Point Cloud Plane Extraction

This tutorial demonstrates how to estimate one 3D plane.

To initialize the module,

dwPointCloudPlaneExtractorHandle_t estimator;
dwPointCloudPlaneExtractorParams params{};
dwPCPlaneExtractor_getDefaultParameters(&params);
dwPCPlaneExtractor_initialize(&estimator, &params, context);
dwPCPlaneExtractor_setCUDAStream(stream, estimator);

The plane estimation is based on ransac fitting and non-linear optimization. User can tailor params.ransacIterationCount and params.optimizerIterationCount to fulfill the expected performance. params.rotation gives the initial guess on the transformation from the point cloud to the 3D plane. params.boxFilterParams controls box filter parameters applied before ransac is executed. User is able to vary size of an area and number of points picked for ransac algorithm to make it working more effectively. If params.cudaPipelineEnabled = true, the module expects CUDA memory type point cloud.

To get the estimated 3D plane,

dwPCPlaneExtractor_bindInput(pointCloud, estimator);
dwPCPlaneExtractor_bindOutput(inliers, outliers, plane, estimator);
dwPCPlaneExtractor_process(estimator);

inliers is a subset of the input point cloud. It indicates the group that is used for plane estimation. outliers is a subset of the input point cloud but mutual exclusive to inliers. It indicates the group that is not used for plane estimation. It is not guaranteed to get estimated plane for each input point cloud. If plane.valid = false, it means the estimated plane should not be used.

Note: the module does not output inliers and outliers if params.cudaPipelineEnabled = true.

To release the module,

dwPCPlaneExtractor_release(estimator);

For more details see Point Cloud Processing Sample