结合通道剪枝和通道注意力的轻量型车辆点云补全网络

doi:10.3778/j.issn.1002-8331.2308-0275

摘要/Abstract

摘要： 针对现有的点云补全网络多关注于补全的精度而忽视补全效率问题，提出了一种轻量型点云补全网络来准确、高效地修复自动驾驶中的不完整车辆点云。为了提高网络推理效率，采用一种高效的一次性通道剪枝技术提高网络的补全效率；在特征提取阶段，网络加入通道注意力模块，将加权特征与全局特征拼接，通过两层多维特征信息提取，得到最终的特征向量；将特征向量传入双解码器结构中，分别通过全连接层和多层感知机生成稠密的粗糙点云和输入点云偏差值；将粗糙点云与输入点云偏差值相加得到最终的精细化完整点云。在PCN数据集和KITTI数据集上进行实验，实验结果表明在补全缺失车辆信息的实时性上有着显著的提升，并且在补全精度上也有不错的表现。

关键词: 点云补全, 通道剪枝, 通道注意力, 轻量型, 深度学习

Abstract: A lightweight point cloud completion network is proposed to accurately and efficiently repair incomplete vehicle point clouds in autonomous driving scenarios, addressing the efficiency issue often overlooked by existing point cloud completion networks that primarily focus on accuracy. Firstly, to enhance inference efficiency, an efficient one-shot channel pruning technique is employed to improve the completion speed of the network. Secondly, a channel attention module is integrated into the network during the feature extraction phase, which combines weighted features with global features. This dual-layered multidimensional feature extraction yields the final feature vector. Subsequently, the feature vector is fed into a dual decoder structure, generating dense coarse point clouds and input point cloud deviation values through fully connected layers and multi-layer perceptron, respectively. Finally, by adding the coarse point cloud and input point cloud deviation values, a refined and complete point cloud is obtained. Experimental evaluations conducted on the PCN dataset and KITTI dataset demonstrate significant improvements in real-time completion of missing vehicle information. Moreover, favorable results are achieved in terms of completion accuracy as well.

Key words: point cloud completion, channel pruning, channel attention, lightweight, deep learning

杨晓文, 冯泊栋, 韩慧妍, 况立群, 韩燮, 何黎刚. 结合通道剪枝和通道注意力的轻量型车辆点云补全网络[J]. 计算机工程与应用, 2025, 61(1): 232-242.

YANG Xiaowen, FENG Bodong, HAN Huiyan, KUANG Liqun, HAN Xie, HE Ligang. Lightweight Vehicle Point Cloud Completion Network Combined with Channel Pruning and Channel Attention[J]. Computer Engineering and Applications, 2025, 61(1): 232-242.

