Integrating Neural Network for Pest Detection in Controlled Environment Vertical Farm

Abstract

Background: An integrated system for creating and maintaining controlled environment ideal for vertical farming prototype is demonstrated. The requirement of optimal artificial light for different growth stages of tomato and chilli plants is studied in detail and CNN model-based method for detection and classification of Leaf disease is also developed. Methods: The artificial environment ensuring adequate artificial lighting, moisture, and minerals was create by implanting various sensors and actuators to the plant beds and connected in network through a cloud based remote server. A CMOS image sensor module was used to monitor the various stages of plant growth. Findings: The duration and intensity requirement for germination, vegetation and flowering of both tomato and chilli plants are relatively lesser with artificial light condition than with sunlight. At the end of fifth epoch the developed convolution neural network model for detection and classification of leaf disease produced training and validation accuracies of 84.8% and 67.2%, respectively. Novelty: For different growth stages of tomato and chilli plants in north eastern India, the requirement of optimal artificial light is studied by exposing them to different light intensities. The study was conducted during summer (May-June) when the average sun exposure in eastern India was ~130- 190 hours. The captured images and data generated were used to monitor the status of the crops and identifying diseases with the application of Deep Learning models. Convolutional Neural Network (CNN) model-based method for detection and classification of leaf disease is presented. Keywords: Convolution Neural Network; Vertical Farming; Artificial Light; Pest Detection; Light Emitting Diode