A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Rapid pavement crack extraction method based on YOLOX and DeepLabv3 plus
The invention discloses a rapid pavement crack extraction method based on YOLOX and DeepLabv3plus, and the method comprises the steps: carrying out the marking of a collected pavement image of a part of a road section, and forming a target detection original data set; the YOLOX-L detection model is optimized, and a C-YOLOX model with a better detection effect is obtained; performing pixel-level labeling on an image in a C-YOLOX detection result rectangular frame to obtain a data set used for segmentation; a backbone network is optimized and improved on the basis of a DeepLabv3plus network model, so that a mobilenetv3-DeepLabv3 + segmentation model with a higher detection speed is obtained. According to the rapid pavement crack extraction method based on the YOLOX and the DeepLabv3plus, the improved C-YOLOX and the MobileNetv3-DeepLabv3 network are adopted, the detection precision in a detection task is effectively improved, the style precision of the crack can still be kept while the speed is greatly improved in a subsequent segmentation task, and the problem that the network consumes too long time in the detection task is effectively solved.
本发明公开了一种基于YOLOX和DeepLabv3plus的快速路面裂缝提取方法，包括对采集后的部分路段路面图像进行标注，形成目标检测原始数据集；对YOLOX‑L检测模型进行优化得到检测效果更优的C‑YOLOX模型；将C‑YOLOX检测结果矩形框内的图像进行像素级标注得到分割所用数据集；基于DeepLabv3plus网络模型对其骨干网络进行优化改进得到检测速度更快的mobilenetv3‑DeepLabv3+分割模型。本发明的基于YOLOX和DeepLabv3plus的快速路面裂缝提取方法采用改进后的C‑YOLOX以及MobileNetv3‑DeepLabv3网络，有效的提升了检测任务中的检测精度，并在后续分割任务中大幅提升速度的同时依旧能保持裂缝的风格精度，有效的解决了网络在检测任务中耗时过长的问题。
Rapid pavement crack extraction method based on YOLOX and DeepLabv3 plus
The invention discloses a rapid pavement crack extraction method based on YOLOX and DeepLabv3plus, and the method comprises the steps: carrying out the marking of a collected pavement image of a part of a road section, and forming a target detection original data set; the YOLOX-L detection model is optimized, and a C-YOLOX model with a better detection effect is obtained; performing pixel-level labeling on an image in a C-YOLOX detection result rectangular frame to obtain a data set used for segmentation; a backbone network is optimized and improved on the basis of a DeepLabv3plus network model, so that a mobilenetv3-DeepLabv3 + segmentation model with a higher detection speed is obtained. According to the rapid pavement crack extraction method based on the YOLOX and the DeepLabv3plus, the improved C-YOLOX and the MobileNetv3-DeepLabv3 network are adopted, the detection precision in a detection task is effectively improved, the style precision of the crack can still be kept while the speed is greatly improved in a subsequent segmentation task, and the problem that the network consumes too long time in the detection task is effectively solved.
本发明公开了一种基于YOLOX和DeepLabv3plus的快速路面裂缝提取方法，包括对采集后的部分路段路面图像进行标注，形成目标检测原始数据集；对YOLOX‑L检测模型进行优化得到检测效果更优的C‑YOLOX模型；将C‑YOLOX检测结果矩形框内的图像进行像素级标注得到分割所用数据集；基于DeepLabv3plus网络模型对其骨干网络进行优化改进得到检测速度更快的mobilenetv3‑DeepLabv3+分割模型。本发明的基于YOLOX和DeepLabv3plus的快速路面裂缝提取方法采用改进后的C‑YOLOX以及MobileNetv3‑DeepLabv3网络，有效的提升了检测任务中的检测精度，并在后续分割任务中大幅提升速度的同时依旧能保持裂缝的风格精度，有效的解决了网络在检测任务中耗时过长的问题。
Rapid pavement crack extraction method based on YOLOX and DeepLabv3 plus
基于YOLOX和DeepLabv3 plus的快速路面裂缝提取方法
WANG CHONGCHANG (author) / YOU JUNYU (author) / SUN SHANGYU (author) / JANG JIN-HYUK (author)
2023-06-23
Patent
Electronic Resource
Chinese
High-precision segmentation and quantification of tunnel lining crack using an improved DeepLabV3+
| Elsevier | 2025
PAVEMENT CRACK ANALYZER, PAVEMENT CRACK ANALYSIS METHOD, AND PAVEMENT CRACK ANALYSIS PROGRAM
| European Patent Office | 2018
Semantic Segmentation of Cracks Using DeepLabv3+
| Springer Verlag | 2023