High-precision size measurement method combining image recognition and parameter calibration

Abstract

The invention relates to the technical field of image recognition, and discloses an image recognition and parameter calibration combined high-precision size measurement method, which comprises the following steps of: acquiring scanning image data of a flame arrester in an initial running state and an initial non-running state; acquiring scanning image data of the flame arrester in the current state; inputting the scanning image data of the flame arrester in the initial operation state and the scanning image data of the flame arrester in the current state into a size measurement model, and outputting a measurement result of the ripple spacing of the flame arrester; according to the method, the image features are separated and calibrated through the deep learning technology, and the ultrasonic scanning image data in the online state can be identified to accurately measure the ripple spacing.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and more specifically, to a high-precision dimension measurement method combining image recognition with parameter calibration. Background Art

[0002] Regarding the image recognition technology for flame arrester measurement, ordinary visible light image recognition technology only requires the flame arrester to be visible and is only applicable to the measurement of flame arresters in an uninstalled state. Although ultrasonic image recognition technology can be applied to the measurement of flame arresters in an uninstalled state, the medium and the flow of the medium in the pipeline in an online state will cause great interference to the ultrasonic wave, causing the image features to distort and lose the expression of the flame arrester corrugation spacing. Therefore, it is generally only applicable to the measurement of flame arresters in an offline state. Summary of the invention

[0003] The present invention provides a high-precision dimension measurement method combining image recognition and parameter calibration, which solves the technical problems in related technologies.

[0004] The present invention provides a high-precision dimension measurement method combining image recognition and parameter calibration, comprising the following steps:

[0005] Step 1, collecting scanning image data of the flame arrester in the initial operating state and the initial non-operating state;

[0006] The initial operating state of the flame arrester refers to the state in which the flame arrester can be ensured to be in normal use and the flame arrester is in the state of conveying medium in the pipeline;

[0007] This state is generally when the flame arrester is installed in the pipeline for the first time. At this time, the pipeline is offline. In the absence of medium, good three-dimensional reconstruction results can be obtained through ultrasonic detection scanning. Then, when the medium is passed into the pipeline, it is the initial operating state of the flame arrester.

[0008] The initial non-operating state means that the flame arrester can be ensured to be in normal use and the flame arrester is in a state where the pipeline does not transport the medium;

[0009] Step 2, collecting the scanned image data of the flame arrester in the current state;

[0010] The current state of the flame arrester means that the state of the flame arrester is unknown and the pipeline is online;

[0011] Step 3, inputting the scanned image data of the flame arrester in the initial operating state and the scanned image data of the flame arrester in the current state into the dimension measurement model, wherein the dimension measurement model at least includes a first image recognition module for identifying the scanned image data of the flame arrester in the initial operating state to obtain a first image feature, a third image recognition module for identifying the scanned image data of the flame arrester in the initial non-operating state to obtain a third image feature, and a third image recognition module for identifying the scanned image data of the flame arrester in the current state to obtain a third image feature;

[0012] The dimensional measurement model also includes a medium feature separation module for identifying the medium and the image feature generated by the movement of the medium from the first image feature and the third image feature;

[0013] The size measurement model also includes a parameter calibration module for removing the media separated by the media feature separation module and the image features generated by the media movement from the third image features;

[0014] The dimension measurement model also includes a measurement module, which is used to input the result of the parameter calibration module processing the third image feature and output the measurement result of the flame arrester corrugation spacing.

[0015] Furthermore, each time it is determined that the flame arrester is in a normal use state, the initial operation state of the flame arrester is updated.

[0016] Further, the measurement result of the corrugation spacing of the flame arrester includes the measurement results of all corrugation spacings of the flame arrester.

[0017] Furthermore, the first image recognition module, the third image recognition module, and the third image recognition module can all adopt convolutional neural networks.

[0018] Furthermore, the scanned image data in the initial operating state, the initial non-operating state, and the current state are all collected in a set time period according to a set cycle, so the collected scanned image data is a sequence data.

