Channel Extraction and Geometric Parameters Measurement Based on Point Clouds

Abstract. In electrified railways, accurately measuring geometric parameters of pre-embedded tunnel channels is crucial for ensuring the installation precision and operational stability of the railway overhead contact network system. However, traditional manual measurement methods face challenges such as complex construction environments, stringent precision demands, and limited technical capabilities, resulting in inefficiency, significant safety risks, and poor repeatability. To address these issues, this paper introduces a novel framework based on total station point clouds. The framework comprises three key modules: point cloud preprocessing, channel extraction, and geometric parameter measurement. During preprocessing, point clouds are aligned using principal component analysis (PCA), and interference points are removed to enhance data quality. For channel extraction, an Otsu-based curvature threshold is first applied to preliminarily identify channel point clouds. Subsequently, a refined extraction process combining statistical denoising and density-based clustering is employed to isolate the channel point clouds with greater precision. In terms of geometric parameter measurement, an arc-based method is utilized for length measurement, while a generalized Gaussian distribution (GGD)-based approach is adopted for depth estimation. Experimental results demonstrate that the proposed method significantly improves channel extraction performance, achieving an F1-score improvement of up to 24.8%. Furthermore, the framework enables millimeter-level depth estimation with a mean absolute error of 1.90 mm.