An improved infrared abnormal transmission line target detection algorithm based on D-FINE

Read full text Summarise with VIRAT
Authors: Tong Wu, Liming Wang, X. George Xu, Wei He
Cited by: 0
References: 3

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Abstract To address the problems of low signal-to-noise ratio, blurry target edges, and complex background interference in infrared images, this study proposes an improved D-FINE-based algorithm for detecting abnormal transmission line targets in infrared imagery. The algorithm enhances high-frequency detail extraction by introducing wavelet transform convolution (WTConv) with a large receptive field. It also constructs a dual-domain selection mechanism to establish dynamic interaction between the spatial and frequency domains for suppressing background interference. In addition, it designs a hybrid dilated residual attention block (HDRAB) to address the problem of missing detection of multi-scale components. Experimental results show that the improved model achieves an AP50 of 85.3%, an AP95 of 48.6%, and an inference speed of 148 FPS on the self-built infrared dataset. Compared with mainstream models such as the YOLO series and RT-DETR, the proposed algorithm significantly improves detection accuracy and robustness while maintaining real-time performance.

Fields of Study & Subjects

Infrared Target Detection Methodologies Advanced Neural Network Applications Power Line Inspection Robots

Publication details

Journal / source
Journal of Physics Conference Series
Volume
3192
Issue
1
Pages
012025–012025
Publication date
1 March 2026
Language
EN