Development of High-Speed Fish Pattern Recognition System with Color Image Multi-DSP

点击翻译按钮获取中文摘要

A new color image processing system with multi-DSP is constructed and applied to the pattern recognition of fish species, the shapes of which are very similar but the colors and/or patterns of which are different. This system has a function of variable partition of picture. Therefore, the size, position and number of processing regions can be varied flexibly and dynamically for each processing task. Also, the proposd system consists of the CCD camera, the interface unit, the pixel processor unit, the system controller and the host computer. The high-speed feature extraction of fish is presented as an example of the application of the proposed system. In this application, the shape features of fish are extracted at first, and then the color features are extracted from several small regions of fish and characterized by the co-occurrence matrix. Further, the classification of fish species is realized by the method of multi-level slice classifier. After discussing the relationship between the number of feature parameters for pattern recognition and the reliability of discrimination, we proceed to clarify the configuration of the parallel image processing system for high-speed fish pattern recognition in detail. Since the proposed approach is suitable for the parallel image processing system, the high-speed recognition time of 55 ms per a fish can be achieved.