FOSS4G-Asia 2023 Seoul

In recent years, aging infrastructure facilities has become an issue in various regions around the world. According to Japan's Ministry of Land, Infrastructure, Transport and Tourism, the lifespan of infrastructure is 50 years after construction, but the number of infrastructure facilities that have exceeded their lifespan is increasing every year. On the other hand, inspections are still conducted visually by facility managers, and the inability to properly conduct detailed inspections due to a lack of engineers has become a social issue. Furthermore, conventional visual inspections lack consistency due to the large variation in results from one inspector to another. To address these issues, we developed a web system that automatically detects cracks, one of the inspection items, by object detection, and manages and visualizes the results using location information. In this study, we focused on cracks located on the apron's top of a quay wall, one of the port facilities. In the inspection process, we first used a small general-purpose drone to automatically photograph the inspection target and create orthophotos. The orthophotos are then uploaded to a crack detection web system. The system segments the uploaded orthophoto into smaller sections and uses object detection to locate cracks in the segmented image. Using the detection results and the orthophoto's location information, the system calculates the detected cracks' location. The results are displayed on the orthophoto and output in GeoJSON format, which has the advantage that it can be easily combined with other geographic information data using QGIS for analysis and comparative analysis based on time series. In addition, the system can perform everything from crack detection to display and download of results centrally on the Web, making it easier and faster for inspectors to proceed with their inspections. This new method, which involves detecting cracks through object detection and acquiring consistent quality aerial images through automated drone flight, improves inspection efficiency while reducing the variability of inspection results.