A method and system for instrument positioning and identification of a nuclear island inspection robot based on dynamic target matching

Abstract

The application belongs to the technical field of intelligent identification, and particularly relates to a nuclear island inspection robot instrument positioning identification method and system based on dynamic target matching. The method comprises the following steps: a task planning module, a target object sampling calling vision device collects an instrument image file, manual and automatic correction preprocessing operations are performed on the collected image file by using a file correction submodule, and the image file and information data after the correction preprocessing operations are called by a target object identification calculation module; a motion posture adjustment executes a position and posture adjustment of an inspection device; a target positioning tool is used to implement a feature point positioning function of an inspection instrument object picture sample; and a 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, special algorithm tools and an inspection device motion axis encoding linkage function module are encapsulated, and the position and posture of the inspection device can be adjusted in real time according to the positioning identification result.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of target intelligent identification of nuclear power plant inspection devices, and particularly relates to a nuclear island inspection robot instrument positioning identification method and system based on dynamic target matching. BACKGROUND

[0002] In the process of nuclear island inspection, various instruments are limited by environmental illumination, complex background and installation position objective conditions, and conventional means and traditional manual inspection are difficult to realize rapid automatic positioning identification; ordinary visual observation method cannot be linked with the movement axis of the inspection device, and the efficiency is low. SUMMARY

[0003] The purpose of the application is to provide a nuclear island inspection robot instrument positioning identification method and system based on dynamic target matching, which realizes the full-automatic instrument inspection task implementation of "initial identification-result feedback-pose adjustment-identification completion" of the nuclear island inspection robot in the process of autonomous navigation by encapsulating special algorithm tools.

[0004] The technical scheme of the application is as follows: a nuclear island inspection robot instrument positioning identification system based on dynamic target matching, comprising:

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

[0006] The target object sampling calls the visual equipment to collect instrument image files, and the corrected pre-processing operation is performed on the collected image files by using the file correction submodule, and the image files and information data after the corrected pre-processing operation are called by the target object identification calculation module;

[0007] The target object identification calculation obtains and analyzes the sample picture files and supporting information through the data transmission interface, calls the target positioning tool and the template comparison submodule to perform LBP transformation, feature point comparison and multi-resolution mobile similarity processing, and finally normalizes and assembles the object positioning identification result;

[0008] The motion posture adjustment executes the pose adjustment of the inspection device, forms the target coding information vector V me of each motion axis, and provides the inspection device with the pose adjustment for the next instrument target object sampling;

[0009] The target positioning tool is used for implementing the feature point positioning function of the inspection instrument object picture sample;

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

[0011] The task planning module comprises 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 comprises a visual equipment access interface, a file correction submodule, and a data transmission submodule.

[0013] The target object recognition calculation comprises a template comparison submodule, a target positioning tool calling interface, a calculation process information repository, 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] The information loading submodule comprises a process data repository, an information analysis operator, and a sample normalization operator; reads a sample object to be recognized, and generates a sample data vector V s-d .

[0017] The sample transformation submodule comprises a process data repository, a sample processing operator, and a processing result normalization operator; reads a sample data information vector V s-d , perfects a sample calculation parameter vector V s-c-p according to a calculation requirement; loads the parameter vector V s-c-p , performs grayscale transformation, LBP calculation, and multi-resolution moving similarity processing on image elements in a template data vector V t-d , and generates a sample transformation result matrix M s-r .

[0018] The positioning result feedback submodule comprises a process data repository, a feature matching operator, and a positioning result normalization operator; loads a feature data matrix M s-d , performs similarity calculation on a sample transformation result matrix M s-r and the feature data matrix M s-d , and generates a 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] The template assembly submodule comprises a process data repository, a data analysis operator, and a template normalization operator; analyzes and assembles an input file to be measured and the information thereof, and generates a template data vector V t-d .

