Biologically Motivated Vergence Control System Based on Stereo Saliency Map Model

Abstract

We proposed a new biologically motivated vergence control method of an active stereo vision system that mimics human-like stereo visual selective attention. We used a trainable selective attention model that can decide an interesting area by the low-level top-down mechanism implemented by Fuzzy ART training model in conjunction with the bottom-up static SM model. In the system, we proposed a landmark selection method using the lowlevel top-down trainable selective attention model and the IOR regions. Also, a depth estimation method was applied for reflecting stereo saliency. Based on the proposed algorithm, we implemented a human-like active stereo vision system. From the computer simulation and experimental results, we showed the effectiveness of the proposed vergence control method based on the stereo SM model. The practical purpose of the proposed system is to get depth information for robot vision with a small computation load by only considering an interesting object but by considering all the area of input image. Depth information of the developed system will operate for avoiding an obstacle in a robotic system. Also, we are considering a look-up table method to reduce the computation load of the saliency map for real-time application. In addition, as a further work, we are now developing an artificial agent system by tracking a moving person as main practical application of the proposed system.