Technology of Classification and Detection of Damage Conditions on the Road Surface

The development of today's technological world is occurring rapidly. allows for comparison with the preparation of a database that is fast, precise, and accurate. Some studies focus on the presence or absence of damage on the road surface, but basically, the data taken should provide informative information to obtain a mapping of the location (cluster), the type of damage, and the classification of road damage. Webbased mapping using Android applications from third parties allows the availability of data in real-time and photos of surface conditions. This study makes three contributions to address this problem. First, a road damage database was created for the first time. This dataset consisted of 9,053 images of road damage taken with a smartphone in the vehicle, and road images contained 15,435 examples of road damage. o produce this data set, the researchers conducted research in collaboration with 7 cities in Japan and obtained street images for more than 40 hours. These pictures were taken in different weather and lighting conditions. In each picture taken, a description of the bounding box representing the location and type of damage is given. In order to obtain the damage detection model from the dataset and evaluate the accuracy and runtime performance of both utilizing a server and smartphone, the second object detection approach uses a CNN. Thirdly, it can be demonstrated that the suggested object detection algorithms can accurately divide the different damage types into eight categories