CN116501304B - Cloud-based model creation method and device for remote sensing product production algorithm - Google Patents

Abstract

The invention provides a cloud-based model creation method and device for a remote sensing product production algorithm. Firstly, defining an S model, wherein the S model comprises one or more of 7 entities; secondly, carrying out algorithm standardized semantic expression, granularity division and semantic assembly on a remote sensing algorithm comprising a basic commonality algorithm and a model algorithm by using an S model; thirdly, carrying out semantic construction and assembly of an algorithm knowledge packet on the remote sensing algorithm by utilizing the proposed S model to form a physical entity of the remote sensing algorithm knowledge packet file, and supporting algorithm and data migration in a cloud environment; and finally, providing a construction device of the algorithm knowledge package file, creating, verifying, assembling, registering, injecting and multiplexing the algorithm knowledge package through the device and the system, realizing complex calculation modeling, maintaining the integrity of process input and process output, providing an interpretation and calculation environment, improving the portability of an algorithm module, and improving the efficiency of understanding and expressing the complex remote sensing calculation task process.

Description

Cloud-based model creation method and device for remote sensing product production algorithm

Technical Field

The invention belongs to the technical field of remote sensing algorithm treatment, and particularly relates to a cloud-based model creation method and device for a remote sensing product production algorithm.

Background

The mass remote sensing products produced by global satellite remote sensing play a great role in the aspects of scientific research of the service earth system, national strategy and the like. In order to better play the functions and values of the remote sensing products serving the national economy and society, the production efficiency and quality of the remote sensing products are urgently required to be further improved.

The remote sensing product production algorithm comprises two algorithms, namely a basic commonality algorithm such as radiation correction, geometric correction, image mosaic, wave band registration, projection transformation, pixel scale conversion, wave band calculation and the like, and a complex model algorithm integrating various basic algorithms and a physical model. The calculation flow of the remote sensing product production algorithm is quite complex, and a unified model is not used for uniformly describing, packaging and executing the remote sensing product production algorithm at present.

For a long time, under the continuous efforts of remote sensing product production, application departments and vast remote sensing users, a large number of remote sensing product production algorithms with application value are generated, the algorithms are often related to specific satellites and loads and are realized by using IDL, python, C ++, fortran, C, JAVA and other different programming languages, the operation environment is various and complex, the shareability, portability and reusability of the algorithms are poor, the sharing and the propagation of algorithms and models in the remote sensing field are not facilitated, and great challenges are caused to the business operation of the remote sensing product production. Therefore, how to effectively treat the remote sensing algorithm becomes a key technical problem to be solved in the field of remote sensing application.

The existing remote sensing algorithm treatment has the following problems:

(1) Algorithm interface normalization miss

The development languages of various remote sensing algorithms and models are different, the interface forms and the calling modes are different, the operation environments of the algorithms are various, and unified algorithm description standards are lacked.

(2) The granularity division of the algorithm is unreasonable

The remote sensing algorithm is often realized according to application requirements, part of the algorithm functions are single, part of the large-scale algorithm comprises a large number of repeated common functions, the common functions are often realized by different users through different programming languages, the performance difference is very large, on one hand, the stability of the algorithm is influenced, and on the other hand, repeated codes of similar functions are caused.

(3) Algorithm modeling lacks semantic expression

The large algorithm can be realized by combining a plurality of fine-grained algorithms by means of a calculation flow modeling tool, but due to the fact that the existing visual modeling tool lacks semantic organization and semantic assembly capabilities, on one hand, algorithm modeling efficiency is low, and on the other hand, complex algorithm models are difficult to understand and reuse, so that a knowledge barrier is formed.

(4) Low computational efficiency

In addition, the large-scale remote sensing algorithm lacks good semantic construction, so that migration capability of the algorithm in a cloud side environment is further limited, and therefore calculation tasks are difficult to be executed in parallel by effectively cooperating with a plurality of calculation nodes, and the calculation efficiency of the large-scale remote sensing algorithm is low.

Disclosure of Invention

In order to solve the problems, the invention provides a cloud-based model creation method and device for a remote sensing product production algorithm. The model creation method of the remote sensing product production algorithm based on the cloud comprises the following steps of:

step one, defining and describing one or more of the following entities of the remote sensing product production algorithm to be modeled according to the structural logic of the remote sensing product production algorithm to be modeled by combining an S model creation method: the method comprises an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator and an algorithm execution entity;

step two, performing up-down Wen Yuyi inference and grammar compliance check on the entity and the semantic relation thereof defined and described in the step one, so as to ensure that the entity meets the S model creation requirement;

step three, carrying out semantic assembly on the entity checked to be compliant in the step two to form an algorithm knowledge packet specification file, constructing an internal directory structure of the algorithm knowledge packet specification file and filling corresponding contents;

uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database, registering and publishing to form an algorithm knowledge package operable instance which can be globally found by a cloud S model interpreter, and searching and downloading the algorithm knowledge package operable instance and a calculation process contained in the algorithm knowledge package operable instance by the cloud S model interpreter through a semantic identifier of the algorithm knowledge package;

Step five, the input data to be processed and the operation configuration information are injected into an algorithm knowledge packet specification file to generate an algorithm knowledge packet instance file, wherein the input data is remote sensing data;

and step six, the cloud S model interpreter reads and analyzes the related definition of the algorithm knowledge package in the algorithm knowledge package instance file, creates an algorithm knowledge package instance and a calculation process instance according to the acquired related data, binds data parameters and drives the execution.

In the first step, according to the structural logic of the remote sensing product production algorithm to be modeled, firstly, creating the top-level algorithm knowledge package entity and a prototype definition file thereof, and filling an algorithm knowledge package semantic identifier, semantic description and definition algorithm knowledge package global variables in the algorithm knowledge package prototype definition file; secondly, one or more computing process entities and corresponding computing process definition files are created, and the computing process definition files are filled with internal semantic identifiers, semantic descriptions and input parameters and output parameters for defining the computing process; next, one or more computing step entities are created in the computing process definition file, including: filling global semantic identifiers of the referenced space operators in the calculating step entity; filling out each input parameter of the referenced space operator and the corresponding real parameter value of the runtime, wherein the space operator comprises a basic operator and a model operator.

