Based On Wavelet Decomposition And Color Information Entropy Plankton Image Recognition Technology

Plankton is an important part of halobios. It is the pivotal tache in the chain of halobios.and also it is one of the most important pivots of zoology balance in the ocean. So developing the research of plankton has important significance for developing,managing and utilizing the resource in ocean of our country ,and so to developing the ocean enterprise of our country.The research for recognition and classification of plankton images is the important content and precondition to the research of zoology, Traditional methods of plankton images recognition is mainly to be recognized under the microscope by professional. But this method has the demerit of lower efficiency and of a sort fallibility. Therefore, bring forward automatic,higher efficiency plankton image recognition arithmetic is very important. At present, with the development of computer technology and aim image recognition technology, many high efficiency aim image recognition methods has been come forth. And at the same time ,those methods has been used in plankton image recognition field which based on image analyses stage by stage.However,with the research of plankton image recognition going deep into step by step. The disadvantage of traditional recognition method has appeared gradually. Therefore ,on the base of guarantying the efficiency,Further improve the efficiency of the recognition algorithms has been to a new direction of plankton image to recognition field.Under this background,the color image information entropy based on wavelet decomposition and the plankton image recognition, Wavelet theory and information theory to describe the plankton and eigenvectors are extracted from the images. achieve a better effectiveness. Concrete progress has been achieved in the following areas:1,Various methods of aim recognition and plankton image recognition are investigated and studied in detail, which are analyzed and compared reciprocally. Then the advantage and disadvantages of these methods are summarized, this paper fully demonstrated the necessity and feasibility of the study.2,A description based on the information entropy of color image is brought forward. Plankton image will be partitioned in quad-tree structure after Pretreatment, and then the overall and local color image information is described as eigenvector in the form of entropy. This method describes the plankton image in the color space while reducing the dimension of the feature vector.3,Wavelet transform theory is to describe the features of plankton image, three tiers of wavelet decomposition has managed on plankton images, the decomposition of the son-channel extracted the coefficients mean, variance, those information constituted the eigenvector of plankton images. Complete details of the plankton image were extracted, and also reduced complexity of the algorithm.4,A classifier of plankton image based on similarity matching and K- neighbor is designed. European distance is using to calculate the similarity between images of test suite and image of the training, and then K-neighbor classification method was used to recognise the plankton images in database. The experimental results show that the recognition rate is higher than others, and at the same time this method is deemed to having high efficiency.