A survey of unmanned aerial vehicles and deep learning in precision agriculture

Abstract

In the wake of significant advances in agronomy, biology, informatics, agricultural robots (Agri-robots), and artificial intelligence, modern agriculture is transforming from labor-intensive to data-driven mode. Precision agriculture (PA) is one of the most practical solutions for bridging the crop yield gap by performing the right treatments in the right place and at the right time. As a rising star among Agri-robots, unmanned aerial vehicles (UAVs) equipped with high-resolution onboard sensors and dedicated application systems are playing an increasingly vital role in collecting multi-scale agricultural information and implementing site-specific treatment. In this process, a large number of images are produced. However, considerable effort is required to extract high-value information from the explosively growing number of images. Over the past decade, deep learning (DL) has demonstrated unparalleled advantages in agricultural analytics, such as crop/weed classification, biotic/abiotic stress detection, crop growth monitoring, yield prediction, natural disaster assessment, etc. The combination of UAVs and DL is of great significance for agricultural information acquisition, processing, analysis, decision-making, and deployment. With the rapid development of UAVs, DL, and PA, this work firstly introduces the key components of PA, UAVs, and DL, respectively, and summarizes their major research progress. Subsequently, we focus on the successful applications of UAVs and DL in PA. Furthermore, based on our extensive literature survey, their current challenges and future development trends are sorted out. Ultimately, we hope this survey can draw more attention to the novel applications of UAVs and DL in PA among multidisciplinary scientists around the world and inspire more exciting and practical studies.