Method for processing data structures

• US 20030115175A1
• Filed: 10/15/2002
• Published: 06/19/2003
• Est. Priority Date: 12/14/1999
• Status: Active Grant

• Current Assignee: Definiens AG (Astrazeneca PLC)
• Original Assignee: Definiens AG (Astrazeneca PLC)
• Inventors: Schäpe, Arno, Schmidt, G?uuml;nter, Baatz, Martin
• Granted Patent: US 7,437,004 B2
• Time in Patent Office: Days
• US Class Current: 707/1
• CPC Class Codes: G06T 2207/10032 Satellite or aerial image; Remote sensing
• CPC Class Codes: G06T 2207/30181 Earth observation
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/174 involving the use of two or more images
• CPC Class Codes: G06V 10/422 for representing the structure of the pattern or shape of an object therefor
• CPC Class Codes: G06V 20/13 Satellite images
• CPC Class Codes: Y10S 707/99931 Database or file accessing

Abstract

What is disclosed is a method for processing data structures with the aid of networked semantic units, which comprises the steps of acquiring a data structure; and generating, modifying, deleting and/or storing semantic structure units and/or networking them on the basis of the acquired data structure while using a knowledge base comprised of a network of semantic knowledge units. Herein semantic structure units and/or their network are classified in iterative steps, and based on this classification a specific processing may be activated which modifies a respective semantic structure unit and/or a particular partial network.

Claims

1. A method for processing data structures with the aid of networked semantic units, said method comprising the following steps:

[a] acquiring a data structure; and

[b] generating, analyzing, modifying, deleting and/or storing semantic structure units and/or networking them on the basis of the acquired data structure while using a knowledge base comprised of a network of semantic knowledge units, wherein

semantic structure units and/or the networking thereof are classified in iterative steps, and based on this classification a specific processing may be activated which modifies a respective semantic structure unit and/or a particular partial network.

2. The method of claim 1, characterized in that an acquired data structure consists of a plurality of data structures.

3. The method of claim 1 or 2, characterized in that processing said acquired data structure involves handling, analyzing, reproducing, describing, compressing and/or storing said acquired data structure.

4. The method of any one of the preceding claims, characterized in that semantic structure units and semantic knowledge units are semantic units.

5. The method of any one of the preceding claims, characterized in that a type of said semantic unit is a linking object linking two arbitrary semantic units with each other such that a type and/or an informational content of said linking objects determines the relationship of said two semantic units among each other.

6. The method of claim 5, characterized in that a type of a linking object is determined by a designation of said linking object, by a location and/or an enumeration in a semantic unit to which said linking object is connected.

7. The method of claim 5 or 6, characterized in that a linking object consists only of a logic link, only of a pointer, or only of an unambiguous designation of a semantic unit.

8. The method of any one of claims 5 to 7, characterized in that a particular partial network is defined as all semantic units of a particular type and/or of a particular classification at a particular distance which, starting out from one or a plurality of particular semantic units, are linked therewith either directly or indirectly via other semantic units of a particular type.

9. The method of any one of claims 5 to 8, characterized in that links generated through linking objects are weighted by entering a weighting into an informational content.

10. The method of claim 9, characterized in that a particular partial network is comprised of all semantic units which, starting out from one or a plurality of particular semantic units, are linked therewith either directly or indirectly, and which may be reached by continuous multiplication of the respective weightings determined in the linking objects along the linking path without exceeding or dropping below a predetermined threshold.

11. The method of any one of claims 5 to 10, characterized in that the following features may be made use of for classification of semantic structure units:

features relating to the property of semantic structure units, without the analysis of other semantic structure units or the relationships with them;

features resulting from analysis of a property of the networked environs of a semantic structure unit;

features resulting from a comparison of a property of a semantic structure unit with a corresponding property of its networked environs; and

features resulting from a comparison of a property of a semantic structure unit with a corresponding property of a particular partial network in said entire network of semantic structure units.

12. The method of any one of claims 5 to 11, characterized in that semantic structure units are structure objects, linking objects linking semantic structure units, or networks/partial networks of semantic structure units.