Further, the second step specifically includes the following steps:

(1) checking compliance of an algorithmic knowledge package entity, comprising: checking whether the algorithm knowledge packet semantic identifiers have conflict or not, and ensuring the uniqueness of the algorithm knowledge packet semantic identifiers; checking whether the global semantic identifier in the spatial operator of the algorithm knowledge package is consistent with the internal semantic identifier of the corresponding computing process entity; checking whether the name, type and variable value of the global variable of the algorithm knowledge packet accord with an S model creation method;

(2) checking compliance of a computing process entity, comprising: checking whether the semantic identifiers in the calculation process have conflict or not, and ensuring the uniqueness of the semantic identifiers in the algorithm knowledge packet; checking whether the name, type, default value and semantic description of each parameter in an input parameter list and an output parameter list defined by a computing process entity accord with an S model creation method or not; checking the reference relation of input and output parameters among all the calculation steps defined in the entity of the calculation process, and deducing whether a dependency relation exists among all the calculation steps and whether a circular dependency problem exists among the steps;

(3) checking compliance of individual computing step entities contained by a computing process entity, comprising: checking whether a global semantic identifier of the spatial operator referenced by the calculating step exists; acquiring prototype definitions of the space operators according to global semantic identifiers of the referenced space operators; checking whether the name of each parameter in the input parameter list and the output parameter list of the calculating step is consistent with the names of the input parameters and the output parameters of the space operator according to the definition of the prototype of the space operator; checking whether the number of parameters in the input parameter list and the output parameter list in the calculation step is consistent with the number of input and output parameters of the space operator according to the definition of the prototype of the space operator; checking whether parameters in an input parameter list and an output parameter list of the calculating step exist in the definition of the space operator prototype according to the definition of the space operator prototype, and finding out the missing parameters of the entity of the calculating step; checking whether the real parameter value of each parameter in the input parameter list of the calculation step is valid or not according to the definition of the space operator prototype, wherein the real parameter value of the calculation step is derived from the output real parameter value, the direct quantity, the expression variable of other calculation steps, the input parameter of the calculation process or the global variable defined in the algorithm knowledge packet; according to the definition of the space operator prototype, checking whether the value of each file type output parameter in the output parameter list of the calculating step is valid or not.

Further, the third step specifically includes the following steps:

(1) copying the algorithm knowledge package prototype definition file created in the first step to a root directory of the algorithm knowledge package compression file, and registering an internal directory structure of the algorithm knowledge package compression file in the algorithm knowledge package prototype definition file, wherein the directory structure comprises: computing a process catalog, an input data catalog, an output result catalog, an algorithm program catalog and an explanation document catalog;

(2) establishing a computing process catalog under the root catalog of the algorithm knowledge package, wherein the catalog is used for storing all computing process definition files defined in the algorithm knowledge package;

(3) an input data catalog is established under the root catalog of the algorithm knowledge packet, and the catalog is used for storing various input data required by the calculation process in the algorithm knowledge packet in operation;

(4) establishing an output result catalog under the root catalog of the algorithm knowledge packet, wherein the catalog is used for storing various output data generated in the calculation process of the algorithm knowledge packet and the space operator in the operation process;

(5) an algorithm program catalog is contained under the root catalog of the algorithm knowledge package, and the catalog is used for storing executable algorithm programs and interface files thereof;

(6) a description document catalog is contained under the root catalog of the algorithm knowledge package and is used for storing description documents, instruction manuals and video courses of the knowledge package.

Further, the fourth step specifically includes the following steps:

(1) uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database, taking the semantic identifier of the algorithm knowledge package as a key, taking the specification file of the algorithm knowledge package as a value, and establishing an algorithm knowledge package and a space operator registry according to semantic description information of the algorithm knowledge package;

(2) when the cloud S model interpreter runs the algorithm knowledge package, if the space operator is found to be absent, the specification file of the algorithm knowledge package corresponding to the required space operator is automatically downloaded from the cloud knowledge package database according to the global semantic identifier of the referenced space operator and is cached in the S model interpreter.

Further, the fifth step specifically includes the following steps:

(1) selecting an algorithm knowledge packet specification file to be operated and input data to be processed, establishing a global variable of the algorithm knowledge packet and a mapping relation between each shape parameter of input parameters and each real parameter of the input data in each calculation process, and generating operation configuration information of the algorithm knowledge packet;

(2) reading an algorithm knowledge packet prototype definition file under a root directory of a compressed file of an algorithm knowledge packet specification file, writing algorithm knowledge packet operation configuration information into the algorithm knowledge packet prototype definition file, storing input data into an input data directory of the knowledge packet specification file, and generating an algorithm knowledge packet instance file.

Further, the sixth step specifically includes the following steps:

(1) the cloud S model interpreter decompresses the algorithm knowledge package compression file, reads and analyzes the algorithm knowledge package prototype definition file in the algorithm knowledge package instance, and extracts the algorithm knowledge package content organization data, the operation configuration data and the global variables;

(2) the cloud S model interpreter organizes data according to the content of the algorithm knowledge packet, sequentially reads the calculation process definition files from the calculation process catalogue, and creates a corresponding calculation process instance and an internal calculation step and a space operator of each calculation process definition file;

(3) the cloud S model interpreter combines the operation configuration data and the global variables to complete the binding of various parameters of the input parameters and various real parameters of the input data of the computing process instance;

(4) the cloud S model interpreter traverses the to-be-operated computing process examples in the algorithm knowledge packet example file in sequence, traverses the computing steps in each computing process example, groups and sequences the computing steps according to the dependency relationship among the computing steps, and forms a series of computing step sets which can be scheduled and executed in parallel;

(5) the cloud S model interpreter sequentially traverses the calculation step sets, creates an execution thread for each calculation step set and invokes an interface program of a space operator referenced by the calculation step in the thread;

(6) After the operation of the algorithm knowledge package instance is finished, the cloud S model interpreter accommodates output results of all computing process instances and stores the output results into an output result catalog of the algorithm knowledge package instance file.

Meanwhile, the invention also provides a model creation device of the remote sensing product production algorithm based on the cloud, which comprises the following components:

(1) creating a module; according to the structural logic of the remote sensing product production algorithm to be modeled, defining and describing one or more of the following entities of the remote sensing product production algorithm to be modeled by combining an S model creation method: the method comprises an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator and an algorithm execution entity;

(2) the verification module is used for carrying out up-down Wen Yuyi inference and grammar compliance check on the defined and described entities and semantic relations among the entities, so as to ensure that the entities meet the S model creation requirement;

(3) the assembly module is used for carrying out semantic assembly on the core entity for checking compliance to form an algorithm knowledge packet specification file, constructing an internal directory structure of the algorithm knowledge packet file and filling corresponding contents;

(4) the registration module is used for uploading the specification file of the algorithm knowledge packet to a cloud algorithm knowledge packet database, registering and publishing the specification file of the algorithm knowledge packet to form an algorithm knowledge packet operable instance which can be globally found by a cloud S model interpreter, and searching and downloading the algorithm knowledge packet operable instance and a calculation process contained in the algorithm knowledge packet operable instance by the cloud S model interpreter through a semantic identifier of the algorithm knowledge packet;

(5) The injection module is used for injecting the input data to be processed and the operation configuration information into the algorithm knowledge packet specification file to generate an algorithm knowledge packet instance file, wherein the input data is remote sensing data;

(6) the interpretation module reads and analyzes the related definition of the algorithm knowledge package in the algorithm knowledge package instance file by the cloud S model interpreter, creates an algorithm knowledge package instance and a calculation process instance according to the acquired related data, binds data parameters and drives the execution.