[0021] The parameter setting submodule comprises a process data repository, a parameter matching operator, and a parameter normalization operator; reads the template data vector Vt-d , based on the preset conversion requirements and real-time computing needs, perfect and update the template parameter vector V t-c ;

[0022] Image processing submodule: contains process data repository, image operation operator, result normalization operator; based on template calculation parameter vector V t-c For template data vector V t-d Image element in the binary processing, region segmentation, gray scale transformation, LBP calculation and other computing operations, generate image processing result matrix M p-r ;

[0023] Feature data generation submodule: contains process data repository, mark processing operator, feature data normalization operator; based on image processing result matrix M p-r , execute center point and feature point calculation, structure feature point marking operation, generate feature data matrix M s-d

[0024] A nuclear island inspection robot instrument positioning identification method based on dynamic target matching, comprising the following steps:

[0025] Step 1: Template generation, including

[0026] Step 11: Image and information loading

[0027] Call the target template generation tool in the target template generation tool call interface, parse and assemble the input test file and the information belonging to it, and generate the template data vector V used in the subsequent steps t-d ;

[0028] Step 12: Image processing calculation

[0029] Read the template data vector V t-d , based on the preset conversion requirements and real-time computing needs, execute binary processing, region segmentation, gray scale transformation, LBP calculation and other operations, perfect and update the template parameter vector V t-c , generate transformation result matrix M tr ;

[0030] Step 13: Template feature data generation

[0031] Based on the image processing result matrix M p-r , execute center point and feature point calculation, structure feature point marking operation, generate feature data matrix M s-d .

[0032] Step 2: Task loading, including

[0033] Step 21: Load basic data

[0034] The task planning module loads the basic data information required for the current task;

[0035] Step 22: Task information generation

[0036] Load feature data matrix M s-d The information is assembled with the task basic data information to generate task information;

[0037] Step 23: Task information storage / call

[0038] The task information is stored in the task information library, and the task information is executed externally sent and called through the data transmission submodule;

[0039] Step 3: Object recognition, including

[0040] Step 31: Instrument object image acquisition

[0041] Use the target object acquisition module to access the camera or other camera device through the visual device access interface to obtain the instrument image file;

[0042] Step 32: Image correction

[0043] The collected image file is manually and automatically corrected by the file correction loading submodule;

[0044] Step 33: Image processing

[0045] Use the target object recognition calculation module to parse the corrected picture file and supporting information, call the target positioning tool and template comparison submodule, and perform gray scale transformation, LBP calculation and multi-resolution moving similarity transformation processing;

[0046] Step 34: Image feature point comparison processing

[0047] The processed image is executed for feature point comparison calculation;

[0048] Step 35: Calculation result storage

[0049] After the image processing, the feature point comparison processing image positioning result information is standardized and assembled and fed back to the process data storage library in the calling positioning result feedback submodule.

[0050] Step 4: Recognition result feedback judgment, including

[0051] Step 41: Data loading

[0052] Use the positioning result feedback submodule to load the stored "step 3 object recognition" calculation result data in the process data storage library;

[0053] Step 42: Comparison judgment

[0054] By calling the data in the process data repository in the positioning result feedback submodule, calling the feature matching operator, the positioning result normalization operator; by loading the feature data matrix M s-d , executing sample transformation result matrix M s-r Similarity calculation with feature data matrix M s-d , generate positioning result matrix M d-r ;

[0055] Step 43: Feedback of comparison results

[0056] After the positioning result matrix M d-r Normalization assembly feedback is generated;

[0057] Step 5: Pose adjustment, including

[0058] Step 51: Instrument positioning identification information analysis: use the motion pose adjustment module, call the object recognition result analysis submodule, and obtain the 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 pose adjustment needs to be executed, and the target code information vector V me is formed to adjust the pose of the inspection device for the next instrument target object sampling;

[0059] Step 52: Motion axis code conversion generation

[0060] Based on the positioning result information after analysis, if the current instrument recognition result does not meet the requirements, the converted target code information and the motion parameters such as axis motion direction and axis motion rate are assembled to form the motion axis code information vector V me ;

[0061] Step 53: Export of motion axis code information

[0062] Through the data transmission submodule, the real-time transmission of the motion axis code information vector is completed, which is used for the inspection device to execute the pose adjustment; then return to "Step 3 object recognition" and execute the object recognition calculation process again;

