P-CSF: Polar coordinate cloth simulation filtering algorithm for multi-type tunnel point clouds

Tunnels are a crucial component of national transportation infrastructure, playing a vital role in social development and urban planning. With the widespread application of 3D laser scanning technology in tunnel engineering, accurately extracting information from vast scanning data and filtering out noise points has become particularly important. To address this challenge, we proposed a Polar coordinate Cloth Simulation Filtering algorithm (P-CSF) to separate lining points from non-lining points in tunnel point cloud data, including tunnels of different shapes and different excavation methods. First, the dual projection method is used to obtain the approximate central axis of the tunnel point cloud. Secondly, a polar coordinate system is established with the roughly determined central axis of the tunnel as the pole, and the simulated cloth is constructed on the outermost part of the section. Subsequently, the gravitational model is used to shrink the cloth particles inward until the distance from the cloth particles to the measured point cloud is less than the specified threshold or the maximum number of iterations is reached. Finally, when the particle motion stops, the points that are in contact with the cloth particles are identified as lining points, while the points that are not in contact are considered as non-lining points. This algorithm was verified in a variety of tunnel scenarios, demonstrating its adaptability and effectiveness. Qualitative analysis indicates that the algorithm can adapt to various scenarios and can adjust the size of simulated cloth details to extract regions of interest as need. Quantitative analysis shows that the overall accuracy of the algorithm exceeded 90% in four typical scenarios, and each scenario obtained a kappa coefficient of nearly 80%, demonstrating its effective extraction capability. In the future, we will continue to optimize the algorithm to cope with more challenging scenarios.