Image matching method based on ultra-high-dimensional data element clustering
A matching method and data element technology, which can be applied to instruments, calculations, character and pattern recognition, etc., can solve problems such as the disaster of dimension, and achieve the effect of overcoming the disaster of dimension and avoiding the disaster of feature dimension.
Active Publication Date: 2021-10-22
GUANGDONG UNIV OF TECH
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
When calculating the density of samples, the distance between samples needs to be calculated. This type of algorithm also has the challenge of dimensionality disaster. In addition, the clustering result is also sensitive to the density threshold parameter.
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0052] Aiming at the problem of dimensionality disaster in image data clustering in existing image matching methods, the present invention provides a clustering algorithm that can effectively avoid ultra-high-dimensional dimensionality disaster. The layer organization structure can directly cluster ultra-high-dimensional sparse data, thereby improving the accuracy of image clustering.
[0053] An image matching method based on ultra-high-dimensional data element clustering, the steps include the following:
[0054] S1, acquire image pixel data S={x 1 ,x 2 ,...,x D}∈R N×D , where x i Represents the i-th feature, D is the data dimension (feature number), N is the number of images; R is a set of real numbers, and in ultra-high-dimensional data, usually N<D.
[0055] S2, using the clustering algorithm of the pyramid paradigm, the number of layers of the algorithm is set to be m layers, the input feature set of the first layer of the pyramid paradigm is the image pixel data S,...
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
Login to View More Abstract
The invention discloses an image matching method based on ultra-high-dimensional data element clustering. According to the method, dimension reduction does not need to be carried out on the ultra-high-dimension data; by using a meta-clustering method, information features of ultra-high-dimensional data are stored as much as possible, and meanwhile, curse of dimensionality is effectively avoided. According to the method for quickly dividing the adaptive feature set in the scheme, firstly, the feature with the maximum similarity difference is divided into the plurality of feature subsets, then, the other features are divided into the feature subsets with the minimum similarity difference, and through the feature set division, the dimension of data is reduced; meanwhile, rich knowledge information can be provided for clustering to obtain accurate clustering labels, and the influence of dimension disasters in the clustering process is avoided.
Description
technical field [0001] The invention relates to the technical field of algorithm optimization and image matching, in particular to an image matching method based on ultra-high-dimensional data element clustering. Background technique [0002] With the development of deep learning, many clustering-based image data mining techniques are used for image matching. However, the pixels of an image are usually huge, and the data is usually high-dimensional and sparse when training a model. When traditional clustering algorithms deal with high-dimensional data of image data mining, due to the curse of dimensionality, the distance between samples is very large, which encounters great challenges and makes the effect of image matching poor. Therefore, there is an urgent need for an efficient ultra-high-dimensional data clustering method to deal with image data clustering to achieve optimal image matching. [0003] At present, there are mainly three types of image clustering algorithms...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62
CPCG06F18/23213
Inventor 辜方清刘浩森
Owner GUANGDONG UNIV OF TECH