Centralized processing and distributed operation framework construction method for spatial data services

Abstract

The invention discloses a centralized processing and distributed operation framework construction method for a spatial data service. The method comprises the following steps: deep analysis of data processing characteristics is performed, and standardized organization is carried out on spatial data and processing services; automatic processing services are integrated into a framework in a centralized parallel computing mode, and interactive editing services are integrated into the framework in a distributed operation mode, so unified regularization processing of spatial data is realized; and meanwhile, a spatial data processing framework is designed based on three principles of middleware, atomization and telescoping, so that the framework can be suitable for various operation scenes in thetechnical field of photogrammetry and remote sensing monitoring application. The invention has the advantages that intelligent production of large-data-volume spatial data is achieved, time redundancy caused by offline function circulation is reduced, connection between an automatic processing service and an interactive operation task is effectively improved, and centralized processing and distributed cooperative operation under various service scenes for spatial data processing are achieved.

Description

technical field [0001] The invention relates to the technical field of photogrammetry and remote sensing monitoring applications, in particular to a centralized processing and distributed operation framework construction method for spatial data services. Background technique [0002] With the rapid development of my country's aerospace technology, surveying and mapping technology, and Internet technology, air and ground data acquisition techniques have been continuously enhanced. Remote sensing data types, spatial resolution, time resolution, and spectrum Both resolution and radiometric resolution have made leaps and bounds, and the amount of space remote sensing data that can be obtained has also increased dramatically. According to incomplete statistics, the daily remote sensing image data obtained nationwide has reached the PB level, and the spatial remote sensing data to be processed has obvious characteristics of large data volume. Currently, GPU hardware acceleration an...

AI Technical Summary

Problems solved by technology

According to incomplete statistics, the daily remote sensing image data obtained nationwide has reached the PB level, and the spatial remote sensing data to be processed has obvious characteristics of large data volume. Currently, GPU hardware acceleration and cluster processing are mainly used for large data volume remote sensing data , cloud computing, high-performance computing and other processing technologies for several business types or processing scenarios, can only solve the batch or distributed processing of local large-scale spatial data business scenarios. Computing resources cannot be intelligently integrated and balancedly scheduled, and interactive operations cannot be planned and coordinated. Currently, there is no set of centralized processing and distributed operation frameworks for large-volume spatial data services, which cannot meet the needs of the increasingly The urgent need for intelligent processing of space remote sensing data with increasing large data volume

[0003] With the continuous improvement of hyperspectral aerospace remote sensing image technology, rapid remote sensing interpretation, automatic change discovery and extraction based on aerospace hyperspectral remote sensing images has gradually become the mainstream operating method in the application direction of remote sensing monitoring. Data processing brings the characteristics of large data volume processing. The current mature commercial software or scientific research software in the application direction of remote sensing monitoring mainly focuses on the research of different element extraction algorithms. By using high-performance GPU hardware, machine learning algorithms, and deep learning Algorithms, semi-automatic change detection and other algorithms carry out information extraction and change discovery of typical elements of remote sensing images, and then manually assign jobs and edit jobs through interactive job software. In terms of job modes, there is still no general centralized processing With the distributed operation framework, it cannot meet the intelligent operation requirements of the remote sensing monitoring application processing business under the current large amount of data

Images