Nuclear island inspection robot instrument positioning identification method and system based on dynamic target matching

Abstract

The invention belongs to the target intelligent identification technical field and relates to a nuclear island inspection robot instrument positioning identification method and system based on dynamic target matching. Comprising a task planning module, a target object sampling module calls a visual device to collect instrument image files, a file correction sub-module is used for conducting manual and automatic correction preprocessing operation on the collected image files, and the image files and information data obtained after correction preprocessing operation are provided for a target object recognition calculation module to call; the motion posture adjustment executes posture adjustment of the inspection device; the target positioning tool is used for implementing a feature point positioning function of an inspection instrument object picture sample; the target template generation tool is a special algorithm tool for configuring instrument object template data. The method has the beneficial effects that a computer vision method based on feature image comparison is adopted, a special algorithm tool is packaged, and the poses of the inspection device can be adjusted in real time according to a positioning recognition result by virtue of an inspection device motion axis coding linkage function module.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent target recognition of nuclear power plant inspection devices, and in particular relates to a method and system for positioning and recognizing instruments of a nuclear island inspection robot based on dynamic target matching. Background Art

[0002] During the nuclear island inspection, various instruments are limited by ambient lighting, complex background and objective conditions of the installation location. Conventional means and traditional manual inspections make it difficult to achieve rapid and automated positioning and identification. Ordinary visual observation methods cannot be linked with the motion axis of the inspection device and are therefore less efficient. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for positioning and identifying instruments of a nuclear island inspection robot based on dynamic target matching. By encapsulating a special algorithm tool, the nuclear island inspection robot can realize the implementation of the fully automatic instrument inspection task of "initial identification-result feedback-posture adjustment-identification completion" during autonomous navigation.

[0004] The technical solution of the present invention is as follows: A nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching, comprising:

[0005] The task planning module is used to realize basic data loading and instrument inspection task information generation;

[0006] The target object sampling calls the visual device to collect the instrument image file, and uses the file correction submodule to perform manual and automatic correction preprocessing operations on the collected image file, and the image file and information data after the correction preprocessing operation are called by the target object recognition calculation module;

[0007] The target object recognition calculation obtains and parses the sample image file and supporting information through the data transmission interface, calls the target positioning tool and template comparison submodule to perform LBP transformation, feature point comparison and multi-resolution moving similarity processing, and finally standardizes and assembles the object positioning and recognition results;

[0008] The motion posture adjustment executes the posture adjustment of the inspection device to form the target encoding information vector V of each motion axis me , so that the inspection device can adjust its posture to carry out the next instrument target object sampling;

[0009] The target positioning tool is used to implement the feature point positioning function of the inspection instrument object image sample;

[0010] The target template generation tool is a special algorithm tool for configuring instrument object template data.

[0011] The task planning module includes a target template generation tool calling interface, a basic configuration information library, a task data information library and a data communication transmission submodule.

[0012] The target object sampling module includes a visual device access interface, a file correction submodule and a data transmission submodule.

[0013] The target object recognition calculation includes: a template comparison submodule, a target positioning tool calling interface, a calculation process information storage library, a recognition result assembly submodule and a data transmission submodule.

[0014] The motion posture adjustment module comprises: an object recognition result information analysis submodule, a motion axis code generation submodule, and a data transmission submodule.

[0015] The target positioning tool comprises an information loading submodule, a sample transformation submodule, and a positioning result feedback submodule;

[0016] Information loading submodule: includes process data repository, information parsing operator, sample normalization operator; reads the instrument sample object to be identified and generates sample data vector V s-d ;

[0017] Sample transformation submodule: includes process data repository, sample processing operator, and processing result normalization operator; reads sample data information vector V s-d , improve the sample calculation parameter vector V according to the calculation requirements s-c-p ; Load parameter vector V s-c-p , for the template data vector V t-d The image elements in the image are subjected to grayscale transformation, LBP calculation and multi-resolution moving similarity processing to generate the sample transformation result matrix M s-r ;

[0018] Positioning result feedback submodule: includes process data repository, feature matching operator, positioning result normalization operator; loads feature data matrix M s-d , execute the sample transformation result matrix M s-r With the feature data matrix M s-d Similarity calculation to generate positioning result matrix M d-r .

