Three-Dimensional Laser Radar Recognition Approaches

Three-dimensional laser radars measure the geometric shape of objects. The shape of an object is a geometric quality that is more intuitively understood than intensity-based sensors, and consequently laser radars are easier to interpret. While the shape contains more salient (and less variable) information, the computational difficulties are similar to those of other common sensor systems. A discussion of common approaches to 3D object recognition, and the technical issues (called operating conditions), are presented. A novel method that provides a straightforward approach to handling articulating object components and multiscale decomposition of complex objects is also presented. Invariants (or more precisely covariants) are a key element of this method. The presented approach is appealing since detection and segmentation processes need not be done beforehand, the object recognition system is robust to articulation and obscuration, and it is conducive to incorporating shape metrics.