The invention has the following beneficial effects:

(1) The S model provided by the invention utilizes a knowledge network formed by semantic identifications of 7 core grammar construction units such as an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator, an algorithm execution entity and the like and association rules thereof, equivalently represents static and dynamic characteristics of an algorithm processing flow, combines reasoning, assumption, approximation, boundary conditions and the like of algorithm model parameters, realizes modeling of a complex calculation process based on semantic knowledge, keeps the integrity of process input and output, provides an interpretation and calculation environment for design, representation and driving of procedural knowledge, and improves the efficiency of understanding and expressing the complex remote sensing calculation task process. And the commonality function can be realized by the same calculation process entity or space operator and basic operator, so that repeated coding is avoided, universality is provided, and a unified standardized interface is provided.

(2) The remote sensing algorithm knowledge package provided by the invention can effectively support granularity division, task assembly and interpretation driving of the remote sensing algorithm, follows the principle of 'approaching data and moving calculation', forms a series of relatively independent and highly cohesive remote sensing algorithm knowledge packages with context semantic perception capability according to context semantic segmentation and reconstruction, registers and publishes the remote sensing algorithm knowledge package specification file to a cloud algorithm knowledge package database to form a knowledge package which can be globally found by a cloud S model interpreter, forms a sharable calculation process and a spatial operator or a basic operator, and can be retrieved and downloaded by a user through a semantic identifier to the existing calculation process and the spatial operator or the basic operator so as to be reused. Meanwhile, the S model interpreter of the cloud end adopts a pipeline arrangement technology to map a remote sensing algorithm knowledge packet set into a movable knowledge packet stream capable of being scheduled and executed in parallel, so that the cooperative execution of remote sensing product production tasks in the cloud end and a polygonal environment is accelerated.

(3) The remote sensing algorithm knowledge packet provided by the invention has good granularity, mobility and context semantic perceptibility, and is convenient for seamless migration and collaborative calculation between cloud edges and polygonal edge nodes.

Drawings

FIG. 1 is an S-model configuration;

FIG. 2 is an algorithmic knowledge package grammar construction;

FIG. 3 is a computational process grammar construct;

FIG. 4 is a computational step grammar construct;

FIG. 5 is a spatial operator grammar construct;

FIG. 6 is an algorithm knowledge package file specification;

fig. 7 is a flowchart of a model creation method of a cloud-based remote sensing product production algorithm.

Detailed Description

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Firstly, defining an S model, wherein the S model comprises a reserved keyword, a modeling method and basic parameter types; secondly, carrying out algorithm standardized semantic expression, granularity division and semantic assembly on a remote sensing algorithm comprising a basic commonality algorithm and a model algorithm by utilizing the proposed S model; thirdly, carrying out semantic construction and assembly of an algorithm knowledge packet on the remote sensing algorithm by utilizing the proposed S model to form a physical entity of the remote sensing algorithm knowledge packet file, and supporting algorithm and data migration in a cloud environment; finally, a construction device and a construction system of the algorithm knowledge package file are provided, and the algorithm knowledge package is created, verified, assembled, registered, injected, multiplexed and the like through the device and the system.

(1) S model

In order to support semantic identification, granularity division and semantic assembly of a remote sensing algorithm and facilitate migration and multiplexing under different computing environments, the invention provides a semantic-supported remote sensing product production task description method, and provides a field specific model for modeling of a remote sensing product production task. The S model configuration is shown in fig. 1.

The S model is characterized in that the S model is a modeling structure for realizing production tasks of remote sensing products, and various complex remote sensing analysis models and the production tasks of the remote sensing products are described through a series of extensible semantic vocabularies and specific grammar structures.

The S model is characterized in that a reserved keyword set is provided, and the reserved keyword set comprises core keywords such as algorithm knowledge package definition, calculation process definition, calculation step definition, space operator reference, semantic identification, input parameter statement, output result statement, entity description, semantic label, conditional branch, circulation and the like.

The S model is characterized by providing basic parameter types, including 8 core parameter types of grids, vectors, integer, floating points, character strings, boolean, enumeration, arrays and the like.

The S model is characterized in that through the self-definition reservation of the key words and the parameter types, a user can define a corresponding S model basic grammar construction unit according to business logic of an algorithm calculation flow, as shown in fig. 1, through a knowledge network formed by semantic identifications of the grammar construction unit and association rules thereof, the modeling of a complex calculation process based on semantic knowledge is realized, and the efficiency of understanding and expressing the complex remote sensing calculation task process is improved.

The S model is characterized by an S model basic grammar construction unit including: the method comprises 7 core grammar construction units of an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator, an algorithm execution entity and the like. The basic construction of the S model is as follows:

< S model > → <1 algorithm knowledge package > → < n computational processes > → < m computational steps > → < k executable space operators identified by identifiers >

The algorithm knowledge package is characterized in that: (1) the algorithm knowledge packet is an object of the algorithm knowledge packet which is interpreted and run by the S model interpreter; (2) the algorithm knowledge package is a self-describing executable entity, is a combination of remote sensing data and an algorithm model, and comprises the following components: an algorithm knowledge package prototype definition file defined by using an S model, one or more calculation process files defined by using the S model, an algorithm execution entity provided by an author of the algorithm knowledge package, an interface file of the algorithm execution entity, an algorithm use instruction, input data, a calculation result, a published public space calculation subset which can be called or referenced by an external user and the like; (3) the algorithm knowledge packet is a basic unit for interpretation and execution of an S model interpreter, and the S model interpreter constructs an algorithm knowledge packet instance object and interprets and executes the object by analyzing an algorithm knowledge packet prototype definition file; (4) the algorithm knowledge packet identifier is used for uniquely identifying a certain algorithm knowledge packet, and in order to avoid the conflict of the knowledge packet identifiers, the S model adopts an identification mode of 'knowledge packet creator domain name, algorithm knowledge packet name'; (5) after being published, the algorithm knowledge packet is used as a provider of a reusable algorithm, and all the internally defined calculation processes can become spatial operators which can be called by external users; (6) the external user may go through "algorithmic knowledge packet identifier: the global semantic identifier of the space operator formed by the internal identifiers in the calculation process calls any space operator in the public space operator set which is externally released by the algorithm knowledge packet; (7) the subset of the spatial computations disclosed by the knowledge package is a list, and each element item stored in the list is a semantic identifier of the computing process to be disclosed. The algorithmic knowledge package grammar construction is shown in fig. 2.

