Machine Vision Algorithm Framework Standardization Platform

Abstract

The present invention relates to a machine vision algorithm framework standardization platform, including a data processing unit, wherein the data processing unit is configured on different PC terminals, and data is transmitted between each data processing unit and between the data processing unit and the client through a communication module; the data processing unit includes an algorithm framework, and the algorithm framework performs algorithm flow processing on the data output by the data acquisition device and outputs the detection result; when the associated output data of a configured algorithm is used as an associated input by the algorithm under other data processing units, the algorithm framework is implemented by a cross-PC call function. The present invention enhances the standardized process of algorithm development and parameter configuration, reduces the difficulty of algorithm development, avoids a large amount of repeated development work, and improves the efficiency of project development; supports algorithm data association under complex visual structures, and provides flexibility and convenience for the implementation of complex projects.

Description

Technical Field

[0001] The present invention relates to the field of visual detection technology, and in particular to a machine vision algorithm framework standardization platform. Background Art

[0002] The main functions of the algorithm integration framework are: (1) to provide a standardized algorithm integration platform for the rapid integration and verification of various functional algorithms; (2) to serve as a functional platform that can connect clients, hardware devices, and algorithms to realize the analysis and transmission of various data.

[0003] The existing algorithm integration framework generally includes four major processes: algorithm library import, template creation, algorithm execution and result display. Algorithm import refers to the design of parameter configuration and dynamic library development that meets the requirements of the corresponding algorithm framework. By creating a template, the algorithm task process is assigned to each data. The framework calls and executes each algorithm according to the task process, and sends the results of the algorithm execution to the client for display. The specific implementation steps are as follows: configure the algorithm parameters to form a standard file; write the algorithm code and generate dynamic library data; the client assigns tasks, sets algorithm parameters, and establishes templates; the algorithm framework obtains template data and assigns calculation tasks; the algorithm framework converts parameter data streams and calls the algorithm dynamic library; the algorithm dynamic library parses parameters, performs calculations, and outputs calculation results to the algorithm framework; the algorithm framework sends the calculation results to the client; the client displays the calculation results.

[0004] However, the existing algorithm integration framework has a low degree of standardization in the configuration and integration of algorithm parameters. The parameter configuration process is complex and time-consuming, and algorithm developers need to parse their own algorithm parameters.

[0005] Moreover, the data information between different algorithms is independent of each other, so complex data structures and complex association templates cannot be shared because the framework has no transmission and synchronization mechanism for data on different PCs. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a standardized platform for a machine vision algorithm framework, which is configurable, extensible, and simulable, has a standard algorithm encapsulation process, supports flexible cross-image data association, and can conveniently implement algorithm parameter configuration, debugging, and result display.

[0007] The technical solution adopted by the present invention to solve its technical problems is: a standardized platform for a machine vision algorithm framework, including a data processing unit, the data processing unit is configured on different PC terminals, and data is transmitted between the data processing units and between the data processing units and the client through a communication module; the data processing unit includes an algorithm framework, and the algorithm framework performs algorithm flow processing on the data output by the data acquisition device and outputs the detection result; when the associated output data of a configured algorithm is used as associated input by the algorithm under other data processing units, the algorithm framework is implemented using a cross-PC call function.

[0008] Furthermore, the algorithm framework described in the present invention includes a data acquisition module, a device module, a camera acquisition and distribution module, a standard algorithm pool module, an image processing management module and a communication module.

[0009] Furthermore, the algorithm framework described in the present invention includes algorithm information configuration, and the algorithm information configuration types include detection parameters, result parameters, simulation parameters, and associated input parameters of the algorithm and associated output parameters of the algorithm; the configuration method of the algorithm information is: parameter information configuration is performed in an Excel table, and an XML configuration file that meets the framework standard is automatically generated through a parameter generation tool; the algorithm information configuration content parameter name, translation name, data type, parameter interface type, parameter range, parameter authority, and result parameter display method on the client, and the client detection result display method supports the permutation and combination of display in the image, display in the result list, and display in the data report.

[0010] Furthermore, the algorithm information configuration described in the present invention does not require parsing the parameter configuration file. After the algorithm framework is parsed, the parameter information is passed to the algorithm. The algorithm obtains the parameters and executes the algorithm process, and compiles and generates a dynamic library and places it in the specified location.

[0011] Furthermore, the algorithm framework described in the present invention includes template creation, sending and data loading. Template creation is completed on the client side through interface configuration, algorithm task allocation and detection parameter setting are performed for each collected data, and the association relationship between the algorithms is set; the algorithm framework includes three kinds of association configurations: the association relationship between the algorithms configured within the same collected data, the association relationship between the algorithms configured within the same data processing unit, and the association relationship between the algorithms configured for data collected by different data processing units.

