A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Road crack detection method
The invention belongs to the technical field of computer vision, and particularly relates to a road crack detection method. Obtaining a picture of a to-be-detected road, and inputting the picture into the road crack detection model to obtain a crack detection result of the to-be-detected road; wherein the road crack detection model is of a U-shaped encoding-decoding structure, feature maps, extracted by an encoder in the U-shaped encoding-decoding structure, of different scales are subjected to layer context information capture and fusion of multiple visual receptive fields through respective ASPP modules, and a fused result is input into a decoder for multi-scale feature fusion processing; and a semantic segmentation result is obtained through the detection head. Considering that the features of the shallow layer have high reference value for final semantic segmentation for crack detection, the method uses a model of a U-shaped coding-decoding structure. And on the basis, an ASPP module is added, and the ASPP module is used for capturing context information in various different ranges, so that the detection precision is improved.
本发明属于计算机视觉技术领域，具体涉及一种道路裂缝检测方法。获取待检测道路的图片，并输入至道路裂缝检测模型中，得到待检测道路的裂缝检测结果；其中，道路裂缝检测模型为U形编码‑解码结构，U形编码‑解码结构中编码器提取的不同尺度的特征图分别通过各自的ASPP模块进行多视觉感受野的图层上下文信息捕获融合，将融合后的结果输入至解码器进行多尺度特征的融合处理，进而通过检测头得到语义分割结果。考虑到浅层的特征对用于裂缝检测的最终的语义分割具有较高的参考价值，因而本发明使用U形编码‑解码结构的模型。并在此基础上增加了ASPP模块，利用ASPP模块进行多种不同范围上下文信息的捕捉，提高了检测精度。
Road crack detection method
The invention belongs to the technical field of computer vision, and particularly relates to a road crack detection method. Obtaining a picture of a to-be-detected road, and inputting the picture into the road crack detection model to obtain a crack detection result of the to-be-detected road; wherein the road crack detection model is of a U-shaped encoding-decoding structure, feature maps, extracted by an encoder in the U-shaped encoding-decoding structure, of different scales are subjected to layer context information capture and fusion of multiple visual receptive fields through respective ASPP modules, and a fused result is input into a decoder for multi-scale feature fusion processing; and a semantic segmentation result is obtained through the detection head. Considering that the features of the shallow layer have high reference value for final semantic segmentation for crack detection, the method uses a model of a U-shaped coding-decoding structure. And on the basis, an ASPP module is added, and the ASPP module is used for capturing context information in various different ranges, so that the detection precision is improved.
本发明属于计算机视觉技术领域，具体涉及一种道路裂缝检测方法。获取待检测道路的图片，并输入至道路裂缝检测模型中，得到待检测道路的裂缝检测结果；其中，道路裂缝检测模型为U形编码‑解码结构，U形编码‑解码结构中编码器提取的不同尺度的特征图分别通过各自的ASPP模块进行多视觉感受野的图层上下文信息捕获融合，将融合后的结果输入至解码器进行多尺度特征的融合处理，进而通过检测头得到语义分割结果。考虑到浅层的特征对用于裂缝检测的最终的语义分割具有较高的参考价值，因而本发明使用U形编码‑解码结构的模型。并在此基础上增加了ASPP模块，利用ASPP模块进行多种不同范围上下文信息的捕捉，提高了检测精度。
Road crack detection method
一种道路裂缝检测方法
CAO YIBING (author) / CUI PENGYU (author) / ZHANG ZHENG (author) / ZHAO XINKE (author) / FAN XINHUA (author) / ZHENG JINGBIAO (author)
2023-05-12
Patent
Electronic Resource
Chinese