mirror of
https://github.com/introlab/rtabmap.git
synced 2026-10-04 09:07:47 +08:00
finished util3d_filtering doc and tests
This commit is contained in:
matlabbe
@@ -1342,4 +1342,596 @@ TEST(Util3dFiltering, frustumFilteringInvalidClipPlaneTriggersAssertion)
|
||||
Transform cameraPose = Transform::getIdentity();
|
||||
|
||||
EXPECT_THROW(util3d::frustumFiltering(cloud, indices, cameraPose, 60.0f, 45.0f, 5.0f, 1.0f, false), UException);
|
||||
}
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, removeNaNFromPointCloud)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// Add valid points
|
||||
cloud->push_back(pcl::PointXYZ(1.0f, 2.0f, 3.0f));
|
||||
cloud->push_back(pcl::PointXYZ(4.0f, 5.0f, 6.0f));
|
||||
|
||||
// Add NaN point
|
||||
cloud->push_back(pcl::PointXYZ(
|
||||
std::numeric_limits<float>::quiet_NaN(),
|
||||
std::numeric_limits<float>::quiet_NaN(),
|
||||
std::numeric_limits<float>::quiet_NaN()));
|
||||
|
||||
cloud->is_dense = false;
|
||||
|
||||
ASSERT_EQ(cloud->size(), 3u);
|
||||
|
||||
auto filtered = util3d::removeNaNFromPointCloud(cloud);
|
||||
|
||||
// Check that the NaN point was removed
|
||||
EXPECT_EQ(filtered->size(), 2u);
|
||||
|
||||
// Check that the remaining points match the original valid points
|
||||
EXPECT_FLOAT_EQ(filtered->points[0].x, 1.0f);
|
||||
EXPECT_FLOAT_EQ(filtered->points[0].y, 2.0f);
|
||||
EXPECT_FLOAT_EQ(filtered->points[0].z, 3.0f);
|
||||
|
||||
EXPECT_FLOAT_EQ(filtered->points[1].x, 4.0f);
|
||||
EXPECT_FLOAT_EQ(filtered->points[1].y, 5.0f);
|
||||
EXPECT_FLOAT_EQ(filtered->points[1].z, 6.0f);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, removeNaNNormalsFromPointCloud)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointNormal>::Ptr(new pcl::PointCloud<pcl::PointNormal>);
|
||||
|
||||
// Add valid point with proper normals
|
||||
pcl::PointNormal pt1;
|
||||
pt1.x = pt1.y = pt1.z = 0.0f;
|
||||
pt1.normal_x = 1.0f; pt1.normal_y = 0.0f; pt1.normal_z = 0.0f;
|
||||
cloud->push_back(pt1);
|
||||
|
||||
// Add point with NaN normal_x
|
||||
pcl::PointNormal pt2 = pt1;
|
||||
pt2.normal_x = std::numeric_limits<float>::quiet_NaN();
|
||||
cloud->push_back(pt2);
|
||||
|
||||
// Add point with NaN normal_y
|
||||
pcl::PointNormal pt3 = pt1;
|
||||
pt3.normal_y = std::numeric_limits<float>::quiet_NaN();
|
||||
cloud->push_back(pt3);
|
||||
|
||||
// Add point with NaN normal_z
|
||||
pcl::PointNormal pt4 = pt1;
|
||||
pt4.normal_z = std::numeric_limits<float>::quiet_NaN();
|
||||
cloud->push_back(pt4);
|
||||
|
||||
ASSERT_EQ(cloud->size(), 4u);
|
||||
|
||||
auto filtered = util3d::removeNaNNormalsFromPointCloud(cloud);
|
||||
|
||||
// Only the valid point should remain
|
||||
ASSERT_EQ(filtered->size(), 1u);
|
||||
EXPECT_FLOAT_EQ(filtered->points[0].normal_x, 1.0f);
|
||||
EXPECT_FLOAT_EQ(filtered->points[0].normal_y, 0.0f);
|
||||
EXPECT_FLOAT_EQ(filtered->points[0].normal_z, 0.0f);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, radiusFilteringWithFewNeighbors)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// Create a simple grid of points: 3 points within radius, 2 isolated
