Accurate identification of multi-rotor UAVs in different environments is crucial in no-fly zones. However, due to the small size and fast flight speed of multi-rotor UAVs, it is difficult to identify them quickly and accurately. This paper proposes a small target detection algorithm, YOLOv7-SPD3, based on YOLOv7 to realize the accurate recognition of multi-rotor UAVs in long-distance states. Firstly, the complexity of the original UAV data set is improved by data enhancement. Secondly, the recognition accuracy of small long-distance targets is enhanced by adding SPD-Conv to the YOLOv7 backbone network. Thirdly, the small target detection ability is further improved by replacing the original loss function with Wise-IOU. Fourthly, the proposed model is fine-tuned to achieve accurate recognition of UAVs. The experimental results obtained by using the workstation in the Python3.8 environment show that the YOLOv7-SPD3 algorithm is superior to YOLOv7, YOLOv7-Tiny, SSD300, and YOLOv3. The maximum precision value is increased by 2.2%; the maximum Recall value is increased by 60.42%, the maximum mAP@0.5 value is increased by 17.83%, and the maximum mAP@0.5:0.95 is increased by 30.1%, which can meet the accurate identification of small long-range targets.
YOLOv7-SPD3: A Small Target Detection Algorithm for Multi-Rotor UAV Based on Improved YOLOv7
Lect. Notes Electrical Eng.
International Conference on Autonomous Unmanned Systems ; 2024 ; Shenyang, China September 19, 2024 - September 21, 2024
Proceedings of 4th 2024 International Conference on Autonomous Unmanned Systems (4th ICAUS 2024) ; Chapter : 3 ; 25-34
2025-03-28
10 pages
Article/Chapter (Book)
Electronic Resource
English
Improved YOLOv7 Target Detection Algorithm Based on UAV Aerial Photography
DOAJ | 2024
|Freeway vehicle detection on improved YOLOv7
SPIE | 2023
|Road pothole detection based on improved YOLOv7
SPIE | 2023
|