Advancements in UAV remote sensing for agricultural yield estimation: A systematic comprehensive review of platforms, sensors, and data analytics

Abstract

Traditional yield estimation approaches are quite tedious, time-consuming, and labor-intensive. Unmanned aerial vehicles (UAVs) present an exciting opportunity to estimate crop yield with high spatial and temporal resolution in agriculture. The objective of this article is to review current studies and research works in agriculture that employ the use of different UAV platforms, sensors, data acquisition, machine learning and photogrammetry techniques, and vegetation indices in UAV-based crop yield prediction. Furthermore, the article also explores the challenges and limitations in yield estimation. Hundred different studies from Google Scholar, Scopus, and Web of Science are presented and reviewed. The result demonstrated that most of the studies are centered on China and USA. Supervised learning models are widely used and exhibit better accuracy in yield estimation. The normalized difference vegetation index (NDVI) is preferred by researchers and emerges as a widely used vegetation index (60 studies). The study concluded that UAV-based crop remote sensing can be an effective method for improving yield estimation. The integration of multimodal data, including textural, structural, thermal, and meteorological features, along with key spectral bands such as near-infrared (NIR) and red-edge (RE), has demonstrated potential for improving the accuracy of yield estimation models. Moreover, supervised models have shown great suitability for cereal crops. Random Forest and linear regression emerge as reliable options for estimating yields of major crops, such as wheat, rice, and maize. However, challenges in yield estimation with UAV-based remote sensing include regulatory constraints, weather conditions, data storage and management, high initial costs, and technical limitations.