Terrestrial laser scanning (TLS) point clouds are increasingly used for monitoring steel grid structures, and accurate instance segmentation is central to their processing. In complex environments, segmenting junction regions is challenging owing to multi-member geometry and incomplete sampling. Existing approaches frequently depend on prior information such as design drawings or Building Information Modeling (BIM), which limits generality and offers few solutions when model priors are unavailable. A hierarchical spherical coordinate segmentation with dual-sphere center refinement method (HSC-DCR) is proposed for geometry-driven junction region instance segmentation. The method uses radial connectivity. A segmentation origin is first located via a grid search driven by directional convergence evaluation, and a spherical coordinate system is then established for initial angular domain clustering. Subsequently, topological correction is guided by multi-dimensional indicators, and instance refinement is achieved through dual-sphere center optimization. The process is geometry-driven and independent of design models. Experiments on a laser-scanned stadium point cloud, covering 22 junction-region types and 521 instances, achieve an F1-score of 0.948 and mIoU of 85.5% under an instance-matched evaluation protocol. The results show that HSC-DCR can reliably obtain node and member instances from TLS point clouds without relying on drawings or BIM, supporting TLS-based monitoring of steel grid structures.
Hairun Chen, Alex Hay-Man Ng, Bo Hu et al.· Remote Sensing· 0 citations
Graph convolutional networks (GCNs) have shown promising performance in real-world applications involving graph-structured data. To enhance the model’s ability to integrate information from neighbor nodes, GCNs typically expand the receptive field by stacking more layers, which can lead to the problem of over-smoothing. To address this limitation, high-order graph convolutional networks (HGCNs) such as Mixhop have been proposed. These high-order models, however, introduce the risk of overfitting due to the increased number of parameters required for designing different order convolutions, which in turn can degrade classification performance. To overcome these issues, we first provide a theoretical analysis of the bottlenecks in conventional GCNs and establish that an effective graph convolutional model should integrate multi-hop neighborhood information in a unified manner. Inspired by this, we propose a novel broad graph convolutional network (BGCN) paradigm, which completely eliminates hidden layers and instead expands the receptive field through network width. We introduce four specific BGCN architectures and provide a detailed analysis of their computational complexity and parameter efficiency. A number of experiments on citation networks demonstrate that the proposed methods achieve state-of-the-art performance in terms of both classification accuracy and stability. In particular, our BGCN with simple attention mechanism (
$$\text {BGCN}_{\text {SA}}$$
BGCN
SA
) achieves top performance on the Citeseer, Cora, and Pubmed datasets, with accuracies of 73.3%, 83.5%, and 81.1%, respectively. The main advantage of this work lies in offering a robust, efficient, and scalable alternative for graph representation learning, effectively circumventing the inherent limitations of over-smoothing and overfitting in existing deep and high-order GCN models.
Alex Hay-Man Ng, Xun Liu, Fangyuan Lei et al.· Complex & Intelligent Sy...· 0 citations