Shape Analysis and Recognition Method of Artworks Based on Computational Geometry

ABSTRACT

Traditional methods for shape analysis and recognition of artworks have many shortcomings in terms of accuracy, efficiency, and complex shape processing. Especially when faced with complex or abstract artworks, it is often difficult to provide accurate classification and recognition results. This paper discusses a method for shape analysis and recognition of artworks based on computational geometry. A unified multistage computational geometry pipeline is used for contour extraction, shape simplification, feature extraction, and classification. First, the main geometric contours of the works are extracted using Canny edge detection and contour tracking techniques, and the shapes are simplified by the convex hull algorithm to remove noise and irrelevant details in the image. Then, for complex shapes, the curvature analysis method is used to finely characterize the bending changes of the curves, and the local features are refined by combining Delaunay triangulation and Voronoi diagram. For three‐dimensional (3D) artworks, structured light scanning and stereo vision reconstruction technology are combined to construct a 3D model and perform shape matching analysis. In the field of shape similarity evaluation, methods such as Procrustes analysis and Hausdorff distance are widely used to achieve precise classification and recognition of artworks. Experimental results demonstrate that the proposed method achieves classification accuracy above 0.78 across different artwork styles. The method also attains precision above 0.75, ensuring reliable recognition performance. These results indicate the effectiveness of the approach in handling complex geometric structures. This provides a practical solution for the digital analysis and automatic identification of artworks.