Image pattern classification for the identification of disease causing agents in plants

摘要: This study reports a machine vision system for the identification of visual symptoms plant diseases, from coloured images. Diseased regions shown in digital pictures cotton crops were enhanced, segmented, and set features extracted each them. Features then used as inputs to Support Vector Machine (SVM) classifier tests performed identify best classification model. We hypothesised that given characteristics images, there should be subset more informative image domain. To test this hypothesis, several models assessed via cross-validation. The results suggested that: texture-related might discriminators when target images do not follow well defined colour or shape domain pattern; systems lead successful discrimination targets fed with appropriate information.