Instrument identification method, industrial equipment identification method and system

Abstract

The invention provides an instrument identification method which is a method for determining a dynamic anchor point through prior information and target information of an inspection area, the prior information comprises prior information of at least two instrument frames, and the prior information of the inspection area comprises position information, a central point or at least two groups of reference coordinates of each prior information instrument frame; the inspection area target information comprises position information, a central point or at least two groups of reference coordinates of each target information instrument frame. Global offset estimation is replaced by dynamic anchor point selection, so that the influence of a single error on overall matching is avoided, and 100% accuracy is still kept in the scene of missing 1 / 3 instruments; the environment robustness is enhanced; global offset, random offset of one instrument and new instrument scenes are supported; the method not only improves the recognition accuracy of a plurality of instruments of the same type or different types, but also greatly increases the time difference between the instrument inspection data and improves the efficiency.

Description

Technical Field

[0001] This invention belongs to the field of computer vision technology, specifically relating to an instrument recognition method, an industrial equipment recognition method, and a system thereof. Background Technology

[0002] In existing industrial settings, instrument inspection and identification employs a "single-meter identification" approach, meaning that only a single instrument is identified during a single photograph or post-photograph processing. However, in industrial settings, an inspection area often contains multiple related instruments, such as pressure, temperature, and flow meters for the same equipment. Existing technologies require multiple photographs of each related instrument, separate identification, and storage of the results, significantly increasing the time lag between instrument inspection data and hindering accurate instrument inspection. Furthermore, existing instrument matching technologies rely solely on instrument location or surface features for matching, leading to frequent mismatches between different types of instruments. Summary of the Invention

[0003] To address the aforementioned problems in the prior art, this invention provides an instrument identification method, an industrial equipment identification method, and a system thereof.

[0004] One instrument identification method includes a method for determining dynamic anchor points based on prior information and target information of an inspection area, wherein the prior information includes prior information of at least two instrument frames. The prior information for the inspection area includes: The location information, center point, or at least two sets of reference coordinates for each prior information instrument frame; The target information for the inspection area includes: The location information, center point, or at least two sets of reference coordinates for each target information instrument frame.

[0005] Preferably, the method for determining the dynamic anchor point includes: Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combinations of instrument frame center point or at least two sets of reference coordinate points; or; Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combination instrument frame center points or at least two sets of reference coordinate points, and its comprehensive error meets the preset comprehensive error threshold. or; Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combination instrument frame center points or at least two sets of reference coordinate points, and which meets the preset screening conditions. or; Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combination instrument frame center points or at least two sets of reference coordinate points, and meet the preset screening conditions, and its comprehensive error meets the preset comprehensive error threshold. The selected candidate combination of benchmarks is the dynamic anchor point.

[0006] Preferably, the method for calculating the comprehensive error of the center point of all reference candidate instrument combination frames or at least two sets of reference coordinate points is as follows: Calculate the individual error of the instrument frame center point or at least two sets of reference coordinates for each of the benchmark candidate combinations and the average matching error of the instrument frame center point or at least two sets of reference coordinates for the other matching candidate combinations. Then calculate the comprehensive error of the instrument frame center point or at least two sets of reference coordinates for each of the benchmark candidate combinations. The comprehensive error of the center point of each benchmark candidate instrument frame or at least two sets of reference coordinate points is the sum of the inherent error of the center point of the benchmark candidate instrument frame or at least two sets of reference coordinate points and the average matching error of the center point of the other matching candidate instrument frame or at least two sets of reference coordinate points, using a preset weighting coefficient.

[0007] Preferably, the method for calculating the inherent error of the center point of the benchmark candidate instrument cluster or at least two sets of reference coordinate points is as follows: First, calculate the relative position vector difference of the center point of the benchmark candidate instrument cluster frame. Then, take the square root of the sum of the squares of the two coordinate values ​​of the relative position vector difference to calculate the error of the center point of the benchmark candidate instrument cluster frame. or, First, calculate the relative position vector difference of at least two sets of reference coordinate points of the benchmark candidate instrument cluster frame. Then, take the square root of the sum of the squares of the two coordinate values ​​of their relative position vector differences to calculate the single-point error of the center point of the benchmark candidate instrument cluster frame. Finally, use a preset weighting coefficient to calculate the sum of the single-point errors of the at least two sets of reference coordinate points, which is the self-error of the at least two sets of reference coordinate points of the benchmark candidate instrument cluster frame.

[0008] Preferably, the method for calculating the average matching error of the center point of the other matching candidate combination instrument frame or at least two sets of reference coordinate points is as follows: Based on the baseline candidate combination, the theoretical position of each instrument frame is predicted according to the relative position vector difference between the center points of other prior information instrument frames and the center points of the baseline candidate combination instrument frames. Each of these positions is then matched with the closest instrument frame in the target information instrument frame, forming several other matching candidate combinations. The matching error of the center points of the instrument frames in each of these other matching candidate combinations is calculated. Finally, the matching errors of the center points of all other matching candidate combinations are weighted and averaged to obtain the average matching error based on the center points of the instrument frames in these other matching candidate combinations. or, Based on the benchmark candidate combination, the theoretical position of each instrument frame is predicted according to the relative position vector difference between the center point of the instrument frame and the center point of the benchmark candidate combination instrument frame, and then matched with the closest instrument frame in the target information instrument frame, forming several other matching candidate combinations. The matching error between at least two sets of reference coordinate points of the instrument frames of the several other matching candidate combinations is calculated. The matching errors between at least two sets of reference coordinate points of all other matching candidate combination instrument frames are weighted and averaged to obtain the average matching error based on at least two sets of reference coordinate points of the instrument frames of the other matching candidate combinations.

[0009] Preferably, the prior information of the inspection area also includes the category information and / or external dimensions of each prior information instrument frame, and the target information of the inspection area also includes the category information and / or external dimensions of each target information instrument frame; the preset screening conditions include size verification and / or category verification.

[0010] Preferably, the size verification is as follows: the difference between the outer dimensions of the two instrument frame components of the benchmark candidate combination is not greater than 20% of the outer dimensions of the instrument frame component in its prior information.

[0011] Preferably, the category verification is: the two instrument frames of the benchmark candidate combination have the same category.

[0012] Preferred options also include: Determine whether the matching error value of other matching candidate combinations corresponding to the dynamic anchor point meets the preset matching error threshold. If it does not meet the threshold, cancel the other matching candidate combination and output an instrument matching abnormality signal. Or / and, Determine whether the difference in the outer dimensions of other matching candidate combinations corresponding to the dynamic anchor point is not greater than 20% of the outer dimensions of the instrument frame in its prior information; if not, cancel the other matching candidate combination and output an instrument size abnormality signal. Or / and, Determine whether the categories of other matching candidate combinations corresponding to the dynamic anchor point are consistent; if not, cancel the other matching candidate combination and output an instrument category abnormality signal.