|
||||
cloud->push_back(pcl::PointXYZ(0.0f, 0.0f, 0.0f)); // Index 0
|
||||
cloud->push_back(pcl::PointXYZ(0.05f, 0.0f, 0.0f)); // Index 1
|
||||
cloud->push_back(pcl::PointXYZ(0.1f, 0.0f, 0.0f)); // Index 2
|
||||
cloud->push_back(pcl::PointXYZ(5.0f, 0.0f, 0.0f)); // Index 3 (isolated)
|
||||
cloud->push_back(pcl::PointXYZ(-5.0f, 0.0f, 0.0f)); // Index 4 (isolated)
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>()); // Empty -> test whole cloud
|
||||
|
||||
// Radius 0.11, minimum 1 neighbor (excluding self, so 2 including self)
|
||||
float radiusSearch = 0.11f;
|
||||
int minNeighbors = 1;
|
||||
|
||||
auto filtered = util3d::radiusFiltering(cloud, indices, radiusSearch, minNeighbors);
|
||||
|
||||
// Only indices 0, 1, 2 are near each other; 3 and 4 are too far
|
||||
ASSERT_EQ(filtered->size(), 3u);
|
||||
EXPECT_EQ(filtered->at(0), 0);
|
||||
EXPECT_EQ(filtered->at(1), 1);
|
||||
EXPECT_EQ(filtered->at(2), 2);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, radiusFilteringProvidedIndicesSubset)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// Add a mix of close and isolated points
|
||||
cloud->push_back(pcl::PointXYZ(0.0f, 0.0f, 0.0f)); // 0 - neighbor
|
||||
cloud->push_back(pcl::PointXYZ(0.05f, 0.0f, 0.0f)); // 1 - neighbor
|
||||
cloud->push_back(pcl::PointXYZ(1.0f, 0.0f, 0.0f)); // 2 - too far
|
||||
|
||||
pcl::IndicesPtr subset(new std::vector<int>{0, 1}); // Only test points 0 and 1
|
||||
|
||||
float radiusSearch = 0.1f;
|
||||
int minNeighbors = 1;
|
||||
|
||||
auto filtered = util3d::radiusFiltering(cloud, subset, radiusSearch, minNeighbors);
|
||||
|
||||
ASSERT_EQ(filtered->size(), 2u);
|
||||
EXPECT_EQ(filtered->at(0), 0);
|
||||
EXPECT_EQ(filtered->at(1), 1);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, proportionalRadiusFilteringBasic)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// Points seen from 10 meters away, so ~5 cm radius
|
||||
cloud->push_back(pcl::PointXYZ(0, 0, 0)); // 0, farthest is 2 (4cm < 10cm) where 10cm is error viewpoint 1 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(0.001, 0, 0)); // 1, farthest is 2 (3.9cm < 10cm) where 10cm is error viewpoint 1 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(0.04f, 0, 0)); // 2, farthest is 0 (4cm < 10cm) where 10cm is error viewpoint 1 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(0.06f, 0, 0)); // 3, farthest is 2 (2cm < 10cm) where 10cm is error viewpoint 1 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(0.15f, 0, 0)); // 4, alone, kept
|
||||
|
||||
// Points seen from 10 meters away, so ~50 cm radius
|
||||
cloud->push_back(pcl::PointXYZ(0, 0, 0)); // 5, reject because of 4 (15cm > 10cm) where 10cm is error viewpoint 1 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(0.4f, 0, 0)); // 6, reject because of 0 (40cm > 10cm) where 10cm is error viewpoint 1 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(0.7f, 0, 0)); // 7, farthest is 6 (30cm < 100cm) where 100cm is error viewpoint 2 times neighborScale
