IMD-Net: Interpretable multi-scale detection network for infrared dim and small objects.

Area of Science:

• Computer Vision
• Infrared Imaging Technology
• Artificial Intelligence

Background:

• Infrared small object detection in complex backgrounds remains challenging.
• Existing methods often suffer from high false alarm and missed detection rates.
• Accurate detection and segmentation are crucial for various applications.

Purpose of the Study:

• To propose an interpretable multi-scale infrared small object detection network (IMD-Net).
• To enhance the precision of infrared small object detection and contour segmentation.
• To address limitations of existing methods in complex backgrounds.

Main Methods:

• Developed a multi-scale object enhancement module converting features into network structures.
• Integrated global object response, channel attention, and multilayer feature fusion modules.
• Constructed a multiple loss constraint module to refine network output.

Main Results:

• IMD-Net achieved superior performance compared to state-of-the-art methods.
• Demonstrated significant improvements in intersection-over-union (IoU) and F-measure values (10.8% and 11.3% higher, respectively).
• Showcased accurate detection and effective segmentation of dim and small objects in diverse infrared complex backgrounds.

Conclusions:

• The proposed IMD-Net effectively improves infrared small object detection and segmentation.
• The network's design enhances feature extraction and information fusion for better accuracy.
• IMD-Net offers a robust solution for challenging infrared imaging scenarios.