Data structures and algorithms for graph based remote sensed image content storage and retrieval

The Image Content Engine (ICE) project at Lawrence Livermore National Laboratory (LLNL) extracts, stores and allows queries of image content on multiple levels. ICE is designed for multiple application domains. The domain explored in this work is aerial and satellite surveillance imagery. The highes...

Full description

Saved in:
Bibliographic Details
Main Author: Grant, C.W.
Format: Conference Proceeding
Language:English
Subjects:
Applied geophysics
Content based retrieval
Data mining
Data structures
Earth sciences
Earth, ocean, space
Engines
Exact sciences and technology
Ice
Image retrieval
Image storage
Information retrieval
Internal geophysics
Laboratories
Remote sensing
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The Image Content Engine (ICE) project at Lawrence Livermore National Laboratory (LLNL) extracts, stores and allows queries of image content on multiple levels. ICE is designed for multiple application domains. The domain explored in this work is aerial and satellite surveillance imagery. The highest level of semantic information used in ICE is graph based. After objects are detected and classified, they are grouped based in their interrelations. The graph representing a locally related set of objects is called a "graphlet". Graphlets are interconnected into a larger graph which covers an entire set of images. Queries based on graph properties are notoriously difficult due to the inherent complexity of the graph isomorphism and sub-graph isomorphism problems. ICE exploits limitations in graph and query structure and uses a set of auxiliary data structures to quickly process a useful set of graph based queries. These queries could not be processed using semantically lower level (tile and object based) queries
DOI:10.1109/IGARSS.2004.1370787