VISUAL TIME-SERIES PATTERN MATCHING METHOD EMPLOYING HOUGH TRANSFORM

Title:

VISUAL TIME-SERIES PATTERN MATCHING METHOD EMPLOYING HOUGH TRANSFORM

Document Type and Number:

WIPO Patent Application WO/2020/199241

Kind Code:

A1

Abstract:

A visual time-series pattern matching method employing the Hough transform, relating to the technical field of visual data analysis. The method comprises: firstly, determining whether historical data to be matched is one-dimensional time-series data or multi-dimensional time-series data, wherein if the historical data is multi-dimensional data, normalization processing is performed first; then, performing time-series selection, that is, selecting, by means of a time window mode, a time-series to be matched from the historical data, and removing the selected time-series from the historical data; then, converting a time-series diagram in original coordinates to a Hough space by means of the Hough transform; determining matching of degree of similarity of the time-series by using a voting mechanism; and finally, selecting a final matching result according to a voting result. The visual time-series pattern matching method employing the Hough transform is easy to implement, is simple and effective, uses an algorithm with high efficiency, is applicable in the field of visualization of industrial real-time data, and is used to implement basic visual functions such as searching for a similar time-series.

Inventors:

DING JINLIANG (CN)
XU QUAN (CN)
XU MEIRONG (CN)
YU XIAORAN (CN)

Application Number:

PCT/CN2019/082423

Publication Date:

October 08, 2020

Filing Date:

April 12, 2019

Export Citation:

Click for automatic bibliography generation Help

Assignee:

UNIV NORTHEASTERN (CN)

International Classes:

G06K9/62

Foreign References:

CN109241077A	2019-01-18
CN1685359A	2005-10-19
CN109409639A	2019-03-01
CN109241190A	2019-01-18
EP2960806A1	2015-12-30
CN106936964A	2017-07-07
CN105046684A	2015-11-11

Other References: