Features Extraction Of Cell Images Of Phytoplankton Appearing Frequently In China's Sea Areas Based On Fractal Dimension

As the most important primary producer in the marine ecosystems, phytoplankton influences the marine environments and the living marine resources enormously. Based on the fractal theory, the analyse and feature extraction of the phytoplankton cell micrograph are studied in this paper mainly. After analyzing and studying the biologic features of the phytoplankton deeply, fractal dimension is used exploringly to extract the shape and texture features of the cell images of the phytoplankton appearing frequently in China's sea areas. Around this problem, the following work is done in this paper.Box-counting method is used widely to estimate the fractal dimension of the gray level image. But the results computed by the method are so coarse that they could hardly be used as the feature for image analysis. So a new fractal dimension estimating method based on area covering is proposed in this paper. The fractal dimensions estimating by this method are continuous almost, which can depict the fractal features of the gray level image accurately.According to the fractal character of the fractal dimension gray map image, an edge detection method is proposed. The edge shape of the cell of the phytoplankton can be detected using this method well from the fractal dimension gray level mapping images obtained by area covering method. The edge can be used to differentiate the different states of the same alga or to classify the alga belong to different family or genus.The edges of the setae in the Chaetoceros cell images are difficult to detect. The blanket method is improved in this paper to solve the problem. The edges of the setae can be detected perfectly by processing the fractal dimension gray level mapping images, which are estimated by improved blanket method, using the proposed edge detection method.For depicting the distributing feature of the Chaetoceros cell setae roundly, an image orientation angle estimating method is proposed. Then a new conception named orientate fractal vector is proposed by introducing the orientation angle into the computation of the fractal vector of the binary image. The orientate fractal vectors are good features for describing the orientate information of the setae, which are benefit for the classification of the Chaetoceros cells.According to the arrangement characteristic of the lucolus on the Coscinodiscus valve, a method of Coscinodiscus features extraction is proposed by using the fractal dimension estimating method based on fractional brown motion. The valve texture features can be described well using this method, which is useful for the classification of the Coscinodiscus cells.By improving the four methods for estimating the fractal dimension, many kinds of the biologic features of the phytoplankton are analyzed ,extracted and described effectively, which lays a foundation for the classification and the recognition of the phytoplankton appearing frequently in China's sea areas.