When detecting uneven deformation of highway subgrade and pavement, the anchor box size of RPN is fixed and relies on rounding operations, resulting in inaccurate alignment between candidate boxes and real deformation areas, especially with poor adaptability to multi-scale deformation. Therefore, a method for detecting uneven deformation of highway subgrade and pavement based on Faster R-CNN algorithm is proposed. Construct a deformation detection network structure based on the Faster R-CNN algorithm, and introduce SE attention mechanism in the feature extraction network module to adaptively adjust the feature channel weights, thereby more accurately capturing the subtle features of uneven deformation and generating a feature map of uneven deformation of highway subgrade and pavement; The regional recommendation network module is based on the uneven deformation feature map of highway subgrade and pavement, and generates a fixed size and more accurate candidate region feature map of uneven deformation of highway subgrade and pavement by introducing multi-scale algorithms; The object detection network module uses ROI Pooling technology to process each candidate area box in the feature map of the uneven deformation candidate area of the highway subgrade and pavement one by one, and classifies them through a fully connected layer to distinguish different types of deformation, thereby achieving uneven deformation detection of the highway subgrade and pavement. The experimental results show that this method exhibits excellent feature representation ability, which can keenly capture the subtle features of uneven deformation of highway subgrade and pavement, and accurately enhance the key information of uneven deformation features; And it has demonstrated extremely high accuracy in the task of detecting uneven deformation of highway subgrade and pavement, with the output of various types of deformation quantities basically matching the actual quantity, indicating that its detection effect is good and its reliability is stronger.
SunJian-cheng, YangShu-han, GongFang-yuan,et al.Pavement crack detection in complex background based on improved YOLOv5[J]. China Sciencepaper,2023,18(7):779-785.
TaoJian, TianLin, ZhangDe-lin,et al.Asphalt pavement crack detection method based on local texture features[J].Computer Engineering and Design,2022,43(2): 517-524.
QiaoPeng, LiangZhi-qian, DuanChang-jiang,et al.Bridge defects detection and quantifying method based on modified Faster R-CNN and U-Net[J].Journal of Southeast University (Natural Science Edition),2024, 54(3): 627-638.
WangXi-chen, PengFu-lun, LiYe-xun, et al. Infrared target detection algorithm based on improved Faster R-CNN[J].Journal of Applied Optics, 2024, 45(2):346-353.
LengYue-feng, LiuZheng, XuBao-yi,et al.Detection of steel surface defect based on improved Faster R-CNN[J]. Mechanical Science and Technology for Aerospace Engineering,2025, 44(1): 75-83.
ChengJia-yu, ChenMiao-jin, LiTong,et al.Detection of peach trees in unmanned aerial vehicle(UAV)images based on improved Faster-RCNN network[J]. Acta Agriculturae Zhejiangensis, 2024, 36(8): 1909-1919.
CuiGuang-yan, WangYan-hui, XuJie,et al.Automatic detection of discrete entity objects in tunnel lining based on improved Faster R-CNN[J]. Journal of the China Railway Society, 2024, 46(2): 171-180.