|
||||
cloud->push_back(pcl::PointXYZ(1.5f, 0, 0)); // 8, alone, kept
|
||||
|
||||
std::vector<int> viewpointIndices = {0, 0, 0, 0, 0, 1, 1, 1, 1};
|
||||
std::map<int, Transform> viewpoints;
|
||||
viewpoints[0] = Transform(-1, 0, 0, 0,0,0);
|
||||
viewpoints[1] = Transform(-10, 0, 0, 0,0,0);
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>()); // process all
|
||||
float factor = 0.05f; // 6 cm radius at 1 meter, 60 cm radius at 10 meters
|
||||
float neighborScale = 2.0f;
|
||||
|
||||
pcl::IndicesPtr result = util3d::proportionalRadiusFiltering(
|
||||
cloud, indices, viewpointIndices, viewpoints, factor, neighborScale
|
||||
);
|
||||
|
||||
std::vector<int> expected = {0,1,2,3,4,7,8};
|
||||
ASSERT_EQ(result->size(), expected.size());
|
||||
for (int idx : *result) {
|
||||
EXPECT_NE(std::find(expected.begin(), expected.end(), idx), expected.end());
|
||||
}
|
||||
|
||||
// If we adjust neighborScale to 4, 5 would be accepted
|
||||
result = util3d::proportionalRadiusFiltering(
|
||||
cloud, indices, viewpointIndices, viewpoints, factor, 4.0f
|
||||
);
|
||||
std::vector<int> expected2 = {0,1,2,3,4,5,7,8};
|
||||
ASSERT_EQ(result->size(), expected2.size());
|
||||
for (int idx : *result) {
|
||||
EXPECT_NE(std::find(expected2.begin(), expected2.end(), idx), expected2.end());
|
||||
}
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, proportionalRadiusFilteringUsingProvidedIndices)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
cloud->push_back(pcl::PointXYZ(0, 0, 0)); // 0
|
||||
cloud->push_back(pcl::PointXYZ(0.05, 0, 0)); // 1
|
||||
cloud->push_back(pcl::PointXYZ(0.1, 0, 0)); // 2
|
||||
cloud->push_back(pcl::PointXYZ(10, 0, 0)); // 3 (far)
|
||||
|
||||
std::vector<int> viewpointIndices = {0, 0, 0, 1};
|
||||
std::map<int, Transform> viewpoints = {
|
||||
{0, Transform(0, 0, -1)},
|
||||
{1, Transform(10, 0, -1)}
|
||||
};
|
||||
|
||||
// Only test indices 0 and 3
|
||||
pcl::IndicesPtr indices(new std::vector<int>{0, 3});
|
||||
float factor = 0.1f;
|
||||
float neighborScale = 1.5f;
|
||||
|
||||
auto result = util3d::proportionalRadiusFiltering(
|
||||
cloud, indices, viewpointIndices, viewpoints, factor, neighborScale);
|
||||
|
||||
// Point 0 has no neighbors in filtered subset, 3 too far
|
||||
ASSERT_EQ(result->size(), 0u);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, proportionalRadiusFilteringAllPointsNoNeighbors)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
cloud->push_back(pcl::PointXYZ(0, 0, 0));
|
||||
cloud->push_back(pcl::PointXYZ(10, 0, 0));
|
||||
cloud->push_back(pcl::PointXYZ(20, 0, 0));
|
||||
|
||||
std::vector<int> viewpointIndices = {0, 1, 2};
|
||||
std::map<int, Transform> viewpoints = {
|
||||
{0, Transform(0, 0, -1)},
|
||||
{1, Transform(10, 0, -1)},
|
||||
{2, Transform(20, 0, -1)}
|
||||
};
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>);
|
||||
float factor = 0.001f; // tiny radius
|
||||
float neighborScale = 1.0f;
|