The calculation process is characterized in that: (1) the calculation process is an executable model algorithm entity defined by using an S model in an algorithm knowledge packet, and one algorithm knowledge packet can define a plurality of calculation processes; (2) the semantic identification of the computation process is defined by the "algorithmic knowledge packet identifier: the internal identifier of the computing process consists of two parts, such as "sys: buffer," wherein "sys" is an identifier of an algorithm knowledge packet, and "buffer" is an internal identifier of the computing process defined in the algorithm knowledge packet where the computing process is located; (3) the computation process can be released through the algorithm knowledge package to become a space operator which can be called by an external user, and the external user can multiplex the computation process by referring to the global semantic identification of the space operator and can multiplex the computation process definition in other algorithm knowledge packages; (4) the computing process must declare its input parameter list and output parameter list, including the name, type, description, and optional parameter defaults for each parameter; (5) the calculation process comprises a plurality of calculation steps, each calculation step must refer to a space operator, and the calculation steps can refer to the space operator through the global semantic identification of the space operator or the user-defined local identifier; (6) the interrelationship between the calculation steps defined in the calculation process is deduced by parameter semantics to establish the dependency relationship between the calculation steps; (7) the definition of the calculation steps in the calculation process is sequentially executed as a whole, but the execution structures such as the reserved keyword expression conditional branches, loops and the like can also be provided by the S model; (8) in order to support the mobility of the calculation, the calculation process is performed by providing input data resources depending on the algorithm knowledge package, namely, acquiring data to be calculated from the algorithm knowledge package, and the calculation result of the calculation process is also stored in the corresponding algorithm knowledge package entity. The syntactic construction of the calculation process is shown in fig. 3.

The calculating step is characterized in that: (1) the calculation step is an executable algorithm step defined by using an S model in the calculation process, is a function statement, and can define a plurality of calculation steps in one calculation process; (2) each calculation step must refer to a space operator to complete a certain space calculation task, and the reference can be carried out through the global semantic identification of the space operator or through an internal identifier to refer to the space operator defined locally by the algorithm knowledge package; (3) different calculation steps may refer to the same spatial operator; (4) the calculation step must declare its input real parameter list and output real parameter list, and the value of the input real parameter of the calculation step can be derived from the output real parameter value, direct quantity, expression variable of other calculation steps in the calculation process, the input parameter of the calculation process or the global variable defined in the algorithm knowledge packet; (5) the parameter names in the input real parameter list and the output real parameter list of the calculation step statement must be consistent with the input and output parameter names of the quoted space operator statement, but the parameter sequence and the parameter type are automatically inferred by the S model interpreter; (6) and calculating the dependency relationship between the steps, and automatically deducing by an S model interpreter according to the reference relationship between the input and output parameters. The grammar construction of the calculation step is shown in fig. 4.

The spatial operator is characterized in that: (1) the space operator is the semantic abstraction of any remote sensing processing algorithm, and essentially carries out certain algorithm task flow processing on the remote sensing image and the related data thereof to obtain one or more output results; (2) the space operator has executable characteristics and is divided into two operators, namely a basic operator and a model operator from the viewpoint of implementation; (3) the space operator defines a common interface of the basic operator and the model operator, and the basic operator and the model operator specifically realize the interface provided by the space operator; (4) the interface of the space operator comprises 4 core components such as a space operator global semantic identifier, a space operator semantic description, an input parameter list, an output parameter list and the like; (5) the spatial operators are reusable and are referenced by "algorithmic knowledge packet identifiers: multiplexing the space operator by using the space operator global semantic identifiers formed by the space operator internal identifiers; (6) the spatial operator must declare its input parameter list and output parameter list, including the name, type, description, and optional parameter defaults for each parameter. The syntactic construction of the spatial operator is shown in fig. 5.

The basic operator is characterized in that: (1) the basic operator has atomic characteristics, is abstraction and encapsulation of a common function in a remote sensing algorithm, and is a minimum executable unit for scheduling and executing of an S model interpreter; (2) the basic operator has executable characteristics and comprises two main components, an algorithm execution entity and an algorithm interface file thereof; (3) the algorithm execution entity is a physical entity of a remote sensing algorithm, is an executable program and can be realized by any programming language; (4) the algorithm interface file is an executable script which is provided for shielding different programming languages and follows the S model space operator interface specification, and the S model interpreter finally realizes the calling of the remote sensing algorithm physical entity by driving the interface file; (5) the basic operator is a space operator with minimum granularity, and can form a more complex coarse-granularity space operator by carrying out semantic organization and multiplexing with other space operators, and the reference or multiplexing mode is the same as that of the space operator.

The model operator is characterized in that: (1) the model operator has a composite characteristic, and is a more complex coarse-grained space operator formed by semantic organization and multiplexing with other basic operators or model operators; (2) the model operator is a space operator which is obtained by the issuing of the algorithm knowledge packet in the calculation process in the algorithm knowledge packet and can be used for the multiplexing of external users; (3) the model operator is a space operator with larger granularity, and can form a more complex coarse-granularity space operator by carrying out semantic organization and multiplexing with other space operators, and the quoting or multiplexing mode is the same as that of the space operator.

(2) Remote sensing algorithm knowledge packet file specification and medium

The remote sensing algorithm knowledge packet is the topmost grammar construction unit of the S model and is the basic unit for interpretation and execution of an S model interpreter. And the S model interpreter constructs an algorithm knowledge package instance object and interprets and executes the object by analyzing the algorithm knowledge package prototype definition file. As an input to the S-model interpreter, the remote sensing algorithm knowledge package file is a removable, shareable disk compressed file with specific specifications. The specification of the remote sensing algorithm knowledge packet file is shown in fig. 6.