[0063] Step 6: Result generation, including

[0064] Step 61: Result judgment

[0065] If the result of "Step 4 identification result feedback judgment" meets the requirements, load the process data repository in the "Step 3.5 calculation result storage" process; if the current instrument recognition result meets the requirements, the inspection device pose adjustment is not executed, and the current instrument target object sampling is available;

[0066] Step 62: Result information generation

[0067] Based on, call "step 4.3 alignment result feedback" generated positioning result matrix M d-r , get normalized assembly after analysis of positioning results information;

[0068] Step 63: result information export

[0069] Through the data transmission sub-module, the final result transmission of "step 6.2" is completed.

[0070] The beneficial effects of the present application are that: the computer vision method based on feature image comparison, encapsulation of special algorithm tool and inspection device motion axis encoding linkage function module can adjust the pose of the inspection device in real time according to the positioning and recognition result, and realize the automation of instrument positioning and recognition operation. BRIEF DESCRIPTION OF DRAWINGS

[0071] Figure 1 Generate a tool structure diagram for the target template;

[0072] Figure 2 Generate a tool structure diagram for the target positioning tool;

[0073] Figure 3 A nuclear island inspection robot instrument positioning and recognition system structure diagram based on dynamic target matching provided by the present application;

[0074] Figure 4 A nuclear island inspection robot instrument positioning and recognition method flow chart based on dynamic target matching provided by the present application. DETAILED DESCRIPTION

[0075] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0076] The present application is aimed at the instrument inspection business characteristics in the nuclear power environment, and around the two key technologies of "target template producing tool (TTPT)" and "target locating tool (TLT)", a complete set of process modules are designed, including "task planning (TP)", "target object acquiring (TA)", "target object recognizing (TR)" and "motion pose adjusting (MPA)", and the following is the detailed content of the technical solution.

[0077] A nuclear island inspection robot instrument positioning and recognition system based on dynamic target matching, comprising:

[0078] The target template generation tool (TTPT) is a special algorithm tool for configuring instrument object template data, and a structural diagram is shown Figure 1 The template generation tool (TGT) comprises a template assembly sub-module, a parameter setting sub-module, an image processing sub-module, and a feature data assembly sub-module.

[0079] The template assembly sub-module comprises a process data repository, a data parsing operator, and a template normalization operator; the input test file and the associated information are parsed and assembled to generate a template data vector V t-d .

[0080] 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 and template attachments) involved in the calculation process; the data parsing operator loads the original file to be processed to obtain image-related parameters such as file size and file capacity, and merges label information such as user-specified feature points and regions of interest; the template normalization operator performs template file configuration operations to form a data template, including a file header, content entries, and type labels; the parsing and assembly function: the data parsing operator is used to perform operations on the test file, and the data template formed by the template planning operator is loaded to perform structure matching operations on the data template (the operation process data is stored in the process data repository), and finally a template data vector V is formed t-d .

[0081] The parameter setting sub-module comprises a process data repository, a parameter matching operator, and a parameter normalization operator; the template data vector V t-d is read, and based on preset conversion requirements and real-time calculation requirements, the template parameter vector V t-c is perfected and updated;

[0082] The parameter matching operator is used to perform data content matching calculation of parameter entries, and specially marks the non-matching results; the parameter normalization operator is used to filter and adjust the target value of the results that do not satisfy the matching calculation; the function of perfecting and updating the template parameter vector: the target data vector V t-d is loaded, content matching calculation, result filtering and adjustment calculation are performed by using the parameter matching operator and the parameter normalization operator, and the calculation result is updated to the template parameter vector V t-c .

[0083] The image processing sub-module comprises a process data repository, an image operation operator, and a result normalization operator; based on the template calculation parameter vector V t-c , the template data vector V t-dThe middle image element performs binaryzation processing, region segmentation, gray scale transformation, LBP calculation and other calculation operations on the image to generate an image processing result matrix M p-r ;

[0084] Image operation operator: used for loading image execution template data vector V t-d The middle image element performs binaryzation processing, region segmentation, gray scale transformation, LBP calculation and other calculation operations on the image to generate an image processing result matrix M p-r .