参考文献

[1] FANG J, ZHOU D F, YAN F L, et al. Augmented LiDAR simulator for autonomous driving[J]. IEEE Robotics and Automation Letters, 2020, 5(2): 1931-1938.
[2] CHEN C, XIANG H, QIU T, et al. A rearend collision prediction scheme based on deep learning in the Internet of vehicles[J]. Journal of Parallel and Distributed Computing, 2018, 117: 192-204.
[3] QI C R, SU H, MO K. PointNet: deep learning on point sets for 3D classification and segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Jul 21-26, 2017, Honolulu, HI, United States. New York: IEEE, 2017: 652-660.
[4] QI C R, SU H, MO K. PointNet++: deep hierarchical feature learning on point sets in a metric space[C]//Proceedings of the 2008 IEEE International Conference on Neural Information Processing Systems. Long Beach, CA, USA, 2017: 5105-5114.
[5] YUAN W T, KHOT T, HELD D, et al. PCN: point completion network[C]//Proceedings of the 2018 International Conference on 3D Vision, Sep 5-8, 2018, Verona, Italy. New York: IEEE, 2018: 728-737.
[6] YANG Y Q, FENG C, SHEN Y, et al. FoldingNet: point cloud auto-encoder via deep grid deformation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2021: 206-215.
[7] HUANG Z T, YU Y K, XU J W, et al. PF Net: point fractal network for 3D point cloud completion[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, Jun 13-19, 2020. Piscataway: IEEE, 2020: 7659-7667.
[8] WEN X, XIANG P, HAN Z Z, et al. PMP-Net: point cloud completion by learning multi-step point moving paths[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Nashville, TN, USA, Jun 20-25, 2021. New York: IEEE, 2021: 7439-7448.
[9] WEN X, XIANG P, HAN Z Z, et al. PMP-Net++: point cloud completion by transformer enhanced multistep point moving paths[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(1): 852-867.
[10] ZHAO H S, JIANG L, JIA J Y, et al. Point transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, Canada, 2021: 16239-16248.
[11] TANG J S, GONG Z J, YI R, et al. Lake-Net: topology-aware point cloud completion by localizing aligned keypoints[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, LA, USA, 2022: 1716-1725.
[12] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the 15th European Conference on Computer Vision, Munich, Germany, 2018: 3-19.
[13] TCHAPMI L P, KOSARAJU V, REZATOFIGHI H, et al. TopNet: structural point cloud decoder[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, CA, United States, Jun 15-21, 2019. New York: IEEE, 2019: 383-392.
[14] DAI A, QI C R, NIE?NER M. Shape completion using 3D-encoder-predictor CNNs and shape synthesis[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, ?Honolulu, HI, USA, 2017: 6545-6554.
[15] STUTZ D, GEIGER A. Learning 3D shape completion from laser scan data with weak supervision[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 1955-1964.
[16] XIE H Z, YAO H X, ZHOU S C, et al. GRNet: gridding residual network for dense point cloud completion[C]//Proceedings of the European Conference on Computer Vision, Glasgow, UK, Aug 23-28, 2020. Cham: Springer, 2020: 365-381.
[17] GROUEIX T, FISHER M, KIM V G, et al. A papier-maché approach to learning 3D surface generation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 216-224.
[18] LITANY O, BRONSTEIN A, BRONSTEIN M, et al. Deformable shape completion with graph convolutional autoencoders[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 1886-1895.
[19] CHEN X, CHEN B, MITRA N J. Unpaired point cloud completion on real scans using adversarial training[C]//Proceedings of the International Conference on Learning Representations, Addis Ababa, Ethiopia, 2020.
[20] GURUMURTHY S, AGRAWAL S. High fidelity semantic shape completion for point clouds using latent optimization[C]//Proceedings of the 2019 IEEE Winter Conference on Applications of Computer Vision, Waikoloa, HI, USA, 2019: 1099-1108.
[21] 李昂, 孙士杰, 张朝阳, 等. 改进YOLOv5s的轨道障碍物检测模型轻量化研究[J]. 计算机工程与应用, 2023, 59(4): 197-207.
LI A, SUN S J, ZHANG C Y, et al. Research on lightweight of improved YOLOv5s track obstacle detection model[J]. Computer Engineering and Applications, 2023, 59(4): 197-207.
[22] JIANG Y, WANG S, VALLS V, et al. Model pruning enables efficient federated learning on edge devices[J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(12): 10374-10386.
[23] SINGH P, VERMA V K, RAI P, et al. Play and prune: adaptive filter pruning for deep model compression[C]//Proceedings of the 28th International Joint Conference on Artificial Intelligence, Macao, China, 2019: 3460-3466.
[24] XIA Y, WANH C, XU Y, et al. RealPoint3D: generating 3D point clouds from a single image of complex scenarios[J]. Remote Sensing, 2019, 11(22): 2644.
[25] FAN H Q, SU H, GUIBAS L. A point set generation network for 3D object reconstruction from a single image[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, ?Honolulu, HI, USA, 2017: 2463-2471.
[26] TATARCHENKO M, RICHTER S R, RANFTL R, et al. What do single-view 3D reconstruction networks learn?[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, ?Long Beach, CA, USA, 2019: 3400-3409.
[27] XIANG P, WEN X, LIU Y S, et al. SnowflakeNet: point cloud completion by snowflake point deconvolution with skip-transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, ?Montreal, QC, Canada, 2021: 5479-5489.
[28] 邱云飞, 赵静, 方立. MF-Net: 结合残差网络的多尺度特征点云补全网络[J]. 计算机工程与应用, 2023, 59(22): 202-212.
QIU Y F, ZHAO J, FANG L. MF-Net: multi-scale feature point cloud completion network combined with residual network[J]. Computer Engineering and Applications, 2023, 59(22): 202-212.
[29] KLOKOV R, BOYER E, VERBEEK J. Discrete point flow networks for efficient point cloud generation[C]//Proceedings of the 16th European Conference on Computer Vision, Glasgow, UK, Aug 23-28, 2020. Cham: Springer, 2020: 694-710.
[30] OUYANG Z C, CUI J H, DONG X Y, et al. SaccadeFork: a lightweight multi-sensor fusion-based target detector[J]. Information Fusion, 2022, 77: 172-183.