Research Of The Algorithm For Statistics Of The Plankton In Ballast Water Based On Image Processing

Today maritime transport is so flourishing. Ballast water is artificially circulated between oceans. Biology invasion and zoology disaster caused by random emmission of ballast water have been announced as one of four hazards by International Maritime Organization. Management of ballast water is not only a difficult international problem admitted,but also a hot spot now.Whether controlling or treatment, detection of ballast water is the key point before management of ballast water. Now research of specialized detection equipment just begins,and few reports exist. Performance criteria of ballast water has been announced in the ship's ballast water and sediment's control and management international covenant. Image processing is the only way which is effective and feasible considering actual detection criteria and detection measure.The international covenant presents the criteria of plankton's size and number. So the paper first introduces classification of plankton in ballast water which is based on image processing according to difference of morphological character,then introduces the size and number's statistics of plankton in ballast water after classification in the steps of image preprocessing,part extraction,character extraction,classification and information statistics.High resolution pictures are used here because the plankton's order of magnitude is micron. It is difficult to process the hign resolution pictures directly which own mass data,so the specisal partition is the first step in image preprocessing. In part extraction,the edge parts and the isolated parts are fully extracted in the way of best threshold combined with mathematical morphology and denoising.Classification here is different with the nomal one,which is not the recognition of plankton's categary,but the body's classification in the purpose of information statistics. In character extraction,character set,which reflects the body feature, contains morphology character and fourier character,and dimension is reduced in the way of principal component analysis. In classification,way of support vector machine based on radial basis function is chosed according to the comparison of different ways of support vector machine.Information statistics is the statistics of size and number,and ways of statistics which fit three classes are proposed here. All disposals are linked by soft platform. In and out of parameter and inspection of result can be easily achieved in platform. According to a mass of experiments,the prospective result is achieved,and soft platform developed here lays the foundations of future research.