Removal of signal artifacts from biomimetic vision sensor based on the common housefly

Abstract

The compound vision system of a common house fly, Musca domestica, is capable of heightened motion detection and tracking capabilities as compared to the camera-eye visual system of mammals. While mimicking the mammalian system makes sense when a human is responsible for viewing and interpreting the image, but in applications requiring obstacle avoidance and quick visual data processing, a fly vision sensor would be far more advantageous. Recent proof-of-concept, fully compound vision sensors have shown that it is possible to mimic the motion hyperacuity characteristic of the fly's visual system while also providing near instantaneous edge detection and motion tracking results; all without the need of a computer processing system. However, since it is currently not feasible to mimic the entire 3000 lens eye of the fly, biomimetic compound sensors must be designed to handle signal artifacts generated by the bioinspired light adaptation system. Previous efforts to handle these artifacts utilized a simple low pass filter architecture that mimics the delay in light adaptation seen in both the fly and humans. However, due to circuit design limitations, the low pass filter solution was only effective when imaging targets move at moderate to high speeds. A more effective sample-and-hold approach is presented in this paper. The approach allows easy customization of the light adaptation delay based on the slowest expected target speed in the application.