Mid Sweden University

Range imaging is often used in classification of objects in process industry. The speed of inspection needs to be high, so it does not become the bottleneck in the process. This paper presents an FPGA based architecture for range imaging. Using centre of gravity it calculates the range positions from 2D images. The results show that the proposed architecture can process range values with a performance up to 150 Msamples per second. Thus, using cheep standard technology we can achieve up to 3 times higher performance than expensive state-of-the-art high performance smart-cameras.

FPGA based implementation of embedded systems has many attractive characteristics such as: flexibility, low cost, high integration, embedded distributed memories and extensive parallelism. One application where there is a significant possible potential for FPGA is for the implementation of real-time video processing. In this paper we present an analysis of a video pre-processor and how this affects the FPGA and RAM resource usage and performance. From these results we indicate the best space-time mapping of operations under different design constraints. These results can be used as a decision base when implementing an FPGA based video enabled display unit.