Motion image sensor with on-chip adaptation and Programmable Filtering

Abstract

We present a fly-inspired integrated motion image sensor for autonomous navigation with on-chip optic flow computation, mismatch compensation, and programmable spatial filters. The sensor computes spatial motion patterns from local motion signals extracted by elementary motion detectors (EMDs). Floating gate transistors are used to cancel fabrication mismatch in the EMDs and also to program spatial filtering coefficients. The sensor is the first reported motion image sensor that extracts multiple motion parameters from the spatial motion pattern, enabled by the mismatch compensation and programmable filters. An array of 19 EMDs provide inputs to 4 weighted spatial filters on a 3 × 3 mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> chip in 0.5 μm CMOS technology. Results from both benchtop tests and closed loop navigation of a ground vehicle are reported.