<p>Similarity retrieval of 3D product models is crucial for promoting the reuse of product design information and enhancing the production efficiency of enterprises. The 3D wheel models, widely used in industries such as automotive, railway, and aviation manufacturing, pose significant challenges in retrieval due to the high similarity between their global and local geometric features. This is because existing methods: (1) fail to effectively represent the geometric features of 3D wheel models, especially the local features that users focus on; (2) lack the flexibility in feature descriptors used for similarity calculation, making it difficult to accurately distinguish and match different models when dealing with highly similar local features. To address these problems, this paper proposes a visualization retrieval framework for 3D wheel models based on customizable geometric features regions. Firstly, a geometric feature extraction method is proposed that visualizes the geometric features of the 3D wheel models as binary geometric feature images, addressing the problem of representing both global and local geometric features. Secondly, building upon these geometric feature images, a customizable geometric features regions selection method is designed, enabling users to select multiple local geometric features to generate combined feature descriptors for similarity calculation in 3D wheel models retrieval. Finally, the effectiveness of the proposed framework is validated through experiments on industrial product datasets and user evaluations by domain experts, demonstrating its ability to promote the effective reuse of 3D wheel models product information.</p> Graphical abstract <p></p>