[0019] The target template generation tool comprises: a template assembly submodule, a parameter setting submodule, an image processing submodule, and a feature data assembly submodule;

[0020] Template assembly submodule: includes process data repository, data parsing operator, and template normalization operator; it parses and assembles the input test file and its associated information to generate a template data vector V t-d ;

[0021] Parameter setting submodule: contains process data repository, parameter matching operator, parameter normalization operator; reads template data vector Vt-d , based on the preset conversion requirements and real-time calculation needs, improve and update the template parameter vector V t-c ;

[0022] Image processing submodule: includes process data repository, image operation operator, result normalization operator; calculates parameter vector V based on template t-c For the template data vector V t-d The image elements in the image are subjected to binarization, region segmentation, grayscale transformation, LBP calculation and other calculation operations to generate the image processing result matrix M p-r ;

[0023] Feature data generation submodule: includes process data repository, label processing operator, feature data normalization operator; based on the image processing result matrix M p-r , perform center point and feature point calculation, structural feature point marking operation, and generate feature data matrix M s-d .

[0025] A method for positioning and identifying instruments of a nuclear island inspection robot based on dynamic target matching comprises the following steps:

[0026] Step 1: Template generation, including

[0027] Step 11: Image and information loading

[0028] Call the target template generation tool calling interface in the target template generation tool to parse and assemble the input test file and its associated information to generate a template data vector V for use in subsequent steps t-d ;

[0029] Step 12: Image Processing Calculations

[0030] Read the template data vector V t-d Based on the preset conversion requirements and real-time computing needs, it performs binarization, region segmentation, grayscale transformation, LBP calculation and other operations to improve and update the template parameter vector V t-c , generate the transformation completion result matrix M tr ;

[0031] Step 13: Template feature data generation

[0032] Based on the image processing result matrix M p-r , perform center point and feature point calculation, structural feature point marking operation, and generate feature data matrix M s-d .

[0033] Step 2: Task loading, including

[0034] Step 21: Loading basic data

[0035] The task planning module loads the basic data information required for the current task;

[0036] Step 22: Generate task information

[0037] Load the feature data matrix M s-d The information is assembled together with the basic data information of the task to generate task information;

[0038] Step 23: Task information storage / call

[0039] The task information is stored in the task information library, and the task information is sent and called externally through the data transmission submodule;

[0040] Step 3: Object recognition, including

[0041] Step 31: Instrument Object Image Collection

[0042] Use the target object acquisition module to call the camera or other imaging devices through the visual device access interface to obtain the instrument image file;

[0043] Step 32: Image Correction

[0044] Perform manual and automatic correction operations on the captured image files through the file correction loading submodule;

[0045] Step 33: Image Processing

[0046] Use the target object recognition calculation module to parse the corrected image file and supporting information, call the target positioning tool and template comparison submodule, and perform grayscale transformation, LBP calculation and multi-resolution moving similarity transformation processing;

[0047] Step 34: Image feature point comparison processing

[0048] Perform feature point comparison calculation on the processed image;

[0049] Step 35: Calculation result storage

[0050] The image positioning result information after image processing and feature point comparison is normalized and assembled and then fed back to the process data repository in the calling positioning result feedback submodule.

[0051] Step 4: Identification result feedback judgment, including

[0052] Step 41: Data loading

[0053] Using the positioning result feedback submodule, load the stored "Step 3 Object Identification" calculation result data in the process data repository;

[0054] Step 42: Comparison and judgment

[0055] By calling the data in the process data repository in the positioning result feedback submodule, the feature matching operator and the positioning result normalization operator are called; by loading the feature data matrix M s-d , execute the sample transformation result matrix M s-r With the feature data matrix M s-d Similarity calculation to generate positioning result matrix M d-r ;

[0056] Step 43: Comparison result feedback

[0057] The positioning result matrix M will be generated d-r Standardize post-assembly feedback;

[0058] Step 5: Pose adjustment, including

[0059] Step 51: Instrument positioning identification information analysis: Use the motion posture adjustment module to call the object recognition result analysis submodule to obtain the positioning result information; judge the feedback result according to step 4. If the current instrument recognition result does not meet the requirements, the inspection device posture adjustment is required to form the target encoding information vector V of each motion axis. me , so that the inspection device can adjust its posture to carry out the next instrument target object sampling;

[0060] Step 52: Motion axis encoding conversion generation

[0061] Based on the positioning result information after analysis, if the current instrument recognition result does not meet the requirements, the converted target encoding information is assembled with the motion parameters such as the axis motion direction and axis motion speed to form the motion axis encoding information vector V me ;

