In indoor wireless positioning systems, non-line-of-sight (NLOS) propagation caused by base station deployment constraints and environmental structures poses a major challenge to positioning accuracy. Mitigating the adverse effects of NLOS propagation without increasing deployment cost remains a fundamental challenge in indoor wireless positioning. To address this challenge, we propose a novel correction framework that utilizing ranging information and structural constraints to construct virtual line-of-sight (LOS) base stations as substitutes for NLOS measurements, thereby enabling effective utilization of NLOS signals. The proposed method models the structural information of the indoor environment in a planar form and, in combination with a location prediction algorithm, enables accurate identification of signal types in dynamic scenarios. Furthermore, an NLOS reconstruction algorithm is developed to infer feasible propagation paths of identified NLOS signals, allowing reliable virtual LOS base stations to be generated for subsequent localization. The entire framework operates rapidly without requiring any prior data collection, and improves positioning stability and system availability using only standard ranging measurements and structural constraints at a low computational cost. To validate the proposed approach, multiple low-cost ultra-wideband (UWB) base stations were deployed in a corridor environment, and pedestrian motion data were collected under dense NLOS conditions. Experimental results demonstrate a substantial improvement in positioning performance: the root mean square error (RMSE) of UWB positioning is reduced from over 4 meters to below 0.4 meters, confirming the effectiveness of the proposed solution.

Understanding objects at the level of their constituent parts is fundamental to advancing computer vision, graphics, and robotics. While datasets like PartNet have driven progress in 3D part understanding, their reliance on untextured geometries and expert-dependent annotation limits scalability and usability. We introduce PartNeXt, a next-generation dataset addressing these gaps with over 23000 high-quality, textured 3D models annotated with fine-grained, hierarchical part labels across 50 categories. We benchmark PartNeXt on two tasks: (1) class-agnostic part segmentation, where state-of-the-art methods (e.g., PartField, SAMPart3D) struggle with fine-grained and leaf-level parts, and (2) 3D part-centric question answering, a new benchmark for 3D-LLMs that reveals significant gaps in open-vocabulary part grounding. Additionally, training Point-SAM on PartNeXt yields substantial gains over PartNet, underscoring the dataset’s superior quality and diversity. By combining scalable annotation, texture-aware labels, and multi-task evaluation, PartNeXt opens new avenues for research in structured 3D understanding.