Path Optimization for Surface Damage Detection on Large Structures

To address the limitations of existing methods in large-scale structural surface damage inspection - including poor path adaptability, insufficient three-dimensional (3D) reconstruction accuracy, and inadequate precision in damage detection and quantification - this study proposes an integrated multimodal inspection framework. An Improved Marine Predators Algorithm (IMPA) is first developed to solve the three-dimensional Open Traveling Salesman Problem (TO-TSP) involved in viewpoint planning. By introducing a hybrid Weibull-distribution-based motion strategy and a nonlinear adaptive step-size mechanism, the algorithm overcomes the limited exploration capability and rigid adjustment characteristics of conventional optimization approaches. The optimized path planning is further integrated with a lightweight deep learning network (LE-YOLOv5), forming a coherent system in which two-dimensional detection facilitates damage localization, while high-quality three-dimensional reconstruction enables accurate physical quantification. A bridge case study demonstrates the superior performance of the proposed approach. Compared with traditional paths, the IMPA-optimized trajectory increases the number of effective feature points by 33.98 % and the valid point cloud count by 25.12 %, while reducing the reprojection error by 17.06 %. Moreover, the system achieves high-precision damage quantification, relative measurement errors for crack length and width were 0.35 % and 8.91 % based on dense point clouds, and 1.30 % and 7.23 % based on triangular mesh models. These results confirm that the proposed framework substantially enhances reconstruction quality, improves damage detection accuracy, and provides a practical solution for structural health assessment of large-scale infrastructures.