Digital Agriculture: Transforming Yield Prediction with Deep Learning and Multispectral Data

Abstract

Automation is becoming essential across diverse professions and sectors, including agriculture. Remote sensing for wheat yield estimation has become a more effective alternative to conventional yield prediction techniques. Traditionally, measuring wheat output required labor-intensive and time-consuming destructive sample methods. Accurate and timely yield estimates are essential for decision-making processes, including crop harvesting plans, milling, marketing, and forward sellingstrategies, thereby improving the efficiency and profitability of the worldwide wheat sector. Currently, growers or productivity officers, frequently financed by mills, utilize destructive or visual sampling methodsto evaluate wheat yield throughout the growing season. There is an increasing demand for rapid and effective problem-solving techniques. This study seeks to demonstrate and compare the efficacy of employingsatellite earth observation data for monitoring agriculture, specifically in wheat production. Various predictor variables are employed to compare multiple regression models. The research includes wheat yield estimation methodologies, including regression models, time series analysis of vegetation indices, remote sensing, phenological observations, and the normalized difference vegetation index (NDVI). Artificial intelligence methods, such as Random Forest and ordinary least squares, are utilized to formulate a proposed method that precisely correlates ground-measured data. This study presents an innovative technique for estimating wheat output, which markedly enhances forecasting precision and offers potential for improving decision-making in wheat cultivation practices.