Experiment study on UAV target detection algorithm based on YOLOv8n-ACW.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Robotics

Background:

• Small target detection with unmanned aerial vehicles (UAVs) faces challenges due to dense and occluded objects.
• Existing algorithms often struggle with the scale and complexity of real-world aerial imagery.

Purpose of the Study:

• To enhance the performance of small target detection algorithms for UAVs.
• To develop a more efficient and accurate detection model for aerial surveillance and monitoring.

Main Methods:

• An enhanced detection algorithm, YOLOv8n-ACW, was developed based on the YOLOv8n model.
• Key modifications include integrating Adown into the Backbone, developing a CCDHead, and implementing WIoU-V3 as the loss function.
• Experiments were conducted on the Visdrone2019 dataset and a self-constructed dataset of G5-Pro drones.

Main Results:

• The YOLOv8n-ACW model achieved a 4.2% increase in mAP50 compared to the baseline YOLOv8n model.
• The enhanced model reduced the parameter count by 36.7%, indicating improved efficiency.
• The model demonstrated robust generalization capabilities on a real-world drone dataset.

Conclusions:

• The YOLOv8n-ACW algorithm offers superior performance for small target detection in UAV applications.
• The model's efficiency and accuracy make it suitable for practical, real-world deployment.
• The research contributes to advancing UAV-based target detection technology and its educational applications.