overview project page about

Outliers Removal
(PCCRemoveOutliers)



Short Description

This tool allows to remove outliers from point clouds using an approach based on Tensor Voting.

Usage

First, the user specifies the point cloud to be used via the pcName string. It specifies the complete scene graph path of the point cloud in the scene graph (e.g. root/data).

The Tensor Voting requires the user to specify an influence radius tvSigma which depends on the input data. Generally spoken, the influence should ensure that for each point contains a sufficient number of neighbors within the radius (e.g. 20) and the radius should significantly exceed the noise in the data. In order to help the user specify this radius, one can check the showTVSigma check box. Then, for a single data point, all neighbors within the influence radius are highlighted using blue spheres. The radius of the spheres is adjusted with the renderScale parameter.

For the actual application of the outlier removal, the user is offered several options to choose from. The first step always has to be that a surfaceness value saliency for each point in the data in computed from the tensor voting. Therefore, the option Compute needs to be choosen in the action combo box during execution. Only after this has been conducted, the points can either be color coded as outliers (option Colorize, attention with this option: original color values are overwritten) or removed from the point cloud (option Delete).

The classifiction into outliers and valid points is done by thresholding using the limit tvThreshold. The effect of the threshold parameter can be checked by the Colorize option.

Notes

The tool requires for the fields position (3f), color (3f) and surfaceness (1f). They are automatically created if they do not exist in the point cloud.


overview project page about