传统的电力走廊点云数据的分割会出现精度低、数据局部特征捕获存在局限性等问题,为此提出了一种基于注意力权重的PointNet++网络场景分割模型.将深度学习中的PointNet++算法用于电力走廊场景分割中,再引入了空间注意力机制,帮助模型更有效地关注重要的空间区域.为此采用自制的数据集,并基于PointNet++网络模型的经典结构,在每个点集抽取(set abstraction,SA)模块中的多层感知机(multi layer perceptron,MLP)加入倒置瓶颈设计,提高对点云数据的处理效率和准确性.研究结果表明,与传统的PointNet++网络相比,改进的PointNet++网络平均交并比(mean intersection over union,mIoU)高出6.3%,加入空间注意力机制的改进模型在自制数据集上表现出更好的分割效果,尤其是在边界划分方面提升明显,验证了该方法在点云语义分割上的有效性.
Research on Power Corridor Point Cloud Semantic Segmentation Based on Attention Weighted PointNet++
The traditional segmentation of point cloud data in power corridors can suffer from problems such as low accuracy and limita-tions in capturing local features of the data.For this reason,a PointNet++network scene segmentation model based on attention weights is proposed.It uses the PointNet++algorithm from deep learning for scene segmentation of power corridors,and then intro-duces a spatial attention mechanism to help the model focus on important spatial regions more effectively.For this purpose,a home-made dataset is used and an inverted bottleneck design is added to the MLP in each SA module based on the classical structure of the PointNet++network model to improve the processing efficiency and accuracy of the point cloud data.The results show that the im-proved PointNet++network has a 6.3%higher average convergence ratio(mIoU).The improved model with spatial attention mecha-nism shows better segmentation effect on self-made data set,especially in boundary division,which verifies the effectiveness of this method in point cloud semantic segmentation.
point cloud semantic segmentationtransmission channelsPointNet++attention mechanism
万方数据
维普
