Detection of pistachio leaf pests (Pistachio Psylla and moth) using image processing and Artificial Neural Networks

In present study, image processing technology and ANNs were used to identify the pistachio leaf pests include Ocneria terebinthina stgr and agonoscena pistaciae. The color images were captured from leaves of Ohadi pistachio variety and, color, texture, morphological and texture-color features were extracted from images in order to detection and classification of pests. To achieve the best models of ANNs, different types of ANNs and extracted features were evaluated and the following choices were selected according to the performance of different developed ANNs as the bests; A:The two-layer back propagation ANNs with two hidden layers with SigmoidAxon transfer function and TanhAxon as transfer function in output layer, by using the six color features (variance, mean, standard deviation, skewness, kurtosis and smoothness) with an accuracy of 93.3%, B: The two-layer back propagation ANNs with two hidden layers with SigmoidAxon transfer function and Linear transfer function in output layer, by using the five texture features (entropy, contrast, correlation, energy, homogeneity) with an accuracy of 95%, C: The two-layer back propagation ANNs with two hidden layers, with TanhAxon transfer function and Linear transfer function in output layer, by using the five morphological features (Area, perimeter, convex hull area, extent and solidity) and also, 11 texture-color features with accuracy of 86.7% and 98.3% respectively. The results showed the image processing technology and ANNs, had excellent ability to identify and classification of pistachio leaf pests.