Intelligent instrument verification method

Abstract

The invention belongs to the field of instrument verification, and provides an intelligent instrument verification method, which comprises the steps of placing an instrument to be detected, shooting an appearance picture, analyzing the appearance picture, matching a target standard file, verifying the instrument to be detected, tracing problems and uploading a result. According to the invention, appearance verification and type matching are completed through the first camera, so that standard matching and an operation process of a specific verification device are facilitated; by means of the traceability model, the defects that an existing intelligent verification device is single in function and has no problem traceability function are overcome, and a reference direction is provided for follow-up instrument problem troubleshooting.

Description

Technical Field

[0001] The invention relates to the field of instrument verification, and in particular to an intelligent instrument verification method. Background Art

[0002] Verification is a quality control and assurance method used to determine whether a product or service meets specified specifications and standards. Verification usually involves testing, measuring and evaluating samples to determine whether they meet specific requirements. The purpose of verification is to ensure that the quality and performance of products or services meet the requirements and provide effective assurance and verification. By conducting verification, any problems can be discovered and corrected in a timely manner to ensure that the product or service reaches the expected quality level. At present, most commonly used verification methods are based on manual operation and recording by staff, which are inefficient and unstable in reliability. Therefore, some intelligent verification equipment has appeared on the market, which can automatically fix the instrument, automatically verify and record, greatly improving the speed and reliability of verification.

[0003] However, most of the existing intelligent quality inspection equipment only calibrates a single type of instrument and has no problem tracing function. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an intelligent instrument calibration method, which completes the appearance calibration and type matching through a first camera to facilitate the matching of standards and the operating procedures of specific calibration devices; through a traceability model, the defects of existing intelligent calibration equipment such as single function and lack of problem traceability function are solved, so as to provide a reference direction for subsequent instrument problem troubleshooting.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] An intelligent instrument verification method, comprising:

[0007] Place the front of the instrument to be tested on the preset conveyor belt;

[0008] Using a preset first camera to take a picture of the instrument to be inspected to obtain an appearance picture;

[0009] Analyze the appearance picture using a preset recognition model to obtain the instrument model and appearance status;

[0010] When the appearance state is normal, the instrument model is matched with a preset calibration standard library to obtain a target standard file;

[0011] Calibrate the instrument to be tested according to the target standard file to obtain a calibration result; the calibration result includes: calibration parameters and qualified status;

[0012] When the qualified state is unqualified, the verification parameters are input into a pre-trained tracing model to obtain a problem tracing result;

[0013] The verification result and the problem tracing result are uploaded to the server.

[0014] Preferably, the training process of the traceability model includes:

[0015] Determine the parameter type and question type;

[0016] Collect the detection parameters and problem causes of the unqualified instruments according to the parameter type and the problem type;

[0017] The parameter type exceeding the parameter standard threshold in each group of the detection parameters is determined as 1, and the parameter type meeting the parameter standard threshold is determined as 0, to obtain an input parameter vector;

[0018] The common items of each group of the problem cause and the problem type are determined as 1, and the different items are determined as 0, to obtain an input type vector;

[0019] According to the input parameter vector and the input type vector, each of the parameter types is mapped to the problem type using a gradient descent method to obtain a plurality of mapping functions;

[0020] All the mapping functions are combined to obtain the traceability model.

[0021] Preferably, the appearance picture is analyzed using a preset recognition model to obtain the instrument model and appearance status, including:

[0022] Denoising the appearance image to obtain a low-noise image;

[0023] Extracting three primary color channel data of the low-noise image;

[0024] The three primary color channel data are input into a recognition model to obtain the instrument model and the appearance status.

