A 3D Steel Coils’ Recognition Method Based on Multi-Scale Features and Pointnet

The recognition of steel coils and saddles from the point cloud data is quite important in complex steel industrial environment. In this work, a deep learning algorithm based on multi-scale features and Pointnet network is proposed. For the preprocessing of raw data, the pass-through filtering method and Ransac algorithm are designed to filter and remove uninteresting point clouds, and then the edge-detection algorithm and MeanShift clustering method are added to segment and obtain valid single point cloud data. Next, an improved Pointnet network is proposed to classify the type of coils and saddles, where a multiscale features’ extraction is introduced to enhance the capability of extracting partial point information. At last, 4500 datasets of steel coils\saddles collected from on-site industry are used for network training. The experimental results demonstrated that the method can achieve high recognition accuracy and meet the real-time requirements.

[1]  Qiang Lu,et al.  PointNGCNN: Deep convolutional networks on 3D point clouds with neighborhood graph filters , 2020, Comput. Graph..

[2]  Dan Niu,et al.  A 3D Vehicle Recognition System Based on Point Cloud Library , 2019, 2019 Chinese Control Conference (CCC).

[3]  Yanmin Wang,et al.  A Segmentation Method for Point Cloud Based on Local Sample and Statistic Inference , 2014, GRMSE.

[4]  Jiaxin Li,et al.  SO-Net: Self-Organizing Network for Point Cloud Analysis , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[5]  Jitendra Malik,et al.  Learning Rich Features from RGB-D Images for Object Detection and Segmentation , 2014, ECCV.

[6]  Michael S. Lew,et al.  Deep learning for visual understanding: A review , 2016, Neurocomputing.

[7]  Victor S. Lempitsky,et al.  Escape from Cells: Deep Kd-Networks for the Recognition of 3D Point Cloud Models , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[8]  Wei An,et al.  Binary Volumetric Convolutional Neural Networks for 3-D Object Recognition , 2019, IEEE Transactions on Instrumentation and Measurement.

[9]  Yang Liu,et al.  O-CNN , 2017, ACM Trans. Graph..

[10]  Jianxiong Xiao,et al.  Deep Sliding Shapes for Amodal 3D Object Detection in RGB-D Images , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[11]  Dong Tian,et al.  Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[12]  Subhransu Maji,et al.  Multi-view Convolutional Neural Networks for 3D Shape Recognition , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[13]  李鑫 Li Xin,et al.  Terrain Classification of LiDAR Point Cloud Based on Multi-Scale Features and PointNet , 2019 .