Classification and Detection of Plant Diseases using higher order Dynamic Conditional Random Fields Through Spatial and Multitemporalimages

Abstract

Agricultural areas should be continuously monitored, because they undergo random changes throughout the year. The problem arises when different crops show similar phenology and backscatter. This occurs when crops are classified based on single date remote detecting pictures. We should consider multitemporal images of crops. In this paper for classification of crops, we design an ensemble classifier which combines both spatial and temporal images of crops. To detect the affected areas of crops we implement first order and higher order dynamic conditional random fields (HDCRF). In order to enhance the diseased area of crop, we use k-means segmentation. All the training datasets are stored in multisvm (support vector machine). By the obtained results from this paper, we consider HDCRF as the best technique when compared with MRF (Markov Random Field) and CRF (Conditional Random Field) techniques.