[0085] Feature data assembly submodule: containing process data repository, label processing operator, feature data normalization operator; based on the image processing result matrix M p-r , performs center point and feature point calculation, structure feature point labeling operation, and generates a feature data matrix M s-d .

[0086] Label 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, performs center point labeling, feature point labeling, and structure feature point labeling operations; feature data normalization operator: based on the processing result of the image operation operator, splices the file header, file body and file label, and encapsulates to form an image processing result matrix M with fixed format p-r .

[0087] Target positioning tool (TLT) for implementing feature point positioning function of patrol instrument object picture sample, structure composition is seen Figure 1 in the middle (6); the target positioning tool (TLT) contains information loading submodule, sample transformation submodule, positioning result feedback submodule.

[0088] Information loading submodule: containing process data repository, information parsing operator, sample normalization operator; reads the instrument sample object to be recognized to generate a sample data vector V s-d .

[0089] Information parsing operator: loads the picture file to be processed, acquires image related parameters such as picture file size and file capacity, and stores the file information into the process data repository; sample normalization operator: performs file header, file body and file label splicing operation on the data processed by the information parsing operator to form a standardized sample file.

[0090] Sample transformation submodule: containing process data repository, sample processing operator, processing result normalization operator; reads the sample data information vector V s-d , and perfects the sample calculation parameter vector V according to the calculation requirements-c-p ; load parameter vector V s-c-p , for template data vector V t-d , perform gray scale transformation, LBP calculation and multi-resolution moving similarity processing on image elements in V s-r .

[0091] Sample processing operator: used to read sample data information vector V s-d ; and perform gray scale transformation, LBP calculation and multi-resolution moving similarity processing calculation on sample data information vector V s-d ; processing result normalization operator: performs file header, file body and file label splicing operation on sample processing data to form standardized sample transformation result matrix M s-r .

[0092] Positioning result feedback submodule: contains process data repository, feature matching operator, positioning result normalization operator; loads feature data matrix M s-d , performs sample transformation result matrix M s-r and feature data matrix M s-d similarity calculation to generate positioning result matrix M d-r .

[0093] Feature matching operator: used to load feature data matrix M s-d , and perform overall similarity calculation on sample transformation result matrix M s-r and feature data matrix M s-d ; positioning result normalization operator: performs file header, file body and file label splicing operation on similarity calculation processed result data to form positioning result matrix M d-r .

[0094] Task planning module (TP): used to realize basic data loading and instrument inspection task information generation; the task planning module includes: target template generation tool (TTPT) calling interface, basic configuration information library, task data information library, data communication transmission submodule; the structural diagram is shown in Figure 3 (1).

[0095] Target object sampling (TA): calling camera and other visual devices to collect instrument image files, and performing manual and automatic correction preprocessing operation on the collected image files by using the file correction submodule, and providing the corrected preprocessed image files and information data for the target object recognition calculation module (TR) to call; the target object sampling (TA) module includes: visual device access interface, file correction submodule, data transmission submodule; the structural diagram is shown in Figure 3 (2).

[0096] Module function implementation: target object sampling (TA) calls visual devices such as cameras, collects instrument image files through the device, and then manually and automatically corrects and preprocesses the collected image files through the file correction submodule inside the module, and provides the image files and information data after the correction and preprocessing operation to the target object recognition calculation module (TR) for calling.

[0097] Target object recognition calculation (TR): through the data transmission interface, obtain and analyze the sample picture files and supporting information, call the target positioning tool (TLT) and the template comparison submodule to perform LBP transformation, feature point comparison, and multi-resolution mobile similarity processing, and finally standardize the object positioning recognition result. Figure 3 (3).

[0098] Module function implementation: target object recognition calculation (TR) obtains and analyzes sample picture files 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 LBP transformation, feature point comparison, and multi-resolution mobile similarity processing, and finally standardizes the object positioning recognition result.