||||
|
||||
auto result = util3d::proportionalRadiusFiltering(
|
||||
cloud, indices, viewpointIndices, viewpoints, factor, neighborScale);
|
||||
|
||||
ASSERT_EQ(result->size(), 0u);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, proportionalRadiusFilteringInvalidViewpointID)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
cloud->push_back(pcl::PointXYZ(0, 0, 0));
|
||||
|
||||
std::vector<int> viewpointIndices = {42}; // invalid ID
|
||||
std::map<int, Transform> viewpoints = {
|
||||
{0, Transform(1, 0, 0)}
|
||||
};
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>);
|
||||
float factor = 0.1f;
|
||||
float neighborScale = 1.0f;
|
||||
|
||||
EXPECT_THROW(util3d::proportionalRadiusFiltering(cloud, indices, viewpointIndices, viewpoints, factor, neighborScale), UException);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, subtractFilteringBasicFiltering)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
cloud->push_back(pcl::PointXYZ(0, 0, 0));
|
||||
cloud->push_back(pcl::PointXYZ(1, 0, 0));
|
||||
cloud->push_back(pcl::PointXYZ(2, 0, 0));
|
||||
pcl::IndicesPtr indices(new std::vector<int>{0, 1, 2});
|
||||
|
||||
auto subtractCloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
subtractCloud->push_back(pcl::PointXYZ(1.05, 0, 0));
|
||||
pcl::IndicesPtr subtractIndices(new std::vector<int>); // Empty = use all
|
||||
|
||||
float radiusSearch = 0.2f;
|
||||
int minNeighborsInRadius = 1;
|
||||
|
||||
auto result = util3d::subtractFiltering(
|
||||
cloud,
|
||||
indices,
|
||||
subtractCloud,
|
||||
subtractIndices,
|
||||
radiusSearch,
|
||||
minNeighborsInRadius);
|
||||
|
||||
// Expected: point 1 has a neighbor within radius -> excluded
|
||||
// Points 0 and 2 do not -> included
|
||||
ASSERT_EQ(result->size(), 2);
|
||||
EXPECT_EQ(result->at(0), 0);
|
||||
EXPECT_EQ(result->at(1), 2);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, subtractAdaptiveFilteringXYZRGB)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZRGB>::Ptr(new pcl::PointCloud<pcl::PointXYZRGB>);
|
||||
auto subtractCloud = pcl::PointCloud<pcl::PointXYZRGB>::Ptr(new pcl::PointCloud<pcl::PointXYZRGB>);
|
||||
|
||||
// Input cloud: 3 points
|
||||
pcl::PointXYZRGB pt;
|
||||
pt.z = 1;
|
||||
cloud->push_back(pt); // (0 ,0, 1)
|
||||
pt.x = 1;
|
||||
cloud->push_back(pt); // (1 ,0, 1)
|
||||
pt.x = 2;
|
||||
cloud->push_back(pt); // (2 ,0, 1)
|
||||
|
||||
// Subtract cloud: one neighbor close to point 1
|
||||
pt.x = 1.05;
|
||||
subtractCloud->push_back(pt); // (1.05 ,0, 1)
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>{0, 1, 2});
|
||||
pcl::IndicesPtr subtractIndices(new std::vector<int>); // use full subtract cloud
|
||||
|
||||
float radiusSearchRatio = 0.2f; // Adaptive search radius
|
||||
int minNeighborsInRadius = 1;
|
||||
Eigen::Vector3f viewpoint(0, 0, 0);
|
||||
|
||||
pcl::IndicesPtr result = util3d::subtractAdaptiveFiltering(
|
||||
cloud, indices, subtractCloud, subtractIndices,
|
||||
radiusSearchRatio, minNeighborsInRadius, viewpoint
|
||||