The specification of the knowledge packet file of the remote sensing algorithm is characterized in that: (1) the root directory of the algorithm knowledge packet comprises a knowledge packet prototype definition file which defines key description information of the algorithm knowledge packet and an organization structure of knowledge packet contents, and an S model interpreter constructs an executable instance of the knowledge packet according to the acquired metadata by reading and analyzing the file; (2) the root directory of the algorithm knowledge packet comprises a calculation process directory which is used for storing all S model calculation process definition files defined in the algorithm knowledge packet; (3) the root directory of the algorithm knowledge packet comprises a data input directory which is used for storing various input data required by the calculation process in the knowledge packet in operation; (4) the root directory of the algorithm knowledge packet comprises an output result directory which is used for storing various output data generated during the operation of the calculation process in the knowledge packet; (5) optionally, the root directory of the algorithm knowledge package comprises an algorithm program directory, and the directory is used for storing executable algorithm programs and interface files thereof provided by a knowledge package creator; (6) optionally, the root directory of the algorithm knowledge package contains a description document directory, and the directory is used for storing the description document, the instruction manual, the video course and other files of the knowledge package.

(3) Remote sensing algorithm knowledge packet creation device

The remote sensing algorithm knowledge packet creation device comprises: (1) the creation module is characterized by defining a core entity such as an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator, an algorithm execution entity and the like according to the structural logic of the remote sensing production algorithm and in combination with an S model creation method; (2) the verification module is characterized by performing up-down Wen Yuyi inference and grammar compliance check on the S model entity and the semantic relation between the entities; (3) the assembly module is characterized by carrying out semantic assembly on the created S model entity, algorithm program, description document and the like to form an algorithm knowledge packet specification file; (4) the registration module is characterized by registering and publishing the specification file of the algorithm knowledge packet to form a knowledge packet operable instance which can be globally found by the S model interpreter, and identifying and searching the knowledge packet and the calculation process contained in the knowledge packet through the semantic identifier; (5) the injection module is characterized by injecting the input data to be processed and the operation configuration information into an algorithm knowledge packet specification file to generate an operational algorithm knowledge packet instance; (6) the interpretation module is characterized in that an S model interpreter reads and analyzes a knowledge packet prototype definition file in an algorithm knowledge packet instance, creates an instance object such as an algorithm knowledge packet and a calculation process according to the acquired metadata, binds data parameters and drives the execution of the calculation process.

FIG. 7 is a flow chart of one embodiment of a model creation method for a cloud-based remote sensing product production algorithm of the present invention. The cloud-based remote sensing product production algorithm model creation method comprises the following steps:

step S701, defining and describing one or more of the following entities of the remote sensing product production algorithm to be modeled according to the structural logic of the remote sensing algorithm in combination with the S model creation method: algorithm knowledge package, calculation process, calculation step, spatial operator, basic operator, model operator and algorithm execution entity.

In the embodiment, according to the structural logic of a remote sensing production algorithm, firstly, a topmost algorithm knowledge package entity and a prototype definition file thereof are created, and algorithm knowledge package semantic identifiers, semantic descriptions and definition algorithm knowledge package global variables are filled in the algorithm knowledge package definition file; secondly, one or more computing process entities and corresponding computing process definition files are created, and the computing process definition files are filled with internal semantic identifiers, semantic descriptions and input parameters and output parameters for defining the computing process; next, one or more computing step entities are created in the computing process definition file, including: filling global semantic identifiers of the referenced space operators in the calculating step entity; filling in each input parameter of the referenced space operator and the corresponding real parameter value in the running process; optionally, the output file name or disk path of each file type output parameter is filled in.

Alternatively, the S-model algorithm knowledge package creator may define one or more base operators that may be invoked by the knowledge package internal or external to the knowledge package. Creating a basic operator, and writing a corresponding interface file by providing an algorithm execution entity on which the basic operator depends, wherein the algorithm execution entity comprises an interface program file and an algorithm program, the algorithm program is a physical entity of a remote sensing algorithm and is an executable program, and the algorithm program can be implemented by any programming language; the algorithm interface file is a script file which is provided for shielding different programming languages and follows the S model space operator interface specification, and the S model interpreter finally realizes the calling of remote sensing algorithm physical entities, namely algorithm programs by driving the interface file.

Optionally, the S-model algorithm knowledge package creator may define and edit the algorithm description document in text, video, etc. format, helping the algorithm knowledge package user to quickly understand, master and use the reusable subset of algorithms disclosed in the algorithm knowledge package.

Optionally, the S-model algorithm knowledge package creator may provide test case data to help the algorithm knowledge package user to quickly verify and learn the reusable subsets disclosed in the algorithm knowledge package, and view their running results and related execution reports.

In step S702, the entity and its semantic relationship defined in step S701 are subjected to up-down Wen Yuyi inference and grammar compliance checking, so as to ensure that the S model entity meets the S model creation requirement.

In this embodiment, the method includes the steps of:

(1) checking compliance of an algorithmic knowledge package entity, comprising: checking whether the algorithm knowledge packet semantic identifiers have conflict or not, and ensuring the uniqueness of the algorithm knowledge packet semantic identifiers; checking whether each element item in the space operator set which is disclosed by the algorithm knowledge packet statement and can be called by the outside is consistent with the semantic identifier of the corresponding computing process entity; and checking whether the name, the type and the variable value of the knowledge package global variable stated by the algorithm knowledge package meet the S model creation requirement.

(2) Checking compliance of a computing process entity, comprising: checking whether the semantic identifiers in the computing process have conflicts or not, and ensuring the uniqueness of the semantic identifiers in the knowledge packet; checking whether the name, type, default value and semantic description of each parameter item in an input parameter list and an output parameter list defined by a computing process entity meet the S model creation requirement; the reference relationship of input and output parameters between each computing step defined inside the computing process entity is checked to infer whether there is a dependency relationship between each computing step and whether there is a circular dependency problem between the steps.

(3) Checking compliance of individual computing step entities contained by a computing process entity, comprising: checking whether a global semantic identifier of the spatial operator referenced by the calculating step exists; acquiring prototype definitions of the space operators according to global semantic identifiers of the referenced space operators; checking whether the name of each parameter in the input parameter list and the output parameter list declared in the calculating step is consistent with the names of the input parameters and the output parameters declared in the space operator prototype according to the definition of the space operator prototype; checking whether the number of parameters in the input parameter list and the output parameter list declared in the calculating step is consistent with the number of input and output parameters declared by the space operator prototype according to the definition of the space operator prototype; checking whether the declaration in the spatial operator prototype definition exists in the input parameter list and the output parameter list declared in the calculating step according to the spatial operator prototype definition, and calculating the missing parameters of the entity in the calculating step; checking whether real parameter values of each parameter in an input parameter list stated in the calculation step are valid or not according to the definition of a space operator prototype, wherein the input real parameter values of the calculation step can be derived from output real parameter values, direct quantity, expression variables of other calculation steps, input parameters of a calculation process or global variables defined in an algorithm knowledge packet; according to the spatial operator prototype definition, checking whether the value of each file type output parameter, such as file name or disk path, in the output parameter list declared by the calculating step is valid.