13. The method of claim 12, characterized in that structure objects are coherent in accordance with the topology of said acquired data structure.

14. The method of claim 12 or 13, characterized in that structure objects are each tied in with a particular topological area of said acquired data structure and represent it and/or represent properties thereof, and/or in that structure objects are not or not exclusively tied in with a particular area of said acquired data structure but supplement information not directly contained in said data structure.

15. The method of any one of claims 12 to 14, characterized in that said network of semantic knowledge units comprises the following types of semantic knowledge units:

linking objects linking said semantic knowledge units among each other;

feature objects containing feature descriptions with regard to a particular feature and describing an evaluation algorithm with regard to a particular feature;

analysis objects containing one or more feature objects through linking;

processing objects each containing algorithms whereby those semantic units to which they are linked and/or a particular partial network may be modified, and sequencing controls with regard to these algorithms; and

class objects each representing a class and containing an analysis object and/or a processing object, wherein semantic units are analyzed, allocated to a class and/or modified with the aid of said class objects.

16. The method of claim 15, characterized in that a pertinence and/or a pertinence value of a semantic structure unit is also usable with regard to a class as a property for said particular feature.

17. The method of claim 16, characterized in that a respective semantic structure unit is linked with a class object via a classifying linking object, wherein said classifying linking object either is an analyzing classifying linking object that determines the pertinence or the pertinence value to the class represented by said class object, or is an allocating classifying linking object that allocates a semantic structure unit of the class represented by said class object through the intermediary of said link to said class object.

18. The method of claim 17, characterized in that processing objects are each linked with particular class objects and modify those semantic units linked with a respective class object via an allocating classifying linking object and/or the respective networked environs thereof.

19. The method of any one of claims 15 to 18, characterized in that a type of a linking object is a feature linking object that links a feature object with an analysis object, wherein said feature linking object applies feature descriptions and evaluation algorithms described in said feature objects to semantic units that are linked with a respective analysis object.

20. The method of any one of claims 15 to 19, characterized in that feature objects within an analysis object are subsumed into groups, and/or that sub-groups of feature objects and/or feature objects are subsumed into super-groups in such a way that evaluation results of said evaluation algorithms of said feature objects and/or of said sub-groups are linked with each other within one group via a unified logic into an evaluation result of said super-group.

21. The method of claim 20, characterized in that linking and processing of the evaluation results of said evaluation algorithms of feature objects and/or groups of feature objects within an analysis object are based on fuzzy logic.

22. The method of any one of claims 15 to 21, characterized in that structure objects may also be referred to as constituents of feature objects for a feature evaluation.

23. The method of any one of claims 15 to 22, characterized in that single structure objects are designated via an interface as typical representatives for one respective class, and with the aid of these typical representatives, the remaining structure objects are automatically classified through a Nearest-Neighbour method in a predetermined feature space.

24. The method of any one of claims 15 to 23, characterized in that semantic structure units are subsumed into groups of similar semantic structure units through a cluster analysis in a predetermined feature space for further processing.

25. The method of any one of claims 15 to 24, characterized in that class objects, together with the feature objects pertaining to them, are automatically generated such that they each describe a cluster of similar semantic structure units within a predetermined feature space.

26. The method of any one of claims 15 to 25, characterized in that class objects are classified on the basis of the feature descriptions of the feature objects contained in them and/or on the basis of the properties of the structure units subsumed in them, and in this manner may be subsumed in super-class objects.

27. The method of any one of claims 19 to 26, characterized in that a weighting with regard to a feature linking object is used to determine a degree in which an evaluation result of said evaluation algorithm contained in a feature object is taken into consideration in an overall evaluation of a plurality of feature objects.

28. The method of any one of claims 15 to 27, characterized in that one type of a linking object is a hierarchical similarity linking object that hierarchically links class objects and transfers analysis objects and/or feature objects and/or processing objects from super-class objects to sub-class objects by heredity.

29. The method of claim 28, characterized in that feature objects are linked with a class object via analysis objects.

30. The method of claim 29, characterized in that heredity depends on it being recorded in an informational content of a respective analysis object and/or processing object whether or not it is transferred by heredity.

