg);
}
} else if (factor == "PLANE_ICP") {
if (reduction == "SERIAL") {
Registration<PointToPlaneICPFactor, SerialReduction> reg;
test_registration(reg);
} else if (reduction == "TBB") {
Registration<PointToPlaneICPFactor, ParallelReductionTBB> reg;
test_registration(reg);
} else if (reduction == "OMP") {
Registration<PointToPlaneICPFactor, ParallelReductionOMP> reg;
test_registration(reg);
}
} else if (factor == "GICP") {
if (reduction == "SERIAL") {
Registration<GICPFactor, SerialReduction> reg;
test_registration(reg);
} else if (reduction == "TBB") {
Registration<GICPFactor, ParallelReductionTBB> reg;
test_registration(reg);
} else if (reduction == "OMP") {
Registration<GICPFactor, ParallelReductionOMP> reg;
test_registration(reg);
}
} else if (factor == "VGICP") {
if (reduction == "SERIAL") {
Registration<GICPFactor, SerialReduction> reg;
test_registration_vgicp(reg);
} else if (reduction == "TBB") {
Registration<GICPFactor, ParallelReductionTBB> reg;
test_registration_vgicp(reg);
} else if (reduction == "OMP") {
Registration<GICPFactor, ParallelReductionOMP> reg;
test_registration_vgicp(reg);
}
} else {
std::cerr << "error: unknown factor type " << factor << std::endl;
}
}