[0062] Step 53: Export the motion axis encoding information

[0063] Through the data transmission submodule, the motion axis encoding information vector is transmitted in real time for the inspection device to perform posture adjustment; then it returns to "Step 3 Object Identification" and performs the object identification calculation process again;

[0064] Step 6: Results generation, including

[0065] Step 61: Result judgment

[0066] If the result of "step 4 recognition result feedback judgment" meets the requirements, load the process data repository in the process of "step 3.5 calculation result storage"; if the current instrument recognition result meets the requirements, the inspection device posture adjustment is no longer performed, and the current instrument target object sampling is available;

[0067] Step 62: Generate result information

[0068] Based on, call the positioning result matrix M generated by "step 4.3 comparison result feedback" d-r , obtain the parsed positioning result information after the standardized assembly;

[0069] Step 63: Export result information

[0070] The final result transmission of "Step 6.2" is completed through the data transmission submodule.

[0071] The beneficial effects of the present invention are: by adopting a computer vision method based on feature image comparison, encapsulating a special algorithm tool and a patrol device motion axis encoding linkage function module, the posture of the patrol device can be adjusted in real time according to the positioning and recognition results, thereby realizing the automation of instrument positioning and recognition operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 Generate a tool structure diagram for the target template;

[0073] Figure 2 This is a schematic diagram of the target positioning tool structure;

[0074] Figure 3 This is a schematic diagram of the structure of a nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching provided by the present invention;

[0075] Figure 4 This is a flow chart of a method for positioning and identifying instruments of a nuclear island inspection robot based on dynamic target matching provided by the present invention. DETAILED DESCRIPTION

[0076] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0077] Aiming at the characteristics of instrument inspection business in nuclear power environment, the present invention focuses on two key technologies, namely "Target Template Producing Tool (TTPT)" and "Target Locating Tool (TLT)", and designs a full set of process "Task Planning (TP)" module, "Target Acquiring (TA)" module, "Target Recognizing (TR)" module and "Motion Pose Adjusting (MPA)" module. The following is the detailed content of the technical solution.

[0078] A nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching, comprising:

[0079] The Target Template Tool (TTPT) is a special algorithm tool for configuring instrument object template data. Figure 1 (5). The template generation tool (TGT) includes: a template assembly submodule, a parameter setting submodule, an image processing submodule, and a feature data assembly submodule;

[0080] Template assembly submodule: includes process data repository, data parsing operator, and template normalization operator; it parses and assembles the input test file and its associated information to generate a template data vector V t-d ;

[0081] The process data repository is used to store structured data (such as instrument reading characteristic values, range value ranges corresponding to instrument types, and category label information) and unstructured data (original image files, template attachments) involved in the calculation process; the data parsing operator loads the original file to be processed, obtains image-related parameters such as file size and file capacity, and merges label information such as feature points and interest area marks specified by the user; the template normalization operator performs template file configuration operations to form a data template, including a file header, content items, and type labels; the parsing and assembly function performs operations on the file to be tested by using the data parsing operator, and loads the data template formed by the template planning operator, performs a structural matching operation of the data template (the operation process data are all stored in the process data repository), and finally forms a template data vector V t-d .

[0082] Parameter setting submodule: contains process data repository, parameter matching operator, parameter normalization operator; reads template data vector V t-d , based on the preset conversion requirements and real-time calculation needs, improve and update the template parameter vector V t-c ;

[0083] Parameter matching operator: used to perform data content matching calculations for parameter entries and to mark mismatched results; parameter normalization operator: used to filter and adjust target values ​​for results that do not satisfy matching calculations; template parameter vector function: load target data vector V t-d , by using parameter matching operators and parameter normalization operators to perform content matching calculations and result screening and adjustment calculations, and the calculated results are updated to the template parameter vector V t-c .

[0084] Image processing submodule: includes process data repository, image operation operator, result normalization operator; calculates parameter vector V based on template t-c For the template data vector V t-dThe image elements in the image are subjected to binarization, region segmentation, grayscale transformation, LBP calculation and other calculation operations to generate the image processing result matrix M p-r ;

[0085] Image operation operator: used to load image execution template data vector V t-d The image elements in the image are used to perform image processing operations such as binarization, region segmentation, grayscale transformation, and LBP calculation on the image; the result normalization operator: for the processing results of the image operation operator, the file header, file body, and file label are spliced ​​to form an image processing result matrix M with a fixed format. p-r .

