Remote sensing based yield estimation of wheat using support vector machine (SVM) in semi-arid environment

Abstract

The increasing demand for food and necessary decision-making on management and security of food crops require prior knowledge of the upcoming yield. The accurate prediction of wheat yield is a hard process that requires information such as location and climatic conditions. In this paper, the accurate prediction of wheat yield is facilitated by integrating both the current and past data of the soil, and climate, along with the spatial features obtained from satellite images. Initially, the normalization of data is carried out to balance the values of different ranges. Then the measurement of current readings of soil characteristics such as soil moisture, air temperature, humidity, and precipitations along with the climatic conditions is performed. These measurements along with the previous historical measurements were considered in order to perform an effective prediction of wheat yield. The multi-kernel-based Support Vector Machine (SVM) is implemented for this purpose. The effectiveness of the proposed approach is validated in terms of performance metrics such as accuracy, precision, recall, and F score. The proposed approach outperforms the existing approaches in predicting the wheat yield with increased accuracy.