[0013] Preferred options also include: Determine whether the matching error value of other matching candidate combinations corresponding to the dynamic anchor point meets the preset matching error threshold. If it does not meet the threshold, cancel the other matching candidate combination and output the instrument matching abnormality signal. Or / and, Determine whether the difference in the outer dimensions of other matching candidate combinations corresponding to the dynamic anchor point is not greater than 20% of the outer dimensions of the instrument frame in its prior information; if not, cancel the other matching candidate combination and output an instrument size abnormality signal. Or / and, Determine whether the categories of other matching candidate combinations corresponding to the dynamic anchor point are consistent; if not, cancel the other matching candidate combination and output an instrument category abnormality signal.

[0014] Preferably, the prior information of the inspection area further includes: the instrument value of each prior information instrument frame; the target information of the inspection area includes: the instrument value of each target information instrument frame.

[0015] Preferably, the prior information of the inspection area further includes: the instrument value of each prior information instrument frame; the target information of the inspection area includes: the instrument value of each target information instrument frame.

[0016] Preferably, the prior information of the inspection area further includes: the instrument value of each prior information instrument frame; the target information of the inspection area includes: the instrument value of each target information instrument frame.

[0017] The present invention also provides a method for identifying industrial equipment with visually identifiable status indicators or parameter display structures based on any of the above-mentioned instrument identification methods, wherein the instrument is replaced by an industrial equipment with a visually identifiable status indicator or parameter display structure, characterized in that the prior information includes prior information of at least two industrial equipment of the same or different types with visually identifiable status indicators or parameter display structures; The prior information for the inspection area includes: Each piece of prior information includes the location information, center point, or at least two sets of reference coordinates of the industrial equipment. The target information for the inspection area includes: Each target information piece contains the location information of the industrial equipment, its center point, or at least two sets of reference coordinates.

[0018] Preferably, the prior information of the inspection area further includes: the status identifier or parameter display value of each prior information industrial equipment; the target information of the inspection area includes: the status identifier or parameter display value of each target information industrial equipment.

[0019] The present invention also provides a system based on any of the instrument recognition methods described above, comprising: a motion positioning and navigation unit, a camera acquisition and control unit, a data storage unit, a main control and computing unit, and a matching and recognition result output unit, wherein: The motion positioning and navigation unit is used to control the movement of the inspection equipment according to the preset inspection path, and after accurately arriving at the preset instrument inspection point, it sends an image acquisition trigger control signal to the camera acquisition control unit. The camera acquisition control unit, upon receiving an image acquisition trigger control signal from the motion positioning and navigation unit, triggers the main control computing unit to determine whether prior information for the inspection area exists in the data storage unit. If not, it acquires images of the inspection area according to preset image acquisition parameters; if so, it acquires images of the inspection area according to initial inspection image acquisition parameters. The acquired inspection area image is then transmitted to the main control computing unit, triggering the main control computing unit to perform a validity check on the inspection area image. When the target detection module of the main control computing unit verifies that the image is valid, the image of the inspected area is stored in the data storage unit. When the target detection module of the main control computing unit verifies that the image is invalid, the re-acquisition signal fed back by the main control computing unit triggers the camera acquisition control unit to readjust the camera focal length and magnification, and re-acquire the image of the inspected area. The main control computing unit includes a target detection module and a multi-table matching module; wherein, The target detection module is used to analyze and process the image transmitted by the camera acquisition control unit to obtain prior information or target information of the inspection area. Then, it determines whether there is an instrument in the inspection area image based on whether the prior information or target information contains the instrument position coordinates, and verifies the validity of the inspection area image: if the prior information or target information contains the instrument, the prior information or target information of the inspection area is output to the data storage unit and the multi-table matching module respectively; otherwise, a signal to re-acquire the image is output to the camera acquisition control unit. At the same time, the main control computing unit marks the inspection area image as abnormal and stores it in the data storage unit. The multi-table matching module is used to determine, based on the target information output by the target detection module and the prior information of the first inspection area corresponding to the data storage unit, the dynamic anchor point and other corresponding matching candidate combinations; or, the dynamic anchor point and other corresponding matching candidate combinations, and verify the other matching candidate combinations corresponding to the dynamic anchor point candidate combinations; to obtain the matching recognition result and store it in the data storage unit. The data storage unit is used to store the following data: valid inspection area images output by the camera acquisition and control unit, invalid inspection area images with anomaly markers output by the main control computing unit, and prior information of the first inspection area, target information of non-first inspection areas, and matching and recognition results output by the main control computing unit. The matching and recognition result output unit is used to push the matching and recognition results in the data storage unit to the remote system platform.

[0020] This invention also provides a system for industrial equipment identification based on any of the above-described methods for having visually identifiable status indicators or parameter display structures, comprising: a motion positioning and navigation unit, a camera acquisition and control unit, a data storage unit, a main control and computing unit, and a matching and identification result output unit, wherein: The motion positioning and navigation unit is used to control the movement of the inspection equipment according to the preset inspection path, and after accurately arriving at the preset industrial equipment inspection point, it sends an image acquisition trigger control signal to the camera acquisition control unit. The camera acquisition control unit, upon receiving an image acquisition trigger control signal from the motion positioning and navigation unit, triggers the main control computing unit to determine whether prior information for the inspection area exists in the data storage unit. If not, it acquires images of the inspection area according to preset image acquisition parameters; if so, it acquires images of the inspection area according to initial inspection image acquisition parameters. The acquired inspection area image is then transmitted to the main control computing unit, triggering the main control computing unit to perform a validity check on the inspection area image. When the target detection module of the main control computing unit verifies that the image is valid, the image of the inspected area is stored in the data storage unit. When the target detection module of the main control computing unit verifies that the image is invalid, the re-acquisition signal fed back by the main control computing unit triggers the camera acquisition control unit to readjust the camera focal length and magnification, and re-acquire the image of the inspected area. The main control computing unit includes a target detection module and a multi-industrial equipment matching module; wherein, The target detection module is used to analyze and process the inspection area image transmitted by the camera acquisition and control unit to obtain prior information or target information of the inspection area. Then, it determines whether there is industrial equipment in the inspection area image based on whether the prior information or target information contains the location coordinates of industrial equipment, and verifies the validity of the inspection area image: if there is industrial equipment in the prior information or target information, the prior information or target information of the inspection area is output to the data storage unit and the multi-industrial equipment matching module respectively; otherwise, a signal to re-acquire the image is output to the camera acquisition and control unit. At the same time, the main control computing unit marks the inspection area image as abnormal and stores it in the data storage unit. The multi-industrial equipment matching module is used to determine, based on the target information output by the target detection module and the prior information of the first inspection area corresponding to the data storage unit, a dynamic anchor point and other corresponding matching candidate combinations; or, a dynamic anchor point and other corresponding matching candidate combinations, and verifying the other matching candidate combinations corresponding to the dynamic anchor point candidate combinations; to obtain the matching recognition result and store it in the data storage unit. The data storage unit is used to store the following data: valid inspection area images output by the camera acquisition and control unit, invalid inspection area images with anomaly markers output by the main control computing unit, and prior information of the first inspection area, target information of non-first inspection areas, and matching and recognition results output by the main control computing unit. The matching and recognition result output unit is used to push the matching and recognition results in the data storage unit to the remote system platform.

