Image analysis for network based Agri Advisory System

Abstract

In recent years, advanced technological solutions, such as an agri advisory systems, wireless sensor network based solutions, have helped in many challenging agriculture-related tasks such as disease prediction and detection, grading of crops, advisory systems, yield prediction, automatic harvesting and storage. In a similar spirit, in this paper, we propose an agri-advisory system developed for analysis of agricultural images, particularly apple images. The image processing tasks considered, are those of super-resolution (SR) and image segmentation for annotation. We develop simple and efficient methods for enhancing the resolution of images, and to automatically segment defects on apples in their images. We propose an example-based SR, which involves simple modules of construction of dictionaries based on local luminance variance, patch selection and weighted reconstruction of patches. The image segmentation algorithm is a combination of chrominance thresholding and low-level morphological operations. Our experimental results demonstrate that such simple and efficient methods suitable for network applications, are also quite effective, given a specific application domain.