3D Model Classification and Retrieval Method Based on Panorama and Multi-Channel CNN

A technology of three-dimensional models and panoramas, applied in biological neural network models, digital data information retrieval, instruments, etc., can solve problems affecting classification and search accuracy, and achieve the effect of improving robustness

Active Publication Date: 2021-08-03
TIANJIN UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In general, all these methods usually focus on structural information or visual information while ignoring the other, affecting classification and search accuracy

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
  • 3D Model Classification and Retrieval Method Based on Panorama and Multi-Channel CNN
  • 3D Model Classification and Retrieval Method Based on Panorama and Multi-Channel CNN
  • 3D Model Classification and Retrieval Method Based on Panorama and Multi-Channel CNN

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] To solve the above problems, you need to extract multi-resolution panoramic views from each 3D model. Studies have shown that the panoramic view of the 3D model can convert the structure information of the 3D model to 2D image information. [13] . Embodiments of the present invention proposes a three-dimensional model view extraction method based on panoramic map and multi-channel CNN, see figure 1 and figure 2 See the description below for details:

[0033] 101: By passing figure 2 The 3D model surface projected onto the side of the cylinder of R, height H = 2R, figure 2 The 3D model origin is center, the axis of the 3D model is parallel to one of the X, Y, and Z spindle of the space to obtain the initial panorama;

[0034] Where the R value is set to 3 * D max , D max It is the maximum distance of the 3D model surface distance.

[0035] 102: Suppose the Z-axis panorama, the embodiment of the present invention uses a set of points To parameterize the initial panorama;

[0...

Embodiment 2

[0042] The following combines the specific network structure, calculating the formula, image 3 , Figure 4 The training of multi-channel and multi-scale CNN networks and the similarity metrics between two different 3D models are introduced, see the description below:

[0043] image 3 The multi-scale network is displayed. This network includes three scales that extract different resolutions of the same input image, assume the input image size of 256 * 256, for the first scale, the image size is the same as the original picture The size is 256 * 256, and the convolutional neural network directly passes through the VGG16 is characterized by normalization, and the 4096-dimensional mapping is obtained.

[0044] For the second scale, the input image conversion scale becomes 128 * 128, so that the generated picture resolution is 1 / 2 of the original figure resolution, then the sample is sampled, and the convolutional neural network obtains a characteristic map of low resolution pictures. ...

Embodiment 3

[0054] Next, the specific data set, Table 1, Table 2 results in the results of the schemes in Examples 1 and 2, as described below with reference to the following description:

[0055] The data set for evaluating the proposed classification method is the Plinston ModelNet large 3DCAD model data set. ModelNet consists of 127,915 CAD models, divided into 662 object categories, divided into two subsets, ModelNet-10 and ModelNet-40, which contains training and test partitions.

[0056] 1) The ModelNet10 consists of 4,899 CAD models, divided into 10 categories. For convenience of processing, these models are adjusted, and the centroid of the model is placed at the origin of the coordinate, and normalized treatment is performed in translation and rotation.

[0057] Among them, the training and test subsets of ModelNet10 consist of 3991 and 908 models, respectively.

[0058] 2) ModelNet-40 includes 12,311 CAD models, divided into 40 categories. For convenience, these models are adjusted,...

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 discloses a method for extracting a 3D model view based on a panorama and a multi-channel CNN, comprising: projecting the 3D model onto the side of a cylinder satisfying preset conditions, centering on the origin of the 3D model, and The axis is parallel to one of the main axes of X, Y, and Z to obtain the initial panorama; respectively sample the angle and y coordinate of the 3D model surface in the three-dimensional space at a certain preset rate, and obtain the position of each point in the initial panorama Two sets of values ​​are used to represent the position characteristics of the 3D model surface in three-dimensional space and the direction characteristics of the 3D model surface; construct a multi-scale network and a multi-channel convolutional neural network to combine the position characteristics of the 3D model surface and the 3D model surface The orientation features of the surface are used as input for network training and similarity measurement between two different 3D models. The invention retains the structure of the three-dimensional model and the local and global information of the vision, automatically calculates the features of the 2D panoramic view, and is used to deal with classification and retrieval problems.

Description

Technical field [0001] The present invention relates to the field of three-dimensional model classification and retrieval, and more particularly to a three-dimensional model classification and retrieval method based on panoramic map and CNN multi-channel CNN. Background technique [0002] With the development of computer vision technology, 3D technology is widely used in the film and television industry, mechanical design, construction, infrastructure, entertainment industry, medical and other fields. More and more people began to upload their own 3D models on some websites, and the number of 3D models showed a growth trend. This causes a 3D model to retrieve hot topics in computer visual fields. Different from the traditional visual manifestations of the two-dimensional image information, the three-dimensional model not only has visual information, but also has structural information. Therefore, the traditional computer vision technology is difficult to represent the 3D model. I...

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 Patents(China)
IPC IPC(8): G06F16/583G06N3/04
CPCG06N3/045
Inventor 梁祺聂为之
Owner TIANJIN 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