Deep Image Processing Based Periodically Leaves Diseases Detection and Classification through Wireless Visual Sensors Network (WVSN)

Abstract

Apples are one of the best sources of nourishment and are packed with various nutrients, including fiber, vitamins, minerals, and antioxidants which are essential for maintaining a healthy body and reducing the risk of chronic diseases. However, many diseases attack apple plants like “Scab”, “Rust”, and “Black rot”. These diseases are responsible for the decrease in the production and cultivation of apples. Identification of these diseases at an early stage can play an important role in their control before spreading into other parts of the plant. This job is challenging, especially in leaves even through an expert’s eye and many imaging methods are applied to identify these diseases from images using machine learning algorithms. To automate the process in real-time monitoring of Apple farms, this paper presents a framework for detecting diseases in apple leaves using a Wireless Visual Sensor Network (WVSN). A WVSN utilizes Convolution Neural Networks (CNN) for cloud-based classification. The framework will periodically capture the images directly from the apple farms through wireless nodes and send the data to the cloud through a gateway for further processing where the trained model classifies the diseases accurately. We tested our proposed model on the developed dataset to benchmark it against other state-of-the-art studies and subsequently deployed it in Apple farms to ensure the best results. The proposed framework gained an accuracy of 96.96% on the developed dataset and 95.1% in the real time with Apple farm images.