An optimal L∞-PLA algorithm for trajectory data compression

With the application and development of global positioning system, huge amounts of real-time trajectory data are collected, which gives a challenge for data transmission, storage and analysis. To attack this issue, data compression technology based on piecewise linear approximation (PLA), which is s...

Full description

Saved in:
Bibliographic Details
Published in:Shenzhen da xue xue bao. Li gong ban Vol. 41; no. 5; pp. 574 - 582
Main Authors: ZHAO Huanyu, SUN Guohao, LI Tongliang, YANG Jian, PANG Chaoyi
Format: Journal Article
Language:English
Published: Science Press (China Science Publishing & Media Ltd.) 01.09.2024
Subjects:
algorithm theory
compression algorithm
maximum error bound
piecewise linear approximation
synchronous error bound
time series
trajectory data
ISSN:1000-2618
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the application and development of global positioning system, huge amounts of real-time trajectory data are collected, which gives a challenge for data transmission, storage and analysis. To attack this issue, data compression technology based on piecewise linear approximation (PLA), which is simple and intuitive, less compression storage and faster data transmission, has been widely researched. Currently, the optimal online PLA algorithm can not effectively compress multi-dimensional trajectory data. This paper presented a novel multi-dimensional compression problem under maximum error bound (mDisPLA∞), and proposed an optimal online PLA algorithm MDisPLA to solve it. MDisPLA used a divide-and-conquer strategy to extend the one-dimensional optimal PLA algorithm for optimizing compression of multi-dimensional trajectory data. It can generate the minimum number of disconnected straight lines in linear time complexity, and these lines are quality-ensured, i.e., the synchronous error between the original point and corresponding recovered one is controlled. With experimental tests by comparing MDisPLA with the state-of-the-art algorithm implemented based on cone intersection using the synchronous Euclidean distance (CISED) on trajectory data sets, it demonstrates that MDisPLA can be stable to generate the quality-ensured lines. It is faster by 14 times than CISED with lower memory requirements, and can reduce the number of segments by 48% and the storage number by 10.5%. MDisPLA significantly improves processing speed and reduces storage while maintaining compression quality.
ISSN:1000-2618
DOI:10.3724/SP.J.1249.2024.05574