Research on a Lightweight Multi-Object Detection Algorithm for Robots Used in Tunnel Inspection

DOI: https://doi.org/10.62517/jbdc.202601202

Author(s)

Jingze Wu1, Yihang He2,*

Affiliation(s)

1Yunnan Yunling Plateau Maintenance Engineering Co., Ltd., Kunming, Yunnan, China 2Southwest Jiaotong University, Chengdu, Sichuan, China *Corresponding Author

Abstract

Tunnel inspection robots face multiple challenges during locomotion, including body vibration, low illumination, and constrained computational resources, which impose dual requirements of lightweight design and high robustness on onboard object detection algorithms. To address these issues, a lightweight multi-object detection algorithm based on improved YOLOv11n is proposed. First, the KT-GLU gated attention enhancement module is introduced, which utilizes gated linear units and depthwise separable convolution to adaptively suppress noise features caused by motion jitter while improving CPU inference efficiency. Second, the KTF-trans attention mechanism is designed, employing asymmetric convolution decomposition and residual connections to enhance directional feature extraction in low-texture tunnel environments. Finally, the SSW-detect lightweight detection head is proposed, which reduces parameter redundancy and improves decision accuracy through cross-scale parameter sharing and group normalization. Experiments are conducted on three public datasets: KITTI, BDD100K-mini, and UA-DETRAC-G2. Results show that the improved model achieves mean average precision (mAP), precision (P), and recall (R) of 63.36%, 66.69%, and 59.31% respectively, representing improvements of 4.6%, 7.25%, and 1.59% over the baseline YOLOv11n, while reducing computational cost to 6.2 GFLOPs. The proposed model effectively balances detection accuracy and computational efficiency, providing a feasible technical solution for real-time object perception on tunnel inspection robots.

Keywords

Tunnel Inspection; Object Detection; YOLOv11; Lightweight Model; Attention Mechanism

References

[1] Girshick R, Donahue J, Darrell T, Malik J. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation. IEEE Computer Society, 2014:115-126. [2] Girshick R. Fast r-cnn// Proceedings of the IEEE international conference on computer vision, 2015:1440-1448. [3] Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks//Advances in Neural Information Processing Systems 28 (NIPS 2015). Montreal, Canada: Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2015: 91-99. [4] Liu W, Anguelov D, Erhan D, et al. SSD: Single Shot MultiBox Detector. Lecture Notes in Computer Science, 2016:21-37. [5] Bochkovskiy A, Wang C, Liao H M. YOLOv4: Optimal Speed and Accuracy of Object Detection. Computing Research Repository, 2020, abs /2004.10934. [6] KHANAM R, HUSSAIN M. YOLOv11: an overview of the key architectural enhancements. KHANAM R, HUSSAIN M. YOLOv11: an overview of the key architectural enhancements. 2410.17725,2024. [7] Geiger A, Lenz P, Stiller C, et al. Vision meets robotics: The KITTI dataset. The International Journal of Robotics Research, 2013, 32(11): 1231-1237. [8] Yu F, Chen H, Wang X, et al. BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 2636-2645. [9] Wen L, Du D, Cai Z, et al. UA-DETRAC: A new benchmark and protocol for multi-object detection and tracking. Computer Vision and Image Understanding, 2020, 193: 102907. [10] Hashempoor H, Hwang Y D. FastTracker: Real-Time and Accurate Visual Tracking. KHANAM R, HUSSAIN M. YOLOv11: an overview of the key architectural enhancements.2508.14370, 2025. [11] Chen J Y, Huang H T, Li Z Y, et al. Feature Enhancement and Metric Optimization for Defect Detection on Steel Surface. Laser & Optoelectronics Progress, 2024, 61(24): 92-101. [12] Tom Fawcett. An introduction to ROC analysis. Pattern Recognition Letters, 2006, 27(8): 861-874. [13] Gao Y, Chen P, Liu Z Q, et al. GUAV-YOLO: A Lightweight Detection Model for Small Target of Unmanned Aerial Vehicles in Grayscale Imag. Acta Optica Sinica, 1-28 [2025-09-25]. [14] Xu J, Chen W C, Jiang M, et al. Lithium Battery X-Ray Defect Detection Based on YOLOv8s. Laser & Optoelectronics Progress, 1-14 [2025-09-25]. [15] WANG C Y, YEH I H, MARK LIAO H Y. YOLOv9: learning what you want to learn using programmable gradient information. European Conference on Computer Vision. Cham: Springer Nature Switzerland, 2024:1-21.