DEVELOPMENT OF AI MODEL FOR DIAGNOSING THE DETERIORATION OF LIGHTING LUMINAIRE IN HIGHWAY TUNNEL

Abstract

In this study, we constructed a deep learning model that outputs deterioration diagnosis results focusing on the corrosion status of bolts and nuts, using images of tunnel luminaire fixture components of expressways as input. As a result, a model that is suitable for judging the deterioration of luminaire fixture components is developed. Even though the training was conducted with an extremely small data set of 333 images, the diagnosis accuracy of the constructed model for the degree of deterioration of luminaire fixture components reached more than 80% by combining data augumentation and optimization of the anchor box size. In the future, we would like to improve the accuracy of the model by increasing the number of image samples under more diverse installation conditions.