A Study On Shallow-Sea-Shoal Recognition Based On Image Sonar Technic

In order to make good use of the ocean resources, protect the environment which fish live in, a study on monitoring and recognizing fishes was developed. The study of shallow-sea-shoal identification is a part of a great project supported by agricultural department in Zhejiang. A monitor in the piscicultural net under the circumstances of thick seawater was developed, aimed at the need of fishery in Zhejiang. A device to obtain the images of shoal using sonar technique was firstly designed in this study, and then the images obtained were processed as follows: intensity transformations, linear and nonlinear spatial filtering, filtering in the frequency domain, image segmentation, region and boundary representation and description, and object recognition. There are mainly two fields in this study: digital image preprocessing and object recognition. The kind and the growing stage of shoal are the two parts of object recognition in this research. A simple software project for user wasbuild. In this software, the kind of shoal and the number of growing stage and the result of whether the fish can be caught is displayed. The efficiency of pisciculture can be improved by using the software.The research orshoal recognizing was based on the theory of digital image processing (DIP). It took the application of the intelligent image processing systems as references, such as fingerprint recognizing, optical character recognition, and remote sensing systems. A system of fish recognition was developed in the study, using DIP toolbox and guide tools in Matlab. After preprocessing the images, the arithmetic to read the length and physical rate of the fish from the image was worked out . Though the datasheet, named repository, built for shoal recognition, which contains the length and physical rate data of fish we tested in the laboratory for the research, shoal can be recognized by match the data prepared and the data read from the processed image. Adapting image was introduced window to the shoal recognition. It was an effective arithmetic for shoal recognizing, a way to optimize the image window was found to reduce the calculation in reading the information from the image.The result of the experiments in practice indicated that the arithmetic we researched in the project is effective for single fish recognition and the low distribution density of shoal recognition. But when the distribution density of shoal is high, it will lose its function. Because the fish is anactive object, its superposition of single fish is the major part of the image if density of shoal is high. The error of recognition will increase, and the arithmetic of recognition is not reliable. So the more efficient arithmetic is expected to reduce the recognizing error when the density of shoal is high.