CN106709989A - Object-oriented image characteristic-based geographic ontology modeling and semantic reasoning method - Google Patents
Object-oriented image characteristic-based geographic ontology modeling and semantic reasoning method Download PDFInfo
- Publication number
- CN106709989A CN106709989A CN201611024360.0A CN201611024360A CN106709989A CN 106709989 A CN106709989 A CN 106709989A CN 201611024360 A CN201611024360 A CN 201611024360A CN 106709989 A CN106709989 A CN 106709989A
- Authority
- CN
- China
- Prior art keywords
- ground
- remote sensing
- codomain
- class
- ontology
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/05—Geographic models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
Abstract
The invention discloses an object-oriented image characteristic-based geographic ontology modeling and semantic reasoning method. The method comprises the following steps of: S1, carrying out preprocessing and multi-scale segmentation on a high-resolution remote sensing image of a researched district so as to obtain an object hierarchical structure; S2, carrying out territory division and carrying out land type division on each territory; carrying out geographic ontology modeling on different land types of each territory according to remote sensing characteristic values of the high-resolution remote sensing image; S3, for geographic ontologies of different territories, calculating a value domain of each remote sensing characteristic value of each land type, and calculating intersected sets and union sets of value domains of different land types; and S4, establishing semantic association of the land types in the territories and the land types between different territories. According to the method, the semantic integration and interoperation between different geographic ontology systems can be realized, so as to push the process of space information socialization, decrease the heavy investment pressure caused by repeated acquisition of basic space data to the greatest extent, and eliminate the islanding effect of space information systems in different territories.
Description
Technical field
The present invention relates to spatial information sharing field, more particularly to a kind of ontology based on object-oriented image feature
Modeling and semantic reasoning method.
Background technology
Space information system is fast-developing to space information grid from standalone version, with carrying for spatial information integration degree
The demand of high and different field spatial information sharing and interoperability is more and more urgent, highlights several problem demanding prompt solutions.Its
In, difficulty maximum is Geospatial Semantic expression inconsistence problems, and the key problem of Semantic Grid of Spatial Information, how it is
Solve the problems, such as the semantic reasoning between the ontology system of different field.
At present in spatial information field, often discuss that ontology is modeled by the research method of field of information system, but
There are the following problems：
(1) in the multiple dimensioned aspect existing defects of expression of space data；
(2) different field is to the difference of same class geographical entity naming method and the characteristic attribute of description ontology concept
The difference of collection, cannot passing through in information science of making that it each obtains carries out semanteme based on concept (feature) inference theory
Interoperability.
In face of two above key issue, it is badly in need of a kind of ontology modeling and semantic reasoning method, realizes based on semanteme
Spatial information integration and interoperability.
The content of the invention
The technical problem to be solved in the present invention is not enough for multi-scale expression in the prior art, it is impossible to by based on general
The inference theory of thought carries out the defect of Semantic Interoperation, there is provided it is a kind of based on object-oriented image feature ontology modeling with
Semantic reasoning method.
The technical solution adopted for the present invention to solve the technical problems is：
The present invention provides a kind of ontology based on object-oriented image feature and models and semantic reasoning method, including with
Lower step：
S1, the high-resolution remote sensing image for obtaining research location, pretreatment and many chis are carried out to high-resolution remote sensing image
Degree segmentation, sets up imaged object hierarchical structure；
S2, field division is carried out according to imaged object hierarchical structure, and land type division is carried out to each field；According to high score
The remote sensing features value of resolution remote sensing image, including spectrum characteristic parameter and shape facility value, the different land types to each field are carried out
Ontology is modeled, and obtains the ontology of different field；
S3, the ontology to different field, calculate the codomain of the remote sensing features value of each ground class；And calculate different land types
Codomain between common factor；If exist between two codomains occuring simultaneously, the union between two codomains is calculated, and common factor is accounted for simultaneously
The percentage of collection；
S4, according to the size occured simultaneously, the semantic association of the ground class set up in field, and ground class between different field
Semantic association.
Further, the method for being pre-processed to high-resolution remote sensing image in step S1 of the invention includes：Geometry
Correction, atmospheric correction and cutting splicing.
Further, the method for ontology modeling being carried out to the different land types in each field in step S2 of the invention
For:
S21, class determine：According to the criteria for classification in field, different land types are divided；
S22, optimum segmentation size selection：Its corresponding optimum segmentation size is selected to each ground class；
S23, class samples selection：Selectively class sample, the ground class sample bag under the optimum segmentation size of each ground class
The remote sensing object for including similar pixel composition and the remote sensing object for being doped with other types pixel, the coverage rate for increasing ground class sample make
The precision of post analysis reaches maximum；
S24, computation attribute statistical form：To each ground class sample, its spectrum and shape Statistics value are calculated, count each attribute
The maximum of value, second largest value, minimum value and sub-minimum, obtain the statistics of attributes table of each ground class sample；
S25, calculating codomain scope：To each property value of each ground class sample, with the model between second largest value and sub-minimum
Enclose the codomain on the attribute as ground class sample；
S26, statistics codomain figure：Codomain range statistics result according to all ground class sample, is counted by attribute, that is, unite
The attribute codomain of different land types on each property value is counted, and makes codomain figure；
S27, codomain figure statistical rules：According to codomain figure, for specific ground class, selected and other in all properties
The maximum attribute of ground class difference, is screened with this property value codomain, rejects part interference, repeats this step, increases attribute
Number, further rejects interference；
S28, set up ontology：According to the attribution rule of statistics, the ontology in the field is set up.
