Prediction of wheat moisture content at harvest time through ANN and SVR modeling techniques

Abstract

The grain moisture content at harvest time is a key factor that limits harvesting windows. The present study aimed to develop a new methodology to predict wheat moisture content by using multi-layer perceptron (MLP) and support vector regression (SVR) techniques. Five input variables included the number of days after sowing, air temperature, air relative humidity, wind speed on an hourly basis, and precipitation on a 6-hour basis. The study area was Sari County located in the north of Iran. Data were collected from field experiments in two crop years (2016/17 and 2017/18). The results indicated that the developed MLP model outperformed the SVR model in determining wheat moisture content by R2 and RMSE value of 0.92 and 2.09% (wet basis) against 0.79 and 3.09%, respectively. In conclusion, the developed MLP model can be considered a useful method to estimate wheat moisture content at harvest time.

Keywords

• Artificial neural network
• Cereal
• Harvest
• Model
• Support vector machine