[0019] Further, the first image features are encoded according to the time sequence of the acquisition of the corresponding scanned image to obtain a first time series, wherein one sequence unit of the first time series corresponds to one first image feature;

[0020] Encode the second image features according to the time sequence of acquisition of the corresponding scanned image to obtain a second time series, wherein one sequence unit of the second time series corresponds to one second image feature;

[0021] The third image features are encoded according to the time sequence of the acquisition of the corresponding scanned images to obtain a third time series, and one sequence unit of the third time series corresponds to one third image feature.

[0022] Furthermore, the medium feature separation module includes two parallel sequence layers, which respectively input the first time series and the second time series. The two sequence layers respectively output the update status to the separation layer, and the separation layer outputs the medium and the image features generated by the medium movement.

[0023] Furthermore, the parameter calibration module includes a timing layer and a calibration layer, wherein the timing layer inputs a third time series and outputs deep features to the calibration layer, and the calibration layer outputs the image features generated by removing the medium separated by the medium feature separation module and the medium movement from the third image features.

[0024] Furthermore, the measurement module comprises a fully connected layer for outputting per unit values ​​of all corrugation spacings of the flame arrester.

[0025] The present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps in each of the above-mentioned high-precision dimension measurement method embodiments combining image recognition with parameter calibration are implemented.

[0026] The beneficial effect of the present invention is that the present invention separates and calibrates image features through deep learning technology, and can identify ultrasonic scanning image data in an online state to perform accurate corrugation spacing measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of a high-precision dimension measurement method combining image recognition and parameter calibration of the present invention;

[0028] Figure 2 is an example of the corrugation spacing of the flame arrester of the present invention;

[0029] Figure 3 It is a module schematic diagram of the computer device of the present invention. DETAILED DESCRIPTION

[0030] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0031] At least one embodiment of the present invention discloses a high-precision dimension measurement method combining image recognition and parameter calibration, such as Figure 1 As shown, the following steps are included:

[0032] Step 1, collecting scanning image data of the flame arrester in the initial operating state and the initial non-operating state;

[0033] The initial operating state of the flame arrester refers to the state in which the flame arrester can be ensured to be in normal use and the flame arrester is in the state of conveying medium in the pipeline;

[0034] This state is generally when the flame arrester is installed in the pipeline for the first time. At this time, the pipeline is offline. In the absence of medium, good three-dimensional reconstruction results can be obtained through ultrasonic detection scanning. Then, when the medium is passed into the pipeline, it is the initial operating state of the flame arrester.

[0035] The initial non-operating state means that the flame arrester can be ensured to be in normal use and the flame arrester is in a state where the pipeline does not transport the medium;

[0036] Of course, it does not rule out the existence of the initial operating state of the flame arrester in other situations. For example, when the system in which the pipeline is located is under maintenance or stops running, the pipeline is also in a state of no medium. The flame arrester is tested to be intact, and then the medium is re-transmitted in the pipeline, which is the initial operating state of the flame arrester.

[0037] The aforementioned medium and pipeline are a general term. The pipeline includes all pipelines that can be used for flame arresters. The medium is the medium that can be transported by the aforementioned pipeline. The medium is generally liquid hydrocarbons, LNG, etc.

[0038] Step 2, collecting the scanned image data of the flame arrester in the current state;

[0039] The current state of the flame arrester does not conflict with the initial operating state of the flame arrester. The current state of the flame arrester means that the state of the flame arrester is unknown and the pipeline is online. Therefore, it may also include the initial operating state of the flame arrester.

[0040] Therefore, each time the flame arrester is judged to be in normal use, the initial operating state of the flame arrester can be updated, so that the initial operating state of the flame arrester compared with the next detection of the current state of the flame arrester is closer;

[0041] Step 3, inputting the scanned image data of the flame arrester in the initial operating state and the scanned image data of the flame arrester in the current state into the dimension measurement model, wherein the dimension measurement model at least includes a first image recognition module for identifying the scanned image data of the flame arrester in the initial operating state to obtain a first image feature, a third image recognition module for identifying the scanned image data of the flame arrester in the initial non-operating state to obtain a third image feature, and a third image recognition module for identifying the scanned image data of the flame arrester in the current state to obtain a third image feature;

[0042] The dimensional measurement model also includes a medium feature separation module for identifying the medium and the image feature generated by the movement of the medium from the first image feature and the third image feature;

[0043] The size measurement model also includes a parameter calibration module for removing the media separated by the media feature separation module and the image features generated by the media movement from the third image features;

[0044] The dimension measurement model also includes a measurement module, which is used to input the result of the parameter calibration module processing the third image feature and output the measurement result of the flame arrester corrugation spacing.