[0025] Preferably, the instrument to be tested is calibrated according to the target standard file to obtain a calibration result, including:

[0026] The instrument to be tested is pushed by a push rod with a soft cushion, and stops running when the resistance of the push rod reaches a preset resistance value;

[0027] Using a preset infrared rangefinder to control the second camera to face the instrument to be detected, and using the second camera to automatically focus and automatically adjust the height;

[0028] Using a preset calibration device to calibrate the instrument to be tested, and recording it through the second camera to obtain calibration video data;

[0029] The data of the verification video data are automatically extracted using a preset parameter recognition program, and the extracted data are compared with a preset parameter standard threshold to obtain the verification result.

[0030] Preferably, according to the input parameter vector and the input type vector, each parameter type is mapped to the problem type using a gradient descent method to obtain a plurality of mapping functions, including:

[0031] Construct the original function;

[0032] Preset that the mapping weights of each parameter type and each question type are the same;

[0033] Inputting the input parameter vector into the primitive function to obtain a prediction vector;

[0034] Calculating the Euclidean distance between the input type vector and the prediction vector using a preset loss function, and determining the calculated Euclidean distance as a loss value;

[0035] Calculating the gradient of the loss function according to the loss value;

[0036] The weight of the original function is updated using the gradient, and the step of "inputting the input parameter vector into the original function to obtain a prediction vector" is returned to iterate, and when the loss value is lower than a preset loss standard, the original function of the last iteration is determined as the mapping function.

[0037] Preferably, the tool for extracting the three primary color channel data is OpenCv.

[0038] Preferably, the parameter recognition program embeds a CRNN image recognition model.

[0039] Preferably, the loss function is a multivariate Euclidean distance calculation formula.

[0040] The present invention discloses the following technical effects:

[0041] The present invention provides an intelligent instrument calibration method, which solves the defects of existing intelligent calibration equipment with single function and no problem tracing function through a traceability model, and realizes the function of providing problem tracing direction according to calibration parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 A schematic diagram of the intelligent instrument calibration process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The purpose of the present invention is to provide an intelligent instrument calibration method, which completes the appearance calibration and type matching through a first camera to facilitate the matching of standards and the operating procedures of specific calibration devices; through a traceability model, it solves the defects of existing intelligent calibration equipment with single function and no problem traceability function, so as to provide a reference direction for subsequent instrument problem troubleshooting.

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] Figure 1 A schematic diagram of the intelligent instrument calibration process provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides an intelligent instrument verification method, comprising:

[0048] Step 100: placing the front face of the instrument to be tested on a preset conveyor belt;

[0049] Step 200: Use a preset first camera to take a picture of the instrument to be inspected to obtain an appearance picture;

[0050] Step 300: Analyze the appearance picture using a preset recognition model to obtain the instrument model and appearance status;

[0051] Step 400: When the appearance state is normal, the instrument model is matched with the preset calibration standard library to obtain the target standard file;

[0052] Step 500: Calibrate the instrument to be tested according to the target standard file to obtain the calibration result; the calibration result includes: calibration parameters and qualified status;

[0053] Step 600: When the qualified state is unqualified, the verification parameters are input into the pre-trained traceability model to obtain the problem traceability result;

[0054] Step 700: Upload the verification results and problem tracing results to the server.

[0055] Specifically, the conveyor belt is connected to a plurality of detection platforms, and different types of instruments will be automatically matched to corresponding detection platforms after being detected by the first camera.

[0056] Furthermore, the training process of the provenance model includes:

[0057] Determine the parameter type and question type;

[0058] Collect the test parameters and problem causes of the unqualified instruments according to the parameter type and problem type;

[0059] The parameter type exceeding the parameter standard threshold in each set of detection parameters is determined as 1, and the parameter type meeting the parameter standard threshold is determined as 0, to obtain an input parameter vector;

[0060] The common items of each set of problem causes and problem types are set to 1, and the different items are set to 0, to obtain an input type vector;

[0061] According to the input parameter vector and the input type vector, each parameter type is mapped to the problem type using the gradient descent method to obtain multiple mapping functions;

[0062] All mapping functions are combined to obtain the traceability model.

