Smart Road Damage Survey System Using Machine Vision and IoT Technology

Road is an essential transportation infrastructure to move people and goods to support economic prosperity. The efficiency, safety, security, and comfort of people have significant consequences by damage and defects on the road surface. A good-condition road ensures the land traffic runs well. This research is a collaboration between the Bandung Institute of Technology and the Department of Public Works and Spatial Planning of West Java Province to develop a smart system for road damage surveys. The system was built by leveraging machine vision and IoT technologies to improve the performance of the road survey activities. The previous road damage detection, classification method, and pavement condition index (IKP) computation were performed semi-automatically. The proposed survey system developed in this work uses the Canny edge Detection algorithm and MobileNet V2 architecture to detect road defects automatically. The pavement condition index computation also improved by using the interpolation reduction factor. The system is designed into a four-layer architecture that consists of several functions and modules, i.e., the sensor system that includes a vision sensor, GPS and odometry, a hardware embedded system using a microcontroller, and machine learning to process the road condition image automatically, and a dashboard application to reveal the road condition on map. The performance resulted in accuracy, precision, recall, and F-score of the road damage detection are 54.5%, 24.2%, 78.9%, and 37%, respectively.