[0021] This invention provides a method for determining dynamic anchor points by using prior information and target information of the inspection area. The prior information includes information from at least two instrument frames. By selecting dynamic anchor points instead of global offset estimation, the invention avoids the impact of a single error on the overall matching, maintaining 100% accuracy even in scenarios where 1 / 3 of the instruments are missing. It also enhances environmental robustness, supports global offset, random offset of one instrument, and scenarios with newly added instruments, and significantly improves the accuracy of identifying multiple instruments of the same or different types. Attached Figure Description

[0022] Figure 1 A schematic diagram of an instrument identification method provided by the present invention; Figure 2 Based on Figure 1 The diagram shows the system structure of the instrument identification method. Figure 3 for Figure 2 The diagram shows the structural composition of the industrial control computing unit. Detailed Implementation

[0023] To better understand the purpose, structure, and function of this invention, we will take as an example an inspection robot inspecting a certain inspection area, in which four industrial equipment instruments with visually identifiable status indicators or parameter display structures are set (the prior information instrument frames are labeled M1, M2, M3, and M4, and the target information instrument frames are labeled 0, 1, 2, and 3, respectively). This will be used to further describe in detail the instrument identification method, industrial equipment identification method, and system provided by this invention. Example 1

[0024] like Figure 1 The diagram shown is a flowchart of an instrument identification method provided in this embodiment. The method is as follows: Step 1: When the inspection robot arrives at an inspection point in a certain inspection area under the control of the motion positioning and navigation unit, the motion positioning and navigation unit triggers the main control computing unit to determine whether prior information (prior_instruments) for the inspection area exists in the data storage unit. If the prior information does not exist, it indicates that this is the first inspection. The robot then receives user-preset parameters or automatically adapts to the inspection point's preset parameters, adjusts the camera's focal length, magnification, and gimbal attitude parameters to determine the acquisition parameters corresponding to the area to be identified, and then starts acquiring images of the inspection area (each containing at least two instrument frames of the inspection area images), and inputs them into the data storage unit and the main control computing unit for processing. In the main control computing unit, a preprocessing module can be used to... The inspection area image is enhanced and denoised, then analyzed and calculated using the target detection module to obtain the prior information (prior_instruments) and its instrument values ​​for the inspection area. This prior information is then added as a new inspection area and stored in the data storage unit, proceeding directly to the end. If prior information exists, it indicates that this is not the first inspection. Based on the gimbal attitude data of the area to be identified corresponding to the inspection point preset during the first inspection, the camera focal length, magnification, and gimbal attitude parameters are controlled and adjusted to ensure that the camera accurately reproduces the preset attitude to align with the inspection area to be identified. Subsequently, the acquisition of the inspection area image is initiated and transmitted to the main control computing unit, triggering the main control computing unit to perform a validity verification of the inspection area image. When the target detection module of the main control computing unit verifies that the image is valid, it means that the image of the inspection area to be identified has been accurately acquired. There is no need to re-acquire the image of the inspection area. The image of the inspection area is then stored in the data storage unit. When the target detection module of the main control computing unit verifies that the image is invalid, it means that the image of the inspection area to be identified has not been accurately acquired. The image acquisition control unit is triggered by the re-acquisition signal fed back by the main control computing unit to readjust the camera focal length and magnification, and re-acquire the image of the inspection area.

[0025] The prior information (prior_instruments) for the inspection area includes: The location information (coordinates of the upper left and lower right corners of the prior information instrument frame), category information, center point coordinates, and external dimensions of each prior information instrument frame; Example: Four instruments are installed in a certain inspection area. The prior information instrument frames of these four instruments are labeled M1, M2, M3, and M4, respectively. 1) Their location information, namely the coordinates of the top left corner and the bottom right corner, and their category information are as follows:

[0026] 2) Their center point coordinates are as follows:

[0027] 3) Their external dimensions are all (50, 50); 4) Examples of the instrument values ​​are as follows: (In order to obtain the final inspection instrument values, the prior information of the inspection area may also include the instrument values.)

[0028] Step 2: The camera acquisition control unit transmits the inspection area image to the main control computing unit. In the main control computing unit, the preprocessing module first enhances and denoises the inspection area image, and the target detection module analyzes and calculates to obtain the prior information (or prior information including its instrument values) or target information (or target information including its instrument values) of the inspection area. Based on whether the prior information or target information contains instrument position coordinates, it determines whether there are instruments in the inspection area image, and then determines whether the inspection area image is a valid image (i.e., inspection area image validity verification): If the prior information or target information contains instrument position coordinate information, the validity verification result is a valid image: The prior information (or prior information including its instrument values) or target information (or target information including its instrument values) of the inspection area is output to the data storage unit and the multi-table matching module respectively; otherwise, the validity verification result is an invalid image: The signal to re-acquire the image is output to the camera acquisition control unit. At the same time, the main control computing unit marks the inspection area image as abnormal and stores it in the data storage unit. The inspection area target information current_instruments includes: The location information (coordinates of the upper left and lower right corners of the target information instrument frame), category information, center point coordinates, and external dimensions of each target information instrument frame; For example: Again, taking the example of setting up 4 instruments in the same inspection area in step 1, the target information instrument frames of these 4 instruments are numbered 0, 1, 2, and 3 respectively: 1) Their location information, namely the coordinates of the top left corner and the bottom right corner, and their category information are as follows:

[0029] 2) Their center point coordinates are as follows:

[0030] 3) Their external dimensions are all (50, 50); 4) Examples of the instrument values ​​are as follows: (To ultimately obtain the inspection instrument values, the target information for the inspection area may also include the values ​​of each instrument.)

[0031] Step 3: The main control computing unit retrieves the prior targets and target information of the inspection area obtained in Steps 1 and 2 from the data storage unit. In its multi-table matching module, it selects a dynamic anchor point: For a target information instrument frame in the inspection area target information obtained in Step 2, it performs a baseline candidate combination with any prior information instrument frame in the corresponding inspection area's prior information. It calculates the relative position vector difference between the center point of the target information instrument frame and the center point of the prior information instrument frame, and calculates the inherent error of the baseline candidate combination. Based on this baseline candidate combination, it predicts the theoretical position of each other prior information instrument frame by adding the relative position vector difference of the baseline candidate combination to the center point of each frame, and matches them with the closest target information instrument frame. It calculates the matching error of each other matching candidate combination. Finally, it weights and averages the matching errors of all other matching candidate combinations to obtain the average error based on the baseline candidate combination. Then, based on the inherent error between the benchmark candidate combinations and the average error based on the benchmark candidate combinations, the comprehensive error of the benchmark candidate combination is calculated. Then, another target information instrument frame in the target information of the inspection area described in step 2 is re-combined with any prior information instrument frame in the prior information of the corresponding inspection area, and the calculation process of the comprehensive error of the benchmark candidate combination is repeated. The comprehensive error of the re-combined benchmark candidate combination is calculated until all target information instrument frames in the target information of the inspection area described in step 2 are combined with any prior information instrument frame in the prior information of the corresponding inspection area, and the comprehensive error of all benchmark candidate combinations is calculated. Finally, the benchmark candidate combination with the smallest comprehensive error is selected from the comprehensive errors of all benchmark candidate combinations, or the benchmark candidate combination with the smallest comprehensive error and meeting the screening conditions is selected as the dynamic anchor point.

