基于空间几何特征融合增强的地铁隧道点云语义分割神经网络模型

Subway Tunnel Point Cloud Semantic Segmentation Network Model Based on Spatial Geometric Feature Fusion

基于深度学习的隧道点云语义分割技术能够对大规模点云数据中的物体对象进行识别与分类,可以实现隧道场景内物体信息的提取与管理。顾及地铁隧道内设施的空间分布特征与几何特点,提出了一种基于空间几何特征融合增强的地铁隧道点云语义分割神经网络模型。设计了隧道点云空间几何特征提取模块,提取了隧道设施点云的相对空间位置与几何分布特征,将其与相应的网络编码层的点云信息进行通道拼接,以增强网络模型对多尺度点云特征信息的感知能力。构建了基于通道注意力机制的特征融合编码层,提取不同通道间特征信息的权重,对不同空间尺度的点云信息进行加权融合,以充分利用不同尺度的信息来提高模型的表示和泛化能力。利用南京某地铁隧道实测点云数据制作语义分割数据集,对所提模型进行了验证。结果表明:模型的训练mIoU值达到0.9556;在测试数据集上的预测结果中,加权平均F1分数为0.9959,加权平均IoU值为0.9631;对于PointNet++模型分割精度较差的通信光缆、管线架、接触吊梁类别,本研究模型的IoU值分别达到0.845、0.825和0.999,有效提高了地铁隧道点云分割的整体准确性,可为实现地铁隧道自动化病害巡检与设施管理提供技术参考。

The tunnel point cloud semantic segmentation technology based on deep learning can recognize and classify objects in large-scale point cloud data,and realize the extraction and management of object information in tunnel scenes.A subway tunnel point cloud semantic segmentation network model based on enhanced spatial geometric feature fusion is proposed in this paper,which takes into account the spatial distribution characteristics and geometric features of subway tunnel facilities.A spatial geometric feature extraction module is designed to extract the relative spatial positions and geometric distribution features of tunnel facilities and combines them with the features of the network encoding layer through channel concatenation,enhancing the network′s perception of multi-scale feature information.A feature fusion encoding layer based on channel attention mechanism is designed to extract the weight of features between different channels,and to perform weighted fusion of the information from different spatial scales,so as to fully utilize information of different scales to improve model representation and generalization abilities.The network model proposed in this paper is verified using a dataset made by the point cloud data of a tunnel in Nanjing Subway.The experimental results demonstrate that the proposed network achieved a training mIoU value of 0.9556.The predicted results on the test dataset showed a weighted average F1 score of 0.9959 and a weighted average IoU value of 0.9631.For the communication cables,pipeline racks,and contact hangers classes with poor IoU values in the Point-Net++network,the IoU values of the proposed network reached 0.845,0.825,and 0.999,respectively,which effectively improved the overall accuracy of point cloud segmentation of subway tunnel and provided a technical reference for realizing automatic disease inspection and facility management of subway tunnel.