[0045] The measurement results of the flame arrester corrugation spacing include the measurement results of all corrugation spacings of the flame arrester, such as Figure 2 The oval mark shown is one corrugation spacing of the flame arrester.

[0046] In some embodiments of the present invention, the first image recognition module, the third image recognition module, and the third image recognition module may all adopt a convolutional neural network (CNN).

[0047] In some embodiments of the present invention, the scanned image data in the initial operating state, the initial non-operating state, and the current state are all collected in a set period of time according to a set cycle, so the collected scanned image data is a sequence data;

[0048] Encode the first image features according to the time sequence of acquisition of the corresponding scanned image to obtain a first time series, wherein one sequence unit of the first time series corresponds to one first image feature;

[0049] Encode the second image features according to the time sequence of acquisition of the corresponding scanned image to obtain a second time series, wherein one sequence unit of the second time series corresponds to one second image feature;

[0050] Encoding the third image features according to the time sequence of acquisition of the corresponding scanned images to obtain a third time series, wherein one sequence unit of the third time series corresponds to one third image feature;

[0051] In one embodiment of the present invention, the medium feature separation module includes two parallel sequence layers, which respectively input a first time series and a second time series, and the two sequence layers respectively output update states to the separation layer, and the separation layer outputs the medium and image features generated by the movement of the medium;

[0052] The expression of the sequence layer is as follows:

[0053] Forget Gate:

[0054] β t =σ(As t-1 +But +D)

[0055] Input Gate:

[0056] α t =σ(Cs t-1 +Du t +e)

[0057] Candidate unit status:

[0058]

[0059] Unit status update:

[0060]

[0061] Output Gate:

[0062] γ t =σ(gs t-1 +Hu t +g)

[0063] Status Update:

[0064]

[0065] β t , α t , γ t : Activation vectors of the forget, input, and output gates.

[0066] z t : The state of the tth unit.

[0067] z t-1 : The state of the t-1th unit.

[0068] The state of the tth candidate unit;

[0069] s t : The tth updated state;

[0070] s t-1 : The t-1th updated state;

[0071] t∈Z + And 1≤t≤n, where n represents the total number of sequence units contained in the first time series or the second time series, Z + represents a set of positive integers, when t=1, s t-1 =T t ; T t represents the t-th sequence unit of the first time series or the second time series;

[0072] A, B, C, D, E, F, G, H: A, B, C, D, E, F, G, H weight matrices.

[0073] d, e, f, g: first, second, third, fourth bias vectors.

[0074] σ: Sigmoid activation function.

[0075] ψ: optional non-linear activation function (e.g. tanh).

[0076] Dot product.

[0077] The expression of the separation layer is as follows:

[0078]

[0079] Where Sep1 represents the image features generated by the medium and the medium motion, U and J represent the Uth and Jth weight matrices respectively; and Represent the nth updated state of the output of the two sequence layers respectively;

[0080] In one embodiment of the present invention, the parameter calibration module includes a time series layer and a calibration layer, wherein the time series layer inputs the third time series and outputs the deep features to the calibration layer, and the calibration layer outputs the image features generated by the medium separated by the medium feature separation module and the medium movement removed from the third image features;

[0081] The expression of the timing layer is as follows:

[0082] h t =tanh(W hh h t-1 +W xh X t +b h )

[0083] where h t represents the t-th deep feature, h t-1 represents the t-1th deep feature, X t represents the tth sequence unit of the third time series, W hh and W xh are the first and second sequence weight parameters, b h is the sequence bias parameter, tanh is the hyperbolic tangent function;

[0084] The expression of the calibration layer is as follows:

[0085] Sep2=σ(K*h n -L*Sep1)

[0086] Where Sep2 represents the image features generated by the medium and medium motion after the medium feature separation module removes the medium feature separation module from the third image features, K and J represent the Kth and Jth weight matrices respectively; h n Represents the nth deep feature output by the temporal layer;

[0087] The measurement module includes a fully connected layer, and the expression of the fully connected layer is as follows:

[0088] Q i =σ(out w *Sep2+out b )

[0089] where out w Represents the weight parameter of the fully connected layer, out b represents the bias parameter of the fully connected layer, Q i Represents the per-unit value of the measured value of the i-th corrugation spacing.

