Research on road waterlogging area detection technology based on image processing
Zhu Fan
Xihua University
DOI: https://doi.org/10.59429/esta.v12i3.11358
Keywords: road water accumulation detection; YOLOv8; MPDIoU
Abstract
With the acceleration of urbanization and the frequent occurrence of extreme weather events, traffic safety hazards and urban waterlogging caused by road waterlogging are becoming increasingly prominent. Traditional detection methods mainly rely on manual inspections or physical sensors, which have problems such as low efficiency, high cost, and poor real-time performance. Based on the YOLOv8 target detection framework, this paper proposes a systematic optimization solution for the challenges of small target missed detection, complex background interference, and multi-scale morphological changes that are common in road waterlogging scenes. In terms of model construction, this paper improves the CSPDarknet53 backbone network, and enhances the model's ability to extract multi-scale features by optimizing the cross-stage partial connection (CSP) structure and the spatial pyramid pooling (SPPF) module; in the neck network, the PANet bidirectional feature pyramid structure is introduced and the dynamic convolution mechanism is integrated, which effectively improves the fusion efficiency of shallow details and deep semantic information; the detection head adopts the Anchor-free design, directly regresses the target center point and width and height parameters, simplifies the model structure, and improves the positioning accuracy of irregular water accumulation areas. To verify the performance of the model, this paper constructs a road waterlogging dataset covering different lighting, rainfall intensities and waterlogging forms, and uses enhancement strategies such as random rotation and reflection simulation to improve generalization. The experimental results show that the optimized YOLOv8 model can achieve real-time detection speed and exhibit high segmentation accuracy and robustness.
References
[1] Liao Yuhong, Guoru Huang. Research on depth monitoring method of urban road waterlogging based on deep learning and ellipse detection[J]. People's Pearl River, 2023, 44 (06): 1-8+17.
[2] Zhao Lili, Chao Xin. Waterlogging detection algorithm for roads based on migration learning[J]. Journal of Texas College, 2023, 39 (02): 31-35+50.
[3] Ding Zhangjie, Jian Huang, Linyuan Xie, et al. Target detection method for engineering scenes based on improved YOLOv8 [J]. China Journal of Construction Machinery, 2025, 23(02): 366-370+376.
[4] Li Xin, Wang Feng, Liu Qiming. Dual-kernel convolutional pavement crack detection algorithm with multidimensional feature fusion [J]. Journal of Lanzhou Jiaotong University, 2025, 44(02): 111-122.
[5] Hua Bing. Research on road detection and target detection technology in intelligent vehicles [D]. Chongqing University, 2008.