Vision Sensing-Driven Tunnel Crack Detection Method Using Particle Filtering-Integrated YOLOv5 Model

Vision computer has been a promising insight for tunnel crack detection. As a consequence, this paper proposes a vision sensing-driven tunnel crack detection method using particle filtering-Integrated YOLOv5 model. First, the tunnel image is preprocessed using Particle Filtering (PF), and the Artistic Lion Optimization (ALO) algorithm is used to extract candidate regions of cracks. Then, the YOLOv5 algorithm is used for deep learning detection of the candidate regions. Subsequently, the preliminary correction of the image and the selection of feature values are completed. Hence, the quantitative indicator calculation method for tunnel crack detection is determined. Finally, an integrated detection algorithm that fuses PF, ALO and YOLOv5 is established, and it is named ALO-PFYO for short. The accuracy and stability of this model are studied through actual testing cases. The experimental results show that ALO-PFYO has achieved good detection results on real datasets. Compared with traditional methods, it has higher accuracy and robustness, and can better cope with complex tunnel environments.