[0086] Feature data assembly submodule: includes process data repository, label processing operator, feature data normalization operator; based on the image processing result matrix M p-r , perform center point and feature point calculation, structural feature point marking operation, and generate feature data matrix M s-d .

[0087] Marking processing operator: based on the image processing result matrix M p-r , based on the image geometric center coordinates and the configured relative feature point coordinates, perform center point marking, feature point marking, and structural feature point marking operations; feature data normalization operator: for the processing results of the image operation operator, splice the file header, file body, and file label, and encapsulate to form an image processing result matrix M with a fixed format p-r .

[0088] The target location tool (TLT) is used to implement the feature point location function of the inspection instrument object image sample. The structure composition is shown in Figure 1 (6); The target positioning tool (TLT) includes an information loading submodule, a sample transformation submodule, and a positioning result feedback submodule.

[0089] Information loading submodule: includes process data repository, information parsing operator, sample normalization operator; reads the instrument sample object to be identified and generates sample data vector V s-d .

[0090] Information parsing operator: loads the image file to be processed, obtains image-related parameters such as image file size and file capacity, and stores the file information in the process data repository; sample normalization operator: performs file header, file body and file label splicing operations on the data processed by the information parsing operator to form a standardized sample file.

[0091] Sample transformation submodule: includes process data repository, sample processing operator, and processing result normalization operator; reads sample data information vector V s-d , improve the sample calculation parameter vector V according to the calculation requirementss-c-p ; Load parameter vector V s-c-p , for the template data vector V t-d The image elements in the image are subjected to grayscale transformation, LBP calculation and multi-resolution moving similarity processing to generate the sample transformation result matrix M s-r .

[0092] Sample processing operator: used to read sample data information vector V s-d , and the sample data information vector V s-d Execute grayscale transformation, LBP calculation and multi-resolution moving similarity processing calculation; Processing result normalization operator: concatenate the file header, file body and file label of the sample processed data to form a standardized sample transformation result matrix M s-r .

[0093] Positioning result feedback submodule: includes process data repository, feature matching operator, positioning result normalization operator; loads feature data matrix M s-d , execute the sample transformation result matrix M s-r With the feature data matrix M s-d Similarity calculation to generate positioning result matrix M d-r .

[0094] Feature matching operator: used to load the feature data matrix M s-d , and transform the sample result matrix M s-r With the feature data matrix M s-d Perform overall similarity calculation; Positioning result normalization operator: perform file header, file body and file label splicing operations on the result data after similarity calculation to form a positioning result matrix M d-r .

[0095] Task Planning Module (TP): used to load basic data and generate instrument inspection task information; the task planning module includes: target template generation tool (TTPT) calling interface, basic configuration information library, task data information library, data communication transmission submodule; structure diagram Figure 3 (1).

[0096] Target object sampling (TA): call the camera and other visual devices to collect instrument image files, use the file correction submodule to perform manual and automatic correction preprocessing operations on the collected image files, and provide the image files and information data after the correction preprocessing operations to the target object recognition calculation module (TR) for calling; the target object sampling (TA) module includes: visual device access interface, file correction submodule, data transmission submodule; structure diagram Figure 3 (2).

[0097] Module function realization: Target object sampling (TA) calls visual devices such as cameras to collect instrument image files through the equipment, and then performs manual and automatic correction preprocessing operations on the collected image files by using the file correction submodule inside the module, and the image files and information data after the correction preprocessing operation are provided to the target object recognition calculation module (TR) for calling.

[0098] Target object recognition calculation (TR): obtain and parse sample image files and supporting information through the data transmission interface, call the target positioning tool (TLT) and template comparison submodule to perform LBP transformation, feature point comparison and multi-resolution moving similarity processing and other calculation operations, and finally assemble the object positioning and recognition results in a standardized manner; Target object recognition calculation (TR) includes: template comparison submodule, target positioning tool calling interface, calculation process information repository, recognition result assembly submodule, data transmission submodule; structure diagram Figure 3 (3).

