Research on Crack Detection Technology Based on Semantic Segmentation
Based on machine vision,this paper designs a tunnel crack detection system based on semantic segmentation.The system u-ses the data acquisition vehicle to collect the tunnel apparent image.Then,based on the classical U-Net segmentation algorithm,VG-GNet16 with deeper network depth is used to replace the original coding network,and the migration learning method is introduced to optimize the robustness of the model and effectively improve the segmentation performance of the segmentation model.The pixel accu-racy and intersection to union ratio of the segmentation model on the test set have reached 98.96%and 0.807 9 respectively.The ex-perimental results show that the system meets the requirements of tunnel crack detection.
NETL
NSTL
万方数据
