Options
2004
Conference Paper
Titel
An embedded neuromorphic vision system suitable for mobile robotics applications
Abstract
An embedded vision system based on a neuromorphic vision sensor is presented. The neuromorphic vision sensor calculates gradient based optical flow with a resolution of 32x32 pixels and provides grey-scale image data as well as estimates of optical flow. The optical flow field estimate is calculated in parallel at the focal plane of the vision sensor itself, so further higher processing can be done with algorithms that use the optical flow field as a starting point. Also, the grey-scale imager input data is available for the analysis, which can be useful for tracking of static visual cues, or for slow moving objects. The vision sensor is paired with an embedded MPC555 PowerPC that is used for on-board post-processing and for the control of the operational mode of the optical flow sensor itself, where the key idea is to combine the distributed analogue processing power of the sensor with the flexibility and programmability of a digital processor. The system is intended for educational use in robotics laboratories where the system could serve as the main processing unit preferably for small mobile robots.