[0012] Furthermore, after the template of the present invention is created, the communication module sends the template data to each data processing unit, the algorithm framework of each data processing unit parses the template data, adjusts the algorithm execution order according to the correlation between the algorithms, and passes the parsed algorithm parameters to the corresponding algorithm modules.

[0013] Furthermore, the data processing unit of the present invention executes the algorithms configured for each collected data according to the detection order adjusted by the algorithm framework, that is, calls the algorithm dynamic library; the algorithm dynamic library performs calculations and associated data outputs according to its own algorithm parameters and associated data input; when the associated output data written by the algorithm is used by algorithms under other data processing units, the algorithm framework sends the associated data to the data processing unit where the algorithm using the associated data is located through the communication module, and then the algorithm framework of the data processing unit sends the associated data to the algorithm using the associated data.

[0014] The present invention carries out a unified modular design and a standard data structure design, provides standardized commands, data interaction interfaces, algorithm execution processes and a series of standardized algorithm input and output interfaces.

[0015] The present invention designs a cross-PC data transmission solution, and solves the problem of cross-PC associated data input and output through a data synchronization mechanism and a communication module.

[0016] The present invention designs an algorithm parameter configuration tool, which can flexibly configure the type, attribute and UI display mode of the algorithm parameters.

[0017] The beneficial effects of the present invention are:

[0018] (1) The present invention enhances the standardized process of algorithm development and parameter configuration, reduces the difficulty of algorithm development, avoids a large amount of repeated development work, and improves project development efficiency.

[0019] (2) The present invention supports algorithmic data association under complex visual structures, providing flexibility and convenience for the implementation of complex projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the visual structure configuration of the present invention;

[0021] Figure 2 It is a structural diagram of the standardized platform of the algorithm framework of the present invention;

[0022] Figure 3 It is a processing flow chart of the algorithm framework of the present invention;

[0023] Figure 4 It is a flow chart of the principle of implementing cross-PC calling of the present invention. DETAILED DESCRIPTION

[0024] The present invention will now be described in further detail in conjunction with the accompanying drawings and preferred embodiments. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0025] A standardized platform for machine vision algorithm framework, the visual structure configuration is as follows Figure 1 As shown in the figure, the process includes data algorithm configuration, establishing and publishing the detection process template, algorithm framework parsing template data, executing the detection process and outputting the results. The algorithm framework can be understood as Figure 1 The data processing unit in the data processing unit. The data processing unit can be configured on different PC terminals, and data is transmitted between the data processing units and between the data processing units and the client through the communication module.

[0026] The algorithm framework includes data acquisition module, equipment module, camera acquisition and distribution module, standard algorithm pool module, image processing management module and communication module. The basic structure is as follows: Figure 2 shown.

[0027] The basic function of the algorithm framework is to process the data output by the data acquisition device through the algorithm flow and output the detection results. The processing flow is as follows: Figure 3 Specific functional features:

[0028] (1) Algorithm information configuration

[0029] Algorithm parameter types include: detection parameters, result parameters, simulation parameters, and association parameters (algorithm association input parameters and association output parameters)

[0030] Algorithm parameter configuration method: configure parameter information in Excel, and automatically generate an XML configuration file that complies with the framework standard through the parameter generation tool

[0031] The parameter information configuration content includes: parameter name, translation name, data type, parameter interface type, parameter range, parameter authority, and the display method of result parameters on the client. The client detection result display method supports the permutation and combination of display in image / display in result list / display in data report.

[0032] Algorithm dynamic library configuration: The algorithm does not need to parse the parameter configuration file. After the algorithm framework completes the parsing, it passes the parameter information to the algorithm. The algorithm obtains the parameters and executes the algorithm process. The compiled dynamic library is placed in the specified location.

[0033] (2) Data algorithm set configuration

[0034] Collect data and select the required algorithms from the algorithm library to form an algorithm set. The algorithm framework supports configuring different algorithm sets for each data.

[0035] (3) Template creation, sending and data loading

[0036] Template creation is completed through interface configuration on the client side, and algorithm tasks are assigned and detection parameters are set for each collected data, and the association between algorithms is set. The algorithm framework supports three types of association configurations: 1) the association between algorithms configured within the same collected data; 2) the association between algorithms configured within the same data processing unit; 3) the association between algorithms configured for data collected by different data processing units.

[0037] After the template is created, the communication module sends the template data to each data processing unit. The algorithm framework of each data processing unit parses the template data, adjusts the algorithm execution order according to the correlation between the algorithms, and passes the parsed algorithm parameters to their corresponding algorithm modules.

