Towards an Ontology for the structuring of Remote Sensing operations shared by different processing chains
Remote sensing is the scientific discipline that brings together all the knowledge and techniques used for observing, analysing, interpreting and managing the environment. This environment is studied from measurements and images obtained with the help of airborne platforms, space, land or sea.
Results are usually obtained following an ordered sequence of operations, which can especially be long. Moreover, each remote sensing engineer has his own experience and preferences in the choice of algorithms and software. This concludes in a fuzzy working environment where people speaking about same things use different words and same words do not necessarily describe the same thing.
What is described in this article is the common language, a common shared formalism, used to describe the processing chains: an ontology. An ontology is a set of concepts and categories in a subject area or domain that shows their properties and the relations between them. An ontology is often represented as a graph. In other words, it is used in support of a knowledge base to structure and define concepts within a specific domain. As an artificial intelligence tool, it is of high-level knowledge but lacks in computations performance.
In a more practical way, considering that, a service, i.e. a processing chain, is made of elementary operations, in red, linking data, in blue; you can represent them simply as a directed sequence of elements.
When a sufficient number of processing chains is defined and structured following the ontology, the different subgraphs are linked into a knowledge graph. This knowledge base is therefore the representation of the different experiences but also the pros and cons of the remote sensing engineers that have created the services.
In this framework, several applications become possible:
First step is to set up consistency rules for error management in the knowledge graph. This is especially possible because every vertices and edges are well described within the ontology using description logic statements. For instance, it is not common that an operation follows itself in processing chains. Some operations may need some type of data and so on.
Because of the graph structure, it is possible to use graph theory algorithm and, for instance, like a GPS in a car find the shortest way between two geographical points, you can find the shortest path between two concepts. Algorithms that are more complex may also recognise patterns in the graph corresponding to pros and cons of remote sensing engineers.
Finally, taking into account the two previous points, comparisons between different services are easy and permit competition automatic detection, what is achievable if the ontology is used in support of a market place (like a remote sensing Amazon for instance).
Future work will study the possibilities to create, in an automatic way, processing chains based on the consistency rules, the common patterns in the knowledge graph but also the detailed operations and data. The interesting part of this idea lies in the fact that the machine will understand by itself the principles, without the help of humans but simply using the hidden knowledge in the data.
This article intends to simplify the content of the following scientific article. It has been simplified and shortened on purpose:
Nys, G.-A., Kasprzyk, J.-P., Hallot, P., and Billen, R.: TOWARDS AN ONTOLOGY FOR THE STRUCTURING OF REMOTE SENSING OPERATIONS SHARED BY DIFFERENT PROCESSING CHAINS, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4, 483-490, https://doi.org/10.5194/isprs-archives-XLII-4-483-2018, 2018.
#Ontology #RemoteSensing #IA #OntologyEngineering #NoSQL #Graph #GraphMining #Knowledge #KnowledgeDiscovery