Log in
Back to #graph neural networks
#graph neural networks

Parallel Dual-Level Attention-Enhanced Semantic Segmentation of 3D Point Clouds in Tunnel Construction

Xin Peng Mingnian Wang Hao Zhong Hong-Lin Shen Bing-Xu Huang
Nov 2026 · Journal of computing in civil engineering · 0 citations · 46 references

TL;DR

Validated on real tunnel data, PDAPointNet++ outperforms models such as PointNet, PointNet++, PointMLP, and dynamic graph convolutional neural network (DGCNN) and the impact of different WDFL balancing coefficients on model performance is discussed.

Abstract

Drill-and-blast is the primary method used for constructing mountain tunnels in China. The tunnel face area contains rich geological and structural information, necessitating high-precision semantic segmentation to support automated measurement during the construction process. This paper proposes PDAPointNet++, a novel network for semantic segmentation of point clouds in tunnel construction scenarios. The method introduces three key innovations: (1) a parallel dual-level attention module (PDAM) to enhance spatial and channel feature focus; (2) integration of PDAM into the set abstraction and feature propagation layers; and (3) a weighted dice focal loss (WDFL) to address data imbalance. Validated on real tunnel data, PDAPointNet++ outperforms models such as PointNet, PointNet++, PointMLP, and dynamic graph convolutional neural network (DGCNN). From the perspective of practical engineering applications, the accuracy of the segmentation results is further verified through rock mass discontinuity measurement on the tunnel face, and the impact of different WDFL balancing coefficients on model performance is discussed.

DOI Semantic Scholar
View source

Similar papers

#graph neural networks Oct 2026

Locality-Preserving Graph Laplacian Manifold Learning Based Model Predictive Control for Three-Phase Inverters

This article presents a model predictive control (MPC) strategy for three-phase inverters based on locality preserving projections (LPPs). Unlike conventional machine learning–based MPC approaches that rely on predefined or high-dimensional input features, the proposed LPP-MPC automatically extracts compact, informative representations by preserving the data’s intrinsic geometric structure. This dimensionality reduction enables fast linear control-law evaluation with computational complexity O(1), making the controller well-suited for real-time implementation. Experimental results demonstrate that the LPP-MPC achieves lower total harmonic distortion (THD) and reduced tracking error compared to quadratic-programming MPC under both linear and nonlinear load conditions, and the proposed controller maintains consistently lower THD throughout load transients than other methods such as two-degree-of-freedom MPC. Compared to existing model-free MPC and deep learning neural network, the LPP-MPC has the lowest THD and root mean square error with the least computational time owing to its efficient linear structure and strong generalization capability.

Jianwu Zeng, Lizheng Cheng, V. Winstead et al. · 1 citation

Related blog posts