Further, the method for optimum segmentation size is selected in step S22 of the invention to be included：Largest face area method, average
Variance method and target function method.
Further, the attribute in step S24 of the invention includes：Normalized differential vegetation index, canopy density, boundary index,
Brightness, length-width ratio, first band spectrum average and shape index.
Further, after the ontology of different field is obtained in step S2 of the invention, the geographical sheet of different field
Body is mapped in same remote sensing image hierarchical structure.
Further, the remote sensing features value in step S3 of the invention includes：Red wave band average, it is maximum spectral differences, red
Band ratio, normalized differential vegetation index, contrast, correlation and energy.
Further, semantic association is set up in step S4 of the invention includes direct semantics association and indirect semantic association.
Further, the method for judgement direct semantics association of the invention is：
If the common factor of two certain remote sensing features values of ground class is empty set, there is no semantic association to close between the two ground classes
System；
If the codomain of two all remote sensing features values of ground class intersects, and there is inclusion relation between codomain, then this two
It is hyponymy between individual ground class；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is less than
10%, then there is no semantic association relation between the two ground classes；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is more than
10% and less than 80%, then it is semantic similarity relation between the two ground classes；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is more than
80%, then it is synonymy between the two ground classes.
Further, the method for the indirect semantic association of judgement of the invention is：
If two ground classes between be synonymy, wherein any one ground class upper strata ground class, with another ground class it
Between be hyponymy.
The beneficial effect comprise that：Ontology modeling and language based on object-oriented image feature of the invention
Adopted inference method, the multi-scale division technology according to object-oriented is mapped to the concept of different field ontology system same
Carry out ontology modeling in individual remote sensing image hierarchical structure, and on this basis by the characteristic attribute codomain of different land types it
Between relation primarily determine that mutual semantic relation, then in hyperspace determine semantic logic contact, realize different geography
Semantic intergration and interoperability between main body system, so as to promote the process of spatial information socialization；And base is reduced to greatest extent
The heavy investment pressure that plinth spatial data repeated acquisition is brought, eliminates the orphan of different field space information system to a certain extent
Island effect.
Brief description of the drawings
Below in conjunction with drawings and Examples, the invention will be further described, in accompanying drawing：
Fig. 1 is the method flow diagram of the embodiment of the present invention；
Fig. 2 is the boundary index attribute codomain figure of the various regions class sample of the embodiment of the present invention；
Fig. 3 is the forestry ground class body tree of the embodiment of the present invention；
Fig. 4 is the ontology concept of the embodiment of the present invention and the relation of multiscale morphology imaged object.
Specific embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the present invention, not
For limiting the present invention.
As shown in figure 1, the modeling of the ontology based on object-oriented image feature of the embodiment of the present invention and semantic reasoning
Method, comprises the following steps：
S1, the high-resolution remote sensing image for obtaining research location, pretreatment and many chis are carried out to high-resolution remote sensing image
Degree segmentation, sets up imaged object hierarchical structure；The method of pretreatment includes：Geometric correction, atmospheric correction and cutting splicing etc..
S2, field division is carried out according to imaged object hierarchical structure, and land type division is carried out to each field；According to high score
The remote sensing features value of resolution remote sensing image, including spectrum characteristic parameter and shape facility value, the different land types to each field are carried out
Ontology is modeled, and obtains the ontology of different field；After obtaining the ontology of different field, the geography of different field
Ontology Mapping is in same remote sensing image hierarchical structure.
The method that the different land types in each field are carried out with ontology modeling is:
S21, class determine：According to the criteria for classification in field, different land types are divided；
S22, optimum segmentation size selection：Its corresponding optimum segmentation size is selected to each ground class；Selection optimum segmentation
The method of size includes：Largest face area method, mean variance method and target function method.
S23, class samples selection：Selectively class sample, the ground class sample bag under the optimum segmentation size of each ground class
The remote sensing object for including similar pixel composition and the remote sensing object for being doped with other types pixel, the coverage rate for increasing ground class sample make
The precision of post analysis reaches maximum；
S24, computation attribute statistical form：To each ground class sample, its spectrum and shape Statistics value are calculated, count each attribute
The maximum of value, second largest value, minimum value and sub-minimum, obtain the statistics of attributes table of each ground class sample；Attribute includes：Normalization
Vegetation index, canopy density, boundary index, brightness, length-width ratio, first band spectrum average and shape index.
S25, calculating codomain scope：To each property value of each ground class sample, with the model between second largest value and sub-minimum
Enclose the codomain on the attribute as ground class sample；
S26, statistics codomain figure：Codomain range statistics result according to all ground class sample, is counted by attribute, that is, unite
The attribute codomain of different land types on each property value is counted, and makes codomain figure；
S27, codomain figure statistical rules：According to codomain figure, for specific ground class, selected and other in all properties
The maximum attribute of ground class difference, is screened with this property value codomain, rejects part interference, repeats this step, increases attribute
Number, further rejects interference；
S28, set up ontology：According to the attribution rule of statistics, the ontology in the field is set up.
