Automatic detection of unstructured elements in 3D scanned scenes

In this work, we present a new tool to assist human operators in the processing of complex scenes obtained from 3D laser scanners in civil engineering applications. Four unsupervised classification methods for detecting unwanted unstructured objects are presented, based on well known filtering techniques adapted to this specific problem. These methods are described in detail and compared using different data sets and scanner locations. As a result, suitable values for each classifier parameter have been obtained giving good results in most common scenarios. These classifiers have been integrated as a tool in scanned data processing software that allows the proper method to be selected depending on the characteristics of the scene. Although this tool cannot be used without the supervision of an operator, it can substantially help to reduce the time required to manually process raw scanned data for its subsequent use.