Herbal Plant Classification and Leaf Disease Identification Using MPEG-7 Feature Descriptor and Logistic Regression

Abstract

Plant disease classification, especially herbal plant disease classification is a prominent problem in the field of botany. It is compelling problem due to the heterogeneity among the plants of the same category and dearth of awareness about the immense medicinal properties of herbal leaf. By not only classifying herbal plant but also identifying the diseased and non-diseased traits among herbal plants will facilitate the naive population as well as herbal product manufacturing industry and pharmaceutical industry to enrich the global economy. In this paper, we have presented how MPEG-7 color and texture feature descriptors are incorporated with the traditional classifiers (for example, Logistic regression and Support Vector Machine, etc.) to yield very impressive results on wide range of classes. A total of two datasets: herbal plant dataset and leaf disease dataset are used to evaluate the results. This classification strategy is not only accurate but also very efficient in terms of number of computations needed and overall performance of the system. Comparison with other traditional features indicates the potential of MPEG-7 feature descriptors.