Three-dimensional object fusion feature representation method based on multi-modal feature fusion

A technology of feature fusion and feature fusion, which is applied in character and pattern recognition, instruments, computer components, etc.

Active Publication Date: 2020-05-22
HANGZHOU DIANZI UNIV
View PDF3 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Despite these issues, grids have stronger 3D shape description capabilities than other types of data

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
  • Three-dimensional object fusion feature representation method based on multi-modal feature fusion
  • Three-dimensional object fusion feature representation method based on multi-modal feature fusion
  • Three-dimensional object fusion feature representation method based on multi-modal feature fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0022] Below in conjunction with accompanying drawing, the present invention is further described;

[0023] A 3D object fusion feature representation method based on multimodal feature fusion, the steps are as follows:

[0024] Step (1), process multi-view through multi-view neural network Figure three dimension information;

[0025] Such as figure 1 As shown, multiple independent CNNs that do not share weights are used to input multi-view information, and then through Max-pooling, the outputs of multiple CNNs are unified into one output, and a discriminator (that is, a non-linear network based on a fully connected layer) is added. classifier) ​​to classify the model.

[0026] First, convert the 3D model data into multi-view data. The specific method is to place 12 cameras evenly around the 3D model (that is, at intervals of 30 degrees) on the middle horizontal plane of the 3D model, and take a group of 12 pictures as the 3D model. multi-view representation of . Then use...

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

No PUM Login to view more

Abstract

The invention provides a three-dimensional object fusion feature representation method based on multi-modal feature fusion. According to the method, the feature representation of a three-dimensional model in three modals is meticulously realized and reproduced, namely, multi-view feature representation, point cloud feature representation and grid feature representation of the three-dimensional model. Multi-modal three-dimensional data is processed, enhanced and fused, and fusion feature representation of a three-dimensional object is extracted on the basis. According to the invention, excellent multi-modal information fusion can be realized, and more robust three-dimensional model feature representation is realized to be used for other three-dimensional object tasks.

Description

technical field [0001] The invention belongs to the technical field of computer images and artificial intelligence, and provides a more efficient three-dimensional object fusion feature representation form that integrates three-dimensional multi-modal information. Background technique [0002] 3D data recognition and analysis is a fundamental and interesting field in multimedia and computer vision, covering a wide range of applications from environment understanding to autonomous driving. How to understand 3D data, such as recognizing 3D shapes, has attracted a lot of attention in recent years. With the development of deep learning, various deep networks are used to process different types of 3D data: point cloud (PointCloud), multi-view (Multi-view) and volume (Volumetric) data. While it is natural and reasonable to extend 2D convolutional neural networks to volumetric data, these methods suffer from large computational complexity and data sparsity, making it difficult to ...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62
CPCG06F18/2453G06F18/253
Inventor 颜成钢龚镖白俊杰孙垚棋张继勇张勇东沈韬
Owner HANGZHOU DIANZI UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap