Automated Form Recognition of Laser Scanned Deformable Objects

Abstract

The use of terrestrial laser scanners is a new and promising way for monitoring the deformations of artificial and natural objects. These new sensors offer distance measurements without artificial reflectors. The data acquisition rates range from 1000 Hz up to 600 kHz, the obtained accuracies are better than 1 cm. The result of such a scan is a highly dense point cloud which enables more precise object models.

Two main problems exist if dense point clouds from laser scanner measurements are used for deformation analysis: the huge data volume which has to be handled, and the lack of fully automated analysis methods. As in a typical deformation scenario the shape of the objects changes due to bending and flexing, a thorough but information preserving parameter reduction is needed by a set of a few characteristic parameters. This is often possible by describing the objects or parts of them using quadratic forms.

A statistically founded identification of the particular quadratic form is presented which solves this task automatically. Thus, a dedicated functional model for the adjustment of the point data can be set up. This procedure is illustrated using numerical examples which are based on synthetic point clouds as well as on real data which were observed with terrestrial laser scanners.