31. The method of claim 28, characterized in that feature objects are linked directly with a class object.

32. The method of claim 31, characterized in that heredity depends on it being recorded in an informational content of a respective feature object and/or processing object whether or not it is transferred by heredity.

33. The method of any one of claims 15 to 32, characterized in that a special form of a structure object is a datapoint object that represents the respective topological location of a single datum of said acquired data structure and said respective single datum.

34. The method of claim 33, characterized in that by means of a hierarchy of structure units contained in said network of semantic units, wherein structure objects representing a smaller area of said acquired data structure are linked in a topologically unambiguous manner, in the sense of an embedding hierarchy, as sub-structure objects with larger super-structure objects representing an area of said acquired data structure having an equal or greater size, said acquired data structure may be represented through structure objects in various degrees of resolution at a same time, with a bottommost plane of such a hierarchy consisting of datapoint objects.

35. The method of claim 34, characterized in that modifications of said networked environs brought about through step [b] are:

a deletion of a semantic structure unit;

a linking of a semantic structure unit with another semantic structure unit;

a modification of an informational content of a semantic structure unit;

a fusioning of two or more already existing structure objects;

a generating of a new structure object as a super-structure object of one or several already existing structure objects and a linking thereof with these;

an exclusion of a sub-structure object from its previous super-structure object by obliterating the linking between them, with a data structure area pertaining to said super-structure object accordingly being reduced in size;

a linking of a structure object as a super-structure object with another structure object as a sub-structure object, with a data structure area pertaining to said super-structure object correspondingly increasing in size;

a dividing of a structure object into two or more structure objects on the basis of said sub-structure object; and

an allocating linking of a semantic unit with a class object, and accordingly an allocation of said semantic unit to the class represented by said class object.

36. The method of claim 34 or 35, characterized in that algorithms contained in a processing object concern all local operations with regard to a respective semantic structure unit and/or the networked environs thereof within the hierarchical network of semantic structure units.

37. The method of any one of claims 34 to 36, characterized in that in said hierarchy of structure units a data structure area represented by each structure object is unambiguously composed of those data structure areas represented by the respective sub-structure objects of said respective structure object.

38. The method of any one of claims 34 to 37, characterized in that said hierarchical network of semantic structure units has a unified hierarchical depth in all locations, with respective entire planes of structure objects being processed, generated, or deleted in this hierarchical network.

39. The method of any one of claims 34 to 38, characterized in that one type of a linking object is a hierarchical linking object that hierarchically interconnects structure objects as super-structure objects and sub-structure objects.

40. The method of claim 39, characterized in that upon obliteration of a hierarchical link by means of the hierarchical linking object of a sub-structure object with its previous super-structure object, a property of coherence of said super-structure object with regard to a topology of said acquired data structure is examined, and in a case of non-coherence said super-structure object is divided into a plurality of structure objects each representing one coherent data structure area of the previous super-structure object brought about through removing said sub-structure object.

41. The method of any one of claims 39 or 40, characterized in that a process of boundary optimization is performed such that all sub-structure objects located at the boundary of a super-structure object are analyzed to the effect whether, in accordance with a criterion to be predetermined, they better match a structure object neighboring said super-structure object than the previous super-structure object, and in the latter case a hierarchical link with the previous super-structure object thereof via said hierarchical linking object is obliterated, and a new hierarchical link with said better-matching super-structure object is generated with the aid of said hierarchical linking object.

42. The method of any one of claims 15 to 41, characterized in that class objects and/or partial networks of class objects within a processing object and/or a feature object in turn are constituents of said algorithm or evaluation algorithm.

43. The method of any one of claims 15 to 42, characterized in that class objects transfer particular attributes described in them to semantic units allocatingly linked with them.

44. The method of any one of the preceding claims, characterized in that said acquired data structure is a digital image information.

45. The method of claim 44, characterized in that said image information exhibits any imaging, any dimensionality, any number of channels and/or temporal sequences.

46. Data carrier usable in a computer system and having a program code recorded on it, wherein under a condition of said data carrier being inserted into said computer system and an activated condition of said program code on said data carrier, said computer system performs a method of any one of the preceding claims.