Article is devoted to a research of a problem of representation of diverse knowledge and data on spatial objects and ways of operating by them at the solution of monitoring problems under control of the multi-purpose system of remote monitoring of ISIT (MpSRM). The review is presented classification of knowledge which systematizes knowledge of semantics, syntax, behavior of a monitoring object and the rule of their interpretation. The approach is considered to representation of knowledge allowing to formalize monitoring problem definition.
The System, in its turn, is targeted at determination of the required measurement rate and information sources; definition of deciphering features of the aspects for the Object to be found on the image with reference to each source; identification of the correspondence between the numeric values as measured on the image and the respective properties of the MO which shall be identified under the Definer query.
Thus, the key task setting parameters to be clarified interactively between the Definer and the System are: Name of the Object (type of objects) and its identities (properties) as known to the Definer for the Object identification by the System among the variety of the surrounding objects; Structural properties of the Object as known to the Definer (when significant for identification); Relations between SO and environment as known to the Definer (when significant for identification); Properties of the Object to be defined; Structure of the Object to be defined; Relations between the SO and environment to be defined. 3
Two knowledge categories have been conceptually identified: MO knowledge; system knowledge. 3.1
MO Knowledge may generally be expressed as below:
= < , , , >,
(1)
where O is a monitoring object, is a set of the Object specialties; R is a set of binary relations with other objects
formalized in [
A descriptor being a future advance in approach (as detailed in [7]) was suggested to align the Definer mental model with the internal System knowledge representation. Statement of tasks is characterized in the below categories: Taxonomy of domain concepts. One or more taxonomies correspond to the specific domain, each taxonomy setting the SO hierarchical classification from a certain position. Specific element (node of a taxonomic tree) may have various specialties. Properties of taxonomy elements. Specific element is described by the set of properties indicative of the respective aspect of the Object. Non-representational (class of properties) and certain properties are identified, each likely matched with the range of values, type of a measurement scale, principles of measurement and possible interpretations. Relations between the taxonomy elements. Specific pair of elements may be described by the set of binary relations over it. Domain Aggregations. Describe the allowable structural combinations of the Object in question as collections of components (elements, parts).
Comparison of SO specialties presented by the set of the properties to be measured, each (specialty or property) being in correlation with the relevant concept taxonomy node and model of the estimated path of the property value being changed in time, peculiar for this SO specialty.
Refer to the collection of the key peculiarities (features) of the Monitoring Object as seen on the image: structural peculiarities, shape features, brightness and textural properties as seen in one or another range or a set of EM-spectrum ranges, spatial attitude, time-to-time variability (in view of each of the above peculiarities).
Knowledge of the interpretation rules being indicative of the semantic values based on the syntactic marker measurements and characteristics of the initial image.
Regulates the ERS data conversion control. The variety of procedures and algorithms support the great many various stages of data processing. It is important to identify the classes of one-function procedures for the classifier being configured in this set, such procedures, in their turn, may be subdivided to atomic actions, if required. System knowledge may be subdivided into the following sub-classes:
Knowledge of the Data Conversion Algorithms may be set in a declarative manner by lists or relational structures
[
Knowledge of the Tools for Interaction with End User describes the principles of the task assignment by the end use, principles of internal task interpretation and principles of the solution results representation. These principles are aimed at description of the structure of “end user - decision system” interaction. Information provided by the user when setting the task shall be of the essence when selecting the specific algorithm or set of algorithms at a particular time.
Knowledge of Principles and Algorithms of Data Conversion. In the context of the user information query resolution it is worth noting the set of algorithms of different difficulty and scope which are gathered in the respective structures united by the logics of the input conversion and input requirements. Algorithmic chaining shall be in the hands of the expert responsible for the data processing and analysis. Such expert shall also associate (assign the R set for each algorithm) the realized chains with the typical tasks of end users.
