<p>Establishing accurate 3D models of ancient buildings can provide important data support for their protection, restoration, and reconstruction after damage. The modeling method based on 3D laser scanning technology is currently the primary method for reconstructing 3D models of ancient buildings. However, this method primarily involves extracting contour feature lines of ancient architectural components and manually modeling them using third-party software. This approach is inefficient, lacks precision, and demands high point cloud integrity. This paper introduces a method for automatically reconstructing detailed models of building components by utilizing the least squares algorithm to extract their geometric parameters. Initially, a ground-based 3D laser scanner is employed to scan ancient buildings. An enhanced region growing method, combined with statistical algorithms, is then utilized to accurately extract point clouds of building components, thereby refining traditional point cloud classification techniques. Subsequently, the point cloud of the building components is projected onto a horizontal plane, enhancing the completeness of the cross-sectional point cloud and significantly improving the accuracy of fitting the cross-sectional shapes of the building components. The least squares principle is applied to fit the cross-sectional shapes of the building components, extract geometric parameters, and enhance the traditional method of extracting contour feature lines of ancient building components. Ultimately, the 3D model is reconstructed using the height and cross-sectional parameters of the building components. The findings indicate that this method enables the high-precision reconstruction of 3D models even with incomplete point cloud data of ancient buildings, thereby enhancing the accuracy and efficiency of traditional modeling techniques. This offers valuable data support for the preservation, restoration, and reconstruction of ancient buildings.</p>