Advancements in UAV-based traffic monitoring: a systematic review of deep learning and edge computing

Abstract

Rapid urbanization necessitates innovative traffic monitoring solutions. Traditional methods (fixed sensors/CCTV) face limitations in coverage, adaptability, and real-time processing. This review examines advancements (2015–2024) in vision-based unmanned aerial vehicle (UAV) traffic monitoring systems, evaluating their effectiveness in vehicle detection, traffic analysis, and congestion management. A systematic preferred reporting items for systematic reviews and meta-analyses (PRISMA)-guided analysis of 2,895 articles from IEEE Xplore, Scopus, Web of Science, and ACM Digital Library identified 49 eligible studies. Quantitative performance metrics (detection accuracy and latency) were standardized for cross-study comparison. Modern systems achieve >94% detection accuracy and <40 ms latency through edge computing and deep learning (e.g., you look only once (YOLO) and Faster region-based convolutional neural network (Faster R-CNN)). Multi-sensor fusion improves robustness by 35% in challenging conditions. However, battery life (reduced by 40% under processing load) and regulatory barriers remain critical constraints. Artificial intelligence (AI)-driven UAV systems enable real-time, high-accuracy traffic monitoring but require solutions for power efficiency and scalability. Future integration of 5G/6G and swarm intelligence holds promise for next-generation smart traffic management.