(4) Optionally, compliance checking is performed on the interface file of the base operator defined by the knowledge package creator.

And step 703, performing semantic assembly on the S model entity checked for compliance in step 702 to form an algorithm knowledge packet specification file. According to the specification of the algorithm knowledge package file shown in fig. 6, an internal directory structure of the algorithm knowledge package file is constructed and corresponding contents are filled, and the method comprises the following steps:

(1) copying the algorithm knowledge package prototype definition file created in step S701 to the root directory of the algorithm knowledge package compression file, and registering the internal directory structure of the knowledge package compression file in the algorithm knowledge package definition file, the directory structure comprising: computing process catalog, input data catalog, output result catalog, algorithm program catalog and description document catalog. And the S model interpreter analyzes the knowledge packet prototype definition file under the knowledge packet compressed file root directory by reading the knowledge packet prototype definition file, and can construct the executable examples of the top-layer object of the algorithm knowledge packet and the internal object thereof according to the acquired metadata.

(2) A computing process catalog is established under the root catalog of the algorithm knowledge package and is used for storing all S model computing process definition files defined in the algorithm knowledge package.

(3) Establishing an input data catalog under the root catalog of the algorithm knowledge packet, wherein the catalog is used for storing various input data required by the calculation process in the knowledge packet in operation; alternatively, the knowledge package creator may provide algorithmic test case data here.

(4) Establishing an output result catalog under the root catalog of the algorithm knowledge packet, wherein the catalog is used for storing various output data generated in the calculation process of the knowledge packet and the space operator in the operation process;

(5) optionally, an algorithm program directory is established under the root directory of the algorithm knowledge packet, and the directory is used for storing the executable algorithm program and the interface file thereof provided by the creator of the algorithm knowledge packet;

(6) optionally, a description document catalog is established under the root catalog of the algorithm knowledge package, and the catalog is used for storing files of description documents, instruction manuals, video courses and the like of the knowledge package.

Step S704, uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database, registering and publishing the specification file of the algorithm knowledge package to form an executable example of the algorithm knowledge package which can be globally found by the cloud S model interpreter, and searching and downloading the algorithm knowledge package and the calculation process contained in the knowledge package through the semantic identifier.

In this embodiment, the method includes the steps of:

(1) uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database, taking the global semantic identifier of the algorithm knowledge package as a key, taking the specification file of the algorithm knowledge package as a value, and establishing an algorithm knowledge package and a space operator registry according to the semantic description information of the algorithm knowledge package.

(2) When the cloud S model interpreter runs the algorithm knowledge package, if the space operator is found to be absent, the specification file of the algorithm knowledge package corresponding to the required space operator is automatically downloaded from the cloud knowledge package database according to the global identifier of the referenced space operator and is cached in the S model interpreter.

Step S705, the input data to be processed and the operation configuration information are injected into the algorithm knowledge package specification file to generate an algorithm knowledge package instance file, wherein the input data is remote sensing data.

In this embodiment, the method includes the steps of:

(1) the method comprises the steps that a knowledge packet user selects an algorithm knowledge packet specification file to be operated and input data to be processed, a mapping relation between overall variables of the algorithm knowledge packet, input parameters of each calculation process and the like and actual parameters of the input data and the like is established, and algorithm knowledge packet operation configuration information is generated;

(2) Reading a knowledge packet prototype definition file under a knowledge packet specification file root directory, writing the algorithm knowledge packet operation configuration information into the knowledge packet prototype definition file, storing input data into an input data directory of the knowledge packet specification file, and generating an algorithm knowledge packet instance file.

Step S706, the cloud S model interpreter reads and analyzes the relevant definitions in the algorithm knowledge packet instance file, creates an algorithm knowledge packet instance and a calculation process instance according to the acquired relevant data, binds data parameters and drives the execution of the calculation process.

In this embodiment, the method includes the steps of:

(1) the cloud S model interpreter decompresses the algorithm knowledge package compression file, reads and analyzes the knowledge package prototype definition file in the algorithm knowledge package instance, and extracts the algorithm knowledge package content organization data, the operation configuration data and the global variables;

(2) the cloud S model interpreter organizes data according to the content of the algorithm knowledge packet, sequentially reads the calculation process definition files from the calculation process catalogue, and creates a corresponding calculation process instance and an internal calculation step and a space operator of each calculation process definition file;

(3) the cloud S model interpreter combines the operation configuration data and the global variable to complete the binding of the input form participation real parameter value of the computing process instance;

(4) The cloud S model interpreter traverses the to-be-operated computing process examples in the knowledge packet examples in sequence, traverses computing step entities in each computing process example, groups and sequences computing steps according to the dependency relationship among the computing steps, and forms a series of computing step sets which can be scheduled and executed in parallel.

(5) The cloud S model interpreter sequentially traverses the calculation step sets, creates an execution thread for each parallelizable calculation step set and invokes an interface program of a space operator entity referenced by the calculation step in the thread;

(6) after the operation of the algorithm knowledge package instance is finished, the cloud S model interpreter accommodates output results of all computing process instances and stores the output results into an output result catalog of the knowledge package instance file.

Alternatively, the knowledge package user may specify a set of computing processes to be run in the run configuration information of the knowledge package definition file.

The application also provides another implementation mode, namely a remote sensing algorithm knowledge packet file specification and medium.

The remote sensing algorithm knowledge packet is the topmost grammar construction unit of the S model and is the basic unit for interpretation and execution of an S model interpreter. And the S model interpreter constructs an algorithm knowledge package instance object and interprets and executes the object by analyzing the algorithm knowledge package prototype definition file. As an input to the S-model interpreter, the remote sensing algorithm knowledge package file is a removable, shareable disk compressed file with specific specifications. The specification of the remote sensing algorithm knowledge packet file is shown in fig. 6.

