RU36907U1 - AUTOMATED KNOWLEDGE BASE SYSTEM FOR PROCESSING, ANALYSIS AND RECOGNITION OF IMAGES - Google Patents

Description

AUTOMATED KNOWLEDGE BASE SYSTEM

ON PROCESSING, ANALYSIS AND RECOGNITION OF IMAGES

The utility model relates to computing and computer science and can be used in automated systems for analysis and recognition of tone, binary and color (photo) images.

The tasks of image analysis and recognition are widespread in industry, medicine, in the automation of scientific research, in ecology, geology, forecasting natural disasters and emergencies, military affairs and other branches of knowledge (see, for example, RU 2163394 C2, 03/29/1999; RU 2178915 C2, 01/27/2002; EP 11844659 A1, 03/06/2000). To solve the problems of processing, analysis and recognition of images, algorithmic-software systems have been created that use modern computing tools (Yu.I. Zhuravlev, IB Gurevich “Pattern Recognition and Image Analysis // Artificial Intelligence: In 3 books, book 2 Models and Methods: Reference Book, Moscow: Radio and Communications, 1990. P. 149 - 191; Yu.I. Zhuravlev, IB Gurevich, Methods and Means of Converting and Processing Information in Image Recognition and Image Analysis Problems // Parallel information processing: V 5 t., T. 5: Problem-oriented s and specialized means of information processing. - Siev: Naukova Dumka, 1990. - S. 218-318); I.B. Gurevich, Yu.I. Zhuravlev, D.M. Murashov and others. “The system of automation of scientific research in the field of analysis and understanding of images based on the accumulation and use of knowledge. Part 1 // Autometry. - 1999. - No. 6. - S. 23-50).

Known expert image processing system (OCAPI) (V. Clement, M. Thonnat, J. Van Den Elst. Supervision of Perception Tasks for Autonomous Systems: The OCAPI Approach, INRIA Rapports de Recherche N2000, 1993, 28 p). The system contains a module, which, together with the interface driver and the interface of the package of routines, can function as an integrated environment designed for image processing. The module contains knowledge on how to provide the following features in the general context of managing an external software system.

IPC7: G06F 17/30, G06T 1/00 h 2. Next, divide the sub-goals into steps that correspond to the basic methods and algorithms of a specific application area. Conceptually, these steps are at the same level as the routines of the software package in question. The authors consider determining the presence of at least three hierarchical levels (goals, subgoals, steps) to have an intermediate level of subtasks between the global goal and subprograms. 3. To allow the solution of two or more related tasks without duplication. Any subtask needed for more than one of the tasks will be solved only once. 4. To allow an iterative control process (for example, in image processing problems, smoothing may require a number of iterations). 5. It is allowed to return to the preliminaries, their steps in case of unsatisfactory results. In this system, this general problem-solving strategy is implemented in approximately 60 rules, some of which use a set of 30 procedures in Prolog. The principles of building knowledge bases for pattern recognition systems using computer technologies, including training, are known (US 5159667, 10.27.1992; US 5748850, 05/05/1998; US 6577757, Bl, 06/10/2003), but they solve particular problems filling the bases with only new reference images. A known method and system for constructing image processing means (WO 01/67389 A1, G06T 1/00, 09/13/2001, the closest analogue), which includes a knowledge base. This system is a multifunctional image analysis tool designed for application development, research and training. The knowledge base includes: -the core of the system, which consists of modules that invoke operators that control input / output operations and translate data objects into the internal language of the system; - data management tools that are responsible for managing the main memory (optimized for image processing), as well as for creating, managing and destroying graphic objects and tuples; - Algorithmic filling tools - a library that includes more than 750 operators for image analysis and data visualization;

