Real-Time Processing of Rasterscan Images
Robotics, industrial automation, automatic inspection and automatic reconnaissance and surveillance include the processing of image sequences. The datarate can be very high in dynamic imaging systems, and often realtime processing requires the use of special hardware. To reduce cost and system size the need for storage should be restricted. As a consequence the amount of data has to be reduced considerably during the scanning of the image. Processing “in the scan” also speeds up the analysis, since total memory access time becomes neglectable.
Some principles are given which allow simultaneous processing and scanning of a TV-like rasterscan image. The video signal flows through special hardware processors which work in parallel at video rate in a parallel-pipeline configuration. Only some postprocessing and the final decisions are left until the completion of the imaging. In most systems a microprocessor can do the final processing during the video flyback time between consecutive frames. This solution will only work for local processors, but relying on knowledge learned from preceding images, the principles can also be extended to most global processes. In a non-static scene this will introduce errors. However, if the image frequency is high compared to the dynamics of the scene, this error is neglectable.
The principles are visualized by examples. The implementation of image segmentation algorithms of various complexities are discussed, and a brief discussion of a complete image analyser is contained. This system allows real time segmentation and recognition of every image in sequences from not too complex out-door scenes.
KeywordsImage Sequence Video Rate Image Segmentation Algorithm Automatic Inspection Feature Extraction Module
Unable to display preview. Download preview PDF.
- 1).Fu, K S “Special Computer Architectures for Pattern Recognition and Image Processing An Overview”, Proc. Nat’l Comp Conf (1978), pp 1003–1013.Google Scholar
- 2).Fu, K S Real-Time/Parallel Computing (Image Analysis]. Ed: M Onoe, K Preston jr, A Rosenfeld, Plenum press, N Y (1978).Google Scholar
- 3).stevold, E Detection and Classification of S Grinaker Military Targets from Infrared Imagery, FFI/RAPPORT-80/40Q6, NDRE, Kjeller, Norway (1980).Google Scholar
- 4).Sevigny, L “Discrimination and Classification G Hvedstrup-Jensen of Operating Military Targets in M Bohner Natural Scenes from Thermal Imag A M Navarro ery”, NATO Report AC/243(Panel III) E østevold D/191, AC243(Panel III)RSG.9 D/4. S Grinaker J DehneGoogle Scholar