The specification of the knowledge packet file of the remote sensing algorithm is characterized in that: (1) the root directory of the algorithm knowledge packet comprises a knowledge packet prototype definition file which defines key description information of the algorithm knowledge packet and an organization structure of knowledge packet contents, and an S model interpreter constructs an executable instance of the knowledge packet according to the acquired metadata by reading and analyzing the file; (2) the root directory of the algorithm knowledge packet comprises a calculation process directory which is used for storing all S model calculation process definition files defined in the algorithm knowledge packet; (3) the root directory of the algorithm knowledge packet comprises a data input directory which is used for storing various input data required by the calculation process in the knowledge packet in operation; (4) the root directory of the algorithm knowledge packet comprises an output result directory which is used for storing various output data generated during the operation of the calculation process in the knowledge packet; (5) optionally, the root directory of the algorithm knowledge package comprises an algorithm program directory, and the directory is used for storing executable algorithm programs and interface files thereof provided by a knowledge package creator; (6) optionally, the root directory of the algorithm knowledge package contains a description document directory, and the directory is used for storing the description document, the instruction manual, the video course and other files of the knowledge package.

The application also provides another implementation mode, namely a model creation device of a remote sensing product production algorithm based on a cloud, which comprises the following components:

(1) creating a module; according to the structural logic of the remote sensing product production algorithm to be modeled, defining and describing one or more of the following entities of the remote sensing product production algorithm to be modeled by combining an S model construction method: the method comprises an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator and an algorithm execution entity;

(2) a verification module; performing up-down Wen Yuyi inference and grammar compliance check on the defined and described entities and semantic relations among the entities to ensure that the entities meet the S model construction requirement;

(3) assembling a module; carrying out semantic assembly on the entity checked for compliance to form an algorithm knowledge packet specification file, constructing an internal directory structure of the algorithm knowledge packet file and filling corresponding contents;

(4) a registration module; uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database for registration and release to form an algorithm knowledge package operable instance which can be globally found by a cloud S model interpreter, and searching and downloading the algorithm knowledge package operable instance and a calculation process contained in the algorithm knowledge package operable instance by the cloud S model interpreter through a semantic identifier of the algorithm knowledge package;

(5) An injection module; the method comprises the steps of injecting input data to be processed and operation configuration information into an algorithm knowledge packet specification file to generate an algorithm knowledge packet instance file, wherein the input data are remote sensing data;

(6) an interpretation module; the cloud S model interpreter reads and analyzes the related definition of the algorithm knowledge package in the algorithm knowledge package instance file, creates an algorithm knowledge package instance and a calculation process instance according to the acquired input data, binds data parameters and drives the execution of the calculation process.

Claims (7)

1. The model creation method of the remote sensing product production algorithm based on the cloud is characterized by comprising the following steps of:

step one, defining and describing the following entities of a remote sensing product production algorithm to be modeled according to the structural logic of the remote sensing product production algorithm to be modeled by combining an S model creation method: the method comprises an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator and an algorithm execution entity;

step two, performing up-down Wen Yuyi inference and grammar compliance check on the entity and the semantic relation thereof defined and described in the step one, so as to ensure that the entity meets the S model creation requirement;

step three, carrying out semantic assembly on the entity checked to be compliant in the step two to form an algorithm knowledge packet specification file, constructing an internal directory structure of the algorithm knowledge packet specification file and filling corresponding contents;

Uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database, registering and publishing to form an algorithm knowledge package operable instance which can be globally found by a cloud S model interpreter, and searching and downloading the algorithm knowledge package operable instance and a calculation process contained in the algorithm knowledge package operable instance by the cloud S model interpreter through a semantic identifier of the algorithm knowledge package;

step five, the input data to be processed and the operation configuration information are injected into an algorithm knowledge packet specification file to generate an algorithm knowledge packet instance file, wherein the input data is remote sensing data;

step six, the cloud S model interpreter reads and analyzes the related definition of the algorithm knowledge package in the algorithm knowledge package instance file, creates an algorithm knowledge package instance and a calculation process instance according to the acquired related data, binds data parameters and drives the execution;

firstly, according to the structural logic of a remote sensing product production algorithm to be modeled, firstly, creating a topmost algorithm knowledge package entity and a prototype definition file thereof, and filling an algorithm knowledge package semantic identifier, semantic description and definition algorithm knowledge package global variables in the algorithm knowledge package prototype definition file; secondly, one or more computing process entities and corresponding computing process definition files are created, and the computing process definition files are filled with internal semantic identifiers, semantic descriptions and input parameters and output parameters for defining the computing process; next, one or more computing step entities are created in the computing process definition file, including: filling global semantic identifiers of the referenced space operators in the calculating step entity; filling out each input parameter of the referenced space operator and the corresponding real parameter value of the runtime, wherein the space operator comprises a basic operator and a model operator.

2. The method according to claim 1, wherein the second step comprises the steps of:

(1) checking compliance of an algorithmic knowledge package entity, comprising: checking whether the algorithm knowledge packet semantic identifiers have conflict or not, and ensuring the uniqueness of the algorithm knowledge packet semantic identifiers; checking whether the global semantic identifier in the spatial operator of the algorithm knowledge package is consistent with the internal semantic identifier of the corresponding computing process entity; checking whether the name, type and variable value of the global variable of the algorithm knowledge packet meet the S model creation requirement;

(2) checking compliance of a computing process entity, comprising: checking whether the semantic identifiers in the calculation process have conflict or not, and ensuring the uniqueness of the semantic identifiers in the algorithm knowledge packet; checking whether the name, type, default value and semantic description of each parameter in an input parameter list and an output parameter list defined by a computing process entity meet the S model creation requirement; checking the reference relation of input and output parameters among all the calculation steps defined in the entity of the calculation process, deducing whether a dependency relation exists among all the calculation steps and whether a circular dependency problem exists among the calculation steps;

(3) checking compliance of individual computing step entities contained by a computing process entity, comprising: checking whether a global semantic identifier of the spatial operator referenced by the calculating step exists; acquiring prototype definitions of the space operators according to global semantic identifiers of the referenced space operators; checking whether the name of each parameter in the input parameter list and the output parameter list of the calculating step is consistent with the names of the input parameters and the output parameters of the space operator according to the definition of the prototype of the space operator; checking whether the number of parameters in the input parameter list and the output parameter list in the calculation step is consistent with the number of input and output parameters of the space operator according to the definition of the prototype of the space operator; checking whether parameters in an input parameter list and an output parameter list of the calculating step exist in the definition of the space operator prototype according to the definition of the space operator prototype, and finding out the missing parameters of the entity of the calculating step; checking whether the real parameter value of each parameter in the input parameter list of the calculation step is valid or not according to the definition of the space operator prototype, wherein the real parameter value of the calculation step is derived from the output real parameter value, the direct quantity, the expression variable of other calculation steps, the input parameter of the calculation process or the global variable defined in the algorithm knowledge packet; according to the definition of the space operator prototype, checking whether the value of each file type output parameter in the output parameter list of the calculating step is valid or not.

