Real-time semantic segmentation of strip mending based on RT PSPNet

Chen Wenke; Li Lecheng; Liu Guoxi; Huang Bi; Dai Fei

doi:10.11959/j.issn.2096-6652.ZNKX26016

您当前的位置：

首页 >

文章列表页 >

Real-time semantic segmentation of strip mending based on RT PSPNet

Papers and Reports | 更新时间：2026-07-29

- Real-time semantic segmentation of strip mending based on RT PSPNet
- Chinese Journal of Intelligent Science and Technology Vol. 8, Issue 2, Pages: 237-249(2026)
- 作者机构：
  
  1.西南林业大学大数据与智能工程学院，云南昆明 650224
  2.西南林业大学林学院，云南昆明 650224
- 作者简介：
- 基金信息：
  
  Foundational Research Special Projects of Yunnan Province(202301AU070223)
- DOI：10.11959/j.issn.2096-6652.ZNKX26016
  CLC： TP391
- Received：30 January 2026，
  
  Revised：2026-05-15，
  
  Accepted：29 May 2026，
  
  Online First：27 July 2026，
  
  Published：15 June 2026
- 稿件说明：
移动端阅览
陈文柯,李乐成,刘国玺等.基于RT PSPNet的路面条状修补实时语义分割研究[J].智能科学与技术学报,2026,08(02):237-249.

Chen Wenke,Li Lecheng,Liu Guoxi,et al.Real-time semantic segmentation of strip mending based on RT PSPNet[J].Chinese Journal of Intelligent Science and Technology,2026,08(02):237-249.
陈文柯,李乐成,刘国玺等.基于RT PSPNet的路面条状修补实时语义分割研究[J].智能科学与技术学报,2026,08(02):237-249. DOI： 10.11959/j.issn.2096-6652.ZNKX26016.

Chen Wenke,Li Lecheng,Liu Guoxi,et al.Real-time semantic segmentation of strip mending based on RT PSPNet[J].Chinese Journal of Intelligent Science and Technology,2026,08(02):237-249. DOI： 10.11959/j.issn.2096-6652.ZNKX26016.

摘要

路面病害中的条状修补检测是路面智能养护的关键技术之一。现有基于深度学习的语义分割模型在此任务中普遍存在模型参数量大、推理速度慢、细节特征丢失等问题，难以同时满足精度与实时性要求。针对上述问题，提出了一种轻量化高精度的实时语义分割模型——RT PSPNet。该模型在PSPNet基础上进行了多维改进：首先，对ResNet50主干进行深度与通道数的结构化剪枝，得到浅层轻量ResNet，在显著减少参数量的同时保留更多细节特征；其次，设计轻量化小型金字塔池化模块（CPPM）替代原PPM，减少池化分支并引入逐深度卷积，以提升特征提取效率；最后，引入多尺度特征融合机制与多分割头复合损失函数，增强模型对细长目标的完整分割能力与训练稳定性。实验结果表明，在自建条状修补数据集上，RT PSPNet相比原PSPNet在分割精度（mIoU）上提升7.06个百分点，参数量减少96.4%，计算量降低76.1%，推理速度达到147 fps，结合混合精度加速策略后可达258 fps，为基准模型的8.6倍。与当前多种主流及轻量化先进模型相比，RT PSPNet在分割精度、推理速度与模型轻量化3个方面表现出综合的领先优势。RT PSPNet的模型代码和权重已在GitHub上供免费下载使用。

Abstract

The detection of strip mending in pavement diseases provides crucial technical support for intelligent road maintenance. Existing deep learning-based semantic segmentation models often suffer from issues such as large model parameters

slow inference speed

and loss of detailed features in strip mending detection tasks

making it difficult to balance accuracy and real-time performance. To address these challenges