[0032] Specific methods for determining dynamic anchor points include: Using the examples from steps 1 and 2 as a basis, and taking the prior information instrumentation frame "M2" and the target information instrumentation frame "0" in the baseline candidate combination M20 as an example, the following explanation will be provided: Step 3.1: Calculate the relative position vector difference C between the center point of the target information instrument frame "0" and the center point of the prior information instrument frame "M2". M20 The specific calculation method is to subtract the center point of the prior information instrument frame "M2" from the center point of the target information instrument frame "0", that is:

[0033] Step 3.2: Calculate the relative position vector difference C based on Step 3.1. M20 Calculate its own error (the smaller the value, the closer the absolute position of the candidate combination is to the prior information instrument frame), which is:

[0034] Step 3.3: Calculate the matching error and its average error for other candidate matching combinations: Predicting the theoretical position of M1: The prior information of M1, the center point of the instrument frame, plus the relative position vector difference C of the benchmark candidate combination M20 calculated in step 3.1. M20 That is, the predicted theoretical position of M1 is (125+115, 125+79) = (240, 204); the closest to (240, 204) in the target information instrument frame is the target information instrument frame "0": the matching error between the center point of M1 and the "0" instrument frame based on the baseline candidate combination M20 is:

[0035] Predicting the theoretical position of M3: The prior information of M3, the center point of the instrument frame, plus the relative position vector difference C of the benchmark candidate combination M20 calculated in step 3.1. M20 That is, the predicted theoretical position of M3 is (325+115, 325+79) = (440, 404); the closest to (440, 404) in the target information instrument frame is frame "1": the matching error between the center point of M3 and the instrument frame "1" based on the baseline candidate combination M20 is:

[0036] Predicting the theoretical position of M4: The prior information of M4, the center point of the instrument frame, plus the relative position vector difference C of the benchmark candidate combination M20 calculated in step 3.1. M20 That is, the predicted theoretical position of M4 is (425+115, 425+79) = (540, 504); the closest to (540, 504) in the target information instrument frame is frame "2": the matching error between the center point of M4 and the instrument frame "2" based on the baseline candidate combination M20 is:

[0037] Then, based on all the matching errors calculated above, the average matching error of other candidate matching combinations is calculated:

[0038] Step 3.4: Based on the self-error of the benchmark candidate combination calculated in Step 3.2 and the average matching error of other matching candidate combinations calculated in Step 3.3, calculate the comprehensive error of the benchmark candidate combination M20. The comprehensive error cost = 0.3 × self-error + 0.7 × average error, that is:

[0039] In the above comprehensive error calculation, the weighting coefficient of 0.3 for the self-error and the weighting coefficient of 0.7 for the matching average error are typical adaptation values ​​determined by the influence weight analysis of "relative position stability of instruments" and "absolute position stability" in industrial inspection scenarios and verified through multiple sets of scenario experiments. By using this coefficient ratio, the optimal matching accuracy can be achieved under most working conditions. At the same time, this coefficient supports dynamic adjustment according to the actual characteristics of the scene. For example, when the relative position between instruments is minimally affected by the environment, it can be adjusted to 0.2 for the self-error and 0.8 for the matching average error. The adjustment logic is based on balancing the influence of the two types of position stability on the matching result. It does not rely on manual subjective setting and can be adaptively adapted through system preset rules or field environmental parameters.

[0040] The calculation method for the self-error of the above-mentioned benchmark candidate combination and the average matching error of other matching candidate combinations calculated in step 3.3 can also be other calculation methods, such as the Manhattan distance calculation method.

[0041] Step 3.5: Using the same method as in Steps 3.1-3.4 for calculating the comprehensive error of the baseline candidate combinations, calculate the comprehensive error of the remaining baseline candidate combinations. For example, the a priori information instrument frame "M1" and the target information instrument frame "0" are baseline candidate combinations M10, M3 and M0 are baseline candidate combinations M30, and M4 and M0 are baseline candidate combinations M40; the a priori information instrument frame "M2" and M1 are baseline candidate combinations M21, M1 and M1 are baseline candidate combinations M11, M3 and M1 are baseline candidate combinations M31, and M4 and M1 are baseline candidate combinations M41; and so on, traversing all baseline candidate combinations to obtain the comprehensive error of all baseline candidate combinations.

[0042] Step 3.6: Method for determining dynamic anchor points: From steps 3.4 and 3.5, select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combinations; or, select the benchmark candidate combination with the smallest comprehensive error whose comprehensive error meets a preset comprehensive error threshold; or, select the benchmark candidate combination with the smallest comprehensive error that meets the screening criteria; or, select the benchmark candidate combination with the smallest comprehensive error that meets the screening criteria and whose comprehensive error meets a preset comprehensive error threshold, and use it as a dynamic anchor point. The filtering criteria include: (1) Dimensional verification: The difference between the outer dimensions of the two instrument frame components of the benchmark candidate combination shall not exceed 20% of the outer dimensions of the instrument frame component in its prior information; or / and (2) Category verification: The two instrument boxes of the benchmark candidate combination are of the same category.

[0043] The comprehensive error threshold is an engineering constraint threshold determined based on the comprehensive error calculation model of instrument vision matching, the existing technical parameter linkage system, and the statistical laws of industrial scenarios. The default optimal threshold for general working conditions is 80. According to the statistical data of industrial tests in multiple scenarios, it is compatible with more than 95% of conventional instrument matching scenarios. This threshold can be dynamically adjusted according to the application scenario: for example, in scenarios with low instrument layout density and high detection accuracy requirements, it can be lowered to 60; when a looser fit is required for complex scenarios such as global offset, it can be raised to 150.

[0044] For example: Continuing with the example in step 3, there are 4 prior information instrument frames and 4 target information instrument frames. Therefore, after traversing all 4×4=16 baseline candidate combinations, it is assumed that the minimum comprehensive error between the prior information instrument frame "M2" and the target information instrument frame is 76.43; the error between the size (50, 50) of frame "0" and the size (50, 50) of the prior information instrument frame "M2" is 0%; and the instrument category of both frame "0" and the prior information instrument frame "M2" is pointer_device1.

[0045] Therefore, the prior information instrument frame "M2" and the target information instrument frame "0" are selected as dynamic anchor points.

