Frequent road inspections are key to maintaining road quality and avoiding casualties associated with poor road conditions. In Taiwan, open contractors conduct inspections of roads and ancillary facilities daily or weekly according to the requirements of the agency awarding the contracts. Unfortunately, the equipment used for inspections the inspection data lacks follow-up applications and numerical conversions, such as the Pavement Condition Index (PCI), to compile a large-scale database to facilitate the long-term conservation of roads. In this study, this paper developed back-end image recognition software using existing road inspection methods and existing equipment. This was aimed at enhancing inspection efficiency by enabling the automatic identification of road damage. Resulting observations can then be converted into PCI values in accordance with ASTM D6433-16 to be exported as a numeric value indicative of road quality. A vehicle-mounted traffic recorder and imaging device with Wi-Fi transmission capability are used as hardware, and the relationship between the captured images and the speed of the car is used to obtain an accurate indication of road conditions across the surface. The simple linear iterative clustering (SLIC) Superpixels algorithm is used to identify areas with pavement damage as patches, potholes, longitudinal cracking, and crocodile cracking. The results of the proposed fully-automated method conform strongly with those obtained using semi-automated pavement inspection software. Despite the restrictions imposed by the limited depth measurement of 2D images, our method achieved results close to those obtained using manual inspection. Future developments will include the application of artificial intelligence to enhance the effectiveness of this software.
- Automatic image recognition
- Pavement distress