Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

High Dimensional Approximate Image Retrieval Method Based on Inverted lsh in Cloud Computing Environment

A cloud computing environment, image retrieval technology, applied in still image data retrieval, still image data query and other directions, can solve problems such as excessive amount of information, inconsistency between required information and displayed pictures, inability to adapt to distributed indexes, etc. time saving effect

Active Publication Date: 2019-05-31
DALIAN UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that existing location-sensitive hash indexes cannot adapt to distributed indexes, the present invention proposes a high-dimensional approximate image retrieval method based on inverted LSH in a cloud computing environment, which can realize location-sensitive hash indexes to adapt to distributed indexes
[0007] Beneficial effects: Since the cloud center service system establishes an inverted position-sensitive hash index, the position-sensitive hash index can be adapted to distributed queries, so that the present invention solves the problems of excessive information volume and mismatch between required information and displayed pictures, etc. , as much as possible to help users save retrieval and query time

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
  • High Dimensional Approximate Image Retrieval Method Based on Inverted lsh in Cloud Computing Environment
  • High Dimensional Approximate Image Retrieval Method Based on Inverted lsh in Cloud Computing Environment
  • High Dimensional Approximate Image Retrieval Method Based on Inverted lsh in Cloud Computing Environment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0013] Example 1: A high-dimensional approximate image retrieval method based on inverted LSH in a cloud computing environment, including steps: a client collects and extracts image features, and communicates with a cloud center service system; the cloud center service system utilizes the powerful distributed computing capabilities of the cloud to establish a Invert the position-sensitive hash index and query the neighboring images corresponding to the captured image.

[0014] Generally, the kNN algorithm is based on a distributed inverted index, but a position-sensitive hash index is not a distributed index. In order to adapt to a distributed index, a distributed inverted index based on a position-sensitive hash is established: in this technical solution, When establishing a position-sensitive hash index, separate out several Hash buckets, use the Hash bucket as the Key, and use the point set in the Hash bucket as the Value, and use MapReduce for distributed solution. This ...

Embodiment 2

[0021] Example 2: This embodiment has the same technical solution as Embodiment 1, and more specifically: this embodiment discloses a specific method for establishing a distributed inverted index based on position-sensitive hash, and stores the data set in the HDFS distributed index in advance. In the file system, when starting a task, read in some configuration files LSH hash function family through the distributed cache mechanism, each Map task reads in the data fragment specified by the JobTracker as input, and then according to the given hash function for each A data object is subjected to hash mapping for dimensionality reduction, and a high-dimensional vector is passed through the hash map to obtain a hash value. This hash value is used as an index value. For example, for a high-dimensional vector v, pass the i-th hash function hi( .) After mapping, the hash value hi(v) is obtained, and finally output in the form of key-value pair. For each high-dimensional data vector...

Embodiment 3

[0045] Embodiment 3: This embodiment has the same technical solution as Embodiment 1 or 2, more specifically: this embodiment discloses a query method, the query is to establish a kNN query based on a position-sensitive hash distributed inverted index, The steps are: set the high-dimensional data set as S, and S is the existing large-scale image library in the image retrieval system, such as a large number of plant image galleries, and each image in the image library is preserved as a 128-dimensional high-dimensional image library. Dimensional features. The set of query objects is Q, and Q is the query image object. For example, a group of images of flowers are taken. After high-dimensional feature extraction is performed first, a feature set is formed. For each query object q belongs to Q, the correlation function h is initialized, h belongs to G, G is a hash family, LSH is a multi-round hash algorithm, different hash functions will get different hash results. h correspond...

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 high-dimensional approximate image retrieval method based on inverted LSH in a cloud computing environment, which belongs to the field of big data and mobile applications. The system establishes a new index structure (LSRP‑tree), which reduces the cost of high-dimensional indexing and improves query efficiency; the new algorithm (H‑c2kNN) formed by combining LSH with MapReduce shows good scalability and efficiency. These two innovative applications solve the approximate retrieval problem in high-dimensional data space. An optimization method based on hash collision counting and sorting is adopted to greatly reduce the amount of intermediate data and speed up data processing. The invention is a system that uses an intelligent mobile platform to search for pictures, including a set of cloud servers and a mobile client. The latter collects and transmits pictures, and the former is responsible for establishing high-dimensional indexes and performing kNN query processing. The invention effectively and advantageously improves the recognition problem of a large number of images and satisfies people's further desire for intelligent mobile information retrieval.

Description

technical field [0001] The invention belongs to the application field of large-scale spatio-temporal data processing and mobile technology, and relates to a high-dimensional approximate image retrieval method based on inverted LSH in a cloud computing environment Background technique [0002] Now the Internet basically covers people's lives, and mobile Internet access has become the main Internet access mode. As of June 2014, among the Internet users in my country, the mobile phone usage rate reached 83.4%, surpassing the overall usage rate of traditional PCs (desktops and notebooks) for the first time at 80.9%, and the position of mobile phones as the largest Internet terminal device has been further consolidated. Nowadays, information technology is developing very rapidly, and the amount of information in various forms is also growing rapidly. With the diversification and complexity of user retrieval requirements, users are no longer satisfied with simple text retrieval, b...

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): G06F16/53
CPCG06F16/53Y02D10/00
Inventor 季长清王宝凤汪祖民宋佳齐
Owner DALIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com