Robust Image Segmentation for Early Plant Diseases Detection on Leaf

Abstract

Image segmentation of plant leaf is an important tool in the early detection of plant diseases from symptoms at the parasite control stage during flowering and maturity. Indeed, in the phase of pest control in order to resist plant diseases in agriculture, the current solution requires the massive use of phytosanitary products that are both dangerous for crops, the agricultural environment, and the health of both farmers and consumers. The main goal of this work is to respond to a new agronomic challenge in the context of agro-ecology through the development of a new method of pest control by using image processing tools. For this, we propose a novel region-based segmentation approach in a credibility context. This approach has the advantage to take into account the segmentation model of all the sources of noise during the acquisition phase. The experimental results and comparison results with similar approaches demonstrated that the proposed method is efficient and very robust to the noises of the image acquired by its genericity.