);
|
||||
|
||||
// Point 1 should be excluded (has a neighbor), others retained
|
||||
ASSERT_EQ(result->size(), 2);
|
||||
EXPECT_EQ(result->at(0), 0);
|
||||
EXPECT_EQ(result->at(1), 2);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, subtractAdaptiveFilteringXYZRGBNormal)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZRGBNormal>::Ptr(new pcl::PointCloud<pcl::PointXYZRGBNormal>);
|
||||
auto subtractCloud = pcl::PointCloud<pcl::PointXYZRGBNormal>::Ptr(new pcl::PointCloud<pcl::PointXYZRGBNormal>);
|
||||
|
||||
// Input point with normal (point 0)
|
||||
pcl::PointXYZRGBNormal pt;
|
||||
pt.x = 0;
|
||||
pt.y = 0;
|
||||
pt.z = 1;
|
||||
pt.normal_x = 0;
|
||||
pt.normal_y = 0;
|
||||
pt.normal_z = 1;
|
||||
cloud->push_back(pt);
|
||||
|
||||
// Neighbor with almost same normal
|
||||
pcl::PointXYZRGBNormal neighbor1 = pt;
|
||||
neighbor1.x = 0.05f;
|
||||
neighbor1.normal_x = cos(89.0f*M_PI/180.0f);
|
||||
neighbor1.normal_z = sin(89.0f*M_PI/180.0f);
|
||||
subtractCloud->push_back(neighbor1);
|
||||
|
||||
// Neighbor with very different normal
|
||||
pcl::PointXYZRGBNormal neighbor2 = pt;
|
||||
neighbor2.x = 0.05f;
|
||||
neighbor2.normal_x = 1;
|
||||
neighbor2.normal_y = 0;
|
||||
neighbor2.normal_z = 0;
|
||||
subtractCloud->push_back(neighbor2);
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>{0});
|
||||
pcl::IndicesPtr subtractIndices(new std::vector<int>); // full cloud
|
||||
|
||||
float radiusSearchRatio = 0.2f;
|
||||
float maxAngle = 20.0f*M_PI/180.0f; // 20 degrees max
|
||||
int minNeighborsInRadius = 1;
|
||||
Eigen::Vector3f viewpoint(0, 0, 0);
|
||||
|
||||
pcl::IndicesPtr result = util3d::subtractAdaptiveFiltering(
|
||||
cloud, indices, subtractCloud, subtractIndices,
|
||||
radiusSearchRatio, maxAngle, minNeighborsInRadius, viewpoint
|
||||
);
|
||||
|
||||
// neighbor2 exceeds angle threshold and should be ignored
|
||||
// neighbor1 should pass if included; result depends on number of valid neighbors
|
||||
ASSERT_EQ(result->size(), 0) << "Point should be excluded due to having a valid neighbor";
|
||||
|
||||
// Now set maxAngle very low so even neighbor1 is excluded
|
||||
result = util3d::subtractAdaptiveFiltering(
|
||||
cloud, indices, subtractCloud, subtractIndices,
|
||||
radiusSearchRatio, 0.5f*M_PI/180.0f, minNeighborsInRadius, viewpoint
|
||||
);
|
||||
|
||||
// No valid neighbors left -> point should be retained
|
||||
ASSERT_EQ(result->size(), 1);
|
||||
EXPECT_EQ(result->at(0), 0);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, normalFilteringBasic)
|
||||
{
|
||||
// Create a flat plane of points along XY plane (normals should be along Z)
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
for (float x = -1.0f; x <= 1.0f; x += 0.5f)
|
||||
{
|
||||
for (float y = -1.0f; y <= 1.0f; y += 0.5f)
|
||||
{
|
||||
cloud->push_back(pcl::PointXYZ(x, y, 0.0f));
|
||||
}
|
||||
}
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>()); // Use all points
|
||||
|
||||
// We expect normals to be [0, 0, 1]
|
||||