S3, the ontology to different field, calculate the codomain of the remote sensing features value of each ground class；And calculate different land types
Codomain between common factor；If exist between two codomains occuring simultaneously, the union between two codomains is calculated, and common factor is accounted for simultaneously
The percentage of collection；Remote sensing features value includes：Red wave band average, maximum spectral differences, red band ratio, normalized differential vegetation index,
Contrast, correlation and energy.
S4, according to the size occured simultaneously, the semantic association of the ground class set up in field, and ground class between different field
Semantic association.
Setting up semantic association includes direct semantics association and indirect semantic association.
Judge that the method that direct semantics is associated is：
If the common factor of two certain remote sensing features values of ground class is empty set, there is no semantic association to close between the two ground classes
System；
If the codomain of two all remote sensing features values of ground class intersects, and there is inclusion relation between codomain, then this two
It is hyponymy between individual ground class；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is less than
10%, then there is no semantic association relation between the two ground classes；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is more than
10% and less than 80%, then it is semantic similarity relation between the two ground classes；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is more than
80%, then it is synonymy between the two ground classes.
The method for judging indirect semantic association is：
If two ground classes between be synonymy, wherein any one ground class upper strata ground class, with another ground class it
Between be hyponymy.
In another specific embodiment of the invention：
Ontology modeling and semantic reasoning method based on object-oriented image feature, comprise the following steps that：
Data acquisition：Obtain the high-resolution remote sensing image in research location.
Data prediction：Remote sensing image data to obtaining is pre-processed, including geometric correction, atmospheric correction, cutting
Splicing etc..
Multi-scale segmentation of remote sensing images：Using the professional software pair of existing support object-oriented classification of remote-sensing images at present
Image carries out multi-scale division, sets up imaged object hierarchical structure.
Ontology modeling is carried out based on spectrum and shape facility：By object-oriented multi-scale division, form by not
A multi_tier architecture is constituted with remote sensing object, on this basis, by the spectrum and shape facility of remote sensing image object to not
The remote sensing object of same type is modeled.Detailed process is as follows：(1) class determines：According to more improve and be widely recognized as in the industry
Domain classification standard, binding region actual conditions, the ground class that is directed to of selection.(2) optimum segmentation scale selection：
According to the ground class studied, its corresponding optimum segmentation yardstick is selected, the selection of optimum segmentation yardstick typically has largest face area method,
Mean variance method, target function method etc..(3) samples selection：Under each ground class optimum segmentation yardstick, various regions class selects it to represent
Property sample, samples selection based on the remote sensing object that is made up of similar pixel completely, while taking into account part is doped with other classes
The remote sensing object of type pixel, in the case where the precision of post analysis is ensured, increases the coverage rate of sample as far as possible.(4) united by class
Meter sample attribute value, obtains the statistics of attributes table of each ground class：For each ground class, its spectrum and shape Statistics value are calculated, united
Count maximum, second largest value, minimum value, the sub-minimum of each property value.(5) each property value codomain scope of each ground class is determined：
With time maximum and time codomain of the minimum value as ground class on the attribute of various regions generic attribute.Selection attribute time maximum and time most
Small value is, in order to prevent from might have artificial error when sample is selected, to elect other ground classes as sample.Generally not
Same ground class SPECTRAL DIVERSITY is larger, if it is the sample for studying ground class that other ground classes are falsely dropped, the spectral value of the sample does not just exist
In the codomain of research ground class.Therefore, in order to reduce error using secondary maximum and time codomain of the minimum value as ground class on attribute.
(6) the attribute codomain of different land types on each property value is counted：Codomain range statistics result according to all ground class, is entered by attribute
Row statistics, that is, count the attribute codomain of different land types on each property value, and make codomain figure.(7) codomain figure statistical rules：Root
According to codomain figure, for specific ground class, the attribute maximum with other ground class differences is selected in all properties, with this property value
Codomain is screened, and rejects part interference, repeats this step, increases attribute number, further rejects interference plot.Until taking
Obtain better effects.(8) ontology is set up：According to the attribution rule of statistics, the field ontology is set up.
Different field ontology system constructing：Using the multi-scale division technology of object-oriented in a width fixed resolution
Remote sensing image on construct it is different segmentation yardsticks under imaged object hierarchical structures, according to above-mentioned ontology model walk
Suddenly, the concept of different field ontology system is mapped in same remote sensing image hierarchical structure.
Image spectrum characteristic parameter is calculated：For the ontology of the different field studied, some are chosen representative
Object, calculate its red wave band average, maximum spectral differences, red band ratio, normalized differential vegetation index, contrast, correlation,
Six image feature values such as energy, finally draw the codomain of each characteristic value.
Set up the common factor of domain object characteristic attribute of the same race：Between the ground class in each field, remote sensing features collection is based on by it
Codomain, the common factor between calculating.If exist in each characteristic value occuring simultaneously, then union between any two is calculated,
And obtain the percentage that common factor codomain accounts for union codomain.
