Infrared Dim Small Target Detection Networks: A Review

Area of Science:

• Computer Vision
• Artificial Intelligence
• Infrared Imaging

Background:

• Deep learning has revolutionized target detection, enabling new methods for infrared dim small target detection.
• Existing approaches face challenges in representation, accuracy, information loss, and complex backgrounds.

Purpose of the Study:

• To provide a comprehensive review of deep learning methods for infrared dim small target detection.
• To summarize recent advancements, datasets, networks, and evaluation metrics in the field.
• To identify key challenges and future research directions.

Main Methods:

• Focuses on deep learning methods developed in the last three years.
• Categorizes methods based on six critical issues in infrared dim small target detection.
• Reviews twelve public datasets and common evaluation metrics.

Main Results:

• Compares the performance of the latest deep learning networks quantitatively.
• Highlights advancements in enhancing small target representation and bounding box regression accuracy.
• Discusses strategies for mitigating information loss and balancing detection performance.

Conclusions:

• Deep learning offers significant advancements for infrared dim small target detection.
• Addressing challenges like complex backgrounds and network efficiency is crucial for practical deployment.
• The review provides a valuable resource for researchers entering or advancing in this field.