[0046] Step 4: Determine whether the matching error value of other matching candidate combinations corresponding to the dynamic anchor point described in Step 3 meets the preset matching error threshold. If not, cancel the candidate combination and output an instrument matching abnormality signal. The preset matching error threshold is determined based on the pixel adaptation characteristics of the system's zoom camera to the actual spatial size, combined with reasonable deviation ranges such as system positioning accuracy and instrument installation deviation in industrial inspection scenarios, and verified by algorithms. It can cover normal deviations in practical applications, avoid mismatches, and ensure recognition accuracy. For example, the preset matching error threshold cost is no greater than 40. Or / and, Determine whether the difference in the outer dimensions of other matching candidate combinations corresponding to the dynamic anchor point described in step 3 is not greater than 20% of the outer dimensions of the instrument frame in its prior information; if not, cancel the other matching candidate combination and output an instrument size abnormality signal. Of these, 20% is determined based on common manufacturing tolerances of industrial instruments, slight deformation after installation, camera zoom imaging deviation, and dimensional measurement deviations caused by missing parts of the instrument imaging due to shaking of the inspection robot when it stops. This is combined with algorithm robustness verification. The 20% value can cover various normal deviations in actual inspection scenarios (including the reasonable impact of partial imaging loss), and avoid misjudging abnormalities due to local loss or minor deviations through a wide adaptation range. At the same time, it reserves sufficient space for dimensional deviation adaptation under different working conditions. Or / and, Determine whether the categories of other matching candidate combinations corresponding to the dynamic anchor point described in step 3 are consistent; if not, cancel the other matching candidate combination and output an instrument category abnormality signal.

[0047] Finally, the multi-table recognition module outputs the obtained matching recognition results (including: prior information of dynamic anchor points and other matching candidate combinations, target information, instrument frame combination matching results, and abnormal signals of instrument matching / instrument size / instrument category) to the data storage unit; in order to finally obtain the inspection instrument values, the matching recognition results also include instrument value matching results.

[0048] For example: Using the prior information instrument frame "M2" obtained in step 3 and the target information instrument frame "0" as dynamic anchor points, the detailed calculation process of the matching and recognition results is as follows: Taking "M3" as an example, in step 3.3 it is known that the center point of the predicted position of "M3" is (440, 404). The object in the target information instrument frame that is closest to the predicted position of "M3" is the prior information instrument frame "1", and its matching error cost is... M20-M31 The result is approximately 3.2. The matching error is less than the preset matching error threshold of 40, and the dimensions of both "M3" and box "1" are (50, 50), with a size error of 0%, and the category information is both pointer_device1. Therefore, "M3" matches box "1". Taking "M1" as an example, in step 3.3 it is known that the center point of the predicted position of "M1" is (240, 204). The closest target information instrument frame to the predicted position of "M1" is the prior information instrument frame "0", and its matching error cost is... M20-M10 The value is approximately 141.4. The matching error exceeds the preset matching error threshold of 40, the matching error verification fails, and an "M1" matching error signal is output. Similarly, we obtain the matching box "2" for "M4"; Based on the above prior information and target information instrument frame combination matching results, assign corresponding values ​​to each instrument: assign the value of 1.4 kPa of target information instrument "0" in value_current of step 3 to instrument "M2"; assign the value of 2.3 kPa of instrument "1" to instrument "M3"; assign the value of 0.8 kPa of instrument "2" to instrument "M4"; "M1" has an abnormal match and is not assigned a value; Finally, the matching and recognition results are as follows: the combined matching result of prior information and target information instrument frames (result_instruments), the matching result of each instrument value (value_result), and the abnormal signals:

[0049]

[0050] Abnormal signal: "M1" mismatch.

[0051] This embodiment also provides a system based on the aforementioned instrument identification method, such as... Figure 2 , 3 As shown. The system includes a motion positioning and navigation unit, a camera acquisition and control unit, a main control and computing unit, a data storage unit, and a matching and recognition result output unit, wherein: The motion positioning and navigation unit (such as the inspection equipment driver and its positioning and navigation controller) is used to control the movement of the inspection equipment according to the preset inspection path, accurately reach the preset inspection point, and send an image acquisition trigger control signal to the camera acquisition control unit. The camera acquisition control unit (such as a camera gimbal control module) is used to trigger the main control computing unit to determine whether prior information (prior_instruments) for the inspection area exists in the data storage unit when it receives an image acquisition trigger control signal from the motion positioning and navigation unit. If the prior information does not exist, it indicates that this is the first inspection. In this case, the system receives user-preset parameters or automatically adapts to the inspection point preset parameters, adjusts the camera focal length, magnification, and gimbal attitude parameters to determine the acquisition parameters corresponding to the area to be identified, and then starts image acquisition. If the prior information exists, it indicates that this is not the first inspection. Based on the gimbal attitude data of the area to be identified corresponding to the inspection point preset during the first inspection, the system controls and adjusts the camera focal length, magnification, and gimbal attitude parameters to ensure that the camera accurately reproduces the preset attitude to align with the inspection area to be identified. Then, the system starts image acquisition and transmits the acquired inspection area image to the main control computing unit, triggering the main control computing unit to perform a validity check on the inspection area image. When the target detection module of the main control computing unit verifies that the image is valid, it means that the image of the inspection area to be identified has been accurately acquired. There is no need to re-acquire the image of the inspection area. The image of the inspection area is then stored in the data storage unit. When the target detection module of the main control computing unit verifies that the image is invalid, it means that the image of the inspection area to be identified has not been accurately acquired. The image acquisition control unit is triggered by the re-acquisition signal fed back by the main control computing unit to readjust the camera focal length and magnification, and re-acquire the image of the inspection area.

[0052] The main control computing unit (such as an embedded industrial control computer) includes a preprocessing module, a target detection module, and a multi-table matching module; wherein, The preprocessing module is used to receive the inspection area image output by the camera acquisition and control unit, enhance and denoise the inspection area image to improve the subsequent target detection accuracy, and output the preprocessed image to the target detection module.

[0053] The target detection module receives the preprocessed image output by the preprocessing module, analyzes and processes it to obtain prior information (or prior information including its instrument values) or target information (or target information including its instrument values) of the inspection area, and determines whether there are instruments in the inspection area image based on whether the prior information or target information contains instrument position coordinates, and then determines whether the inspection area image is a valid image (i.e., inspection area image validity verification): if instrument position coordinate information is present in the prior information or target information, the validity verification result is a valid image: the prior information (or prior information including its instrument values) or target information (or target information including its instrument values) of the inspection area is output to the data storage unit and the multi-table matching module respectively; otherwise, the validity verification result is an invalid image: the signal to re-acquire the image is output to the camera acquisition control unit, and at the same time, the inspection area image is marked as abnormal by the main control computing unit and stored in the data storage unit.

[0054] The multi-table matching module is used to receive target information (or target information including its instrument values) output by the target detection module, and at the same time read the prior information of the first inspection area corresponding to the data storage unit (or the prior information including its instrument values) to determine the dynamic anchor point candidate combination. Then, for other matching candidate combinations corresponding to the dynamic anchor point candidate combination, through size verification and / or category verification and / or matching error verification, the matching identification result (including: prior information of dynamic anchor points and other matching candidate combinations and target information instrument frame combination matching result, instrument value matching result, and abnormal signals of instrument matching / instrument size / instrument category) is obtained and stored in the data storage unit.

[0055] The data storage unit (such as a solid-state drive or memory card) is used to store the following data: valid inspection area images output by the camera acquisition and control unit, invalid inspection area images with abnormal markers output by the main control computing unit, and prior information of the first inspection area (or prior information including its instrument value), target information of non-first inspection areas (or target information including its instrument value), and matching and recognition results output by the main control computing unit.