[0090] In one embodiment of the present invention, the fully connected layer includes multiple layers, which are output in parallel, and each layer outputs a per-unit value of a measurement value of the ripple spacing.

[0091] In one embodiment of the present invention, the fully connected layer outputs serially, and the ith time step outputs the per-unit value of the measured value of the ith ripple spacing.

[0092] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

[0093] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 3As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to celestial body information. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a high-precision dimension measurement method combining image recognition and parameter calibration is realized. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0094] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0095] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps in each of the above-mentioned high-precision dimension measurement method embodiments combining image recognition and parameter calibration are implemented.

[0096] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in each of the above-mentioned high-precision dimension measurement method embodiments combining image recognition and parameter calibration are implemented.

[0097] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

Claims

1. A high-precision dimension measurement method combining image recognition and parameter calibration, characterized in that: The following steps are involved: Collecting scanning image data of the flame arrester in an initial operating state and an initial non-operating state; Collecting scanned image data of the flame arrester in its current state; Inputting the scanned image data of the flame arrester in the initial operating state and the scanned image data of the flame arrester in the current state into the dimension measurement model, the dimension measurement model at least comprising a first image recognition module for identifying the scanned image data of the flame arrester in the initial operating state to obtain a first image feature, a third image recognition module for identifying the scanned image data of the flame arrester in the initial non-operating state to obtain a third image feature, and a third image recognition module for identifying the scanned image data of the flame arrester in the current state to obtain a third image feature; The dimensional measurement model also includes a medium feature separation module for identifying the medium and the image feature generated by the movement of the medium from the first image feature and the third image feature; The size measurement model also includes a parameter calibration module for removing the media separated by the media feature separation module and the image features generated by the media movement from the third image features; The dimension measurement model also includes a measurement module, which is used to input the result of the parameter calibration module processing the third image feature and output the measurement result of the flame arrester corrugation spacing.

2. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 1 is characterized in that: After each time it is determined that the flame arrester is in a normal use state, the initial operation state of the flame arrester is updated.

3. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 1 is characterized in that: The measurement results of the corrugation spacing of the flame arrester include the measurement results of all corrugation spacings of the flame arrester.

4. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 1 is characterized in that: The first image recognition module, the third image recognition module, and the third image recognition module can all adopt convolutional neural networks.

5. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 1 is characterized in that: The scanned image data in the initial operating state, the initial non-operating state, and the current state are all collected in a set period of time according to a set cycle, so the collected scanned image data is a sequence data.

6. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 5 is characterized in that: Encode the first image features according to the time sequence of acquisition of the corresponding scanned image to obtain a first time series, wherein one sequence unit of the first time series corresponds to one first image feature; Encoding the second image features according to the time sequence of acquisition of the corresponding scanned image to obtain a second time series, wherein one sequence unit of the second time series corresponds to one second image feature; The third image features are encoded according to the time sequence of the acquisition of the corresponding scanned images to obtain a third time series, and one sequence unit of the third time series corresponds to one third image feature.

7. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 6 is characterized in that: The medium feature separation module includes two parallel sequence layers, which respectively input the first time series and the second time series. The two sequence layers respectively output the update status to the separation layer, and the separation layer outputs the medium and the image features generated by the medium movement.

8. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 6, characterized in that: The parameter calibration module includes a timing layer and a calibration layer, wherein the timing layer inputs a third time series and outputs deep features to the calibration layer, and the calibration layer outputs the image features generated by the medium separated by the medium feature separation module and the medium movement in the third image features.

9. The high-precision dimension measurement method combining image recognition and parameter calibration according to claim 8, characterized in that: The measurement module includes a fully connected layer to output the per-unit values ​​of all corrugation spacings of the flame arrester.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by the processor, the steps in each of the above-mentioned high-precision dimension measurement method embodiments combining image recognition with parameter calibration are implemented.

Images