[0099] Module function implementation: The target object recognition calculation (TR) obtains and parses the sample image file and supporting information through the data transmission interface, and then calls the target positioning tool (TLT) and the template comparison submodule inside the module to perform calculation operations such as LBP transformation, feature point comparison and multi-resolution moving similarity processing, and finally standardizes and assembles the object positioning and recognition results.

[0100] Motion Posture Adjustment (MPA): Perform posture adjustment of the inspection device to form the target encoding information vector V of each motion axis me , for the inspection device to adjust its posture for the next instrument target object sampling; the motion posture adjustment (MPA) module includes: object recognition result information analysis submodule, motion axis code generation submodule, data transmission submodule; structure diagram Figure 3 (4).

[0101] Module function realization: Use the object recognition result information parsing submodule to obtain the instrument recognition result information. If the recognition result meets the requirements, the detailed motion axis target coding information to the next point is generated through the motion axis coding generation submodule; if the recognition result does not meet the requirements, the target coding value information of each motion axis of the robot required to perform instrument sampling again is generated; the generated motion axis target coding values ​​are all in the form of target coding information vector V me Storage form.

[0102] A method for positioning and identifying instruments of a nuclear island inspection robot based on dynamic target matching comprises the following steps:

[0103] Step 1: Template generation, including

[0104] Step 11: Image and information loading

[0105] Call the target template generation tool calling interface in the target template generation tool to parse and assemble the input test file and its associated information to generate a template data vector V for use in subsequent steps t-d ;

[0106] Step 12: Image Processing Calculations

[0107] Read the template data vector V t-d Based on the preset conversion requirements and real-time computing needs, it performs binarization, region segmentation, grayscale transformation, LBP calculation and other operations to improve and update the template parameter vector V t-c , generate the transformation completion result matrix M tr ;

[0108] Step 13: Template feature data generation

[0109] Based on the image processing result matrix M p-r , perform center point and feature point calculation, structural feature point marking operation, and generate feature data matrix M s-d .

[0110] Step 2: Task loading, including

[0111] Step 21: Loading basic data

[0112] The task planning module (TP) loads the basic data information required for the current task;

[0113] Step 22: Generate task information

[0114] Load the feature data matrix M s-d The information is assembled together with the basic data information of the task to generate task information;

[0115] Step 23: Task information storage / call

[0116] The task information is stored in the task information library, and the task information is sent and called externally through the data transmission submodule.

[0117] Step 3: Object recognition, including

[0118] Step 31: Instrument Object Image Collection

[0119] Use the target object acquisition module (TA) to call the camera or other imaging devices through the visual device access interface to obtain the instrument image file;

[0120] Step 32: Image Correction

[0121] Perform manual and automatic correction operations on the captured image files through the file correction loading submodule;

[0122] Step 33: Image Processing

[0123] Use the target object recognition calculation module (TR) to parse the corrected image file and supporting information, call the target location tool (TLT) and template comparison submodule, and perform grayscale transformation, LBP calculation and multi-resolution moving similarity transformation processing;

[0124] Step 34: Image feature point comparison processing

[0125] Perform feature point comparison calculation on the processed image;

[0126] Step 35: Calculation result storage

[0127] The image positioning result information after image processing and feature point comparison is normalized and assembled and then fed back to the process data repository in the calling positioning result feedback submodule.

[0128] Step 4: Identification result feedback judgment, including

[0129] Step 41: Data loading

[0130] Using the positioning result feedback submodule, load the stored "Step 3 Object Identification" calculation result data in the process data repository;

[0131] Step 42: Comparison and judgment

[0132] By calling the data in the process data repository in the positioning result feedback submodule, the feature matching operator and the positioning result normalization operator are called; by loading the feature data matrix M s-d , execute the sample transformation result matrix M s-r With the feature data matrix M s-d Similarity calculation to generate positioning result matrix M d-r ;

[0133] Step 43: Comparison result feedback

[0134] The positioning result matrix M will be generated d-r Normalize post-assembly feedback.