RT PSPNet

a lightweight and high-precision real-time semantic segmentation model

was proposed. Based on PSPNet

RT PSPNet incorporated multi-dimensional improvements. Firstly

structured pruning was applied to the ResNet50 backbone in terms of depth and channel number

resulting in a shallow lightweight ResNet that significantly reduced parameters while preserving more detailed features. Secondly

a lightweight compact pyramid pooling module (CPPM) was designed to replace the original PPM

which reduced pooling branches and introduced depth-wise convolutions to further enhance feature extraction efficiency. Finally

a multi-scale feature fusion mechanism and a multi-head composite loss function were introduced to improve the model's ability to segment slender targets completely and enhance training stability. Experimental results on a self-built strip mending dataset show that RT PSPNet improves the segmentation accuracy (mIoU) by 7.06 percentage points compared to the original PSPNet

while reducing the parameter count by 96.4%

computational cost by 76.1%

and achieving an inference speed of 147 fps

and reaching 258 fps while using a mixed precision speeding strategy

which is 8.6 times that of the baseline model. Compared with many mainstream and lightweight state-of-the-art models

RT PSPNet demonstrates comprehensive advantages in accuracy

speed

and model light-weighting. RT PSPNet's code and weights are available for free accessing and downloading on GitHub.

关键词

Keywords

references

王双喜. 沥青混凝土路面病害分析及应对措施[J]. 黑龙江交通科技, 2011(11): 29-36.

Wang S X. Analysis of asphalt concrete pavement diseases and countermeasures[J]. Communications Science and Technology Heilongjiang, 2011(11): 29-36.

王飞跃. 智能的本质与意义[J]. 智能科学与技术学报, 2026, 8(1): 1-11.

Wang F Y. On intelligence: the origin and the goal[J]. Chinese Journal of Intelligent Science and Technology, 2026, 8(1): 1-11.

王飞跃, 缪青海, 张军平, 等. 探讨AI for Science的影响与意义: 现状与展望[J]. 智能科学与技术学报, 2023, 5(1): 1-6.

Wang F Y, Miao Q H, Zhang J P, et al. The DAOs to AI for science by DeSci: the state of the art and perspective[J]. Chinese Journal of Intelligent Science and Technology, 2023, 5(1): 1-6.

严毅, 邓超, 李琳, 等. 深度学习背景下的图像语义分割方法综述[J]. 中国图象图形学报, 2023, 28(11): 3342-3362.

Yan Y, Deng C, Li L, et al. Survey of image semantic segmentation methods in the deep learning era[J]. Journal of Image and Graphics, 2023, 28(11): 3342-3362.

田萱, 王亮, 丁琪. 基于深度学习的图像语义分割方法综述[J]. 软件学报, 2019, 30(2): 440-468.

Tian X, Wang L, Ding Q. Review of image semantic segmentation based on deep learning[J]. Journal of Software, 2019, 30(2): 440-468.

徐文远, 李帅, 纪泳丞, 等. RTFormerF10: 基于RTFormer-slim模型融合红外与可见光图像的裂缝分割研究[J]. 中国公路学报, 2025, 38(4): 69-84.

Xu W Y, Li S, Ji Y C, et al. RTFormerF10: research on crack segmentation based on RTFormer-slim model fusion of infrared and visible light images[J]. China Journal of Highway and Transport, 2025, 38(4): 69-84.

唐源, 董绍江, 刘超, 等. 基于改进SegFormer网络的路面裂缝分割算法[J]. 陕西科技大学学报, 2024, 42(3): 166-173.

Tang Y, Dong S J, Liu C, et al. An improved pavement crack segmentation algorithm based on SegFormer network[J]. Journal of Shaanxi University of Science & Technology, 2024, 42(3): 166-173.

韩建峰, 张静, 宋丽丽, 等. 改进U-Net的无人机航拍路面破损检测方法[J]. 电子测量技术, 2024, 47(19): 137-145.

Han J F, Zhang J, Song L L, et al. Improved pavement damage detection method of UAV based on U-Net[J]. Electronic Measurement Technology, 2024, 47(19): 137-145.

杨萍, 张汐. 改进DeepLabv3+的道路表面裂缝检测方法[J]. 计算机工程, 2025, 51(4): 262-268.

Yang P, Zhang X. Improved DeepLabv3+ road surface crack detection method[J]. Computer Engineering, 2025, 51(4): 262-268.

Yu C Q, Gao C X, Wang J B, et al. BiSeNet V2: bilateral network with guided aggregation for real-time semantic segmentation[J]. International Journal of Computer Vision, 2021, 129(11): 3051-3068.

