关键词: 小波包分解, 特征提取, 模糊支持向量机, 红虫识别, 信息熵

Abstract: The chironomid larvaes enter the watersupply system through transmission pipeline. Although there are no indications that chironomid larvae pose a threat to public health, their presence is still not appreciated because most people associate the organisms with low hygiene. In accordance with the characteristics of the organisms, this paper studied chironomid larvae images recognition method based on wavelet packet decomposition and Fuzzy Support Vector Machine (FSVM). The energy features of subgraph decomposed by wavelet packet, color information entropy, and the shape features of plankton were selected to construct feature vector for FSVM that is used to classify the image. The experimental result shows the method is effective for freshwater plankton images, such as chironomid larvae, cyclops and harpacticoida, and provides the basis for the prevention and treatment of chironomid larvaes in water plant.

Key words: wavelet packets decompositions, feature extraction, Fuzzy Support Vector Machine (FSVM), chironomid larvae recognition, information entropy