Application Of Shape And Texture Analysis In Marine Phytoplankton Automatic Identification System

Under present condition, the identification and classificatin of marine phytoplankton is usually conducted by people with high espertise. The task is laborious and time-consuming. Yet in our country, the experienced experts are confined to a few universities or institutes of oceanography and there is a lack of succeeding experts.Therefore, recent days witnesses an urgent need to develop a software to recognize and identify marine phytoplankton automatically. The current microscopic image processing software, whose image analysis function is based on the geometric parameters measurement and counting function ,cannot contribute to the research of the morphological classification of the marine phytoplankton. And the shape and texture indeed a vast interspecific difference between different marine phytoplankton's shell surface. Thus the need to develop a computer expression of the shape and texture of marine phytoplanktona becomes all the more urgent. Moreover, in a society where internet has already been an indispensable part of our life, it is worth exploring possible means of applying our research findings to internet technologies so as to promote and popularize the morphological classification of marine phytoplankton.This thesis centers on phytoplanktonic cells in China's offshore area. According to the features of these cells, the study is conducted from perspectives as follows:1. On the basis of an effectively extracted cell edge, cells' disparity is analyzed from the perspective of the geometric shape of the cell outline in terms of basic geometric parameters and moments features such as similar circle degree, similar rectangle degree, eccentricity, etc.2. The texture feature of similai circular Phytoplankton cells is analyzed through statics texture methods as well as gray-basic dollar co-occurrence matrix.3. A dendriform classifier is designed. Its first layer categorizes shapes so as to narrow down the identification range. In its second layer, a texture matching classifier is designed which based on slightest distance.4. Finally but not lastly, after an anatomy of all modules in the framework of the system, an automatic identification system is completed by establishing a dedicated marine phytoplankton database and constructing a module for marine phytoplankton automatic identification.The Innovation of this thesis :1. Innovation Based on statistics texture analysis algorithm.In this thesis, when calculating the texture features of Marine Phytoplankton,the tradition texture analysis methods have got large change.The texture analysis is availability only to cells.2. Selection of texture basic dollar. The basic Matrix is suitable for most similar circle Marine Phytoplankton's texture structure.