[0056] The matching and identification result output unit (such as a display / communication module) is used to push the matching and identification results in the data storage unit to the remote system platform (including mobile terminals) according to dimensions such as inspection points and instrument types. If there are abnormal matching / size / category signal information in the matching and identification results, the remote system platform will issue corresponding voice or light warning signals to remind relevant personnel to handle the situation in a timely manner.

[0057] This embodiment has the following advantages: it can improve the recognition accuracy of multiple instruments of the same or different types. By using dynamic anchor point selection instead of global offset estimation, it avoids the impact of a single error on the overall matching, maintaining 100% accuracy even in scenarios where 1 / 3 of the instruments are missing; it enhances environmental robustness; and it supports global offset, random offset of one instrument, and scenarios with newly added instruments.

[0058] Example 2 Since the matching stability is poor when calculating the self-error, matching error and average matching error by using a single center point of the instrument frame in Example 1 to obtain prior information and target information, if environmental factors such as light change significantly, this embodiment improves upon Example 2. The main difference is that the single center point of the instrument frame in Example 1 is replaced with at least two sets of reference coordinate points of the instrument frame with different environmental stability.

[0059] This embodiment is merely a further improvement on the technical solution in step 3 of embodiment 1; all other steps remain unchanged. The difference from the technical solution in step 3 of embodiment 1 is as follows: When calculating the self-error of the two sets of reference coordinate points of the benchmark candidate combination and the matching error and average matching error of the two sets of reference coordinate points of other matching candidate combinations, the center coordinate point of the benchmark candidate combination instrument frame in step 3.1 of Example 1 is replaced with at least two sets of reference coordinate points (the selection of reference coordinate points may or may not include the center point of the instrument frame). One set of reference coordinate points is selected from points such as the center axis of the dial of the pointer instrument, the center point of the digital display area of ​​the digital instrument, and the rotation axis of the valve handle. These types of reference coordinate points are minimally affected by light. The other set of reference coordinate points is selected from points such as the pointer tip of the pointer instrument, the dynamic digital edge of the digital instrument, and the end point of the free end of the valve handle. These types of reference coordinate points are greatly affected by light.

[0060] The method for determining dynamic anchor points in this embodiment includes: Using the examples from steps 1 and 2 in Example 1 as a basis, and taking the prior information instrument box "M2" and the target information instrument box "0" in the baseline candidate combination M20 as examples, the explanation will proceed as follows: Step 3.1A (same as step 3.1 in Example 1): Calculate the relative position vector difference C between the center point of the target information instrument frame "0" and the center point of the prior information instrument frame "M2". M20 The specific calculation method is to subtract the center point of the prior information instrument frame "M2" from the center point of the target information instrument frame "0", that is:

[0061] Simultaneously, from the prior information instrument frame "M2" and the target information instrument frame "0", the coordinates of a set of reference coordinate points (x1, y1) of the prior information instrument frame "M2" and the coordinates of the reference coordinate point (x2, y2) corresponding to the position of the target information instrument frame "0", and the coordinates of another set of reference coordinate points (x3, y3) of the prior information instrument frame "M2" and the coordinates of the reference coordinate point (x4, y4) corresponding to the position of the target information instrument frame "0" are collected respectively.

[0062] Thus, steps 3.2 and 3.3 in Example 1 are replaced as follows: Step 3.2A: Calculate the relative position vector difference C based on Step 3.1A. M20 The single-point error of the above two sets of reference coordinate points is calculated respectively:

[0063] The final calculated error of the two sets of reference coordinate points for the benchmark candidate combination is as follows:

[0064] Specifically, high-weight coordinate points are assigned a high-weight coefficient w1=0.8, and low-weight coordinate points are assigned a low-weight coefficient w2=0.2. The weight depends on the stability of the coordinate points; the higher the stability of the coordinate points, the larger the weight coefficient, and the lower the stability of the coordinate points, the smaller the weight coefficient. (This weight coefficient is determined based on the environmental anti-interference characteristics and error calculation reliability requirements of the reference coordinate points of industrial equipment: the error data of high-stability coordinate points (which are minimally affected by environmental factors such as light) are more valuable for reference, and assigning them a high weight of 0.8 can highlight their core decision-making role and reduce the impact of environmental interference on the calculation results; the error data of low-stability coordinate points (which are greatly affected by environmental factors such as light) are more volatile, and assigning them a low weight of 0.2 can take into account their auxiliary correction value while avoiding the introduction of additional errors. After verification of the algorithm's robustness, this weight ratio can maximize the accuracy and anti-interference of error calculation.)

[0065] Step 3.3A: Calculate the matching error and average matching error of the two sets of reference coordinate points for other matching candidate combinations: For example: Predicting the theoretical position of M1: The prior information of M1, the center point of the instrument frame, plus the relative position vector difference C of the baseline candidate combination M20 calculated in step 3.1A. M20 Obtain the closest prior information instrument frame "0": and select a set of high-weight coordinate points from both the prior information instrument frame "M2" and the target information instrument frame "0". and a set of low-weight coordinate points .

[0066] Calculate the single-point matching error for each of the two sets of reference coordinate points:

[0067] Therefore, the matching error between M1, which is based on the benchmark candidate combination M20, and the center point of the "0" instrument frame is:

[0068] High-weight coordinate points are assigned a high-weight coefficient w1=0.8, and low-weight coordinate points are assigned a low-weight coefficient w2=0.2. The principle for determining this weight coefficient is consistent with that in step 3.2A, that is, it is determined based on the environmental anti-interference characteristics of the reference coordinate points and the reliability requirements of error calculation. After verification of algorithm robustness, this ratio can maximize the accuracy and anti-interference of error calculation.

[0069] Similarly, the matching errors between M3 and the target information's closest instrument frame center point, and between M4 and the target information's closest instrument frame center point, are calculated using the same method. Finally, the average matching error of the two sets of reference coordinate points for other matching candidate combinations is calculated.

[0070] Step 3.4A: Calculate the individual errors of the two sets of reference coordinate points for the benchmark candidate combination based on Step 3.2A and the average matching errors of the two sets of reference coordinate points for other matching candidate combinations based on Step 3.3A. Calculate the comprehensive error of the two sets of reference coordinate points for the benchmark candidate combination M20. For specific methods, please refer to Step 3.4 of Example 1, which will not be repeated here.

[0071] Step 3.5A calculates the combined error of the two sets of reference coordinate points for all other candidate benchmark combinations. For the specific method, please refer to step 3.4 of Example 1, which will not be repeated here.

[0072] Step 3.6A, which determines the dynamic anchor point, is the same as step 3.6 in Example 1.

[0073] The instrument identification method and system provided in this embodiment not only utilize the anti-interference characteristics of highly stable reference points to determine the core position benchmark, but also weaken the environmental interference influence of low-stability reference points through weight allocation. This greatly avoids the problem of error amplification when the single center point of the instrument frame is affected by light interference, and improves the stability of the instrument identification calculation process.

