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

Feature base library construction method, feature retrieval method and related device

A construction method and bottom database technology, applied to instruments, character and pattern recognition, computer components, etc., can solve problems such as low retrieval recall rate, unstable connectivity of index structure, and long time-consuming, so as to improve recall rate effect

Active Publication Date: 2021-08-13
ZHEJIANG DAHUA TECH CO LTD
View PDF4 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, HNSW (hierarchical navigable small world graph) is connected to all, and there will be connectivity and instability of the index structure, which will lead to low retrieval recall rate and long time-consuming

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
  • Feature base library construction method, feature retrieval method and related device
  • Feature base library construction method, feature retrieval method and related device
  • Feature base library construction method, feature retrieval method and related device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0026] The existing technology provides an approximate nearest neighbor search method based on a hierarchical navigable small world graph. The layered method improves the search speed and stability by solving the problem of unbalanced HNSW index structure. Specifically, in the process of building the HNSW index, this method checks whether any feature exists in the upper layer of the current inserted feature in the neighbor features of the current layer to confirm whether to promote the current feature to the upper layer. It can effectively compensate for the imbalance of the index structure constructed by the HNSW algorithm and the instability of the search speed, but it only considers the HNSW index construction stage and does not consider the subsequent search stage, and its method still has room for improvement. Its search recall rate and performance There is also the possibility of further improvement.

[0027] The prior art also provides a processing method and device fo...

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 provides a feature base library construction method, a feature retrieval method and a related device, and the feature base library construction method comprises the steps: determining a first entrance feature of a highest target layer from a layer above the highest target layer of a to-be-inserted feature in a feature base library; searching in the highest target layer based on the first entrance feature to obtain a first matching feature of the to-be-inserted feature; and in response to the situation that the number of the first matching features in the previous layer of the highest target layer does not reach a preset number, determining a second entrance feature of the previous layer of the highest target layer, and searching in the previous layer of the highest target layer based on the second entrance feature to obtain a second matching feature of the to-be-inserted feature. The method can improve the recall rate of feature retrieval.

Description

technical field [0001] The invention relates to the technical field of image retrieval, in particular to a method for constructing a feature base database, a method for feature retrieval and related devices. Background technique [0002] The feature retrieval technology is based on the similarity comparison of the feature vectors obtained by the feature extraction technology, so as to find the similar features of the features to be retrieved, and then find the corresponding photos according to the similar features. In general, when there is not much data in the bottom database, linear brute force search can be used for feature retrieval. However, with the development of science and technology, especially in the field of security and monitoring, hundreds of billions of data have been accumulated, and the vectorized feature dimensions are as high as hundreds or thousands. Linear search has been unable to meet current needs due to resource occupation and efficiency issues. 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
IPC IPC(8): G06K9/00G06K9/62
CPCG06V40/161G06V40/168G06F18/22G06V10/454G06V10/761G06F16/583
Inventor 杨凯航
Owner ZHEJIANG DAHUA TECH CO LTD