Set up the semantic relation between the concept in domain body and between body：Its codomain has intersecting or wraps on characteristic attribute
Be present semantic association in the imaged object that can map out different scale containing relation, occur simultaneously with reference to the attribute of above-mentioned foundation, set up neck
The semantic relation between concept in the body of domain and between body.Semantic association is divided into direct semantics association and indirect semantic association two
Kind, specific inference step is as follows：
(1) direct semantics association：There is following semantic reasoning rule：
If it is empty set that 1. a ground classes in A fields are occured simultaneously with the b classes in B fields on certain spectrum characteristic parameter, this can be determined
A, b both ground classes there is no semantic relation；
If 2. class B field b ground class in A fields a ground intersects on all spectrum characteristic parameters, if there is bag between feature set
Containing relation, then hyponymy can be determined that it is；If there is the proportion for accounting for union that occurs simultaneously between feature set less than 10%, can
To determine a, there is no semantic relation between b both ground classes；If exist between feature set occur simultaneously account for the proportion of union more than 10% and
Less than 80%, then a can be determined, be semantic similarity relation between b both ground classes；If existing to occur simultaneously between feature set and accounting for union
Proportion is more than 80%, then be synonymy.
(2) indirect semantic association：According to direct semantics related reasoning rule, if inferring a ground classes and the B fields in A fields
B ground class be synonymy；Then upper strata object c and b the ground class of a ground class are b ground class in hyponymy, and B fields in A fields
Upper strata object d and a ground class be also hyponymy.
In another specific embodiment of the invention, ontology modeling and semanteme based on object-oriented image feature
Inference method, comprises the following steps that：
Step S1, data acquisition：The high-resolution remote sensing image in research location is obtained, this case-based system is Hangzhou west
The high resolution image of lake region is experimental data.
Step S2, data prediction：Remote sensing using GIS major software (ARCGIS, ENVI etc.) to obtaining
Image data is pre-processed, including geometric correction, atmospheric correction, cutting splicing etc..
Step S3, multi-scale segmentation of remote sensing images：The special of object-oriented classification of remote-sensing images is supported using existing at present
Industry software part (such as eCognication, ENVI, Erdas) carries out multi-scale division to image, sets up imaged object hierarchical structure,
The eCognication softwares that this example is used.
Step S4, ontology modeling is carried out based on spectrum and shape facility：By object-oriented multi-scale division, formed
By different remote sensing objects a multi_tier architecture is constituted, on this basis, the spectrum and shape by remote sensing image object are special
Levy and different types of remote sensing object is modeled.Detailed process is as follows：(1) class determines：According to more perfect and wide in the industry
The domain classification standard of general accreditation, binding region actual conditions, the ground class that selection is directed to, this example is led with forestry
It is introduced as a example by the criteria for classification of domain, forest department's land type has：Forest land, shrub land, arbor forest land, opening, city
Villeggiatura people construction land, Land Use for Transport Construction (road), lake, the river water surface, other lands used etc..(2) optimum segmentation chi
Degree selection：According to the ground class studied, its corresponding optimum segmentation yardstick is selected, the selection of optimum segmentation yardstick typically has maximum
Area-method, mean variance method, target function method etc..This example uses the optimum segmentation that various regions class is determined based on point dimension method
Yardstick.Table 1 below is the corresponding best segmental scale sample table of field of forestry ground class：
The corresponding best segmental scale sample table of the field of forestry of table 1 ground class
(3) samples selection：Under each ground class optimum segmentation yardstick, various regions class selects its representational sample, sample choosing
Select based on the remote sensing object to be made up of similar pixel completely, while taking into account the remote sensing pair that part is doped with other types pixel
As in the case where the precision of post analysis is ensured, the coverage rate of sample being increased as far as possible.(4) by class statistical sample property value：
To the statistics of attributes table of each ground class；For each ground class, the maximum of each property value, second largest value, minimum value, secondary small is counted
Value.Attribute has NDVI (normalized differential vegetation index), canopy density, boundary index, brightness, length-width ratio, first band light in this example
Spectrum average, shape index.This example first distinguishes forest land and non-forest land by normalized differential vegetation index, in statistical attribute value
When the scope of domain, forest department's sample is divided into two major classes：Forest land and non-forest land.Table 2 below is various regions class boundary index system
Evaluation (numeral is various regions class numbering after the class of various regions, corresponding with following codomain figure)：
BorderIndex (boundary index) statistical value of the various regions class sample of table 2
(5) each property value codomain scope of each ground class is determined：With time maximum and time minimum value conduct of various regions generic attribute
Codomain of the ground class on the attribute.(6) the attribute codomain of different land types on each property value is counted：According to the codomain of all ground class
Range statistics result, is counted by attribute, that is, count the attribute codomain of different land types on each property value, and makes codomain
Figure.Accompanying drawing 2 is BorderIndex (boundary index) attribute codomain figure of various regions class sample, the digitized representation upper table of transverse axis in figure
In corresponding ground class, the longitudinal axis is property value.(7) codomain figure statistical rules：According to codomain figure, for specific ground class, all
