Early detection of diseases in tomato crops: An Electronic Nose and intelligent systems approach

Abstract

Sensor arrays also known as Electronic Noses (ENs) have been used to analyse the Volatile Organic Compounds (VOCs) of both healthy and infected tomato (Solanum lycopersicum) crops. Statistical and intelligent systems techniques were employed to process the data collected by an EN. Principal Component Analysis (PCA), K-Means clustering and Fuzzy C-Mean (FCM) clustering were applied to visualise any clusters within the dataset. Furthermore, Multi-Layer Perceptron (MLP), Learning Vector Quantization (LVQ) and Radial Basis Function (RBF) based Artificial Neural Network (ANNs) were used to learn to classify and hence categorise the datasets. Using the RBF, MLP and LVQ techniques we achieved 94, 96 and 98% classification accuracy for the healthy, powdery mildew (Oidium lycopersicum) and spider mite infected plants respectively. From these results it is evident that EN is capable of discriminating between the healthy and artificially infected tomato plants and hence may be deployed as a potential early disease detection tool for tomato crops in commercial greenhouses.