A PARALLEL IMPLEMENTATION OF A MULTISENSOR FEATURE-BASED RANGE-ESTIMATION METHOD

There are many proposed vision based methods to perform obstacle detec tion and avoidance for autonomous or semi-autonomous vehicles. A syste m capable of supporting autonomous helicopter navigation will need to extract obstacle information from imagery at rates varying from ten im ages per second to thirty or more images per second depending on the v ehicle speed. This paper describes an efficient and flexible parallel implementation of a multisensor feature based range-estimation algorit hm, targeted for automated helicopter flight. The algorithm can track hundreds of features in multiple image sensors using an extended Kalma n filter to estimate the feature's location in a master sensor coordin ate frame. The feature-tracking algorithm has reached relative maturit y in the laboratory and is now being ported to several real-time archi tectures to support autonomous helicopter navigation research. The foc us of this paper is nob the core theory of the vision algorithm itself , but those aspects of it that affect the method of parallelization. T he performance of the parallel algorithm is analyzed, with respect to three load balancing schemes, on bath a distributed-memory and shared- memory parallel computer.