[0135] Step 5: Posture adjustment, including

[0136] Step 51: Instrument positioning identification information analysis: Use the motion posture adjustment module (MPA) to call the object recognition result analysis submodule to obtain the positioning result information; judge the feedback result according to step 4. If the current instrument recognition result does not meet the requirements, the inspection device posture adjustment is required to form the target encoding information vector V of each motion axis. me , so that the inspection device can adjust its posture to carry out the next instrument target object sampling;

[0137] Step 52: Motion axis encoding conversion generation

[0138] Based on the positioning result information after analysis, if the current instrument recognition result does not meet the requirements, the converted target encoding information is assembled with the motion parameters such as the axis motion direction and axis motion speed to form the motion axis encoding information vector V me ;

[0139] Step 53: Export the motion axis encoding information

[0140] Through the data transmission submodule, the motion axis encoding information vector is transmitted in real time for the inspection device to perform posture adjustment; then it returns to "Step 3 Object Identification" and performs the object identification calculation process again;

[0141] Step 6: Results generation, including

[0142] Step 61: Result judgment

[0143] If the result of "step 4 recognition result feedback judgment" meets the requirements, load the process data repository in the process of "step 3.5 calculation result storage"; if the current instrument recognition result meets the requirements, the inspection device posture adjustment is no longer performed, and the current instrument target object sampling is available;

[0144] Step 62: Generate result information

[0145] Based on, call the positioning result matrix M generated by "step 4.3 comparison result feedback" d-r , obtain the parsed positioning result information after the standardized assembly;

[0146] Step 63: Export result information

[0147] The final result transmission of "Step 6.2" is completed through the data transmission submodule.

Claims

1. A nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching, characterized in that: include: The task planning module is used to realize basic data loading and instrument inspection task information generation; The target object sampling calls the visual device to collect the instrument image file, and uses the file correction submodule to perform manual and automatic correction preprocessing operations on the collected image file, and the image file and information data after the correction preprocessing operation are called by the target object recognition calculation module; The target object recognition calculation obtains and parses the sample image file and supporting information through the data transmission interface, calls the target positioning tool and template comparison submodule to perform LBP transformation, feature point comparison and multi-resolution moving similarity processing, and finally standardizes and assembles the object positioning and recognition results; The motion posture adjustment executes the posture adjustment of the inspection device to form the target encoding information vector V of each motion axis me , so that the inspection device can adjust its posture to carry out the next instrument target object sampling; The target positioning tool is used to implement the feature point positioning function of the inspection instrument object image sample; The target template generation tool is a special algorithm tool for configuring instrument object template data.

2. The nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching as claimed in claim 1 is characterized by: The task planning module includes a target template generation tool calling interface, a basic configuration information library, a task data information library and a data communication transmission submodule.

3. The nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching as claimed in claim 1 is characterized by: The target object sampling module includes a visual device access interface, a file correction submodule and a data transmission submodule.

4. The nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching as claimed in claim 1, characterized in that: The target object recognition calculation includes: a template comparison submodule, a target positioning tool calling interface, a calculation process information storage library, a recognition result assembly submodule and a data transmission submodule.

5. The nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching as claimed in claim 1, characterized in that: The motion posture adjustment module comprises: an object recognition result information analysis submodule, a motion axis code generation submodule, and a data transmission submodule.

6. The nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching as claimed in claim 1, characterized in that: The target positioning tool comprises an information loading submodule, a sample transformation submodule, and a positioning result feedback submodule; Information loading submodule: includes process data repository, information parsing operator, sample normalization operator; reads the instrument sample object to be identified and generates sample data vector V s-d ; Sample transformation submodule: includes process data repository, sample processing operator, and processing result normalization operator; reads sample data information vector V s-d , improve the sample calculation parameter vector V according to the calculation requirements s-c-p ; Load parameter vector V s-c-p , for the template data vector V t-d The image elements in the image are subjected to grayscale transformation, LBP calculation and multi-resolution moving similarity processing to generate the sample transformation result matrix M s-r ; Positioning result feedback submodule: includes process data repository, feature matching operator, and positioning result normalization operator; Load the feature data matrix M s-d , execute the sample transformation result matrix M s-r With the feature data matrix M s-d Similarity calculation to generate positioning result matrix M d-r .

7. The nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching as claimed in claim 1 is characterized by: The target template generation tool comprises: a template assembly submodule, a parameter setting submodule, an image processing submodule, and a feature data assembly submodule; Template assembly submodule: includes process data repository, data parsing operator, and template normalization operator; it parses and assembles the input test file and its associated information to generate a template data vector V t-d ; Parameter setting submodule: contains process data repository, parameter matching operator, parameter normalization operator; reads template data vector V t-d , based on the preset conversion requirements and real-time calculation needs, improve and update the template parameter vector V t-c ; Image processing submodule: includes process data repository, image operation operator, result normalization operator; calculates parameter vector V based on template t-c For the template data vector V t-d The image elements in the image are subjected to binarization, region segmentation, grayscale transformation, LBP calculation and other calculation operations to generate the image processing result matrix M p-r ; Feature data generation submodule: includes process data repository, label processing operator, feature data normalization operator; based on the image processing result matrix M p-r , perform center point and feature point calculation, structural feature point marking operation, and generate feature data matrix M s-d .