The attribute maximum with other ground class differences is selected in attribute, is screened with this property value codomain, reject part interference, weight
Multiple this step, increases attribute number, further rejects interference plot.Until obtaining better effects.This example passes through NDVI first
Value, two major classes are divided into by forest department's sample：(forest land NDVI value minimum values are greater than non-forest land sample NDVI for forest land and non-forest land
It is worth maximum, is made a distinction with this)；Non-forest land is divided into by waters, construction land, arable land and grassplot by NDVI values again
(the NDVI values in waters are negative value, and construction land, arable land and grassplot are on the occasion of and increase successively)；Pass through brightness value area again
The spinney divided in forest land and other forest lands；Again by canopy density (about in standard clear stipulaties Different forest stands it is strongly fragrant
Degree of closing defines value) (high forest and bamboo grove, bamboo grove sample is less in this example, therefore not right for the forest land of distinguishing in other forest lands
The ground class is studied.), opening, without standing tree forest land；Lake and the river water surface are distinguished by length-width ratio again；Pass through again
As shown in Figure 2, the shape index interval of unused land, other lands used and industrial and mineral construction land is all compared in non-forest land ground class
Small, its codomain upper limit is also more much smaller than other ground classes, if taking the upper limit of these three ground classes as constraints, can be built with other
If land used makes a distinction；Unused land, other lands used and industrial and mineral construction land are distinguished by shape index again；In the same way,
Urban and rural residents' construction land and Land Use for Transport Construction are distinguished by first band spectrum average again.(8) ontology is set up：According to
The attribution rule of statistics, sets up the field ontology.The ontology that this example is set up is as shown in Figure 3.
Step S5, different field ontology system constructing：Multi-scale division technology using object-oriented is solid in a width
Determine to construct the imaged object hierarchical structure under different segmentation yardsticks on the remote sensing image of resolution ratio, according to above-mentioned ontology
Modeling procedure, the concept of different field ontology system is mapped in same remote sensing image hierarchical structure.According to step
4, what this example built is the ontology system on forest department's forestry and agricultural sector grassland, ontology concept and multiple dimensioned
The ontology system on the relation of remote sensing image object and forest department's forestry and agricultural sector grassland is shown in accompanying drawing 4.
Step S6, image feature value is calculated：For the ontology of the different field studied, choosing some has representative
Property object, calculate its red wave band average, maximum spectral differences, red band ratio, normalized differential vegetation index, contrast is related
Property, six image feature values such as energy finally draw the codomain of each characteristic value.This example forest department suitable for afforestationly and agricultural
Six image feature value codomain such as table 3 below on the hill covered with grass grass slope of department：
Table 3 suitable for afforestationly and hill covered with grass grass slope six image feature value codomains
Step S7, sets up the common factor of domain object characteristic attribute of the same race：Between the ground class in each field, remote sensing is based on by it
The codomain of feature set, the common factor between calculating.If exist in each characteristic value occuring simultaneously, then calculate between any two
Union, and obtain the percentage that common factor codomain accounts for union codomain.
Table 4 below for forest department suitable for afforestationly and Land_use change department hill covered with grass grass six, slope characteristic attribute common factor situation：
Table 5 be advisable forest land and hill covered with grass grass slope six characteristic attributes occur simultaneously account for union percentage：
Six characteristic attributes of the table 4 suitable for afforestationly with hill covered with grass grass slope occur simultaneously
Table 5 occurs simultaneously with six characteristic attributes on hill covered with grass grass slope and accounts for union percentage suitable for afforestationly
Step S8, sets up the semantic relation between the concept in domain body and between body：There is its codomain on characteristic attribute
Be present semantic association in the imaged object that intersecting or inclusion relation can map out different scale, handed over reference to the attribute of above-mentioned foundation
Collection, sets up the semantic relation between the concept in domain body and between body.Semantic association is divided into direct semantics and associates and indirect
Two kinds of semantic association, specific inference step is as follows:
(1) direct semantics association：There is following semantic reasoning rule:
If it is empty set that 1. a ground classes in A fields are occured simultaneously with the b classes in B fields on certain spectrum characteristic parameter, this can be determined
A, b both ground classes there is no semantic relation；
If 2. class B field b ground class in A fields a ground intersects on all spectrum characteristic parameters, if there is bag between feature set
Containing relation, then hyponymy can be determined that it is；If there is the proportion for accounting for union that occurs simultaneously between feature set less than 10%, can
To determine a, there is no semantic relation between b both ground classes；If exist between feature set occur simultaneously account for the proportion of union more than 10% and
Less than 80%, then a can be determined, be semantic similarity relation between b both ground classes；If existing to occur simultaneously between feature set and accounting for union
Proportion is more than 80%, then be synonymy.
(2) indirect semantic association：According to direct semantics related reasoning rule, if inferring a ground classes and the B fields in A fields
B ground class be synonymy；Then upper strata object c and b the ground class of a ground class are b ground class in hyponymy, and B fields in A fields
Upper strata object d and a ground class be also hyponymy.
