Real-Time Detection for Small UAVs: Combining YOLO and Multi-frame Motion Analysis

Juanqin Liu,Leonardo Plotegher,Eloy Roura,Cristino de Souza Junior,Shaoming He

2024-10-10

Abstract:Unmanned Aerial Vehicle (UAV) detection technology plays a critical role in mitigating security risks and safeguarding privacy in both military and civilian applications. However, traditional detection methods face significant challenges in identifying UAV targets with extremely small pixels at long distances. To address this issue, we propose the Global-Local YOLO-Motion (GL-YOMO) detection algorithm, which combines You Only Look Once (YOLO) object detection with multi-frame motion detection techniques, markedly enhancing the accuracy and stability of small UAV target detection. The YOLO detection algorithm is optimized through multi-scale feature fusion and attention mechanisms, while the integration of the Ghost module further improves efficiency. Additionally, a motion detection approach based on template matching is being developed to augment detection capabilities for minute UAV targets. The system utilizes a global-local collaborative detection strategy to achieve high precision and efficiency. Experimental results on a self-constructed fixed-wing UAV dataset demonstrate that the GL-YOMO algorithm significantly enhances detection accuracy and stability, underscoring its potential in UAV detection applications.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to conduct efficient and accurate detection of small unmanned aerial vehicles (UAVs) at long distances. Traditional methods face significant challenges in identifying small - pixel targets at long distances, resulting in a decline in detection accuracy. Specifically, UAVs usually occupy less than 0.1% of the area in an image, which leads to insufficient feature information. Especially under complex background conditions, the detection difficulty is further increased.

To solve this problem, the paper proposes a detection algorithm named Global - Local YOLO - Motion (GL - YOMO). This algorithm combines YOLO object detection technology and multi - frame motion detection technology, aiming to significantly improve the accuracy and stability of small - target UAV detection. The main improvements include:

1. **Multi - scale feature fusion and attention mechanism**: Optimize the YOLO detection algorithm through multi - scale feature fusion and attention mechanism to enhance the detection ability of small targets.
2. **Ghost module**: Integrate the Ghost module to improve efficiency and reduce computational complexity and the number of parameters.
3. **Template - matching motion detection**: Develop a template - matching - based motion detection method to enhance the detection ability of tiny targets.
4. **Global - local collaborative detection strategy**: Utilize the global - local collaborative detection strategy to achieve high - precision and high - efficiency detection.

Experimental results show that the GL - YOMO algorithm significantly improves detection accuracy and stability on the self - built fixed - wing UAV data set, demonstrating its potential in UAV detection applications.