[0099] Motion posture adjustment (MPA): perform the position adjustment of the inspection device to form the target encoding information vector V me of each motion axis, so that the inspection device adjusts the position for the next instrument target object sampling; the motion posture adjustment (MPA) module includes: object recognition result information analysis submodule, motion axis encoding generation submodule, and data transmission submodule. Figure 3 (4).

[0100] Module function implementation: use the object recognition result information analysis submodule to obtain the instrument recognition result information, and if the recognition result meets the requirements, generate the detailed motion axis target encoding information of the next point position through the motion axis encoding generation submodule; if the recognition result does not meet the requirements, generate the target encoding value information of each motion axis of the robot required for re-executing instrument sampling; the generated motion axis target encoding values are stored in the form of target encoding information vector V me .

[0101] A nuclear island inspection robot instrument positioning and recognition method based on dynamic target matching, comprising the following steps:

[0102] Step 1: template generation, including

[0103] Step 11: image and information loading

[0104] The target template generation tool in the target template generation tool calling interface is called to parse and assemble the input test file and the information thereof, and a template data vector V is generated for subsequent steps t-d ;

[0105] Step 12: image processing calculation

[0106] Reading the template data vector V t-d , based on the preset conversion requirements and real-time calculation requirements, performing binaryzation processing, region segmentation, gray scale transformation, LBP calculation and other operations, perfecting and updating the template parameter vector V t-c , generating a transformation completion result matrix M tr ;

[0107] Step 13: template feature data generation

[0108] Based on the image processing result matrix M p-r , performing center point and feature point calculation, structure feature point marking operation, generating a feature data matrix M s-d .

[0109] Step 2: task loading, including

[0110] Step 21: basic data loading

[0111] The task planning module (TP) loads the basic data information required for the current task;

[0112] Step 22: task information generation

[0113] Loading the feature data matrix M s-d The information is assembled with the task basic data information to generate task information;

[0114] Step 23: task information storage / calling

[0115] The task information is stored in the task information library, and the task information is externally sent and called through the data transmission submodule.

[0116] Step 3: object recognition, including

[0117] Step 31: instrument object image acquisition

[0118] The target object acquisition module (TA) is used to access the camera or other camera devices through the visual device access interface to obtain instrument image files;

[0119] Step 32: image correction

[0120] Manual and automatic correction operations are performed on the collected image files through the file correction loading submodule;

[0121] Step 33: image processing

[0122] The corrected picture file and the supporting information are parsed using a target object recognition calculation module (TR), a target location tool (TLT) and a template comparison sub-module are called, and a gray scale transformation, LBP calculation and multi-resolution moving similarity transformation processing are performed;

[0123] Step 34: image feature point comparison processing

[0124] Feature point comparison calculation is performed on the processed image;

[0125] Step 35: calculation result storage

[0126] The image processing, feature point comparison processing and image positioning result information are standardized and assembled and then fed back to the process data storage library in the calling positioning result feedback sub-module.

[0127] Step 4: recognition result feedback judgment, including

[0128] Step 41: data loading

[0129] The "step 3 object recognition" calculation result data stored in the process data storage library is loaded using the positioning result feedback sub-module;

[0130] Step 42: comparison judgment

[0131] The process data storage library data in the calling positioning result feedback sub-module is called, a feature matching operator and a positioning result standardization operator are called, a feature data matrix M s-d is loaded, a sample transformation result matrix M s-r is generated, and a similarity calculation is performed between the feature data matrix M s-d and the positioning result matrix M d-r ;

[0132] Step 43: comparison result feedback

[0133] The generated positioning result matrix M d-r is standardized and assembled and then fed back.