Occur simultaneously with reference to the attribute of above-mentioned foundation, set up the semantic relation between the concept in domain body and between body.By
Above-mentioned to understand, existing on six characteristic attributes with the hill covered with grass grass slope of agricultural sector suitable for afforestationly for forest department is occured simultaneously, and is deposited
The feature set of the proportion more than 80% of union is accounted in common factor, it is suitable for afforestation according to the Article 2 semantic reasoning rule that direct semantics is associated
Ground and hill covered with grass grass slope are synonymy.Further according to indirect semantic association：In agriculture field, the upper strata object on hill covered with grass grass slope is natural
Grassland (see accompanying drawing 4), then the natural grasslands of agriculture field and field of forestry is suitable for afforestationly hyponymy.
Ontology modeling of the present invention based on object-oriented image feature is as follows with the specific innovative point of semantic reasoning method：
(1) propose carries out ontology modeling method based on spectrum and shape facility：By multiple dimensioned point of object-oriented
Cut, form and constitute a multi_tier architecture by different remote sensing objects, on this basis, by the spectrum of remote sensing image object and
Shape facility is modeled to different types of remote sensing object.
(2) different field ontology system constituting method is proposed：Using the multi-scale division technology of object-oriented one
The imaged object hierarchical structure under different segmentation yardsticks is constructed on the remote sensing image of width fixed resolution, according to institute of the present invention
The concept of different field ontology system, is mapped to same remote sensing image level knot by the ontology modeling procedure of proposition
In structure.
(3) propose set up in domain body and body between concept between semantic relation correlation rule：Semanteme is closed
Connection is divided into two kinds of direct semantics association and indirect semantic association, and specific inference step is as follows：
(3.1) direct semantics association：There is following semantic reasoning rule：
If it is empty set that 1. a ground classes in A fields are occured simultaneously with the b classes in B fields on certain spectrum characteristic parameter, this can be determined
A, b both ground classes there is no semantic relation；
If 2. class B field b ground class in A fields a ground intersects on all spectrum characteristic parameters, if there is bag between feature set
Containing relation, then hyponymy can be determined that it is；If there is the proportion for accounting for union that occurs simultaneously between feature set less than 10%, can
To determine a, there is no semantic relation between b both ground classes；If exist between feature set occur simultaneously account for the proportion of union more than 10% and
Less than 80%, then a can be determined, be semantic similarity relation between b both ground classes；If existing to occur simultaneously between feature set and accounting for union
Proportion is more than 80%, then be synonymy.
(3.2) indirect semantic association：According to direct semantics related reasoning rule, if a ground classes for inferring A fields are led with B
The b ground class in domain is synonymy；Then upper strata object c and b the ground class of a ground class are b ground in hyponymy, and B fields in A fields
Upper strata object d and a the ground class of class are also hyponymy.
It should be appreciated that for those of ordinary skills, can according to the above description be improved or converted,
And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (10)
1. it is a kind of based on object-oriented image feature ontology modeling with semantic reasoning method, it is characterised in that including with
Lower step：
S1, the high-resolution remote sensing image for obtaining research location, pretreatment and multiple dimensioned point are carried out to high-resolution remote sensing image
Cut, set up imaged object hierarchical structure；
S2, field division is carried out according to imaged object hierarchical structure, and land type division is carried out to each field；According to high-resolution
The remote sensing features value of remote sensing image, including spectrum characteristic parameter and shape facility value, the different land types to each field carry out geography
Ontology Modeling, obtains the ontology of different field；
S3, the ontology to different field, calculate the codomain of the remote sensing features value of each ground class；And calculate the value of different land types
Common factor between domain；If exist between two codomains occuring simultaneously, the union between two codomains is calculated, and common factor accounts for union
Percentage；
S4, the semanteme according to the size occured simultaneously, the semantic association of the ground class set up in field, and the ground class between different field
Association.
2. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 1, its
It is characterised by, the method pre-processed to high-resolution remote sensing image in step S1 includes：Geometric correction, atmospheric correction and sanction
Cut splicing.
3. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 1, its
It is characterised by, the method that in step S2 the different land types in each field are carried out with ontology modeling is:
S21, class determine：According to the criteria for classification in field, different land types are divided；
S22, optimum segmentation size selection：Its corresponding optimum segmentation size is selected to each ground class；
S23, class samples selection：The selectively class sample under the optimum segmentation size of each ground class, the ground class sample is included together
The remote sensing object of class pixel composition and the remote sensing object for being doped with other types pixel, the coverage rate for increasing ground class sample make the later stage
The precision of analysis reaches maximum；
S24, computation attribute statistical form：To each ground class sample, its spectrum and shape Statistics value are calculated, count each property value
Maximum, second largest value, minimum value and sub-minimum, obtain the statistics of attributes table of each ground class sample；
S25, calculating codomain scope：To each property value of each ground class sample, made with the scope between second largest value and sub-minimum
It is codomain of the ground class sample on the attribute；
S26, statistics codomain figure：Codomain range statistics result according to all ground class sample, is counted by attribute, that is, count every
The attribute codomain of different land types on individual property value, and make codomain figure；
S27, codomain figure statistical rules：According to codomain figure, for specific ground class, selected in all properties and other ground classes
Maximum attribute is distinguished, is screened with this property value codomain, reject part interference, repeat this step, increase attribute number,
Further reject interference；
S28, set up ontology：According to the attribution rule of statistics, the ontology in the field is set up.
4. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 3, its
It is characterised by, selects the method for optimum segmentation size to include in step S22：Largest face area method, mean variance method and object function
Method.
5. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 3, its
It is characterised by, the attribute in step S24 includes：Normalized differential vegetation index, canopy density, boundary index, brightness, length-width ratio, first
Band spectrum average and shape index.
6. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 1, its
It is characterised by, after the ontology of different field is obtained in step S2, the ontology of different field is mapped to same distant
In sense image hierarchical structure.
7. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 1, its
It is characterised by, the remote sensing features value in step S3 includes：Red wave band average, maximum spectral differences, red band ratio, normalization are planted
By index, contrast, correlation and energy.
8. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 1, its
It is characterised by, semantic association is set up in step S4 includes direct semantics association and indirect semantic association.
9. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 8, its
It is characterised by, the method for judging direct semantics association is：
If the common factor of two certain remote sensing features values of ground class is empty set, there is no semantic association relation between the two ground classes；
If the codomain of two all remote sensing features values of ground class is intersecting, and there is inclusion relation between codomain, then the two ground
It is hyponymy between class；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is less than 10%,
Then there is no semantic association relation between the two ground classes；
If the codomain of two all remote sensing features values of ground class intersects, and account for the percentage of union in the presence of occuring simultaneously more than 10% and
Then it is semantic similarity relation between the two ground classes less than 80%；
If the codomain of two all remote sensing features values of ground class intersects, and the percentage for accounting for union in the presence of occuring simultaneously is more than 80%,
It is then synonymy between the two ground classes.
10. ontology modeling and semantic reasoning method based on object-oriented image feature according to claim 9, its
It is characterised by, the method for judging indirect semantic association is：
If being synonymy between two ground classes, wherein it is between class, with another ground class the upper strata of any one ground class
Hyponymy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611024360.0A CN106709989B (en) | 2016-11-18 | 2016-11-18 | Geographic ontology modeling and semantic reasoning method based on object-oriented image features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611024360.0A CN106709989B (en) | 2016-11-18 | 2016-11-18 | Geographic ontology modeling and semantic reasoning method based on object-oriented image features |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106709989A true CN106709989A (en) | 2017-05-24 |
| CN106709989B CN106709989B (en) | 2020-04-21 |
Family
ID=58941190
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611024360.0A Active CN106709989B (en) | 2016-11-18 | 2016-11-18 | Geographic ontology modeling and semantic reasoning method based on object-oriented image features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106709989B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107291918A (en) * | 2017-06-28 | 2017-10-24 | 武汉大学 | A kind of visual mining methods of point of interest POI with bit pattern |
| CN108280220A (en) * | 2018-02-08 | 2018-07-13 | 广州地理研究所 | A kind of regional geography space-data integration |
| CN108428220A (en) * | 2018-03-05 | 2018-08-21 | 武汉大学 | Satellite sequence remote sensing image sea island reef region automatic geometric correction method |
| CN113095303A (en) * | 2021-06-04 | 2021-07-09 | 成都数之联科技有限公司 | Model training method, forest land change detection system, forest land change detection device and forest land change detection medium |
| US20210390351A1 (en) * | 2020-06-15 | 2021-12-16 | Toyota Research Institute, Inc. | Systems and methods for improving the classification of objects |
| CN114119575A (en) * | 2021-11-30 | 2022-03-01 | 二十一世纪空间技术应用股份有限公司 | Spatial information change detection method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102279929A (en) * | 2010-06-13 | 2011-12-14 | 中国科学院电子学研究所 | Remote-sensing artificial ground object identifying method based on semantic tree model of object |
| CN102810158A (en) * | 2011-05-31 | 2012-12-05 | 中国科学院电子学研究所 | High-resolution remote sensing target extraction method based on multi-scale semantic model |
| CN102902978A (en) * | 2012-08-31 | 2013-01-30 | 电子科技大学 | Object-oriented high-resolution remote-sensing image classification method |
-
2016
- 2016-11-18 CN CN201611024360.0A patent/CN106709989B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102279929A (en) * | 2010-06-13 | 2011-12-14 | 中国科学院电子学研究所 | Remote-sensing artificial ground object identifying method based on semantic tree model of object |
| CN102810158A (en) * | 2011-05-31 | 2012-12-05 | 中国科学院电子学研究所 | High-resolution remote sensing target extraction method based on multi-scale semantic model |
| CN102902978A (en) * | 2012-08-31 | 2013-01-30 | 电子科技大学 | Object-oriented high-resolution remote-sensing image classification method |
Non-Patent Citations (2)
| Title |
