期刊论文

Light-weight Bridge Crack Images Segmentation Algorithm Based on GhostNet

中文

公路

0451-0712

2023

68

04

246-255

湖北省交通运输厅科技项目计划,项目编号2020-2-6-1；

本校为第一机构

针对现有的桥梁裂缝检测语义分割网络参数量过于庞大、计算时间长、难以部署在移动端等资源受限的设备上,而且存在裂缝细节识别困难、分割结果不连续的问题,提出了一种优化PSPNet网络结构的改进型轻量级裂缝语义分割模型Ghost...展开更多 针对现有的桥梁裂缝检测语义分割网络参数量过于庞大、计算时间长、难以部署在移动端等资源受限的设备上,而且存在裂缝细节识别困难、分割结果不连续的问题,提出了一种优化PSPNet网络结构的改进型轻量级裂缝语义分割模型Ghost-PSPNet。首先利用GhostNet替换ResNet-50作为主干网络,以更小的计算代价生成冗余特征,其次在模型中引入一条支路结构融合多层级特征,得...

In view that the existing bridge crack detection semantic segmentation network parameters are too large,the calculation time is long,it is difficult to deploy on resource-constrained devices such as mobile terminals,and there are difficulties in identifying crack detail...MORE In view that the existing bridge crack detection semantic segmentation network parameters are too large,the calculation time is long,it is difficult to deploy on resource-constrained devices such as mobile terminals,and there are difficulties in identifying crack details and discontinuous segmentation results,an improved...