Automated remote sensing system for crops monitoring and irrigation management, based on leaf color change and piecewise linear regression models for soil moisture content predicting

Abstract

Plants can serve as biological sensors if their “readings” and the feedback they provide us through changes in the colour of their leaves can be correctly interpreted. The study aims to predict soil moisture and, as such, the need for irrigation, using nonlinear mathematical models, describing the relationship between RGB and HSL colour model components and soil moisture and temperature. Nonlinear mathematical models used in the study are based on piecewise linear regression with breakpoint and soil moisture prediction using colour components and soil temperature with a deviation of +-6%. A system for automated irrigation was created and its control program was made, the basic control law of which is based on non-linear piecewise linear models. The automated irrigation management system includes a remote crop monitoring subsystem and an irrigation management subsystem. The program processes the photo received from the camera and activates the actuators when watering is needed. Compared to manual data collection in the first part of the study, the program calculates the average RGB model values from images in the studied row of tomato plantations with an accuracy of over 99% for the R and G components and over 92% for the B component. The program also predicts soil moisture with 98% accuracy. The practical significance of the water-saving efforts of this study lies in the development of a program-controlled automated irrigation system that utilizes plants as biological sensors, employing nonlinear mathematical models based on leaf colour changes to accurately predict soil moisture