In recent years, UAV (Unmanned Aerial Vehicle) has developed rapidly and has been widely used in various fields, which has also posed a great threat to the security of military bases, government departments and other sensitive areas. Considering the problem of low accuracy and insufficient datasets of current algorithm in the target detection of “low, and slow small” UAV, and in order to improve the real-time detection performance of “low, and slow small” UAV, this paper proposes an improvement strategy of anchor, and obtains the anchor size suitable for UAV dataset through K-means++ clustering algorithm, making the size of anchor more targeted. Meanwhile, the different loss functions and IoU related threshold parameters for research, analysis of different thresholds, and the influence of different loss functions for UAV detection precision, thus by selecting the best loss function and the threshold is set to YOLOv4 target detection algorithm is improved, further, enhancing the YOLOv4 for “low, and slow small” UAV detection accuracy.
UAV Target Detection Algorithm Based on Improved YOLOv4
Lect. Notes Electrical Eng.
Proceedings of 2021 5th Chinese Conference on Swarm Intelligence and Cooperative Control ; Chapter : 27 ; 271-282
2022-07-29
12 pages
Article/Chapter (Book)
Electronic Resource
English