Precision irrigation for Sunagoke moss production using intelligent image analysis

There are many methods for sensing water condition in Sunagoke moss Rachomitrium canescens. The direct measurement of canopy is considered to be relatively inefficient and destructive to the plant. One alternative is the use of indirect measurement and non-destructive techniques such as machine vision. This study investigated the use of machine vision for monitoring water content in Sunagoke moss. The goal of this paper was to propose and investigate a combined genetic-neural algorithm to find the most significant image features or the sets of image features suitable for predicting Sunagoke moss water content. We extracted 50 features consisting of color, textural (Gray Level Co-occurrence Matrix and RGB Color Co-occurrence Matrix textural features) and morphological features. Ten textural features were calculated, including Entropy, Energy, Contrast, Homogeneity, Sum Mean, Variance, Correlation, Maximum Probability, Inverse Difference Moment and Cluster Tendency. The specificity of this problem was that we were not looking for single feature but several associations of features that may be involved in determining water content. The genetic algorithm, was able to select features with 27 selected features and artificial neural network was able to predict water content according to the selected features with minimum error of MSE 0.0021.