A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Computer Vision and Deep Learning for Real-Time Pavement Distress Detection
Abstract Despite of the increased level of automation in vehicles, the detection of pavement distress, such as cracks and potholes, is mostly performed manually. We propose a methodology for automated pavement distress detection based on computer vision. Thereby, images obtained by cameras installed in common passenger vehicles are analyzed in real time, resulting in cost savings and a reduced amount of stored data. For this purpose, the wavelet transform was implemented on Graphics Processing Units (GPU). In addition, median filtering and top-hat transform were also implemented on GPU to enable real-time noise removal and correction of non-uniform background illumination. To distinguish between surface types, we incorporated textural features into our methodology and deep learning was utilized to determine the distress type (cracks, potholes or patches). Results obtained by different vehicles were aggregated to improve the reliability of the methodology. Case studies were conducted for validation and tests achieved promising results.
Computer Vision and Deep Learning for Real-Time Pavement Distress Detection
Abstract Despite of the increased level of automation in vehicles, the detection of pavement distress, such as cracks and potholes, is mostly performed manually. We propose a methodology for automated pavement distress detection based on computer vision. Thereby, images obtained by cameras installed in common passenger vehicles are analyzed in real time, resulting in cost savings and a reduced amount of stored data. For this purpose, the wavelet transform was implemented on Graphics Processing Units (GPU). In addition, median filtering and top-hat transform were also implemented on GPU to enable real-time noise removal and correction of non-uniform background illumination. To distinguish between surface types, we incorporated textural features into our methodology and deep learning was utilized to determine the distress type (cracks, potholes or patches). Results obtained by different vehicles were aggregated to improve the reliability of the methodology. Case studies were conducted for validation and tests achieved promising results.
Computer Vision and Deep Learning for Real-Time Pavement Distress Detection
Doycheva, Kristina (author) / Koch, Christian (author) / König, Markus (author)
2018-10-04
7 pages
Article/Chapter (Book)
Electronic Resource
English
Pavement distress detection , Graphics processing units , Textural features , Deep learning Engineering , Building Construction and Design , Data Mining and Knowledge Discovery , Building Repair and Maintenance , Computer-Aided Engineering (CAD, CAE) and Design , Light Construction, Steel Construction, Timber Construction , Construction Management
GPU-enabled real-time pavement distress detection based on computer vision and deep learning
| BASE | 2020
| British Library Online Contents | 2017
| British Library Online Contents | 2017
| British Library Online Contents | 2017
| British Library Online Contents | 2017