Object segmentation based on improved PointNet

Zongyi Wang; Qikun Wang

doi:10.1117/12.2631813

14 April 2022 Object segmentation based on improved PointNet

Zongyi Wang, Qikun Wang

Proceedings Volume 12178, International Conference on Signal Processing and Communication Technology (SPCT 2021); 121781B (2022) https://doi.org/10.1117/12.2631813
Event: International Conference on Signal Processing and Communication Technology (SPCT 2021), 2021, Tianjin, China

Abstract

As a classic point cloud segmentation network, PointNet is widely used in point cloud segmentation and classification due to its quick speed and stability. But there are still many shortcomings that can be improved. Firstly, the difficulty and numbers of different types in point cloud varies, which lead to poor results on difficult samples. In addition, the PointNet network using the maximum value operation to compress the features of the global point cloud, a lot of detailed information will be lost. Aiming at the problem of segmentation and classification of objects, this paper improves on the original PointNet network and adds a self-encoding network to replace the maximum pooling function to encode global information. Then design a loss function that combines focal loss, self-encoding loss and regullizer loss, which effectively solves the problem of uneven training difficulty. Comparative experiments prove that the method proposed in this paper improves the accuracy of prediction while retaining the speed performance of PointNet.

Citation Download Citation

Zongyi Wang and Qikun Wang "Object segmentation based on improved PointNet", Proc. SPIE 12178, International Conference on Signal Processing and Communication Technology (SPCT 2021), 121781B (14 April 2022); https://doi.org/10.1117/12.2631813

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Neural networks

Convolution

Feature extraction

3D modeling

3D vision

Image segmentation

Network architectures

Show All Keywords

Keywords/Phrases

Search In:

Publication Years