Mehta S, Rastegari M. MobileViT: light-weight, general-purpose, and mobile-friendly vision transformer[PP]. V2. (2022-03-04)[2026-01-30]. arXiv: arXiv.2110.02178.

Xu Z Z, Wu D Y, Yu C Q, et al. SCTNet: single-branch CNN with transformer semantic information for real-time segmentation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(6): 6378-6386.

Zhao H S, Shi J P, Qi X J, et al. Pyramid scene parsing network[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE Press, 2017: 6230-6239.

He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE Press, 2016: 770-778.

Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic segmentation[C]//Proceedings of the IEEE Transactions on Pattern Analysis and Machine Intelligence. Piscataway: IEEE Press, 2016: 640-651.

Chen L C, Zhu Y K, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Computer Vision - ECCV 2018. Cham: Springer, 2018: 833-851.

Sun K, Zhao Y, Jiang B R, et al. High-resolution representations for labeling pixels and regions[PP]. V1. (2019-04-09)[2026-01-30]. arXiv: arXiv.1904.04514.

Xie E, Wang W, Yu Z, et al. SegFormer: simple and efficient design for semantic segmentation with transformers[C]//Advances in Neural Information Processing Systems. Massachusetts: MIT Press, 2021: 12077-12090.

Howard A G, Zhu M L, Chen B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[PP]. V1. (2017-04-17)[2026-01-30]. arXiv: arXiv.1704.04861.

王国忠, 陈明星, 姚辉, 等. 图像灰度处理对路面裂缝病害检测影响分析[J]. 市政技术, 2024, 42(4): 270-277.

Wang G Z, Chen M X, Yao H, et al. Influence of image gray process on the detection of pavement cracks[J]. Journal of Municipal Technology, 2024, 42(4): 270-277.

Li X Y, Sun X F, Meng Y X, et al. Dice loss for data-imbalanced NLP tasks[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Kerrville: Association for Computational Linguistics, 2020: 465-476.

Badrinarayanan V, Kendall A, Cipolla R. SegNet: a deep convolutional encoder-decoder architecture for image segmentation[PP]. V3. (2016-10-10)[2026-01-30]. arXiv: arXiv.1511.00561.

Zhou Z W, Rahman Siddiquee M M, Tajbakhsh N, et al. UNet++: a nested U-Net architecture for medical image segmentation[C]//Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. Cham: Springer International Publishing, 2018: 3-11.

Paszke A, Chaurasia A, Kim S, et al. ENet: a deep neural network architecture for real-time semantic segmentation[PP]. V1. (2016-06-07)[2026-01-30]. arXiv: arXiv.1606.02147.

Zhao H S, Qi X J, Shen X Y, et al. ICNet for real-time semantic segmentation on high-resolution images[C]//Computer Vision - ECCV 2018. Cham: Springer, 2018: 418-434.

Yu C Q, Wang J B, Peng C, et al. BiSeNet: bilateral segmentation network for real-time semantic segmentation[C]//Computer Vision - ECCV 2018. Cham: Springer, 2018: 334-349.

Wang L B, Li R, Zhang C, et al. UNetFormer: a UNet-like transformer for efficient semantic segmentation of remote sensing urban scene imagery[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2022, 190: 196-214.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Feature-enhanced visual-inertial SLAM method for low-texture dynamic environment

Path planning for indoor service robots using an improved ACO-DWA hybrid algorithm

Review of deep learning-based foreign object intrusion detection for transmission lines via UAV aerial photography

The knowledge origin of artificial intelligence

Ternary collaborative model of occupational tasks: a tangram occupational analysis framework for digital humans-robots-biological humans

Related Author

Fei Dai

Bi Huang

Guoxi Liu

Lecheng Li

Wenke Chen

XIE Mingyang

XU Xin

YANG Chen

Related Institution

College of big data and intelligent engineering of Southwest Forestry University

College of Automation Engineering, Nanjing University of Aeronautics and Astronautics

Postal code：100079
Tel：（010）53879206 Email：cjist@bjxintong.com.cn
Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19 京公网安备11010802024621
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
Total Visits：185511 Visits Today：506

⁰