Object recognition method based on convolutional restricted Boltzmann machine combining Centering Trick

A Boltzmann machine and object recognition technology, applied in the field of image recognition, can solve the problems of high noise in the input image and ignoring the two-dimensionality of the image.

Active Publication Date: 2014-09-10
BEIJING UNIV OF TECH
View PDF4 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems of excessive input image noise and ignoring the two-dimensionality of the image itself and the relationship between them in the prior art, the...

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
  • Object recognition method based on convolutional restricted Boltzmann machine combining Centering Trick
  • Object recognition method based on convolutional restricted Boltzmann machine combining Centering Trick
  • Object recognition method based on convolutional restricted Boltzmann machine combining Centering Trick

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0067] The object recognition method based on the Centering Trick convolution restricted Boltzmann machine, including the training phase and the testing phase, specifically includes the following steps:

[0068] 1. In the robot training stage, the robot constructs a training database through its own visual system, and the computer collects and inputs N V ×N V The pixel-sized object images are divided into N categories according to the object categories, and the category numbers are 1 to N. Each type of object image contains T training images, and the training image set is constructed. Use P train Indicates that the total number is: N×T=Q images.

[0069] 2. To P train Each image in is reconstructed. Represent the original 2D image as a 4D image. The first and second dimensions represent the height and width respectively. Since the height and width of the image used are equal, they are both recorded as N V ; The third dimension represents the color of the image, which is r...

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 relates to an object recognition method based on a convolutional restricted Boltzmann machine combining Centering Trick. The method includes: structuring a training database, reconstructing an image, performing pre-whitening, using a CRBM model combining Centering Trick to perform feature extraction for a training set, transforming a 3-dimensional eigenmatrix into a one-dimensional eigenvector, using a Softmax classifier to classify the features, obtaining a test set, recombining and whitening the test set, computing the assumed value Si corresponding to output of a pool layer, and determining whether classification is correct or not according to a fact that whether Si is equal to the label of the test set or not. During a feature extraction process, a memory needed for computation is reduced and the computing speed is increased by making full use of two dimensionality of an image itself and the relations between pixels and also by using a CRBM model to achieve weight sharing. By using Centering Trick, the noise of inputting ofeach layer is reduced, and the accuracy of calculation and the stability of the model are improved.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and introduces an object recognition method of a convolutional restricted Boltzmann machine (Convolutional Restricted Boltzmann Machine, CRBM) model fused with a centering trick (a data preprocessing method using a parameter center factor). When using CRBM to extract the global features of the image, the centering trick is fused, and the original CRBM energy function is reconstructed to reduce the noise of each input, thereby reducing the noise in the entire generation model calculation process, and using the greedy algorithm as a learning mechanism , making the model more stable and having better generative properties. technical background [0002] As one of the greatest inventions of mankind in the 20th century, robotics has been developed for more than 50 years since it came out in the early 1960s, and has become one of the representative strategic technologies in the high-tech field...

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): G06K9/00G06K9/66
Inventor 杨金福高晶钰张珊珊李明爱张济昭
Owner BEIJING UNIV OF TECH
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