Precision spray modeling using image processing and artificial neural network

Abstract

This study employed artificial neural network method for predicting the sprayer drift under different conditions using image processing technique. A wind tunnel was used for providing air flow in different velocities. Water Sensitive Paper (WSP) was used to absorb spray droplets and an automatic algorithm processed the images of WSPs for measuring droplet properties including volume median diameter (D_v0.5) and Surface Coverage Percent (SCP). Four Levenberg-Marqurdt models were developed to correlate the sprayer drift (output parameter) to the input parameters (height, pressure, wind velocity and D_v0.5). The ANN models were capable of predicting the output variables in different conditions of spraying with a high performance. Both models predicted the output variables with R² values higher than 0.96 indicating the accuracy of the selected networks. Therefore, the developed predictor models can be used in precision agriculture for decreasing spray costs and losses and also environmental contamination.