Title: RDBL-Net: detection of foreign objects on transmission lines based on positional encoding multiscale feature fusion

Authors: Xiaoli Guo; Yifan Bao; Hao Jiang; Zichong Feng; Yuhan Sun

Addresses: Northeast Electric Power University, Jilin City, Jilin Province 132012, China ' Northeast Electric Power University, Jilin City, Jilin Province 132012, China ' Liaoning University of Technology, Jinzhou City, Liaoning Province, 121001, China ' China Logistics Co., Ltd., Harbin Branch, Heilongjiang 100073, China ' Jilin Investment Promotion Centre, Jilin City, Jilin Province 132011, China

Abstract: Timely detection of foreign objects on transmission lines is the key to ensuring transmission lines' safe and stable operation. This paper takes the dataset provided by the Guangzhou Pazhou Algorithm Competition as the data basis, proposing a transmission line foreign object detection model. First, a residual connection module combined with deformable convolution is introduced into the backbone, aiming to improve the feature extraction capability of the model for transmission line foreign objects. Second, a multiscale feature fusion structure is designed to enhance the fusion effect of multiscale features. Subsequently, a learnable position encoding is introduced into the multiscale feature interaction module to enhance the model's ability to cope with complex environments' interference. Finally, the effectiveness of the proposed method is demonstrated through experiments.

Keywords: detection of foreign objects on transmission lines; RT-DETR; deformable convolution network; multiscale feature fusion.

DOI: 10.1504/IJSNET.2025.144556

International Journal of Sensor Networks, 2025 Vol.47 No.2, pp.61 - 71

Received: 14 Oct 2024
Accepted: 28 Oct 2024
Published online: 19 Feb 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article