A Novel Dataset and Detection Method for Unmanned Aerial Vehicles Using an Improved YOLOv9 Algorithm

Area of Science:

• Computer Vision
• Artificial Intelligence
• Aerospace Engineering

Background:

• Unmanned aerial vehicles (UAVs) present detection challenges due to interference from visually similar objects like planes, helicopters, and birds.
• Existing UAV detection systems often lack robust anti-interference capabilities, leading to misclassifications.

Purpose of the Study:

• To enhance the anti-interference performance of UAV detection systems.
• To develop a robust method for distinguishing UAVs from other flying objects.

Main Methods:

• Construction of a novel anti-interference dataset with 5062 images, including UAVs, planes, helicopters, and birds.
• Proposal of a UAV detection method utilizing YOLOv9-C, incorporating dot distance for sample assignment to improve small target detection.

Main Results:

• The developed dataset aids in training detectors to differentiate UAVs from non-target objects.
• The proposed YOLOv9-C based method demonstrated superior anti-interference performance compared to existing algorithms.
• Improved detection accuracy for small UAV targets was achieved through optimized sample assignment.

Conclusions:

• The novel dataset and detection method significantly improve UAV detection's anti-interference capabilities.
• This research provides a valuable resource for developing and validating advanced UAV detection technologies.
• The findings contribute to safer and more reliable UAV monitoring and management systems.