[0074] It is understood that this invention is illustrated by taking an inspection robot inspecting a specific inspection area, where four industrial equipment instruments with visually identifiable status indicators or parameter display structures are installed (the prior information instrument frames are labeled M1, M2, M3, and M4, and the target information instrument frames are labeled 0, 1, 2, and 3, respectively). The invention provides a method for determining dynamic anchor points based on the prior information and target information of the inspection area, wherein the prior information includes information from at least two instrument frames. Those skilled in the art will understand that, without departing from the spirit and scope of this invention, the prior information of an inspection area can include the number of two or more instruments. The instrument in this invention can be modified or replaced with at least two of the same type of industrial equipment or at least one of at least two different types of equipment, such as instruments, control devices, valves, etc., that have visually identifiable status indicators or parameter display structures. For example, an inspection area may have at least one instrument and one control device (at least one of at least two different types of equipment), or an inspection area may have at least two control devices (at least two of the same type of equipment), or an inspection area may have at least two valves (at least two of the same type of equipment), or an inspection area may have at least one instrument and one valve (at least one of at least two different types of equipment), or an inspection area may have at least one control device and one valve (at least one of at least two different types of equipment), etc. Of course, the inspection robot in this invention can also be an inspection drone, a fixed monitoring camera device installed near the inspection area, or other inspection equipment. In addition, under the teachings of this invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of this invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of this invention.

Claims

1. A method for instrument identification, characterized in that, The method includes determining dynamic anchor points based on prior information and target information of an inspection area, wherein the prior information includes prior information of at least two instrument frames; The prior information for the inspection area includes: The location information, center point, or at least two sets of reference coordinates for each prior information instrument frame; The target information for the inspection area includes: The location information, center point, or at least two sets of reference coordinates for each target information instrument frame.

2. The instrument identification method according to claim 1, characterized in that, The method for determining dynamic anchor points includes: Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combinations of instrument frame center point or at least two sets of reference coordinate points; or; Select the benchmark candidate combination with the smallest comprehensive error among all the comprehensive errors of the instrument frame center point or at least two sets of reference coordinate points, and its comprehensive error meets the preset comprehensive error threshold. or; Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combination instrument frame center points or at least two sets of reference coordinate points, and which meets the preset screening conditions. or; Select the benchmark candidate combination with the smallest comprehensive error among all benchmark candidate combination instrument frame center points or at least two sets of reference coordinate points, and meet the preset screening conditions, and its comprehensive error meets the preset comprehensive error threshold. The selected candidate combination of benchmarks is the dynamic anchor point.

3. The instrument identification method according to claim 2, characterized in that, The method for calculating the combined error of the center point of all candidate reference instrument clusters or at least two sets of reference coordinate points is as follows: Calculate the individual error of the instrument frame center point or at least two sets of reference coordinates for each of the benchmark candidate combinations and the average matching error of the instrument frame center point or at least two sets of reference coordinates for the other matching candidate combinations. Then calculate the comprehensive error of the instrument frame center point or at least two sets of reference coordinates for each of the benchmark candidate combinations. The comprehensive error of the center point of each benchmark candidate instrument frame or at least two sets of reference coordinate points is the sum of the inherent error of the center point of the benchmark candidate instrument frame or at least two sets of reference coordinate points and the average matching error of the center point of the other matching candidate instrument frame or at least two sets of reference coordinate points, using a preset weighting coefficient.

4. The instrument identification method according to claim 3, characterized in that, The method for calculating the inherent error of the center point of the reference candidate instrument frame or at least two sets of reference coordinate points is as follows: First, calculate the relative position vector difference of the center point of the benchmark candidate instrument cluster frame. Then, take the square root of the sum of the squares of the two coordinate values ​​of the relative position vector difference to calculate the error of the center point of the benchmark candidate instrument cluster frame. or, First, calculate the relative position vector difference of at least two sets of reference coordinate points of the benchmark candidate instrument cluster frame. Then, take the square root of the sum of the squares of the two coordinate values ​​of their relative position vector differences to calculate the single-point error of the center point of the benchmark candidate instrument cluster frame. Finally, use a preset weighting coefficient to calculate the sum of the single-point errors of the at least two sets of reference coordinate points, which is the self-error of the at least two sets of reference coordinate points of the benchmark candidate instrument cluster frame.

5. The instrument identification method according to claim 4, characterized in that, The method for calculating the average matching error of the center point of the other candidate instrument frame or at least two sets of reference coordinate points is as follows: Based on the baseline candidate combination, the theoretical position of each instrument frame is predicted according to the relative position vector difference between the center points of other prior information instrument frames and the center points of the baseline candidate combination instrument frames. Each of these positions is then matched with the closest instrument frame in the target information instrument frame, forming several other matching candidate combinations. The matching error of the center points of the instrument frames in each of these other matching candidate combinations is calculated. Finally, the matching errors of the center points of all other matching candidate combinations are weighted and averaged to obtain the average matching error based on the center points of the instrument frames in these other matching candidate combinations. or, Based on the benchmark candidate combination, the theoretical position of each instrument frame is predicted according to the relative position vector difference between the center point of the instrument frame and the center point of the benchmark candidate combination instrument frame, and then matched with the closest instrument frame in the target information instrument frame, forming several other matching candidate combinations. The matching error between at least two sets of reference coordinate points of the instrument frames of the several other matching candidate combinations is calculated. The matching errors between at least two sets of reference coordinate points of all other matching candidate combination instrument frames are weighted and averaged to obtain the average matching error based on at least two sets of reference coordinate points of the instrument frames of the other matching candidate combinations.

6. The instrument identification method according to claim 2, 3, 4, or 5, characterized in that, The prior information of the inspection area also includes the category information and / or external dimensions of each prior information instrument frame, and the target information of the inspection area also includes the category information and / or external dimensions of each target information instrument frame; the preset screening conditions include size verification and / or category verification.

7. The instrument identification method according to claim 6, characterized in that, The size verification is as follows: the difference between the outer dimensions of the two instrument frames of the benchmark candidate combination is not greater than 20% of the outer dimensions of the instrument frame in its prior information.

8. The instrument identification method according to claim 6, characterized in that, The category verification is that the two instrument frames of the benchmark candidate combination have the same category.

9. The instrument identification method according to claim 3, 4, 5, 7, or 8, characterized in that, Also includes: Determine whether the matching error value of other matching candidate combinations corresponding to the dynamic anchor point meets the preset matching error threshold. If it does not meet the threshold, cancel the other matching candidate combination and output an instrument matching abnormality signal. Or / and, Determine whether the difference in the outer dimensions of other matching candidate combinations corresponding to the dynamic anchor point is not greater than 20% of the outer dimensions of the instrument frame in its prior information; if not, cancel the other matching candidate combination and output an instrument size abnormality signal. Or / and, Determine whether the categories of other matching candidate combinations corresponding to the dynamic anchor point are consistent; if not, cancel the other matching candidate combination and output an instrument category abnormality signal.

10. The instrument identification method according to claim 6, characterized in that, Also includes: Determine whether the matching error value of other matching candidate combinations corresponding to the dynamic anchor point meets the preset matching error threshold. If it does not meet the threshold, cancel the other matching candidate combination and output the instrument matching abnormality signal. Or / and, Determine whether the difference in the outer dimensions of other matching candidate combinations corresponding to the dynamic anchor point is not greater than 20% of the outer dimensions of the instrument frame in its prior information; if not, cancel the other matching candidate combination and output an instrument size abnormality signal. Or / and, Determine whether the categories of other matching candidate combinations corresponding to the dynamic anchor point are consistent; if not, cancel the other matching candidate combination and output an instrument category abnormality signal.

