An optical flow sensor realized by retinal resistive network

This paper describes a new optical flow velocity sensor which is based on a local velocity estimation algorithm and realized by analog circuits including resistive networks. It is proved that velocity can be calculated from five kinds of spatiotemporal differential coefficients. It is also proved that these coefficients can be approximated by finite alternating series derived from Shannonpsilas sampling theorem and Eulerpsilas transformation. Comparing these series and property of the resistive networks, it is realized that these coefficients can be derived everywhere in the network, which is suitable for velocity distribution sensors. In experiments, a monolithic photo diode array which consists of 16 times 16 cells and three layers of resistive networks are used to develop an analog motion sensor. From simulation of the resistive networks, it is also shown that the implemented system will be extended to measure velocity distribution everywhere on the image sensor.