8. A method for positioning and identifying instruments of a nuclear island inspection robot based on dynamic target matching, comprising the following steps: Step 1: Template generation, including Step 11: Image and information loading Call the target template generation tool calling interface in the target template generation tool to parse and assemble the input test file and its associated information to generate a template data vector V for use in subsequent steps t-d ; Step 12: Image Processing Calculations Read the template data vector V t-d Based on the preset conversion requirements and real-time computing needs, it performs binarization, region segmentation, grayscale transformation, LBP calculation and other operations to improve and update the template parameter vector V t-c , generate the transformation completion result matrix M tr ; Step 13: Template feature data generation Based on the image processing result matrix M p-r , perform center point and feature point calculation, structural feature point marking operation, and generate feature data matrix M s-d . Step 2: Task loading, including Step 21: Loading basic data The task planning module loads the basic data information required for the current task; Step 22: Generate task information Load the feature data matrix M s-d The information is assembled together with the basic data information of the task to generate task information; Step 23: Task information storage / call The task information is stored in the task information library, and the task information is sent and called externally through the data transmission submodule; Step 3: Object recognition, including Step 31: Instrument Object Image Collection Use the target object acquisition module to call the camera or other imaging devices through the visual device access interface to obtain the instrument image file; Step 32: Image Correction Perform manual and automatic correction operations on the collected image files through the file correction loading submodule; Step 33: Image Processing Use the target object recognition calculation module to parse the corrected image file and supporting information, call the target positioning tool and template comparison submodule, and perform grayscale transformation, LBP calculation and multi-resolution moving similarity transformation processing; Step 34: Image feature point comparison processing Perform feature point comparison calculation on the processed image; Step 35: Calculation result storage The image positioning result information after image processing and feature point comparison is normalized and assembled and then fed back to the process data repository in the calling positioning result feedback submodule. Step 4: Identification result feedback judgment, including Step 41: Data loading Use the positioning result feedback submodule to load the stored "Step 3 Object Identification" calculation result data in the process data repository; Step 42: Comparison and judgment By calling the data in the process data repository in the positioning result feedback submodule, the feature matching operator and the positioning result normalization operator are called; by loading the feature data matrix M s-d , execute the sample transformation result matrix M s-r With the feature data matrix M s-d Similarity calculation to generate positioning result matrix M d-r ; Step 43: Comparison result feedback The positioning result matrix M will be generated d-r Standardize post-assembly feedback; Step 5: Posture adjustment, including Step 51: Instrument positioning identification information analysis: Use the motion posture adjustment module to call the object recognition result analysis submodule to obtain positioning result information; According to the feedback result of step 4, if the current instrument recognition result does not meet the requirements, the inspection device posture adjustment needs to be performed to form the target encoding information vector V of each motion axis. me , so that the inspection device can adjust its posture to carry out the next instrument target object sampling; Step 52: Motion axis encoding conversion generation Based on the positioning result information after analysis, if the current instrument recognition result does not meet the requirements, the converted target encoding information is assembled with the motion parameters such as the axis motion direction and axis motion speed to form the motion axis encoding information vector V me ; Step 53: Export the motion axis encoding information Through the data transmission submodule, the motion axis encoding information vector is transmitted in real time for the inspection device to perform posture adjustment; then it returns to "Step 3 Object Identification" to perform the object identification calculation process again; Step 6: Results generation, including Step 61: Result judgment If the result of "Step 4 Identification Result Feedback Judgment" meets the requirements, load the process data repository in the process of "Step 3.5 Calculation Result Storage"; If the current instrument recognition result meets the requirements, the inspection device posture adjustment is no longer performed, and the current instrument target object sampling is available; Step 62: Generate result information Based on, call the positioning result matrix M generated by "step 4.3 comparison result feedback" d-r , obtain the parsed positioning result information after the standardized assembly; Step 63: Export result information The final result transmission of "Step 6.2" is completed through the data transmission submodule.

Images