[0134] Step 5: pose adjustment, including

[0135] Step 51: instrument positioning recognition information parsing: the object recognition result parsing sub-module is called using the motion pose adjustment module (MPA), the positioning result information is obtained, and according to the judgment feedback result of step 4, if the current instrument recognition result does not meet the requirements, the inspection device pose adjustment needs to be performed, a target encoding information vector V me is formed for the inspection device to adjust the pose for the next instrument target object sampling;

[0136] Step 52: Motion axis encoding conversion generation

[0137] Based on the parsed positioning result information, if the current instrument recognition result does not meet the requirements, the converted target encoding information and the axis motion direction, axis motion rate and other motion parameters are assembled to form the motion axis encoding information vector V me ;

[0138] Step 53: Motion axis encoding information export

[0139] Through the data transmission submodule, the real-time transmission of the motion axis encoding information vector is completed, and the inspection device is executed for pose adjustment; then return to "Step 3 object recognition" to execute the object recognition calculation process again;

[0140] Step 6: Result generation, including

[0141] Step 61: Result judgment

[0142] If the "Step 4 recognition result feedback judgment" result meets the requirements, load the process data storage library in the "Step 3.5 calculation result storage" process; if the current instrument recognition result meets the requirements, the pose adjustment of the inspection device is not executed, and the current instrument target object sampling is available;

[0143] Step 62: Result information generation

[0144] Based on, call the positioning result matrix M generated by "Step 4.3 comparison result feedback" d-r , get the normalized parsed positioning result information after assembly;

[0145] Step 63: Result information export

[0146] Through the data transmission submodule, the final result transmission of "Step 6.2" is completed.

Claims

1. A nuclear island inspection robot instrument positioning and identification system based on dynamic target matching, characterized in that, Comprise: Task planning module is used to realize basic data loading, instrument inspection task information generation; Target object sampling module calls visual equipment to collect instrument image file, and through using file correction submodule, manual and automatic correction preprocessing operation is carried out on collected image file, and corrected preprocessing operation image file and information data are provided for target object recognition calculation module to call; Target object recognition calculation module obtains and analyzes sample picture file and supporting information through data transmission interface, calls target positioning tool and template comparison submodule to execute LBP transformation, feature point comparison and multi-resolution mobile similarity processing, and finally object positioning recognition result is standardized assembled; The motion posture adjustment module performs the patrol device pose adjustment to form a target code information vector of each motion axis for the patrol device to adjust the pose for the next instrument target object sampling; The target positioning tool is used to implement the feature point positioning function of the instrument object picture sample; The target positioning tool comprises an information loading submodule, a sample transformation submodule and a positioning result feedback submodule; Information loading submodule: contains process data repository, information parsing operator, sample normalization operator; reads the instrument sample object to be identified to generate a sample data vector ; Sample transformation sub-module: contains process data repository, sample processing operator, processing result normalization operator; reads sample data information vector , according to the computing requirements to improve the sample computing parameter vector ; load parameter vector , for template data vector image element execution gray scale transformation, LBP calculation and multi-resolution mobile similar processing, generate sample transformation result matrix ; The positioning result feedback submodule comprises a process data storage library, a feature matching operator and a positioning result standardization operator; loading feature data matrix , performing sample transformation result matrix with feature data matrix similarity calculation, generating positioning result matrix ; The target template generation tool is a special algorithm tool for configuring instrument object template data.

2. A nuclear island inspection robot instrument positioning and identification system based on dynamic target matching as claimed in claim 1, characterized in that: The task planning module comprises 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 identification system based on dynamic target matching of claim 1, wherein: The target object sampling module comprises a visual equipment access interface, a file correction submodule and a data transmission submodule.

4. The nuclear island inspection robot instrument positioning and identification system based on dynamic target matching of claim 1, wherein: The target object recognition calculation module comprises a template comparison submodule, a target positioning tool calling interface, a calculation process information storage library, an identification result assembly submodule and a data transmission submodule.

5. A nuclear island inspection robot instrument positioning and identification 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 encoding generation submodule and a data transmission submodule.

6. A nuclear island inspection robot instrument positioning and identification system based on dynamic target matching as claimed in claim 1, characterized in that: 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 assembling submodule: including process data repository, data parsing operator, template normalization operator; parsing and assembling the input test file and the information belonging to it to generate template data vector ; Parameter setting submodule: including process data repository, parameter matching operator, parameter normalization operator; reading template data vector , based on the preset conversion requirements and real-time computing needs, perfect and update the template parameter vector ; Image processing sub-module: contains process data repository, image operation operator, result normalization operator; calculate parameter vector based on template For template data vector The image elements perform binaryzation processing, region segmentation, gray scale transformation, LBP calculation and other calculation operations, and generate an image processing result matrix ; Feature data generation submodule: contains process data repository, label processing operator, feature data normalization operator; based on image processing result matrix , execute center point and feature point calculation, structure feature point label operation, generate feature data matrix .