h debugging tools; - a knowledge base of operators, which contains reference information for each operator of the system (names and number of operators, parameter types, proposed parameter values). The disadvantage of the above knowledge bases is that none of them poses and solves the problem of image analysis and recognition based on the automation of information extraction from images. At the same time, building a knowledge base on image analysis and recognition with automation of extracting information from images, accumulating and using knowledge on image analysis and recognition, can significantly increase the efficiency of solving problems in this area. The objective of the present invention is the construction of a complex of technical means of data processing, which provides the automated formation and replenishment of a knowledge base for analysis and recognition of images in the form of a search-oriented set of information arrays - task blocks - based on various criteria. Each of these blocks is a structured description of the problem, formed using a hierarchical classification of tasks of analysis, processing and recognition of images, and a thesaurus for analysis, processing and recognition of images. The technical result formulated in the task of the present invention is achieved by the fact that the knowledge base for image analysis and recognition contains a central control unit for the processing and data transfer processes, a user interface, an external communication interface and an integrated subsystem of organized knowledge storage. The complex subsystem of organized knowledge storage includes a task module connected to a common management information highway, a module for supporting planning, management and decision-making processes, an algorithm module, an intermediate module for solving problems, a reference and test image module, a module for extracting, structuring and recording new knowledge and module of the thesaurus, glossary, bibliography and reference. The first input-output of the central data processing and transmission control unit is connected to the common information-control highway, the second input-output of which is connected to the user interface, and the third to the internal input-output of the external communication interface. The external input-output of the external communication interface is the input-output of the knowledge base for analysis and recognition of images with the ability to connect to the image analysis and processing system. The central control unit for the processing and data transmission contains interconnected search and control subsystems. The knowledge base can also be characterized by the fact that the modules, which are part of the integrated subsystem of organized knowledge storage, are made in the form of autonomous relational databases. The knowledge base can also be characterized in that the autonomous relational database contains at least one memory block oriented to storing spreadsheets, a specialized controller, and a communication unit with a common information-control highway, the first input-output of which is the input-output of the module, and the second input-output is connected to the general input-output of the controller, the information and control inputs and outputs of which are associated with the information and control inputs and outputs of the memory block, respectively. The knowledge base may also be characterized in that the autonomous relational database contains; at least one memory block oriented to storing spreadsheets, a specialized controller, a unit for checking the consistency of input knowledge, and a communication unit with a common information and control line, the first input-output of which is the input-output of the module, and the second input-output is connected to a common controller input-output, the information and control inputs and outputs of which are associated with the information and control inputs and outputs of the memory unit and the unit for checking the consistency of input knowledge, respectively. The invention is illustrated in the drawings, where: in FIG. 1 is a block diagram of a patentable knowledge base; In FIG. 2, 3 is a block diagram of an autonomous relational database; in FIG. 4 - algorithm for finding solutions in the knowledge base; in FIG. 5 - an algorithm for planning a solution to a problem. The knowledge base for image analysis and recognition (see Fig. 1) contains a central unit 10 for controlling the processes of data processing and transmission, at least one user interface 12, an external communication interface 14, and an integrated subsystem 20 of organized knowledge storage. The user interface 12 is intended for user interaction with the knowledge base and is used for input, output and visualization of information. .

