A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
A Semantic Segmentation Method of Front-View Pavement Distress Based on SegFormer
Pavement distress segmentation and recognition is an important research direction of applying computer vison technology to intelligent traffic inspection. At present, most of the existing studies have used the perspective of overlooking to segment and detect pavement distress, and the types are single, most of them are segmentation and identification of crack distress. In addition, for the semantic segmentation of pavement distress in the front-view, there is a lack of public multi-type distress datasets for experiments. Aiming at the above situations, this study uses Encoder-Decoder to build SegFormer network structure to realize multi-category segmentation and recognition system for front-view pavement distress, and made a self-made front-view pavement distress Chongqing datasets applicable to domestic road traffic situation. Chongqing datasets consists of 1015 images with pixels of 1920x1080, which have been collected from real-word road sections with 11 different distress categories. SegFormer has achieved MIoU of 69.75% and the distress categories segmentation accuracy of 97.72% in the training and testing of the Chongqing data set. The successful application of this study not only provides a new perspective for the research in the field of front-view pavement distress segmentation, but also provides strong support for the application and further exploration in related fields. Chongqing Dataset will be released at: https://gitee.com/orangeYJ/chongqing-pavement-dataset
A Semantic Segmentation Method of Front-View Pavement Distress Based on SegFormer
Pavement distress segmentation and recognition is an important research direction of applying computer vison technology to intelligent traffic inspection. At present, most of the existing studies have used the perspective of overlooking to segment and detect pavement distress, and the types are single, most of them are segmentation and identification of crack distress. In addition, for the semantic segmentation of pavement distress in the front-view, there is a lack of public multi-type distress datasets for experiments. Aiming at the above situations, this study uses Encoder-Decoder to build SegFormer network structure to realize multi-category segmentation and recognition system for front-view pavement distress, and made a self-made front-view pavement distress Chongqing datasets applicable to domestic road traffic situation. Chongqing datasets consists of 1015 images with pixels of 1920x1080, which have been collected from real-word road sections with 11 different distress categories. SegFormer has achieved MIoU of 69.75% and the distress categories segmentation accuracy of 97.72% in the training and testing of the Chongqing data set. The successful application of this study not only provides a new perspective for the research in the field of front-view pavement distress segmentation, but also provides strong support for the application and further exploration in related fields. Chongqing Dataset will be released at: https://gitee.com/orangeYJ/chongqing-pavement-dataset
A Semantic Segmentation Method of Front-View Pavement Distress Based on SegFormer
Yang, Yuanji (author) / Ning, Zhipeng (author) / Li, Shenglin (author)
2023-08-25
4948593 byte
Conference paper
Electronic Resource
English
Evidential transformer for pavement distress segmentation
| Wiley | 2023
Critical Assessment of Pavement Distress Segmentation Methods
| Online Contents | 2010
Quantitative Performance Evaluation Algorithms for Pavement Distress Segmentation
| British Library Online Contents | 2010
Analysis of Segmentation Algorithms for Pavement Distress Images.
| Online Contents | 1993