Detection and classification of various pest attacks and infection on plants using recursive backpropagation neural network with GA based particle swarm optimization algorithm

Abstract

<p>Machine learning methodologies are commonly used in the field of<br />precession farming. It prospects greatly in the plant safety measure like<br />disease detection and classification of pest attacks. It highly influences the<br />crop production and management. The venture of our system is to produce<br />healthy plantation. The proposed system involves Enhanced Feature Fractal<br />Texture Analysis, Statistical Feature Selection and Machine Learning<br />methodology for classification. Hence more than ever there is a need for<br />such a tool that combines image processing methodologies and the Neural<br />network concepts and that is supported by huge cloud of structured data<br />which makes the diagnosis and classification part much easier and<br />convenient. The proposed system recognizes and classifies the plant<br />taxonomy and the infection based on the selected statistical features. The<br />neural network concept followed in our proposed system is focused on<br />Artificial Neural Network which uses Recursive Back Propagation Neural<br />network to speed up the training process as well as reduce multiclass<br />problem in the network and optimize the weights on hidden layers of the<br />Network using Genetic Algorithm based Particle Swarm Optimization<br />technique. We have used MATLAB to implement the concept and obtained<br />better accuracy in disease/pest detection and proved to be an efficient<br />method.</p>