[0038] (4) Execution of the testing process

[0039] During the detection process, the algorithm framework executes the algorithm configured under each data of the current data processing unit. The specific functional features are as follows:

[0040] The data processing unit executes the algorithm configured for each collected data according to the detection sequence adjusted by the algorithm framework, that is, calling the algorithm dynamic library

[0041] The algorithm dynamic library performs calculations and outputs associated data based on its own algorithm parameters and associated data input.

[0042] 3) When the associated output data of an algorithm configured in the template is used as associated input by the algorithm under other data processing units, the algorithm framework uses the cross-PC call function to implement this requirement. The implementation principle is as follows: Figure 4 As shown: when the associated output data written by the algorithm is used by the algorithms under other data processing units, the algorithm framework sends the associated data to the data processing unit where the algorithm using the associated data is located through the communication module, and then the algorithm framework of the data processing unit sends the associated data to the algorithm using the associated data.

[0043] 4) The algorithm framework takes the collected data as a unit, and after executing a set of collected data algorithms, the test results are sent to the client. The client displays the test results on the interface according to the display mode configuration of each test result.

[0044] Aiming at the problem that algorithm parameter files need to be manually generated and have complex configuration, the present invention develops an automatic generation tool for algorithm XML configuration files. Algorithm developers only need to fill in the corresponding parameter information and attributes in an Excel table to automatically generate an XML configuration file that meets the requirements of the algorithm framework. The framework will automatically parse the template data and decompose and send it to each algorithm module. This simplifies the parameter configuration process of algorithm developers, and it only takes about 10 minutes at most to configure a set of algorithm parameters, which improves the efficiency of algorithm development.

[0045] In order to solve the problem that data cannot be shared between algorithms or algorithm data under different PCs cannot be shared, the present invention designs a cross-PC data calling solution, and realizes data interaction between different algorithms by using communication interaction and data synchronization methods. This can conveniently and quickly realize the application of algorithm data association under complex visual structures.

[0046] The above description only describes the specific implementation mode of the present invention. Various examples do not limit the essential content of the present invention. After reading the description, ordinary technicians in the relevant technical field can modify or deform the specific implementation modes described above without departing from the essence and scope of the invention.

Claims

1. A machine vision algorithm framework standardization platform, characterized by: It includes a data processing unit, which is configured on different PC terminals. Data is transmitted between the data processing units and between the data processing units and the client through a communication module. The data processing unit includes an algorithm framework, which processes the data output by the data acquisition device through an algorithm flow and outputs a detection result. When the associated output data of a configured algorithm is used as an associated input by an algorithm under another data processing unit, the algorithm framework is implemented by a cross-PC call function. The algorithm framework includes algorithm information configuration, and the algorithm information configuration types include detection parameters, result parameters, simulation parameters, and algorithm-related input parameters and algorithm-related output parameters; the algorithm information configuration method is: parameter information configuration is performed in an Excel table, and an XML configuration file that meets the framework standard is automatically generated by a parameter generation tool; the algorithm information configuration content parameter name, translation name, data type, parameter interface type, parameter range, parameter authority, and the display method of the result parameter on the client side, and the client detection result display method supports the permutation and combination of display in an image, display in a result list, and display in a data report; The algorithm information configuration does not require parsing the parameter configuration file. After the algorithm framework is parsed, the parameter information is passed to the algorithm. The algorithm obtains the parameters and executes the algorithm process. The dynamic library is compiled and placed in the specified location. The algorithm framework includes a data acquisition module, a device module, a camera acquisition and distribution module, a standard algorithm pool module, an image processing management module and a communication module; the acquired data selects the required algorithm from the algorithm library to form an algorithm set, and the algorithm framework supports configuring different algorithm sets for each data; The algorithm framework includes template creation, sending and data loading. Template creation is completed on the client through interface configuration. Algorithm tasks are assigned and detection parameters are set for each collected data, and the association relationship between algorithms is set. The algorithm framework includes three kinds of association configurations: the association relationship between algorithms configured within the same collected data, the association relationship between algorithms configured within the same data processing unit, and the association relationship between algorithms configured for data collected by different data processing units. After the template is created, the communication module sends the template data to each data processing unit. The algorithm framework of each data processing unit parses the template data, adjusts the algorithm execution order according to the correlation between algorithms, and passes the parsed algorithm parameters to the corresponding algorithm modules. The data processing unit executes the algorithms configured for each collected data according to the detection order adjusted by the algorithm framework, that is, calls the algorithm dynamic library; the algorithm dynamic library performs calculations and associated data outputs based on its own algorithm parameters and associated data input; when the associated output data written by the algorithm is used by algorithms under other data processing units, the algorithm framework sends the associated data to the data processing unit where the algorithm using the associated data is located through the communication module, and then the algorithm framework of the data processing unit sends the associated data to the algorithm using the associated data.

Images