More than sixty thousand bone sticks excavated from Building No. 3 of the Weiyang Palace site in the Han Chang’an City are predominantly elongated scapular bone fragments, whose large quantity, similar morphology, and complex information content make systematic classification essential for archaeological data organization. However, long-term burial has caused extensive fragmentation and dispersion, rendering manual classification inefficient and increasing the risk of artifact damage, thereby necessitating automated solutions. To address this challenge, this study proposes an intelligent bone stick classification method based on dual-view parallel Vision-RWKV feature modeling. High-precision segmentation of bone stick regions is first achieved by incorporating channel attention mechanisms, lightweight feature generation strategies, and edge-guided constraints into the UNet++ framework, and the segmented regions are used as inputs to emphasize key structural features. Subsequently, two parallel Vision-RWKV models are employed to extract features from the front and back views of bone sticks, and a cooperative fusion strategy is adopted to unify dual-view information and enhance discriminative capability. Experimental results demonstrate that the proposed method achieves an accuracy of 94.76%, outperforming existing image classification models and providing effective technical support for archaeological data organization.