A Reconfigurable Architecture for Stereo-Assisted Detection of Point-Features for Robot Mapping

Abstract

A hardware-friendly procedure is presented for the extraction of point-features from stereo image pairs for the purpose of real-time robot motion estimation and 3-D environmental mapping. The procedure is implemented in reconfigurable hardware and is developed as a set of custom HDL library components ready for integration in a system-on-a-programmable-chip. The main hardware stages are a stereo accelerator, a left and right image corner detector and a stage performing left-right consistency check. For the stereo-processor stage we have implemented and tested a SAD-based component for local area-matching and a global-matching component based on a Maximum-Likelihood dynamic programming technique. The system includes a Nios II processor for data control and a USB 2.0 interface for host communication. Resource usage and 3D-mapping results are reported for different versions of the reconfigurable system.