Eigen::Vector4f refNormal(0, 0, 1, 0);
|
||||
float angleMax = 10.0f*M_PI/180.0f; // Allow 10 degrees of deviation
|
||||
int normalKSearch = 5;
|
||||
Eigen::Vector4f viewpoint(0, 0, 1, 0);
|
||||
float groundNormalsUp = 0.0f; // No flipping
|
||||
|
||||
auto result = util3d::normalFiltering(
|
||||
cloud,
|
||||
indices,
|
||||
angleMax,
|
||||
refNormal,
|
||||
normalKSearch,
|
||||
viewpoint,
|
||||
groundNormalsUp);
|
||||
|
||||
// Expect all points to pass (normals pointing up)
|
||||
ASSERT_EQ(result->size(), cloud->size());
|
||||
|
||||
for (int idx : *result)
|
||||
{
|
||||
EXPECT_GE(idx, 0);
|
||||
EXPECT_LT(idx, static_cast<int>(cloud->size()));
|
||||
}
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, normalFilteringRejectNormalsWithLargeDeviation)
|
||||
{
|
||||
// Construct a simple slanted surface so normals will not be vertical
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
for (float x = -1.0f; x <= 1.0f; x += 1.0f)
|
||||
{
|
||||
cloud->push_back(pcl::PointXYZ(x, 0, x)); // Diagonal line: normals ≈ 45 degrees
|
||||
}
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>()); // Use all
|
||||
|
||||
Eigen::Vector4f refNormal(0, 0, 1, 0); // Want normals close to Z
|
||||
float angleMax = 10.0f*M_PI/180.0f; // Tight filter: only near vertical normals
|
||||
int normalKSearch = 3;
|
||||
Eigen::Vector4f viewpoint(0, 0, 0, 0);
|
||||
float groundNormalsUp = 0.0f;
|
||||
|
||||
auto result = util3d::normalFiltering(
|
||||
cloud,
|
||||
indices,
|
||||
angleMax,
|
||||
refNormal,
|
||||
normalKSearch,
|
||||
viewpoint,
|
||||
groundNormalsUp);
|
||||
|
||||
// Expect points to be rejected due to large angle from reference normal
|
||||
ASSERT_EQ(result->size(), 0);
|
||||
}
|
||||
|
||||
// Test for PointXYZ type
|
||||
TEST(Util3dFiltering, extractClustersBasic)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// Generate 3 distinct clusters: centered at (0,0), (5,0), and (10,0)
|
||||
for (float i = 0; i < 3; ++i)
|
||||
{
|
||||
float cx = i * 5.0f;
|
||||
for (int j = 0; j < 10; ++j)
|
||||
{
|
||||
cloud->push_back(pcl::PointXYZ(cx + static_cast<float>(rand()) / RAND_MAX * 0.1f,
|
||||
static_cast<float>(rand()) / RAND_MAX * 0.1f,
|
||||
0.0f));
|
||||
}
|
||||
}
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>()); // Use all points
|
||||
float clusterTolerance = 1.0f;
|
||||
int minClusterSize = 3;
|
||||
int maxClusterSize = 20;
|
||||
int biggestClusterIdx = -1;
|
||||
|
||||
auto clusters = util3d::extractClusters(
|
||||
cloud, indices, clusterTolerance,
|
||||
minClusterSize, maxClusterSize, &biggestClusterIdx);
|
||||
|
||||
// We should find 3 clusters
|
||||
ASSERT_EQ(clusters.size(), 3);
|
||||
|
||||
// Each cluster should have 10 points
|
||||
for (const auto& cluster : clusters)
|
||||
{
|
||||
ASSERT_EQ(cluster->size(), 10);
|
||||
}
|
||||
|
||||
// Largest cluster index should be valid
|
||||
ASSERT_GE(biggestClusterIdx, 0);
|
||||
ASSERT_LT(biggestClusterIdx, static_cast<int>(clusters.size()));
|
||||