[0063] Specifically, due to the small amount of instrument failure data in actual production, it is difficult to collect training data. In order to solve this problem, this embodiment adopts a method of artificially creating problems and conducting calibration. That is, the staff sets the faults in advance, records the corresponding problems, and then calibrates the corresponding problems, and records the obtained data. In the actual implementation process, in order to avoid randomness, different personnel are required to calibrate the same fault at least 10 times to different degrees to ensure that the obtained data has reference value.

[0064] Preferably, the appearance picture is analyzed using a preset recognition model to obtain the instrument model and appearance status, including:

[0065] De-noise the appearance image to obtain a low-noise image;

[0066] Extract the three primary color channel data of low-noise images;

[0067] Input the three primary color channel data into the recognition model to obtain the instrument model and appearance status.

[0068] Specifically, the instrument to be tested is calibrated according to the target standard file to obtain the calibration results, including:

[0069] Use a push rod with a soft cushion to push the instrument to be tested, and stop running when the resistance of the push rod reaches the preset resistance value;

[0070] Use the preset infrared rangefinder to control the second camera to face the instrument to be inspected, and use the second camera to automatically focus and adjust the height;

[0071] The instrument to be tested is calibrated using a preset calibration device, and recorded by a second camera to obtain calibration video data;

[0072] The data of the calibration video data is automatically extracted using a preset parameter recognition program, and the extracted data is compared with the preset parameter standard threshold to obtain the calibration result.

[0073] Optionally, in order to realize the function of automatically fixing the instrument, the present embodiment adopts 6 small push rods, 4 of which are used to fix the instrument panel and are cross-distributed; and the other two are used to fix the instrument handle and are parallel and staggered.

[0074] Specifically, during the actual testing process, the staff can preset a solution library, and the equipment will automatically match the corresponding solution based on the type and parameter standards.

[0075] Furthermore, according to the input parameter vector and the input type vector, each parameter type is mapped to the problem type using the gradient descent method to obtain multiple mapping functions, including: (This process will add formulas after the customer's feedback)

[0076] Construct the original function;

[0077] The mapping weights for each parameter type and question type are preset to be the same;

[0078] Input the input parameter vector into the original function to obtain the prediction vector;

[0079] The Euclidean distance between the input type vector and the prediction vector is calculated using a preset loss function, and the calculated Euclidean distance is determined as a loss value;

[0080] Calculate the gradient of the loss function based on the loss value;

[0081] The weight of the original function is updated using the gradient, and the step "inputting the input parameter vector into the original function to obtain the prediction vector" is returned to iterate. When the loss value is lower than the preset loss standard, the original function of the last iteration is determined as the mapping function.

[0082] Specifically, in the process of constructing the mapping function, the inherent weight of each parameter type may be different, resulting in different degrees of influence of different parameter types on the same problem type. Therefore, in the actual construction process, a mapping index (additional formula) can be added to avoid this risk. In the iterative process, the weight and mapping index are used as variables together to find the optimal ratio using the gradient descent method.

[0083] Specifically, the tool for extracting the three primary color channel data is OpenCv.

[0084] Furthermore, the image can be directly extracted by inputting it into OpenCV. Before the extraction is completed, the image can be segmented or divided into regional grids. When dividing the grid, the grid can be refined for the locations with large grayscale value changes (boundary locations or text and other parts) to increase the recognition effect of the model.

[0085] Preferably, the parameter identification program embeds a CRNN image recognition model.

[0086] Optionally, the loss function is a multivariate Euclidean distance calculation formula.

[0087] Specifically, the Euclidean distance is used to calculate the spatial distance between two vectors, and its size can reflect the deviation of the prediction. The larger the calculated value of the Euclidean distance, the worse the prediction effect of the proof function, and the larger its gradient. When the gradient descent method is actually applied for optimization, the learning rate can be changed to increase the optimization speed and accuracy. The larger the learning rate is set, the faster the speed, but the worse the final accuracy. Therefore, a variable learning rate can be used to set the learning rate as a linear function of the loss value. This can ensure that the optimization speed is not too slow and has a better optimization effect.

