A reconfigurable architecture for robotic stereo vision

Abstract

A reconfigurable architecture for dense stereo is presented as an observation framework for a real-time implementation of the simultaneous localization and mapping problem in robotics. The reconfigurable sensor detects point features as corners from stereo image pairs, in order to use them at the measurement update stage of the procedure. The main hardware blocks are a dense depth stereo accelerator, a left and right image corner detector and a stage performing left-right consistency check for the detected features. For the stereo-processor stage we have implemented and tested both a localmatching method based on the Sum of Absolute Differences (SAD) and a global-matching component based on a maximum-likelihood dynamic programming technique (MLDP). The system includes a Nios II processor for data control and a USB 2.0 interface for host communication. The proposed hardware is applied as the sensor part in a real-time robotic localization and mapping experiment with the help of a small guided vehicle.