A Fungus Detection System for Greenhouses Using Wireless Visual Sensor Networks and Machine Learning

Abstract

Greenhouses are proliferating across Canada. Greenhouse crop production requires considerable attention. The only way to maintain the production growth is by controlling the greenhouse atmosphere and monitoring the plants so that they remain healthy in the greenhouse. In this paper, we utilize a Wireless Visual Sensor Network (WVSN) with machine learning and image processing to observe any deficiency, pest, or disease presenting on the leaves of the plants. We distribute camera sensors throughout the greenhouse. Each camera sensor node captures an image from inside the greenhouse and use machine learning and image processing techniques to detect the presence of fungus. When a fungus is detected, the camera sensor node sends a message to the sensor node via the wireless sensor network to measure the humidity and then send a message to the actuator to re-set accordingly. This paper demonstrates how Hough forest machine learning and image processing can be successful in detecting fungus present on crop plant leaves from the images taken from camera sensors in the greenhouse. Cross-validation was applied to measure the performance of the system. The results are highly promising. There was a 94% success rate in detecting the fungus.