|---|
| 高丽萍: "基于面向对象遥感技术的地理本体推理研究", 《中国优秀硕士学位论文全文数据库基础科学辑》 * |
| 鲁伟等: "基于本体的城市遥感影像语义模型", 《遥感技术与应用》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107291918A (en) * | 2017-06-28 | 2017-10-24 | 武汉大学 | A kind of visual mining methods of point of interest POI with bit pattern |
| CN108280220A (en) * | 2018-02-08 | 2018-07-13 | 广州地理研究所 | A kind of regional geography space-data integration |
| CN108428220A (en) * | 2018-03-05 | 2018-08-21 | 武汉大学 | Satellite sequence remote sensing image sea island reef region automatic geometric correction method |
| CN108428220B (en) * | 2018-03-05 | 2020-12-01 | 武汉大学 | Automatic geometric correction method for ocean island reef area of remote sensing image of geostationary orbit satellite sequence |
| US20210390351A1 (en) * | 2020-06-15 | 2021-12-16 | Toyota Research Institute, Inc. | Systems and methods for improving the classification of objects |
| US11710324B2 (en) * | 2020-06-15 | 2023-07-25 | Toyota Research Institute, Inc. | Systems and methods for improving the classification of objects |
| CN113095303A (en) * | 2021-06-04 | 2021-07-09 | 成都数之联科技有限公司 | Model training method, forest land change detection system, forest land change detection device and forest land change detection medium |
| CN114119575A (en) * | 2021-11-30 | 2022-03-01 | 二十一世纪空间技术应用股份有限公司 | Spatial information change detection method and system |
| CN114119575B (en) * | 2021-11-30 | 2022-07-19 | 二十一世纪空间技术应用股份有限公司 | Spatial information change detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106709989B (en) | 2020-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106709989A (en) | Object-oriented image characteristic-based geographic ontology modeling and semantic reasoning method | |
| Zhao et al. | Mapping watershed-level ecosystem service bundles in the Pearl River Delta, China | |
| CN111709379B (en) | Remote sensing image-based hilly area citrus planting land plot monitoring method and system | |
| Muriithi | Land use and land cover (LULC) changes in semi-arid sub-watersheds of Laikipia and Athi River basins, Kenya, as influenced by expanding intensive commercial horticulture | |
| Teferi et al. | Understanding recent land use and land cover dynamics in the source region of the Upper Blue Nile, Ethiopia: Spatially explicit statistical modeling of systematic transitions | |
| Rallings et al. | Towards multifunctional land use in an agricultural landscape: A trade-off and synergy analysis in the Lower Fraser Valley, Canada | |
| Kiptala et al. | Land use and land cover classification using phenological variability from MODIS vegetation in the Upper Pangani River Basin, Eastern Africa | |
| CN106548141A (en) | A kind of object-oriented farmland information extraction method based on the triangulation network | |
| CN105447274B (en) | A method of seashore wetland drawing being carried out to intermediate resolution remote sensing images using object oriented classification technology | |
| Lisita et al. | Identifying potential areas of Cannabis sativa plantations using object-based image analysis of SPOT-5 satellite data | |
| Gichenje et al. | An analysis of the drivers that affect greening and browning trends in the context of pursuing land degradation-neutrality | |
| CN107194857B (en) | Available land resource analysis system based on high-resolution satellite remote sensing data | |
| Mirzaei et al. | Trans-boundary land cover changes and its influences on water crisis: Case study of the Aras River | |
| Lu et al. | Examining the influencing factors of forest health, its implications on rural revitalization: A case study of five forest farms in Beijing | |
| Esbah* et al. | Changing land cover characteristics of a developing coastal town: a case study of Didim, Turkey | |
| Wang et al. | Wetland mapping in East Asia by two-stage object-based Random Forest and hierarchical decision tree algorithms on Sentinel-1/2 images | |
| Liao et al. | Identifying critical area of ecosystem service supply and demand at different scales based on spatial heterogeneity assessment and SOFM neural network | |
| Badjana et al. | Assessment of land-cover changes in a sub-catchment of the Oti basin (West Africa): A case study of the Kara River basin | |
| Combalicer et al. | Changes in the forest landscape of Mt. Makiling forest reserve, Philippines | |
| Zhang et al. | Semantic urban vegetation modelling based on an extended CityGML description | |
| Dong et al. | Fine mapping of key soil nutrient content using high resolution remote sensing image to support precision agriculture in Northwest China | |
| Nelson et al. | Spatial statistical techniques for aggregating point objects extracted from high spatial resolution remotely sensed imagery | |
| Lu et al. | Mapping the terraces on the Loess Plateau based on a deep learning-based model at 1.89 m resolution | |
| Ri et al. | Health assessment of natural larch forest in arxan guided by forestry remote sensing integrated with canopy feature analysis | |
| Ma et al. | Quantitative Analysis of Spatio-Temporal Evolution of Ecosystem Structure in the Water-Land Interaction Area of Nansi Lake Based on Remote Sensing Technology. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211111 Address after: 315201 5th floor, Qihang building, Ningbo National University Science Park, No. 777, Zhongguan West Road, Zhuangshi street, Zhenhai District, Ningbo City, Zhejiang Province Patentee after: Zhejiang spacetime Zhizi big data Co., Ltd Address before: 430070 No. 122, Luoshi Road, Hongshan District, Wuhan City, Hubei Province Patentee before: Wuhan University of Technology |
|
| TR01 | Transfer of patent right |