ASSERT_EQ(clusters[biggestClusterIdx]->size(), 10);
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, extractClustersMinClusterSize)
|
||||
{
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// One big cluster and one tiny cluster
|
||||
for (int i = 0; i < 10; ++i)
|
||||
cloud->push_back(pcl::PointXYZ(i * 0.05f, 0, 0)); // Close points
|
||||
|
||||
cloud->push_back(pcl::PointXYZ(10.0f, 0, 0)); // Isolated point
|
||||
|
||||
pcl::IndicesPtr indices(new std::vector<int>());
|
||||
float clusterTolerance = 0.2f;
|
||||
int minClusterSize = 2;
|
||||
int maxClusterSize = 20;
|
||||
int biggestClusterIdx = -1;
|
||||
|
||||
auto clusters = util3d::extractClusters(
|
||||
cloud, indices, clusterTolerance,
|
||||
minClusterSize, maxClusterSize, &biggestClusterIdx);
|
||||
|
||||
// Should only find one cluster (ignoring the single point)
|
||||
ASSERT_EQ(clusters.size(), 1);
|
||||
ASSERT_EQ(clusters[0]->size(), 10);
|
||||
ASSERT_EQ(biggestClusterIdx, 0);
|
||||
}
|
||||
|
||||
// Test case: Extract the points corresponding to the specified indices
|
||||
TEST(Util3dFiltering, extractIndices) {
|
||||
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
cloud->resize(5);
|
||||
|
||||
// Assigning values to the cloud points
|
||||
for (size_t i = 0; i < cloud->points.size(); ++i) {
|
||||
cloud->points[i].x = static_cast<float>(i);
|
||||
cloud->points[i].y = static_cast<float>(i * 2);
|
||||
cloud->points[i].z = static_cast<float>(i * 3);
|
||||
}
|
||||
|
||||
// Creating a set of indices to extract (e.g., indices 1 and 3)
|
||||
pcl::IndicesPtr indices(new std::vector<int>({1, 3}));
|
||||
|
||||
pcl::IndicesPtr output = util3d::extractIndices(cloud, indices, false);
|
||||
|
||||
// Verify the output contains the correct indices
|
||||
ASSERT_EQ(output->size(), 2);
|
||||
EXPECT_EQ(output->at(0), 1); // Index 1 should be extracted
|
||||
EXPECT_EQ(output->at(1), 3); // Index 3 should be extracted
|
||||
|
||||
// Verify the points corresponding to those indices
|
||||
EXPECT_FLOAT_EQ(cloud->points[output->at(0)].x, 1.0);
|
||||
EXPECT_FLOAT_EQ(cloud->points[output->at(1)].x, 3.0);
|
||||
|
||||
output = util3d::extractIndices(cloud, indices, true);
|
||||
|
||||
// Verify the output contains the correct indices (everything except 1 and 3)
|
||||
ASSERT_EQ(output->size(), 3); // There should be 3 points left (0, 2, 4)
|
||||
EXPECT_EQ(output->at(0), 0); // Index 0 should be included
|
||||
EXPECT_EQ(output->at(1), 2); // Index 2 should be included
|
||||
EXPECT_EQ(output->at(2), 4); // Index 4 should be included
|
||||
|
||||
// Verify that the excluded points are not in the output
|
||||
EXPECT_NE(output->at(0), 1); // Index 1 should not be in the output
|
||||
EXPECT_NE(output->at(1), 3); // Index 3 should not be in the output
|
||||
}
|
||||
|
||||
TEST(Util3dFiltering, extractPlane) {
|
||||
|
||||
auto cloud = pcl::PointCloud<pcl::PointXYZ>::Ptr(new pcl::PointCloud<pcl::PointXYZ>);
|
||||
|
||||
// Define some points that should form a plane
|
||||
cloud->push_back(pcl::PointXYZ(0.0, 0.0, 0.001));
