An Approach to Develop a Robotic Arm for Identifying Tomato Leaf Diseases using Convolutional Neural Network

Abstract

Tomato plants play a vital function within the agricultural attitude but their production is greatly hampered by various leaf diseases. This research paper aims to develop a robotic arm regarding the detection and classification of several types of tomato leaf diseases through an experimental setup. Because of this research activity, a sequential model applying Convolutional Neural Network (CNN) has been implemented. Real-time images of diseased tomato leaf dataset were collected from Bangladesh Agricultural Development Corporation (BADC), Nowdapara, Rajshahi, and Department of Agricultural Extension (DAE), Godagari Upazila, Rajshahi, Bangladesh. The sample images have been processed before the final training program. The total number of training and testing images was 21902 and 5805, respectively. Eventually, after the training process, a system has been developed within a structure of a robotic arm. The system was intelligent enough to detect and recognize tomato plant diseases. The obtained classification training accuracy was about 98.27% & validation accuracy was 90.80%. In a real-time interface, the device was able to reach the plant, locate and classify nine forms of tomato leaf diseases as well as healthy leaves.