Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Point cloud completion method based on local covariance optimization

A local covariance and point cloud technology, which is applied in the field of point cloud completion, can solve problems such as ignoring local structure and shape features, failing to restore local structure, and failing to complete the second task of point cloud. The effect of structural optimization

Pending Publication Date: 2022-04-12
HANGZHOU DIANZI UNIV
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, since these methods only consider reconstructing the point cloud from the global feature vector, ignoring the local structure and shape features, they can only recover the basic shape of the point cloud in the end, but cannot complete the second task of point cloud completion, and cannot restore fine details. local structure

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Point cloud completion method based on local covariance optimization
  • Point cloud completion method based on local covariance optimization
  • Point cloud completion method based on local covariance optimization

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0043] The present invention provides a point cloud completion method based on local covariance optimization, such as figure 1 shown, including the following steps:

[0044] S1, data set acquisition, obtain the public data set ShapeNet, complete the construction of the point cloud data set described in the 3D model training, wherein the point cloud data set contains 55 different objects

[0045] S2, data preprocessing, perform surface sampling on each 3D model in the data set to obtain a complete point cloud, then simulate the process of shape loss to generate corresponding incomplete point clouds for each complete point cloud, and standardize all point cloud data, which is used for The preprocessed data set selects 16 common objects from 55 different objects in the data set, thus greatly reducing the amount of calculation;

[0046] Understandably, all point cloud data include complete point cloud and incomplete point cloud.

[0047] S3. The construction of neural network mo...

Embodiment 2

[0090] The difference between this embodiment and Embodiment 1 is that after the training of the neural network model CovNet is completed, the selected aircraft point cloud is selected to reduce the number of points of the incomplete point cloud from N=2048 to 35% when the missing ratio is 35%. 256. Such as Figure 5 As shown, the method proposed by the present invention still has excellent completion results.

Embodiment 3

[0092] The difference between this embodiment and Embodiment 1 is that after the training of the neural network model CovNet is completed, when the point quantity of the selected aircraft point cloud is fixed (N=2048) in the incomplete point cloud, the missing ratio is gradually increased from 0.25 to 0.75. Such as Figure 6 As shown, the method proposed by the present invention still has better completion results.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

PropertyMeasurementUnit
Learning rateaaaaaaaaaa
Login to View More

Abstract

The invention discloses a point cloud completion method based on local covariance optimization. The method comprises the following steps: S1, acquiring a data set; s2, data preprocessing; s3, constructing a neural network model; s4, constructing the loss of the neural network model; s5, training and optimizing a neural network model; and S6, saving the model and model parameters, and by adopting the technical scheme, taking the incomplete point cloud as input, and outputting the complete point cloud which has a complete shape and is finer. In the feature coding stage, a disordered and complex topological relation between local points is analyzed by using covariance, and local geometric structure information is extracted by using different convolution kernels; the integrity of the shape and the structural similarity are considered, features of a missing structure are inferred by fusing multi-scale hierarchical features, and a global feature vector of the point cloud under the complete shape is obtained; in the decoding stage, not only can the expansion of the number of the point clouds be realized, but also the local geometric structure of the point clouds can be optimized, so that the finer complete point clouds are generated.

Description

technical field [0001] The invention relates to the technical field of point cloud completion, in particular to a point cloud completion method based on local covariance optimization. Background technique [0002] With the increasing availability of scanning equipment, people can obtain a large amount of high-precision point cloud data from the surface of 3D objects in a short time. As a collection of points in 3D space, point cloud retains more details and rich semantics, has gradually become a popular representation of 3D models, and is widely used in autonomous driving, robot control, 3D design and AR / VR and other fields. However, due to the complexity of the shape and the influence of the shooting environment, such as light intensity, occlusion, etc., the point cloud acquired by scanning is usually incomplete. An incomplete point cloud model not only fails to represent the actual shape of the object and affects recognition, but also affects subsequent point cloud proc...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06T5/00G06T9/00G06T19/20G06N3/04G06N3/08
Inventor 吴向阳渠冲冲
Owner HANGZHOU DIANZI UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com