Unlock instant, AI-driven research and patent intelligence for your innovation.

Complex surface object classification method based on point cloud VFH (Vector Field Histogram) descriptor and neural network

A neural network and complex surface technology, which is applied in the field of classification of complex surface objects based on point cloud VFH descriptors and neural networks, can solve the problems of huge search data and weakened classification recognition effect, so as to reduce calculation time and improve classification accuracy The effect of improving the rate and recognition accuracy

Active Publication Date: 2017-06-13
HARBIN INST OF TECH
View PDF3 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to solve the problem that if a large amount of collected data is built to build a Kd tree, the search data will be too large, and if the collected data is too small, the classification and recognition effect will be weakened, and a point cloud based VFH descriptor is proposed. Classification method of complex surface object with neural network

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
  • Complex surface object classification method based on point cloud VFH (Vector Field Histogram) descriptor and neural network
  • Complex surface object classification method based on point cloud VFH (Vector Field Histogram) descriptor and neural network
  • Complex surface object classification method based on point cloud VFH (Vector Field Histogram) descriptor and neural network

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0022] Specific embodiment one: a kind of complex curved surface object classification method based on point cloud VFH descriptor and neural network of this embodiment, specifically prepare according to the following steps:

[0023] Introducing Neural Networks from Machine Learning into the Classification Process

[0024] Step 1. In the training phase, according to the actual situation of the object, divide each object into M angles evenly, collect the point cloud at the angle of view of each of the M angles and calculate the vfh (Viewpoint FeatureHistogram) feature corresponding to the point cloud Descriptor x=(x 1 ,x 2 ,...,x i ,...,x M ) T like figure 1 and figure 2 ;

[0025] Step 2. According to the vfh (Viewpoint Feature Histogram) feature descriptor x obtained in step 1 i Compute for each vector x i and the difference vector d of the mean vfh vector Ψ i ;

[0026] Step 3. Calculate the vector d by using the principal component analysis method i The feature ...

specific Embodiment approach 2

[0039] Specific embodiment two: the difference between this embodiment and specific embodiment one is: in step two, according to the vfh (Viewpoint Feature Histogram) feature descriptor x of step one i Compute for each vector x i and the difference vector d of the mean vfh vector Ψ i Specifically:

[0040] Step 21. Calculate the average vfh vector Ψ according to the vfh feature descriptor in step 1:

[0041]

[0042] Among them, M=200;

[0043] Step 22. Calculate each vector x i and the difference vector d of the mean vfh vector Ψ i :

[0044] d i =x i- Ψ,i=1,2...M. Other steps and parameters are the same as in the first embodiment.

specific Embodiment approach 3

[0045] Specific embodiment three: the difference between this embodiment and specific embodiment one or two is that in step three, the principal component analysis method is used to calculate the vector d i The feature vfh describes the subspace w specifically as:

[0046] Step 31, constructing the covariance matrix C;

[0047]

[0048] Among them, A is the vector d i collection of

[0049] Step 32, find A T The eigenvalue λ of Ai and the orthonormalized eigenvector ν i ; Use Singular Value Decomposition (Singular Value Decomposition, SVD) theorem to select A T A eigenvalue λ i contribution rate The largest first p eigenvalues ​​and the eigenvectors corresponding to the p eigenvalues;

[0050] Step 33: Calculate the eigenvector u of the covariance matrix C i ;

[0051]

[0052] Step three and four, then the feature vfh describes the subspace w as:

[0053] w=(u 1 ,u 2 ,...,u p ). Other steps and parameters are the same as those in Embodiment 1 or Embodimen...

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 a complex surface object classification method based on a point cloud VFH (Vector Field Histogram) descriptor and a neural network, and relates to a complex surface object classification method. The complex surface object classification method based on the point cloud VFH descriptor and the neural network is provided in order to solve the problems that too big search data is caused by establishing a Kd tree by using a large amount of acquisition data and the weakened classification recognition effect is caused by less acquisition data. The method comprises the following steps: firstly, calculating a vfh feature descriptor corresponding to a point cloud; secondly, calculating a difference vector di; thirdly, calculating the feature vector space of the vector di; fourthly, calculating coordinates projecting to a descriptor space; fifthly, determining input dimension and output dimension; sixthly, determining the output of corresponding angle of the projected vfh descriptor; seventhly, obtaining a BP neural network library; eighthly, determining the point cloud of the view of an object of the current bp neural network; finally, determining the final result. The complex surface object classification method based on the point cloud VFH descriptor and the neural network provided by the invention is applied to the field of complex surface object classification.

Description

technical field [0001] The invention relates to a method for classifying complex curved surface objects, in particular to a method for classifying complex curved surface objects based on point cloud VFH descriptors and neural networks. Background technique [0002] With the emergence of 3D cameras, three-dimensional images that introduce depth information have become a new focus in the field of robot vision. How to make robots classify and register objects is of great significance. There are significant problems in the existing classification methods based on point cloud vfh descriptor histograms. If a large amount of data is collected to build a Kd tree, the search data will be too large, and if the collection of data is too small, the effect of classification and recognition will be weakened. Contents of the invention [0003] The purpose of the present invention is to solve the problem that if a large amount of collected data is established to build a Kd tree, the searc...

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): G06K9/62G06N3/08
CPCG06N3/08G06F18/2415
Inventor 高会军毕程林伟阳李湛杨学博于兴虎邱剑彬
Owner HARBIN INST OF TECH