CaSAM: Boosting SAM2 for Power Cable Segmentation in Complex Scenarios

View at publisher Summarise with VIRAT
Authors: Rui Zou, Kai Zhang, Xiaoliang Wu, JunJie Xu, LinFei Ding, YiJun Yang, Shuai Li
Cited by: 0
References: 12

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

With the development of smart grids, the need for precise cable segmentation in power robot live-line operations has become increasingly urgent. Existing deep neural networks rely heavily on large quantities of labeled samples, while visual foundation models such as SAM2 exhibit limited accuracy when directly applied to cable segmentation in complex environments. To address this, we propose the CaSAM architecture, which integrates dual low-rank adaptation (DLoRA) by fine-tuning both the encoder and decoder of SAM2, enabling efficient and accurate segmentation with a minimal number of trainable parameters. Additionally, we introduce the Cable-RJ dataset, which contains multimodal data for model training and evaluation. Experimental results demonstrate that CaSAM outperforms SAM, SAM2, and other models on various segmentation metrics within the Cable-RJ dataset, thereby validating the effectiveness of the proposed method and providing reliable perception support for power robot operations.

Fields of Study & Subjects

Power Line Inspection Robots Advanced Neural Network Applications Infrastructure Maintenance and Monitoring

Publication details

Journal / source
International Conference on Electrical Automation and Artificial Intelligence (ICEAAI)
Pages
622–626
Publication date
16 January 2026
Language
EN