Sunagoke moss water content sensing using machine vision- Texture analysis and bio-inspired algorithms

One of the primary determinants of Sunagoke moss Rachomitrium japonicum growth is water availability. Too much water or too little water can cause water stress in plants. Non-destructive sensing (machine vision using texture analysis) was developed for sensing water content of Sunagoke moss to realize automation and precision irrigation to stabilize the water content at optimum condition. The goal of this study is to propose and investigate bio-inspired algorithms i.e. Neural-Genetic Algorithms (N-GAs) and Neural-Ant Colony optimization (N-ACO) to find the most significant set of textural image features suitable for predicting cultured Sunagoke moss water content in a close bio-production system. Textural features consisted of 90 textural features included grey level co-occurrence matrix, RGB, HSV and HSL colour co-occurrence matrix textural features. Non-linear relationships between textural features and water content were identified by Back-Propagation Neural Network (BPNN). The lowest average prediction Mean Square Error (MSE) based on average testing-set data was 4.79×10^-3 when using HSL co-occurrence matrix textural features as the input of BPNN. Based on testing-set data, N-ACO had better performance for predicting Sunagoke moss water content than N-GAs with the average testing-set MSE of 1.43×10^-3.