PointViewSet RadiusOutlierFilter::run(PointViewPtr input)
{
bool logOutput = log()->getLevel() > LogLevel::Debug1;
if (logOutput)
log()->floatPrecision(8);
log()->get(LogLevel::Debug2) << "Process RadiusOutlierFilter...\n";
// convert PointView to PointXYZ
typedef pcl::PointCloud<pcl::PointXYZ> Cloud;
Cloud::Ptr cloud(new Cloud);
BOX3D bounds;
input->calculateBounds(bounds);
pclsupport::PDALtoPCD(input, *cloud, bounds);
pclsupport::setLogLevel(log()->getLevel());
// setup the outlier filter
pcl::RadiusOutlierRemoval<pcl::PointXYZ> ror(true);
ror.setInputCloud(cloud);
ror.setMinNeighborsInRadius(m_min_neighbors);
ror.setRadiusSearch(m_radius);
pcl::PointCloud<pcl::PointXYZ> output;
ror.setNegative(true);
ror.filter(output);
// filtered to return inliers
pcl::PointIndicesPtr inliers(new pcl::PointIndices);
ror.getRemovedIndices(*inliers);
PointViewSet viewSet;
if (inliers->indices.empty())
{
log()->get(LogLevel::Warning) << "Requested filter would remove all points. Try a larger radius/smaller minimum neighbors.\n";
viewSet.insert(input);
return viewSet;
}
// inverse are the outliers
std::vector<int> outliers(input->size()-inliers->indices.size());
for (PointId i = 0, j = 0, k = 0; i < input->size(); ++i)
{
if (i == (PointId)inliers->indices[j])
{
j++;
continue;
}
outliers[k++] = i;
}
if (!outliers.empty() && (m_classify || m_extract))
{
if (m_classify)
{
log()->get(LogLevel::Debug2) << "Labeled " << outliers.size() << " outliers as noise!\n";
// set the classification label of outlier returns as 18
// (corresponding to ASPRS LAS specification for high noise)
for (const auto& i : outliers)
{
input->setField(Dimension::Id::Classification, i, 18);
}
viewSet.insert(input);
}
if (m_extract)
{
log()->get(LogLevel::Debug2) << "Extracted " << inliers->indices.size() << " inliers!\n";
// create new PointView containing only outliers
PointViewPtr output = input->makeNew();
for (const auto& i : inliers->indices)
{
output->appendPoint(*input, i);
}
viewSet.erase(input);
viewSet.insert(output);
}
}
else
{
if (outliers.empty())
log()->get(LogLevel::Warning) << "Filtered cloud has no outliers!\n";
if (!(m_classify || m_extract))
log()->get(LogLevel::Warning) << "Must choose --classify or --extract\n";
// return the input buffer unchanged
viewSet.insert(input);
}
return viewSet;
}
PointViewSet StatisticalOutlierFilter::run(PointViewPtr input)
{
bool logOutput = log()->getLevel() > LogLevel::Debug1;
if (logOutput)
log()->floatPrecision(8);
log()->get(LogLevel::Debug2) << "Process StatisticalOutlierFilter...\n";
// convert PointView to PointXYZ
typedef pcl::PointCloud<pcl::PointXYZ> Cloud;
Cloud::Ptr cloud(new Cloud);
BOX3D bounds;
input->calculateBounds(bounds);
pclsupport::PDALtoPCD(input, *cloud, bounds);
// PCL should provide console output at similar verbosity level as PDAL
int level = log()->getLevel();
switch (level)
{
case 0:
pcl::console::setVerbosityLevel(pcl::console::L_ALWAYS);
break;
case 1:
pcl::console::setVerbosityLevel(pcl::console::L_ERROR);
break;
case 2:
pcl::console::setVerbosityLevel(pcl::console::L_WARN);
break;
case 3:
pcl::console::setVerbosityLevel(pcl::console::L_INFO);
break;
case 4:
pcl::console::setVerbosityLevel(pcl::console::L_DEBUG);
break;
default:
pcl::console::setVerbosityLevel(pcl::console::L_VERBOSE);
break;
}
// setup the outlier filter
pcl::StatisticalOutlierRemoval<pcl::PointXYZ> sor(true);
sor.setInputCloud(cloud);
sor.setMeanK(m_meanK);
sor.setStddevMulThresh(m_multiplier);
pcl::PointCloud<pcl::PointXYZ> output;
sor.setNegative(true);
sor.filter(output);
// filtered to return inliers
pcl::PointIndicesPtr inliers(new pcl::PointIndices);
sor.getRemovedIndices(*inliers);
log()->get(LogLevel::Debug2) << inliers->indices.size() << std::endl;
PointViewSet viewSet;
if (inliers->indices.empty())
{
log()->get(LogLevel::Warning) << "Requested filter would remove all points. Try increasing the multiplier.\n";
viewSet.insert(input);
return viewSet;
}
// inverse are the outliers
std::vector<int> outliers(input->size()-inliers->indices.size());
for (PointId i = 0, j = 0, k = 0; i < input->size(); ++i)
{
if (i == (PointId)inliers->indices[j])
{
j++;
continue;
}
outliers[k++] = i;
}
if (!outliers.empty() && (m_classify || m_extract))
{
if (m_classify)
{
log()->get(LogLevel::Debug2) << "Labeled " << outliers.size() << " outliers as noise!\n";
// set the classification label of outlier returns as 18
// (corresponding to ASPRS LAS specification for high noise)
for (const auto& i : outliers)
{
input->setField(Dimension::Id::Classification, i, 18);
}
viewSet.insert(input);
}
if (m_extract)
{
log()->get(LogLevel::Debug2) << "Extracted " << inliers->indices.size() << " inliers!\n";
// create new PointView containing only outliers
PointViewPtr output = input->makeNew();
for (const auto& i : inliers->indices)
{
output->appendPoint(*input, i);
}
viewSet.erase(input);
viewSet.insert(output);
}
}
else
{
if (outliers.empty())
log()->get(LogLevel::Warning) << "Filtered cloud has no outliers!\n";
if (!(m_classify || m_extract))
log()->get(LogLevel::Warning) << "Must choose --classify or --extract\n";
// return the input buffer unchanged
viewSet.insert(input);
}
return viewSet;
}