|
||||
cloud->push_back(pcl::PointXYZ(1.0, 0.0, 0.002));
|
||||
cloud->push_back(pcl::PointXYZ(0.0, 1.0, -0.001));
|
||||
cloud->push_back(pcl::PointXYZ(1.0, 1.0, 0.002)); // All points are close to z = 0 plane
|
||||
|
||||
// Optionally, create a subset of indices to test
|
||||
pcl::IndicesPtr indices(new std::vector<int>{0, 1, 2, 3});
|
||||
|
||||
pcl::ModelCoefficients coefficients;
|
||||
pcl::IndicesPtr inliers = util3d::extractPlane(cloud, indices, 0.01, 100, &coefficients);
|
||||
|
||||
// Check if the extracted plane coefficients are correct
|
||||
EXPECT_LT(fabs(coefficients.values[0]), 0.01); // Plane normal x = 0
|
||||
EXPECT_LT(fabs(coefficients.values[1]), 0.01); // Plane normal y = 0
|
||||
EXPECT_GT(fabs(coefficients.values[2]), 0.99); // Plane normal z = 1 (for the z = 0 plane)
|
||||
EXPECT_LT(fabs(coefficients.values[3]), 0.01); // Plane offset (z = 0)
|
||||
|
||||
// Check that all the indices were identified as inliers
|
||||
EXPECT_EQ(inliers->size(), 4); // All points should be inliers
|
||||
|
||||
coefficients = pcl::ModelCoefficients();
|
||||
pcl::IndicesPtr subsetIndices = pcl::IndicesPtr(new std::vector<int>{0, 1, 2}); // Subset of points
|
||||
inliers = util3d::extractPlane(cloud, subsetIndices, 0.01, 100, &coefficients);
|
||||
|
||||
// Check if the extracted plane coefficients are correct
|
||||
EXPECT_LT(fabs(coefficients.values[0]), 0.01); // Plane normal x = 0
|
||||
EXPECT_LT(fabs(coefficients.values[1]), 0.01); // Plane normal y = 0
|
||||
EXPECT_GT(fabs(coefficients.values[2]), 0.99); // Plane normal z = 1 (for the z = 0 plane)
|
||||
EXPECT_LT(fabs(coefficients.values[3]), 0.01); // Plane offset (z = 0)
|
||||
|
||||
// Check that the correct number of inliers are found
|
||||
EXPECT_EQ(inliers->size(), 3); // Only 3 points were used
|
||||
|
||||
// Set a very strict distance threshold to ensure no inliers
|
||||
inliers = util3d::extractPlane(cloud, indices, 0.0001, 100, &coefficients);
|
||||
|
||||
// Check that no inliers are found
|
||||
EXPECT_EQ(inliers->size(), 3); // less than 4 inliers with such a small threshold
|
||||
|
||||
pcl::IndicesPtr emptyIndices; // Empty indices
|
||||
coefficients = pcl::ModelCoefficients();
|
||||
|
||||
// When indices are empty, the function should use the whole cloud
|
||||
inliers = util3d::extractPlane(cloud, emptyIndices, 0.01, 100, &coefficients);
|
||||
|
||||
// Check if the coefficients correspond to a plane in the z = 0 plane
|
||||
EXPECT_LT(fabs(coefficients.values[0]), 0.01); // Plane normal x = 0
|
||||
EXPECT_LT(fabs(coefficients.values[1]), 0.01); // Plane normal y = 0
|
||||
EXPECT_GT(fabs(coefficients.values[2]), 0.99); // Plane normal z = 1 (for the z = 0 plane)
|
||||
EXPECT_LT(fabs(coefficients.values[3]), 0.01); // Plane offset (z = 0)
|
||||
|
||||
// Check if all points are inliers (since we are using the whole cloud)
|
||||
EXPECT_EQ(inliers->size(), 4); // All points should be inliers
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user