11. The instrument identification method according to claim 1, 2, 3, 4, 5, 7, 8, or 10, characterized in that, The prior information of the inspection area also includes: the instrument value of each prior information instrument frame; the target information of the inspection area also includes: the instrument value of each target information instrument frame.

12. The instrument identification method according to claim 6, characterized in that, The prior information of the inspection area also includes: the instrument value of each prior information instrument frame; the target information of the inspection area also includes: the instrument value of each target information instrument frame.

13. The instrument identification method according to claim 9, characterized in that, The prior information of the inspection area also includes: the instrument value of each prior information instrument frame; the target information of the inspection area also includes: the instrument value of each target information instrument frame.

14. A method for identifying industrial equipment with a visually identifiable status indicator or parameter display structure, based on the instrument identification method according to any one of claims 1-13, wherein the instrument is replaced by an industrial equipment with a visually identifiable status indicator or parameter display structure, characterized in that, The prior information includes prior information of at least two industrial devices of the same or different types that have visually identifiable status markers or parameter display structures. The prior information for the inspection area includes: Each piece of prior information includes the location information, center point, or at least two sets of reference coordinates of the industrial equipment. The target information for the inspection area includes: Each target information piece contains the location information of the industrial equipment, its center point, or at least two sets of reference coordinates.

15. The industrial equipment identification method according to claim 14, characterized in that, The prior information of the inspection area also includes: the status identifier or parameter display value of each prior information industrial equipment; the target information of the inspection area also includes: the status identifier or parameter display value of each target information industrial equipment.

16. A system based on the instrument identification method according to any one of claims 1-13, characterized in that: include: The system includes a motion positioning and navigation unit, a camera acquisition and control unit, a data storage unit, a main control and computing unit, and a matching and recognition result output unit, among which: The motion positioning and navigation unit is used to control the movement of the inspection equipment according to the preset inspection path, and after accurately arriving at the preset instrument inspection point, it sends an image acquisition trigger control signal to the camera acquisition control unit. The camera acquisition control unit, upon receiving an image acquisition trigger control signal from the motion positioning and navigation unit, triggers the main control computing unit to determine whether prior information for the inspection area exists in the data storage unit. If not, it acquires images of the inspection area according to preset image acquisition parameters; if so, it acquires images of the inspection area according to initial inspection image acquisition parameters. The acquired inspection area image is then transmitted to the main control computing unit, triggering the main control computing unit to perform a validity check on the inspection area image. When the target detection module of the main control computing unit verifies that the image is valid, the image of the inspected area is stored in the data storage unit. When the target detection module of the main control computing unit verifies that the image is invalid, the re-acquisition signal fed back by the main control computing unit triggers the camera acquisition control unit to readjust the camera focal length and magnification, and re-acquire the image of the inspected area. The main control computing unit includes a target detection module and a multi-table matching module; wherein, The target detection module is used to analyze and process the image transmitted by the camera acquisition control unit to obtain prior information or target information of the inspection area. Then, it determines whether there is an instrument in the inspection area image based on whether the prior information or target information contains the instrument position coordinates, and verifies the validity of the inspection area image: if the prior information or target information contains the instrument, the prior information or target information of the inspection area is output to the data storage unit and the multi-table matching module respectively; otherwise, a signal to re-acquire the image is output to the camera acquisition control unit. At the same time, the main control computing unit marks the inspection area image as abnormal and stores it in the data storage unit. The multi-table matching module is used to determine, based on the target information output by the target detection module and the prior information of the first inspection area corresponding to the data storage unit, the dynamic anchor point and other corresponding matching candidate combinations; or, the dynamic anchor point and other corresponding matching candidate combinations, and verify the other matching candidate combinations corresponding to the dynamic anchor point candidate combinations; to obtain the matching recognition result and store it in the data storage unit. The data storage unit is used to store the following data: valid inspection area images output by the camera acquisition and control unit, invalid inspection area images with anomaly markers output by the main control computing unit, and prior information of the first inspection area, target information of non-first inspection areas, and matching and recognition results output by the main control computing unit. The matching and recognition result output unit is used to push the matching and recognition results in the data storage unit to the remote system platform.

17. A system for identifying industrial equipment based on the method for identifying status indicators or parameter displays with visual recognition capabilities as described in claim 14 or 15, characterized in that: include: The system includes a motion positioning and navigation unit, a camera acquisition and control unit, a data storage unit, a main control and computing unit, and a matching and recognition result output unit, among which: The motion positioning and navigation unit is used to control the movement of the inspection equipment according to the preset inspection path, and after accurately arriving at the preset industrial equipment inspection point, it sends an image acquisition trigger control signal to the camera acquisition control unit. The camera acquisition control unit, upon receiving an image acquisition trigger control signal from the motion positioning and navigation unit, triggers the main control computing unit to determine whether prior information for the inspection area exists in the data storage unit. If not, it acquires images of the inspection area according to preset image acquisition parameters; if so, it acquires images of the inspection area according to initial inspection image acquisition parameters. The acquired inspection area image is then transmitted to the main control computing unit, triggering the main control computing unit to perform a validity check on the inspection area image. When the target detection module of the main control computing unit verifies that the image is valid, the image of the inspected area is stored in the data storage unit. When the target detection module of the main control computing unit verifies that the image is invalid, the re-acquisition signal fed back by the main control computing unit triggers the camera acquisition control unit to readjust the camera focal length and magnification, and re-acquire the image of the inspected area. The main control computing unit includes a target detection module and a multi-industrial equipment matching module; wherein, The target detection module is used to analyze and process the inspection area image transmitted by the camera acquisition and control unit to obtain prior information or target information of the inspection area. Then, it determines whether there is industrial equipment in the inspection area image based on whether the prior information or target information contains the location coordinates of industrial equipment, and verifies the validity of the inspection area image: if there is industrial equipment in the prior information or target information, the prior information or target information of the inspection area is output to the data storage unit and the multi-industrial equipment matching module respectively; otherwise, a signal to re-acquire the image is output to the camera acquisition and control unit. At the same time, the main control computing unit marks the inspection area image as abnormal and stores it in the data storage unit. The multi-industrial equipment matching module is used to determine, based on the target information output by the target detection module and the prior information of the first inspection area corresponding to the data storage unit, a dynamic anchor point and other corresponding matching candidate combinations; or, a dynamic anchor point and other corresponding matching candidate combinations, and verifying the other matching candidate combinations corresponding to the dynamic anchor point candidate combinations; to obtain the matching recognition result and store it in the data storage unit. The data storage unit is used to store the following data: valid inspection area images output by the camera acquisition and control unit, invalid inspection area images with anomaly markers output by the main control computing unit, and prior information of the first inspection area, target information of non-first inspection areas, and matching and recognition results output by the main control computing unit. The matching and recognition result output unit is used to push the matching and recognition results in the data storage unit to the remote system platform.

Images