Development of a novel machine vision procedure for rapid and non-contact measurement of soil moisture content

Soil moisture measurement is one of the essential management components to decrease water consumption and prevent water stresses in plants. In this study, a fast and non-contact method using machine vision and artificial intelligence was developed so as to make operators capable of having an estimate of soil moisture by taking only one image. Three soil textures along with three levels of added organic matter were applied. Mean comparison and the subsequent stepwise multiple regression were applied to find superior features from different color spaces. ANFIS and stepwise multiple regression were used to predict the soil moisture. Results indicated that the general model could predict the soil moisture with mean absolute error of less than 1.1%. This value reached to 0.3% for some sub-models belonging to the texture-organic matter group. Application of the present method is highly recommended for soil moisture measurement because of simple implementation and potential for online measurements.