Automatic identification of cassava leaf diseases utilizing morphological hidden patterns and multi-feature textures with a distributed structure-based classification approach

Developing an automatic identification system for plant diseases caused by pathogens is of utmost significance regarding establishing an effective agricultural production. This study aims to tackle disease identification by uncovering the differences in morphological characteristics existing in the patterns of image texture features and multi-feature textures (LBP, TEM, HOG, GLCM and GABOR) of the cassava plants. To this end, a wide image database of the cassava plant has been classified by the proposed distributed structure-based classification system that is based on revealing hidden patterns existing in the image texture features. The results show that the proposed approach can uncover disease symptoms by examining the morphological distinctness of the employed image features. The diseases were classified by a distributed structured k -NN classifier that was employed in the binary approach. The system reached satisfactory accuracy rates for all classes considering the dataset contains images captured in natural conditions that have complex backgrounds. Moreover, the study can be implemented by utilizing additional image texture to obtain a more suitable pattern that has a higher capability of representing the symptoms of a wider class of diseases more efficiently. Graphical abstract