PURPOSE: harmful microalgae recognition methods, which uses, means that feature is supplied to building identification candidate image group by creation learning image as meaning feature image NMF (based on Non-negative Matrix Factorization). ;CONSTITUTION: study harmful microalgae normalizes and is configured to image vector group. A kind of extraction of significance characteristic is the harmful microalgae (210) by a kind of for handling NMF. Candidate harmful microalgae group is configured to the characteristic by reference to meaning, passes through which red tide recognition candidate's harmful microalgae group (220) of input picture. Candidate harmful microalgae group is identified in the image of red tide. The study of the harmful microalgae and original image of the input of circularity is calculated (310). Proximity image selects circularity. Image is similar with entropy be considered as red tide image, by the entropy (320) for calculating the image. ;The 2013 of copyright KIPO submissions;[Reference numerals] (100) pre-process; (200) candidate harmful microalgae is searched; (210) Non-negative Matrix Factorization; (220) the identification candidate's harmful microalgae organized; (300) harmful microalgae is identified; (310) circularity is calculated; (320) entropy is calculated; (AA, BB) learning image; (CC) meaning characteristic
展开▼