. Subsystem 20 includes connected to a common management information highway 22: task module 24, module 26 for supporting planning, management and decision-making processes, module 28 for algorithms, module 30 for intermediate results of problem solving, module 32 for reference and test images, module 34 for thesaurus, glossary , bibliography and reference, and module 36 extraction, structuring and recording of new knowledge. The first input-output of the central control subsystem 10 is connected to the common control information highway 22, the second input-output 52 of which is connected to the user interface 12. The third input-output 54 is connected to the internal input-output of the external communication interface 14, the external input-output 56 of which is input-output of the knowledge base on image analysis and recognition. This input-output 56 of interface 14 is intended for connection to image analysis and processing systems and exchanges data, in particular, it can be connected to the Black Square system according to the application RU 2003107587, 03.20.2003 or patent RU 30206 U1, 03.20.2003. The central unit 10 for controlling the processes of data processing and transmission contains interconnected search 102 and control 104 subsystems. Module 24 of tasks, module 26 of support for planning, management and decision-making processes, module 28 of algorithms, module 30 of intermediate results of solving problems, module 32 of reference and test images and module 34 of the thesaurus, glossary, bibliography and reference book are implemented as autonomous relational databases 240- 340 data (see FIG. 2). Each of the autonomous relational databases 200 of the modules 24-32 contains one or more memory blocks 202 oriented to storing spreadsheets, a specialized controller 203 and a communication unit 204 with a common information and control highway 22. The first input-output 205 of the communication unit 204 is an input -output of the module, and the second input-output 206 is connected to a common input-output of the controller 203. The information and control inputs-outputs 207 of the controller 203 are connected respectively with the information control inputs-outputs 207 of the memory unit 202. The module 36 for extracting, structuring and recording new knowledge is also implemented in the form of an autonomous relational base 360, the structure of which is similar to that described above, is shown in FIG. 3. It contains a memory block 362 oriented to storing spreadsheets, a specialized controller 363, a communication block 364 with a common information and control highway 22, and a block 368 for checking the consistency of input knowledge. The first input-output 365 of the communication unit 364 x is the input-output of the module 360, and the second input-output 366 is connected to a common input-output of the controller 363. The information and control inputs and outputs 367, 369 of the controller 363 are connected respectively with the information and control inputs - outputs of blocks 362 and 368. Module 24 of the tasks contains knowledge about classifications of tasks, knowledge of typical and solved problems in the form of structured descriptions, task blocks. The module 26 for supporting planning, management and decision-making processes contains general information about solving problems using computer software, general knowledge on processing, analysis and recognition of images, knowledge about the application area, as well as decision rules of various types corresponding to the selected model for solving problems. Algorithm module 28 contains knowledge about the image processing library used: information about the algorithms, their input parameters, results and methods for evaluating the quality of the algorithms. Module 30 of the intermediate results of problem solving serves to save and subsequently use the results of processing and recognition algorithms. Module 32 reference and test images is used for testing algorithms and training. Module 34 contains a thesaurus, glossary, bibliography, and reference. The thesaurus and the glossary are used in the preparation of task descriptions, and are also the semantic base of the model used to solve problems. This module 34 also functions as a Help user directory. The module 36 for extracting, structuring, and recording new knowledge is intended for introducing new task descriptions into the knowledge base, automated detection of dependencies, characteristics, and formalized expression of image properties, as well as checking the consistency of input knowledge with existing in the database. Each of the modules consists of one or a group of knowledge base tables, a control subsystem (table controller) and can be connected (if it is functionally necessary) with the corresponding user interface. The table controller processes and exchanges data between the knowledge base tables and the control subsystem.

by user. These procedures include: finding a solution to a typical or similar problem among the knowledge in the database, as well as planning a solution. Both procedures are discussed in detail below.

A) Search for a solution to a typical or similar problem in the analysis, processing or recognition of images among the available knowledge base. The algorithm is shown in FIG. 4.

Ethan 1. The user enters the source data. The system analyzes the presented image (s) and extracts useful information about this image, which will be used in setting the problem and finding a solution (image format, number of gradations of brightness, etc. signs).

Stage 2. The user through the user interface 12 creates an object corresponding to the formulation of a specific task, for which the maximum possible amount of data about the task is entered. The information received in the 1st stage is taken into account.

Stage 3. Based on the created object, a complex request is generated. The query conditions take into account the properties of the elements of the problem statement object. For the properties of elements that reflect semantic content, descriptors are preliminarily selected from the thesaurus module 34, which are included in the conditional part of the request.

Stage 4. The query result is a set of descriptions of problems (solutions) containing algorithms (operators) of the solution. Descriptions are ranked according to the selected proximity measure.

Step 5. The algorithm corresponding to the first of the ranked descriptions is selected.

Ethan 6. The algorithm is launched.