3. The method according to claim 2, characterized in that step three comprises in particular the steps of:

(1) copying the algorithm knowledge package prototype definition file created in the first step to a root directory of the algorithm knowledge package compression file, and registering an internal directory structure of the algorithm knowledge package compression file in the algorithm knowledge package prototype definition file, wherein the internal directory structure comprises: computing a process catalog, an input data catalog, an output result catalog, an algorithm program catalog and an explanation document catalog;

(2) establishing a computing process catalog under the root catalog of the algorithm knowledge package, wherein the catalog is used for storing all computing process definition files defined in the algorithm knowledge package;

(3) an input data catalog is established under the root catalog of the algorithm knowledge packet, and the catalog is used for storing various input data required by the calculation process in the algorithm knowledge packet in operation;

(4) establishing an output result catalog under the root catalog of the algorithm knowledge packet, wherein the catalog is used for storing various output data generated in the calculation process of the algorithm knowledge packet and the space operator in the operation process;

(5) an algorithm program catalog is contained under the root catalog of the algorithm knowledge package, and the catalog is used for storing executable algorithm programs and interface files thereof;

(6) A description document catalog is contained under the root catalog of the algorithm knowledge package and is used for storing description documents, instruction manuals and video courses of the knowledge package.

4. A method according to claim 3, characterized in that step four comprises in particular the steps of:

(1) uploading the specification file of the algorithm knowledge package to a cloud algorithm knowledge package database, taking the semantic identifier of the algorithm knowledge package as a key, taking the specification file of the algorithm knowledge package as a value, and establishing an algorithm knowledge package and a space operator registry according to semantic description information of the algorithm knowledge package;

(2) when the cloud S model interpreter runs the algorithm knowledge package, if the space operator is found to be absent, the specification file of the algorithm knowledge package corresponding to the required space operator is automatically downloaded from the cloud knowledge package database according to the global semantic identifier of the referenced space operator and is cached in the S model interpreter.

5. The method of claim 4, wherein step five specifically comprises the steps of:

(1) selecting an algorithm knowledge packet specification file to be operated and input data to be processed, establishing a global variable of the algorithm knowledge packet and a mapping relation between each shape parameter of input parameters and each real parameter of the input data in each calculation process, and generating operation configuration information of the algorithm knowledge packet;

(2) Reading an algorithm knowledge packet prototype definition file under a root directory of a compressed file of an algorithm knowledge packet specification file, writing algorithm knowledge packet operation configuration information into the algorithm knowledge packet prototype definition file, storing input data into an input data directory of the algorithm knowledge packet specification file, and generating an algorithm knowledge packet instance file.

6. The method according to claim 5, wherein step six specifically comprises the steps of:

(1) the cloud S model interpreter decompresses the algorithm knowledge package compression file, reads and analyzes the algorithm knowledge package prototype definition file in the algorithm knowledge package instance, and extracts the algorithm knowledge package content organization data, the operation configuration data and the global variables;

(2) the cloud S model interpreter organizes data according to the content of the algorithm knowledge packet, sequentially reads the calculation process definition files from the calculation process catalogue, and creates a corresponding calculation process instance and an internal calculation step and a space operator of each calculation process definition file;

(3) the cloud S model interpreter combines the operation configuration data and the global variables to complete the binding of various parameters of the input parameters and various real parameters of the input data of the computing process instance;

(4) the cloud S model interpreter traverses the to-be-operated computing process examples in the algorithm knowledge packet example file in sequence, traverses the computing steps in each computing process example, groups and sequences the computing steps according to the dependency relationship among the computing steps, and forms a series of computing step sets which can be scheduled and executed in parallel;

(5) The cloud S model interpreter sequentially traverses the calculation step sets, creates an execution thread for each calculation step set and invokes an interface program of a space operator referenced by the calculation step in the thread;

(6) after the operation of the algorithm knowledge package instance is finished, the cloud S model interpreter accommodates output results of all computing process instances and stores the output results into an output result catalog of the algorithm knowledge package instance file.

7. A cloud-based model creation device for a remote sensing product production algorithm, comprising:

(1) creating a module; according to the structural logic of the remote sensing product production algorithm to be modeled, the following entities of the remote sensing product production algorithm to be modeled are defined and described by combining an S model creation method: the method comprises an algorithm knowledge packet, a calculation process, a calculation step, a space operator, a basic operator, a model operator and an algorithm execution entity;

(2) the verification module is used for carrying out up-down Wen Yuyi inference and grammar compliance check on the defined and described entities and semantic relations among the entities, so as to ensure that the entities meet the S model creation requirement;

(3) the assembly module is used for carrying out semantic assembly on the entity for checking compliance to form an algorithm knowledge packet specification file, constructing an internal directory structure of the algorithm knowledge packet file and filling corresponding contents;

(4) The registration module is used for uploading the specification file of the algorithm knowledge packet to a cloud algorithm knowledge packet database, registering and publishing the specification file of the algorithm knowledge packet to form an algorithm knowledge packet operable instance which can be globally found by a cloud S model interpreter, and searching and downloading the algorithm knowledge packet operable instance and a calculation process contained in the algorithm knowledge packet operable instance by the cloud S model interpreter through a semantic identifier of the algorithm knowledge packet;

(5) the injection module is used for injecting the input data to be processed and the operation configuration information into the algorithm knowledge packet specification file to generate an algorithm knowledge packet instance file, wherein the input data is remote sensing data;

(6) the interpretation module reads and analyzes the related definition of the algorithm knowledge package in the algorithm knowledge package instance file by the cloud S model interpreter, creates an algorithm knowledge package instance and a calculation process instance according to the acquired related data, binds data parameters and drives the execution;

firstly, according to the structural logic of a remote sensing product production algorithm to be modeled, firstly, creating a topmost algorithm knowledge package entity and a prototype definition file thereof, and filling an algorithm knowledge package semantic identifier, semantic description and definition algorithm knowledge package global variables in the algorithm knowledge package prototype definition file; secondly, one or more computing process entities and corresponding computing process definition files are created, and the computing process definition files are filled with internal semantic identifiers, semantic descriptions and input parameters and output parameters for defining the computing process; next, one or more computing step entities are created in the computing process definition file, including: filling global semantic identifiers of the referenced space operators in the calculating step entity; filling out each input parameter of the referenced space operator and the corresponding real parameter value of the runtime, wherein the space operator comprises a basic operator and a model operator.