A novel multiprocessor system for road traffic applications

Abstract

In this paper we describe a novel image processing approach for road traffic analysis. We use combined background differencing and edge detection techniques to detect vehicles and measure various traffic parameters such as the vehicle count and queue parameters. Since the computational requirement for the vision based road traffic algorithms is beyond the capacity of conventional computers, we designed a RISC based array processor system to execute the image processing algorithms in real-time. To expand the performance of the array processor system, we chose a combinatorial architecture for the processing elements; i.e., we add some functional units to a traditional RISC architecture. The image processing algorithms for traffic analysis have been implemented on the array processor system and the results are reported online.