Enhancing UAV object detection with an efficient multi-scale feature fusion framework.

Area of Science:

• Remote Sensing
• Computer Vision
• Artificial Intelligence

Background:

• Unmanned Aerial Vehicle (UAV) technology enables high-resolution remote sensing for various applications.
• Detecting small objects in UAV imagery is challenging due to scale variations and environmental factors.
• Existing methods often struggle with high computational costs or loss of fine-grained features for small target detection.

Purpose of the Study:

• To develop an improved object detection model for small targets in UAV remote sensing images.
• To enhance the YOLOv5n model for better accuracy and efficiency in detecting minute objects.
• To address the limitations of current methods in preserving crucial features for small object identification.

Main Methods:

• Proposed SRD-YOLOv5, an enhanced YOLOv5n model featuring a novel multi-scale feature fusion framework.
• Introduced Scale Sequence Feature Fusion Module (SSFF) and Multi-Scale Feature Extraction Module (MSFE) for capturing global context and detailed semantics.
• Incorporated an Extremely Small Target Detection Layer (ESTDL) to retain high-resolution features and a Decoupled Head for optimized detection.

Main Results:

• SRD-YOLOv5 demonstrated superior performance in detecting small objects in UAV remote sensing images compared to existing methods.
• The model achieved higher accuracy in small target detection.
• Maintained low computational demands, suitable for real-time UAV applications.

Conclusions:

• SRD-YOLOv5 effectively overcomes the challenges of small object detection in UAV remote sensing.
• The proposed model offers a balance between high accuracy and computational efficiency.
• SRD-YOLOv5 is a promising solution for real-time small object detection in UAV applications.