Cracks in the Road
Problem being addressed
Road maintenance is a complex infrastructure task. But given the size and number of roads that need to maintained, its hard to know what needs attention. How can deep learning and imaging be used to detect cracks in the road?
A deep learning approach is used to study image data from roads and pavements. Given a pavement image, the object of the algorithm is to determine from the pixel structure if there is a crack in the road or pavement. The deep learning approach is then compared to other classification approaches from machine learning such as support vector machines and boosting methods. It is found that the deep learning approach is very robust, and works with low resolution images such as those take from a cellphone camera. This gives readily deployable road monitoring system.
Advantages of this solution
Infrastructure maintenance is an important government function. This research makes the problem of road maintenance simpler by using AI to find those areas that need attention.
Solution originally applied in these industries
Possible New Application of the Work
Insurers often have difficulty in assessing properties and can sometimes take risks on buildings that are not sound. Using this methodology perhaps a system could be designed to check for cracks and soundness in buildings for insurance purposes.
Mining equipment needs to be checked for wear and tear on a regular basis in order to keep people safe. This methodology could be reapplied toward automatically checking equipment for damage via image scanners.
Source DOI: #############
Source URL: #############