Classification Model of Point Cloud Along Transmission Line Based on Group Normalization

This article proposes a point cloud classification model based on group normalization to increase the classification accuracy when the computing power of the terminal device is limited. This model groups and normalizes the features of point cloud during inference and increases the classification acc...

Full description

Bibliographic Details
Main Authors: Zhimin Yin, Shichao Ji, Xuyong Zhang, Jianhua Dai, Weiyong Yu, Song Wu
Format: Article
Language:English
Published: Frontiers Media S.A. 2022-05-01
Series:Frontiers in Energy Research
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fenrg.2022.839273/full
_version_ 1818268809452060672
author Zhimin Yin
Shichao Ji
Xuyong Zhang
Jianhua Dai
Weiyong Yu
Song Wu
author_facet Zhimin Yin
Shichao Ji
Xuyong Zhang
Jianhua Dai
Weiyong Yu
Song Wu
author_sort Zhimin Yin
collection DOAJ
description This article proposes a point cloud classification model based on group normalization to increase the classification accuracy when the computing power of the terminal device is limited. This model groups and normalizes the features of point cloud during inference and increases the classification accuracy when the computing power is limited. The group normalization first groups the features of point cloud by their channel, then computes their statistic metrics and normalizes them. Also, one-dimensional convolution layers are used to replace the fully connected layers to decrease the model parameters and keep the model's performance when the computing power is limited. In the experiment, PointNet++ is used to pretrain on ModelNet40 and then fine-tune on the point cloud data of transmission lines. The result shows that the proposed method can effectively increase the classification accuracy and help the 3D modeling process of the transmission line.
first_indexed 2024-12-12T20:44:23Z
format Article
id doaj.art-fed717511f1241578677fb47f95b2e8b
institution Directory Open Access Journal
issn 2296-598X
language English
last_indexed 2024-12-12T20:44:23Z
publishDate 2022-05-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Energy Research
spelling doaj.art-fed717511f1241578677fb47f95b2e8b2022-12-22T00:12:37ZengFrontiers Media S.A.Frontiers in Energy Research2296-598X2022-05-011010.3389/fenrg.2022.839273839273Classification Model of Point Cloud Along Transmission Line Based on Group NormalizationZhimin YinShichao JiXuyong ZhangJianhua DaiWeiyong YuSong WuThis article proposes a point cloud classification model based on group normalization to increase the classification accuracy when the computing power of the terminal device is limited. This model groups and normalizes the features of point cloud during inference and increases the classification accuracy when the computing power is limited. The group normalization first groups the features of point cloud by their channel, then computes their statistic metrics and normalizes them. Also, one-dimensional convolution layers are used to replace the fully connected layers to decrease the model parameters and keep the model's performance when the computing power is limited. In the experiment, PointNet++ is used to pretrain on ModelNet40 and then fine-tune on the point cloud data of transmission lines. The result shows that the proposed method can effectively increase the classification accuracy and help the 3D modeling process of the transmission line.https://www.frontiersin.org/articles/10.3389/fenrg.2022.839273/fullgroup normalizationPointNet++ModelNet40batch normalizationpoint cloud of transmission line
spellingShingle Zhimin Yin
Shichao Ji
Xuyong Zhang
Jianhua Dai
Weiyong Yu
Song Wu
Classification Model of Point Cloud Along Transmission Line Based on Group Normalization
Frontiers in Energy Research
group normalization
PointNet++
ModelNet40
batch normalization
point cloud of transmission line
title Classification Model of Point Cloud Along Transmission Line Based on Group Normalization
title_full Classification Model of Point Cloud Along Transmission Line Based on Group Normalization
title_fullStr Classification Model of Point Cloud Along Transmission Line Based on Group Normalization
title_full_unstemmed Classification Model of Point Cloud Along Transmission Line Based on Group Normalization
title_short Classification Model of Point Cloud Along Transmission Line Based on Group Normalization
title_sort classification model of point cloud along transmission line based on group normalization
topic group normalization
PointNet++
ModelNet40
batch normalization
point cloud of transmission line
url https://www.frontiersin.org/articles/10.3389/fenrg.2022.839273/full
work_keys_str_mv AT zhiminyin classificationmodelofpointcloudalongtransmissionlinebasedongroupnormalization
AT shichaoji classificationmodelofpointcloudalongtransmissionlinebasedongroupnormalization
AT xuyongzhang classificationmodelofpointcloudalongtransmissionlinebasedongroupnormalization
AT jianhuadai classificationmodelofpointcloudalongtransmissionlinebasedongroupnormalization
AT weiyongyu classificationmodelofpointcloudalongtransmissionlinebasedongroupnormalization
AT songwu classificationmodelofpointcloudalongtransmissionlinebasedongroupnormalization