Knowledge of the Data to be Converted. The System shall be capable of searching the relevant and actual data for successive resolution of the user information query. Such data include satellite images with significant scattering of spatial resolution as well as radiometric, spectral and time resolution. Therefore, restrictions for the possible ERS data sources must be identified for various classes of tasks. 4
The initial design of the knowledge base shall be based on two major activities: 1) preparation and structuring of the Object knowledge in the specialty of choice , which includes attribution of the elements of R, P and sets; 2) export of the ready structure to the data base. Software environment which supports execution of the above activities and includes such key components as Protégé OWL-ontology editor which builds up the RDF data structure describing the graph of objects and their relations in XML format; local web-interface which allows decomposing the XML data structures and export them to MySQL database, the structure of which is given on Fig 1.
class KnowledgeSchema
Classes «column» *PK id_class * name_class «PK» + PK_Classes()
Subclasses «column» *PK id_relate * name_superclass * name_subclass «PK» + PK_Subclasses()
Individual «column» *PK id_individual * name_individual «PK» + PK_Individual()
Data_property «column» *PK id_data_property * name_data_property
datatype «PK» + PK_Data_property()
Object_property «column» *PK id_object_property * name_object_property domen range «PK» + PK_Object_property()
Fig 1. Database Scheme
Logical data structure includes such basic entities as Object Classes’ Guide (Classes), Instance Guide for all classes (Individuals), Guide of Possible Intra-Instances’ Relations (Object_property), Guide of Possible Object Properties (Data_properties), description of Class-Subclass relations (Subclasses).
The reviewed concept of the knowledge structure representation found its practical use in the automated agricultural
monitoring system created by the authors [
Queries can be divided into occasional and cyclic. Occasional queries are, as a rule, commands to be immediately performed. Metric calculations, for instance, area, perimeter and etc., set the example of such queries in MpSRM.
Cyclic queries imply the regular queries to change the set parameter for the given Object thus complying with the monitoring tasks. The typical examples may be monitoring of the vegetation structure, surface temperature, moisture and etc.
Fig 2. Typical queries of users to the decision system on the example of the agricultural land appraisal (AL) As an example, let’s consider the monitoring task for the vegetation heterogeneity within the agricultural contour, which model may be given as an expression (2): ОМ = П, , , , , , , , , > (2)
In semantic space the monitoring Object is defined by the crop ∈ typified by the model of its development in time Tm, such model being determined by the sequence of phenophases = { } and Plan of agrotechnical activities as given in (2) as an aggregate of events = { }.
A variety of Tm time ranges for the change of the Object F states is descriptive of the Object behavior and target path of its lifecycle.
Syntactic description includes a metric set describing the form (a set of coordinates П, thickness T) and dimensions (area NS, perimeter NP) of the Object as well as the spectral NDVI index.
Rules of averaged NDVI values’ interpretation – ´ allow to compare the status ∈ of each heterogeneous region
and the Object in whole [
Knowledge base area representative of the bunch of taxonomy elements and their correlation when solving the given task is shown on Fig.3 as a direct graph.
Fig 3. Correlation of the objects of MpSRM ISIT knowledge base when solving the tasks of the vegetation heterogeneity monitoring
The central graph node is “Field No 157” which belongs to the agricultural enterprise, “Uchkhoz Minderlinskoe” LLC. Wheat was grown in that field from 2013 till 2018 which is shown by a set of arcs connecting the “Field No 157” object with the “Wheat” object. Agrotechnological activities, such as dragging, seeding, top-dressing and so on, are executed at the “Field” object. Instances of the “Agrotechnological Activities” class are defined by the recommended time execution interval.
Wheat is characterized by the set of phenological development stages (sprouts, tillering, stem elongation and so on),
which is shown on the graph by the “Wheat” object ties with the instances of the “Phenological Stage” class.
Phenological stages, in their turn, are characterized by the reference time interval, duration and range of NDVI values [
The approach to the management of the satellite monitoring knowledge reviewed herein implies use of the suggested knowledge classification by setting the sets of concepts and relations which simulate the domain the user is interested in, as well as the all-system representations. The created knowledge base was tested with resolution of some typical SPARQL and SQL queries which simulated the information queries of users. Software module which is responsible for the “end user - decision system” chat and which allows defining the tasks of occasional or cyclic measurement of the set parameters in order to assess the state of the given object in various specialties is being prototyped.
The research was financially supported by the Russian Fund of Fundamental Research (Project No 18-47-242002 r_mk), Government of the Krasnoyarsk Territory and the Krasnoyarsk Territory Science Fund as part of the Scientific Project named “Development of the technology to build up the intellectual information systems for the object-oriented monitoring of the areas as per the RS data”.
LITERATURE [1] E.A. Lupyan, I.V. Balashov, M.A. Burtsev, V.Yu. Efremov, A.V. Kashnitski and others. Development of technologies to build up the remote monitoring information systems//Modern issues of the Earth Remote Sensing from space. 2015. V. 12. No5. P.53–75.