An Automatic Plant Disease Symptom Segmentation Concept Based on Pathological Analogy

Abstract

This paper proposes an automatic disease symptom segmentation algorithm using a simple pathological pattern recognition concept to segment plant disease visual symptoms on digital leaf images. The novelty of the algorithm is in the use of pathological analogy of diseases caused by pathogens, distinct homogeneous patterns relative to the disease progression, to segment individual images into symptomatic, necrotic, and blurred regions. Applying the pathological concept allow for actual disease lesion areas to be quantized in accordance with their true analogy. As a result, individual pattern characteristics of each lesion along the leaf surface can be tracked and features can later be extracted for characterization using machine learning. By employing the concept, the proposed algorithm applies a fusion of simple color space manipulation HSV and CIElab with deltaE (ΔE) color relativity equation to compute each lesion type pixels color. The obtained results are encouraging, successfully localizing and quantifying individual disease lesions. This also indicates the applicability of the proposed approach in discriminating plant diseases based on their analogical dissimilarity. Moreover, it provides opportunities for early identification and detection of fine changes in plant growth, disease stage and severity estimation to assisting crop diagnostics in precision agriculture.