Abstract

How do animals, e.g. insects, detect meaningful visual motion cues involving directional and locational information of moving objects in visual clutter accurately and efficiently? This open question has been very attractive for decades. In this paper, with respect to latest biological research progress made on motion detection circuitry, we conduct a novel hybrid visual neural network, combining the functionality of two bio-plausible, namely motion and position pathways explored in fly visual system, for mimicking the tracking and fixation behaviors. This modeling study extends a former direction selective neurons model to the higher level of behavior. The motivated algorithms can be used to guide a system that extracts location information on moving objects in a scene regardless of background clutter, using entirely low-level visual processing. We tested it against translational movements in synthetic and real-world scenes. The results demonstrated the following contributions: (1) Compared to conventional computer vision techniques, it turns out the computational simplicity of this model may benefit the utility in small robots for real time fixating. (2) The hybrid neural network structure fulfills the characteristics of a putative signal tuning map in physiology. (3) It also satisfies with a profound implication proposed by biologists: visual fixation behaviors could be simply tuned via only the position pathway; nevertheless, the motion-detecting pathway enhances the tracking precision.