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

Rapid KNN retrieval method and system for high-dimensional metric spatial data

A spatial data, high-dimensional technology, applied in the field of high-dimensional spatial data retrieval, can solve the problems of large data volume and low query efficiency, and achieve the effect of high recall rate and fast retrieval speed

Pending Publication Date: 2022-04-29
湖南视觉伟业智能科技有限公司
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The invention provides a fast KNN retrieval method and system for high-dimensional dimensional spatial data, which is used to solve the technical problems of excessive data volume and low query efficiency in the existing nearest neighbor retrieval query

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
  • Rapid KNN retrieval method and system for high-dimensional metric spatial data
  • Rapid KNN retrieval method and system for high-dimensional metric spatial data
  • Rapid KNN retrieval method and system for high-dimensional metric spatial data

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0029] see figure 1 , the fast KNN retrieval method of the high-dimensional dimensional space data of the present embodiment, comprises the following steps:

[0030] S110. Obtain a query instruction, including query data, nearest neighbor K value, and default query radius r.

[0031] S120. Acquire a current node, where the current node includes at least one child node.

[0032] S130. Prune the child nodes of the current node from top to bottom until leaf nodes.

[0033] S140. Traverse the leaf nodes, obtain K nearest neighbor data, obtain the distance d of the Kth data farthest from the queried data, and update the query radius r=d.

[0034] S150. Take the parent node of the current node as the new current node, and execute step S120.

[0035] S160. The current node has no sibling nodes satisfying the pruning condition, sorting from small to large and returning K nearest neighbor results.

Embodiment 2

[0037] see figure 2 , the fast KNN retrieval method of the high-dimensional dimensional space data of the present embodiment, comprises the following steps:

[0038] S201. Input query data q and default query radius r.

[0039] S202. Calculate the distance from the query data q to the supporting point sequence p[0,...,n-1], and obtain the distance sequence Pd[0,...,n-1] between q and each supporting point; Pivot Point , can be understood as a reference point, and each data will get its position in multidimensional space relative to multiple support points.

[0040] S203. Sort Pd[0,...,n-1] according to increasing distance, and obtain the storage path Ps[0,...,n-1] where the query data q belongs to the leaf node. The data storage structure is as follows image 3 As shown; for example, when the calculated sequence is Ps[1,0,11,10,...], since there are 4 layers of B+Tree in the figure, the corresponding leaf can be found by matching the first four numbers in the sequence Node...

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 rapid KNN retrieval method and system for high-dimensional metric space data, and the method comprises the steps: obtaining a query instruction which comprises queried data q, a nearest neighbor K value and a default query radius r; acquiring a current node; pruning child nodes of the current node from top to bottom until leaf nodes; traversing leaf nodes to obtain K nearest neighbor data, and assigning a distance d of the Kth data farthest from the queried data q to update a query radius r; taking a father node of the current node as a new current node; returning to the steps of repeatedly acquiring the current node, pruning to the leaf nodes, traversing the leaf nodes and updating the query radius r until the current node has no brother node and meets the pruning condition; and returning K nearest neighbor results which are ranked from small to large according to the distances to the current node. According to the method, the KNN retrieval of the high-dimensional data with the scale of tens of millions can be realized on the stand-alone server, and the method has the advantages of high retrieval speed and high recall rate.

Description

technical field [0001] The present invention relates to the retrieval of high-dimensional metric space data, in particular to a fast KNN (KNN, k-Nearest Neighbor, nearest neighbor retrieval) retrieval method and system for high-dimensional dimensional spatial data. Background technique [0002] For a long time, nearest neighbor retrieval has been a research area that researchers have paid attention to in the field of data retrieval. With the rise of big data research, nearest neighbor retrieval has become an important research hotspot at present. Nearest Neighbor Retrieval or Approximate Nearest Neighbor Retrieval is the main method in the current data retrieval field, one of which is based on the tree structure for index storage to improve retrieval performance, and the other is based on the processing of the data itself, such as hashing and vector quantization etc. [0003] Nearest neighbor retrieval is divided into two categories, one is exact neighbor retrieval, and the...

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): G06F16/901G06F16/9032G06N20/00
CPCG06F16/9024G06F16/90332G06N20/00
Inventor 夏东
Owner 湖南视觉伟业智能科技有限公司