Abstract: The extraction of railway track point clouds from complex railroad environments constitutes a critical technology for intelligent railway construction, autonomous train operation, and safety inspection. With the rapid expansion of railway networks and increasingly stringent requirements for track extraction, the accurate and efficient acquisition of track point clouds from extended curved segments remains a significant challenge. This study proposes a refined extraction method for railway track point clouds based on equidistant projection transformation. The proposed methodology involves three key phases: First, an equidistant projection transformation converts the planar alignment of curved track point clouds into linear configurations. Subsequently, track point clouds are extracted through an integrated approach combining data rasterization with geometric feature analysis. Finally, Laplacian filtering is applied to optimize the extracted results, achieving refined point cloud extraction. Experimental validation was conducted using two distinct datasets containing linear and curved track alignments. Results demonstrate that the method achieves segmentation detection rates exceeding 94% and accuracy rates above 93% for both alignment types, while maintaining high computational efficiency. The proposed approach exhibits robust performance and effectiveness for track extraction across varying alignment geometries, providing a novel solution for high-precision railway point cloud processing.