| Previous research efforts in mimicking the compound eye of Musca domestica, the common house fly, in hardware have primarily been focused on optical designs to achieve the fast, analog, parallel processing and motion hyperacuity properties of the biological system. While these optical platforms have shown promising results, nearly all proposed advantages of utilizing a fly-inspired sensor have been discovered through the use of software simulation with very little or no effort being made to validate these simulations using the actual fly-inspired optics. In effect, these simulations can be viewed as the first step into mimicking the function of the intermediate neural layer structure (the lamina) of the fly's visual system. The lamina is believed to perform preprocessing on image data, extracting motion and object information from the scene before sending it to the brain.;The work presented in this dissertation advances the frontier of compound vision research by mimicking the function of the lamina structure in hardware. The hardware designs were kept as biomimetic as possible: no system clocks were used, system-wide parallel operation was never violated, and discrete-time (digital) solutions were not considered unless absolutely necessary. In addition, the unique motion hyperacuity characteristic of Musca's visual system was preserved in each circuit design. The specific lamina functions that were mimicked in device-level electronics are the light adaptation process and a software proposed edge detection process. It will be shown that an effective light adaptation system can be designed that is both analog and parallel by directly mimicking the fly's visual system. It will also be shown that the edge detection system can be designed to utilize the underlying structure that allows for motion hyperacuity to improve the accuracy of locating small linear objects such as power lines. The proof-of-concept designs presented in this dissertation are an important first step in turning the novel compound optical designs of the past decade into useful sensors that can function in dynamic real-world environments and that are capable of on-board biomimetic image processing. |
