A motion detection algorithm, based on biological and computational models, for focal plane implementation has been developed. This Temporal Domain Optical Flow Measurement (TDOFM) algorithm uses computational components which have direct and compact electronic counterparts. It uses a binary image of zero-crossings, 2 level analog signals, the signs of spatiotemporal derivatives, 1-b multiplication and pulse widths to measure image velocity. Compared to other IC visual motion detectors, this sensor represents the first instance of a robust, wideband and general purpose 2-D motion sensor which reports speed and direction explicitly, has a wide dynamic range and has a compact IC implementation. The front-end of the motion cells is an edge detection circuit which responds to 5–6 orders of magnitude of light intensity and produces near maximum outputs for contrasts as low as 40% in bright and dim ambient conditions. The theoretical velocity measurement dynamic range of the sensor is 4–5 orders of magnitude, and motion ranging over three orders of magnitude has been measured. The variation in the measured speed is less than 15% across 1- and 2-D implementations, multiple chips, cells and directions. The complete system, including the photoreceptors and edge detection circuits, consumes less than 0.4 mW per cell at ± 3.5 V.
Date of this Version
motion sensor, temporal domain optical flow
Date Posted: 27 June 2007
This document has been peer reviewed.