[0088] The beneficial effects of the present invention are as follows:

[0089] The present invention completes the appearance verification and type matching through the first camera, which facilitates the matching of standards and the operation process of specific verification devices; through the traceability model, it solves the defects of the existing intelligent verification equipment with single function and no problem traceability function, and provides a reference direction for subsequent instrument problem troubleshooting.

[0090] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0091] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. An intelligent instrument verification method, characterized in that: include: Place the front of the instrument to be tested on the preset conveyor belt; Using a preset first camera to take a picture of the instrument to be inspected to obtain an appearance picture; Analyze the appearance picture using a preset recognition model to obtain the instrument model and appearance status; When the appearance state is normal, the instrument model is matched with a preset calibration standard library to obtain a target standard file; Calibrate the instrument to be tested according to the target standard file to obtain a calibration result; the calibration result includes: calibration parameters and qualified status; When the qualified state is unqualified, the verification parameters are input into a pre-trained tracing model to obtain a problem tracing result; The verification result and the problem tracing result are uploaded to the server.

2. The intelligent instrument verification method according to claim 1 is characterized in that: The training process of the traceability model includes: Determine the parameter type and question type; Collect the detection parameters and problem causes of the unqualified instruments according to the parameter type and the problem type; The parameter type exceeding the parameter standard threshold in each group of the detection parameters is determined as 1, and the parameter type meeting the parameter standard threshold is determined as 0, to obtain an input parameter vector; The common items of each group of the problem cause and the problem type are determined as 1, and the different items are determined as 0, to obtain an input type vector; According to the input parameter vector and the input type vector, each of the parameter types is mapped to the problem type using a gradient descent method to obtain a plurality of mapping functions; All the mapping functions are combined to obtain the traceability model.

3. The intelligent instrument verification method according to claim 1 is characterized in that: The appearance picture is analyzed using a preset recognition model to obtain the instrument model and appearance status, including: Denoising the appearance image to obtain a low-noise image; Extracting three primary color channel data of the low-noise image; The three primary color channel data are input into a recognition model to obtain the instrument model and the appearance status.

4. The intelligent instrument verification method according to claim 1 is characterized in that: The instrument to be tested is calibrated according to the target standard file to obtain a calibration result, including: The instrument to be tested is pushed by a push rod with a soft cushion, and stops running when the resistance of the push rod reaches a preset resistance value; Using a preset infrared rangefinder to control the second camera to face the instrument to be detected, and using the second camera to automatically focus and automatically adjust the height; Using a preset calibration device to calibrate the instrument to be tested, and recording it through the second camera to obtain calibration video data; The data of the verification video data are automatically extracted using a preset parameter recognition program, and the extracted data are compared with a preset parameter standard threshold to obtain the verification result.

5. The intelligent instrument verification method according to claim 2 is characterized in that: According to the input parameter vector and the input type vector, each parameter type is mapped to the problem type using a gradient descent method to obtain a plurality of mapping functions, including: Construct the original function; Preset that the mapping weights of each parameter type and each question type are the same; Inputting the input parameter vector into the primitive function to obtain a prediction vector; Calculating the Euclidean distance between the input type vector and the prediction vector using a preset loss function, and determining the calculated Euclidean distance as a loss value; Calculating the gradient of the loss function according to the loss value; The weight of the original function is updated using the gradient, and the step of "inputting the input parameter vector into the original function to obtain a prediction vector" is returned to iterate, and when the loss value is lower than a preset loss standard, the original function of the last iteration is determined as the mapping function.

6. The intelligent instrument verification method according to claim 3 is characterized in that: The tool for extracting the three primary color channel data is OpenCv.

7. The intelligent instrument verification method according to claim 4 is characterized in that: The parameter recognition program embeds a CRNN picture recognition model.

8. The intelligent instrument verification method according to claim 5 is characterized in that: The loss function is a multivariate Euclidean distance calculation formula.

Images