Neural-Discrete Hungry Roach Infestation Optimization to select informative textural features for determining water content of cultured Sunagoke moss

In this paper, a method to extract textural features from Colour Co-occurrence Matrix (CCM) and also a method to select relevant textural features for predicting water content of Sunagoke moss (Rhacomitrium japonicum) are proposed. The aim of this paper is to construct machine vision-based precision irrigation system. The objective of this paper is to propose Neural-Discrete Hungry Roach Infestation Optimization (N-DHRIO) algorithm to find the most significant set of textural features suitable for predicting water content of cultured Sunagoke moss using machine vision. N-DHRIO is an optimization algorithm for feature selection that is inspired by the social behaviour of cockroaches. The performance of the proposed feature selection method here was compared with Neural-Genetic Algorithms (N-GAs), Neural-Discrete Particle Swarm Optimization (N-DPSO) and Neural-Simulated Annealing (N-SA). Textural features consisted of 120 textural features extracted from gray, RGB, HSV, HSL and L^*a^*b^* colour spaces. Non-linear relationships between textural features and water content were identified by Back-Propagation Neural Network (BPNN). The results showed significant statistical improvement between methods using feature selection and methods without feature selection. Experimental results also indicated the superiority of N-DHRIO among other feature selection methods, since it achieved better prediction performance as the objective of this research.