A Load Balancing Approach for Silicon Retina Based Asynchronous Temporal Data Processing

Abstract

In this paper we present a load balancing approach for Silicon Retina technology based computer vision applications. The Silicon Retina technology is a new kind of bio-inspired analogue sensor that is derived from the human vision system. In comparison to conventional imagers that provide frame information captured by a fixed frame-rate, Silicon Retina imagers only provide information of pixels with variations of intensity in a scene. The high amount of pixels without intensity variations need neither be transmitted nor processed. Due to these special characteristics, the imager delivers asynchronous data with data-rates up to a peak of 6M events per second (Meps) per channel and a time resolution of 10ns. A distributed embedded system consisting of a single-core processor for data acquisition and load balancing and a multi-core processor for data processing is used. We discuss both the Silicon Retina technology, the principles of the computer vision algorithms being used, and the load balancing approach.