7. A nuclear island inspection robot instrument positioning and recognition method based on dynamic target matching, comprising the following steps: Step 1: template generation, comprising Step 11: image and information loading The target template generation tool calls the target template generation tool calling interface in the target template generation tool to parse and assemble the input test file and the information thereof, and generates a template data vector for subsequent steps ; Step 12: image processing calculation Reading template data vector , based on the preset conversion requirements and real-time computing needs, performing binary processing, region segmentation, gray scale conversion, LBP calculation and other operations, perfecting and updating the template parameter vector , generating a transformation completion result matrix ; Step 13: template feature data generation Based on image processing result matrix , execute center point and feature point calculation, structure feature point marking operation, generate feature data matrix ; Step 2: task loading, comprising Step 21: basic data loading The task planning module loads the basic data information required for the current task; Step 22: task information generation Loading feature data matrix The information is assembled with the task basic data information to generate task information; Step 23: task information storage / calling The task information is stored in the task information library, and the data transmission submodule is used to execute external sending and calling of the task information; Step 3: object recognition, comprising Step 31: instrument object image collection Using the target object collection module, the camera or other camera equipment is called through the visual equipment access interface to obtain the instrument image file; Step 32: image correction Through the file correction loading submodule, manual and automatic correction operation is carried out on the collected image file; Step 33: image processing Using the target object recognition calculation module, the corrected picture file and the supporting information are analyzed, the target positioning tool and the template comparison submodule are called, and the gray scale transformation, LBP calculation and multi-resolution mobile similarity transformation processing are executed; Step 34: image feature point comparison processing Feature point comparison calculation is performed on the processed image; Step 35: calculation result storage After the image processing, the feature point comparison processing image positioning result information is standardized and assembled, and then fed back to the process data storage in the calling positioning result feedback submodule; Step 4: recognition result feedback judgment, including Step 41: data loading Using the positioning result feedback submodule, the stored "step 3 object recognition" calculation result data in the process data storage is loaded; Step 42: comparison judgment By calling the data in the process data repository in the positioning result feedback submodule, calling the feature matching operator, the positioning result normalization operator; by loading the feature data matrix , performing similarity calculation of the sample transformation result matrix and the feature data matrix , and generating the positioning result matrix ; Step 43: comparison result feedback Generating a positioning result matrix Normalized post-assembly feedback; Step 5: pose adjustment, including Step 51: instrument positioning identification information analysis: using the motion pose adjustment module, the object recognition result analysis submodule is called to obtain the positioning result information; According to the feedback result of step 4, if the current instrument recognition result does not meet the requirements, the position of the inspection device needs to be adjusted, and the target code information vector of each motion axis is formed for the inspection device to adjust the position for 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 identification result does not meet the requirements, the converted target code information and the motion parameters such as the shaft motion direction and the shaft motion rate are assembled to form a motion shaft code information vector ; Step 53: motion axis encoding information export Through the data transmission submodule, the real-time transmission of the motion axis encoding information vector is completed, and the inspection device is executed for pose adjustment; then return to "step 3 object recognition" to execute the object recognition calculation process again; Step 6: result generation, including Step 61: result judgment If the result of "step 4 recognition result feedback judgment" meets the requirements, load the process data storage in "step 3.5 calculation result storage" process; If the current instrument recognition result meets the requirements, the pose adjustment of the inspection device is not executed, and the current instrument target object sampling is available; Step 62: result information generation Based on, call "step 4.3 comparison results feedback" generated positioning results matrix , get normalized after assembly after analysis of positioning results information; Step 63: result information export Through the data transmission submodule, the final result of "step 6.2" is transmitted.

Images