A View-Dependent Adaptive Matched Filter for LADAR-Based Vehicle Tracking
This provides a key sensing component for autonomous ground vehicles, addressing the specific problem of vehicle tracking in on-road and off-road driving scenarios, though it appears incremental as an improvement over existing tracking methods.
The paper tackled the problem of inferring the location and orientation of nearby vehicles from raw LADAR data on mobile platforms, proposing a view-dependent adaptive matched filter that enabled tracking over 50 targets at 10 Hz with fast and precise pose measurements.
LADARs mounted on mobile platforms produce a wealth of precise range data on the surrounding objects and vehicles. The challenge we address is to infer from these raw LADAR data the location and orientation of nearby vehicles. We propose a novel view-dependent adaptive matched filter for obtaining fast and precise measurements of target vehicle pose. We derive an analytic expression for the matching function which we optimize to obtain target pose and size. Our algorithm is fast, robust and simple to implement compared to other methods. When used as the measurement component of a tracker on an autonomous ground vehicle, we are able to track in excess of 50 targets at 10 Hz. Once targets are aligned using our matched filter, we use a support vector-based discriminator to distinguish vehicles from other objects. This tracker provides a key sensing component for our autonomous ground vehicles which have accumulated hundreds of miles of on-road and off-road autonomous driving.