Step 7. The resulting sample of task descriptions is analyzed. If the selection does not satisfy the user (the set is empty or there are too many descriptions), then changes are made to the problem statement and a new request is made.

B) The procedure for planning a solution to a problem includes the steps shown in FIG. 5:

Stage 1. In the search subsystem, according to the found description of the typical task (see the procedure under item A), a query is generated to search for descriptions of (typical) subtasks. h algorithms sends a request. Stage 3. For each of the typical subtasks, a request is generated to search for descriptions of the constituent subtasks and other components. Stage 4. Based on the results of a series of queries, a scheme (plan) for solving the problem is formed and displayed on the user interface 12 (on the monitor screen in the editor window). The control subsystem 104 monitors the algorithms included in the solution scheme for compatibility by input and output data types, and also assigns the initial values of the algorithm parameters. Stage 5. The user adjusts the generated decision scheme and the parameters of the algorithms. As a result, a program is compiled compiled in macro language of the image processing and analysis system. Step 6. The user compiles and runs the program for execution. Step 7. The user evaluates the results of solving the problem, if necessary, adjusts the solution scheme and parameters of the algorithms. Based on the results of the assessment, the user introduces new task descriptions into the knowledge base and checks the consistency of the entered knowledge already in the database. The system includes, in particular, the following basic means for image processing: affine transforms, Fourier transforms, Gabor transforms, wavelet transforms; matrix filtering, embedding brightness transitions and contours, segmentation methods, morphological operations, texture analysis methods, factor analysis algorithms, recognition and decision-making algorithms. Using plug-in technology, new algorithms for processing, converting, analyzing and recognizing images, visualization tools, and advanced database search tools can be added to the system during operation. This is done without changing the basis of the system (control connections and architecture) while preserving the already accumulated knowledge in the databases (their number can also be increased). For example, a successful sequence of image analysis algorithms can be added as a new algorithm with a description indicating recommended areas of competence. The patented object is implemented in the Black Square. Version 1.1, designed to automate scientific research in the field of processing, analysis and recognition of images based on PC.

Claims (4)

1. An automated knowledge base system for analysis and recognition of images, containing a central control unit for the processing and data transfer processes, a user interface, an external communication interface and a comprehensive subsystem of organized knowledge storage, characterized in that the integrated subsystem of organized knowledge storage includes those connected to a common information and management highway module of tasks, module of planning, management and decision-making processes support, algorithm module, module p the interim results of solving problems, a module of reference and test images, a module for extracting, structuring and recording new knowledge and a module for a thesaurus, glossary, bibliography and reference, the first input-output of the central control unit for processing and transmitting data is connected to the common information and control line, the second the input-output of which is connected with the user interface, and the third - with the internal input-output of the external communication interface, the external input-output of which is the input-output of the knowledge base for nalysis and pattern recognition can be connected to an image analysis and processing system, said central processing unit processes the control and data retrieval comprising interlocking and control subsystem. 2. The system according to claim 1, characterized in that the modules that make up the integrated subsystem of organized knowledge storage are made in the form of autonomous relational databases. 3. The system according to claim 1 or 2, characterized in that the autonomous relational database contains at least one memory block oriented to the storage of spreadsheets, a specialized controller and a communication unit with a common information-control highway, the first input-output which is the input-output of the module, and the second input-output is connected to a common input-output of the controller, the information and control inputs and outputs of which are associated with the information and control inputs and outputs of the memory block, respectively. 4. The system according to claim 1 or 2, characterized in that the autonomous relational database contains at least one memory block oriented to the storage of spreadsheets, a specialized controller, a unit for checking the consistency of input knowledge, and a communication unit with a common information-management a highway, the first input-output of which is the input-output of the module, and the second input-output is connected to the general input-output of the controller, the information and control inputs and outputs of which are connected respectively with the information and control their inputs and outputs of the memory unit and the unit for checking the consistency of input knowledge.