Ladar-Based Vehicle Detection and Tracking in Cluttered Environments

Detecting and tracking vehicles is crucial for safe operation of Unmanned Ground Vehicles UGVs, but is challenging in cluttered, real-world environments. Here we present a method for discriminating vehicles from clutter found in natural terrain such as foliage, steep slopes, rock-outcrops, etc. Our method relies on a scanning LADAR and combines an obstacle detector and tracker, a vehicle modeling scheme, and a Support Vector-based discriminator. The output of our real-time system is a list of labeled obstacles and vehicles along with their positions, sizes and velocity estimates. This is used by a planner to enable autonomous navigation in the presence of other vehicles and significant clutter. We provide a quantitative analysis of the performance of our algorithm.