Volumetric measuring device online detection system and method for instrument calibration
By using an online detection system combining camera and level sensor in flow meter calibration, the problems of inaccurate manual recording and low accuracy are solved, high-precision automated calibration and consistency detection are realized, and the intelligence of instrument detection is improved. .
Patent Information
- Application Number
- CN202411431806.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-14
- Publication Date
- 2024-12-06
AI Technical Summary
The prior art has problems in the calibration of flow meters that are inaccurate manual recording, low measurement accuracy and difficult to ensure calibration consistency, especially when high-precision calibration and large-scale calibration.
An online detection system for volumetric measurement device including camera shooting module, liquid level sensor acquisition module, data processing module, image recognition module, human-computer interaction module, alarm module, data transmission module and cloud service module was designed to pass image recognition Compare the data collected with the liquid level sensor to achieve automatic calibration, and use the Adam optimization algorithm to optimize the image recognition model parameters.
It realizes online recording of measurement data under high-precision calibration, provides consistency and traceability for the detection process, improves the intelligence level of instrument detection, and solves the problems of manual recording and accuracy.
Smart Images

Figure CN119085806A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of flow calibration devices, and in particular to an online detection system and method for a volumetric measuring device for instrument calibration. Background Art
[0002] There are many types of flow measurement instruments, and their measurement principles, instrument structures, operating conditions and installation methods are different. In order to unify the flow values of various instruments produced by various manufacturers and achieve a certain measurement accuracy, it is necessary to calibrate and calibrate the newly manufactured or in-use flow meters.
[0003] Commonly used calibration methods include standard flowmeters, mass flow standard devices, and volume flow standard devices. Among them, standard flowmeters usually have a wide calibration range, but limited accuracy, so auxiliary calibration is usually required for high-precision calibration; mass flow standard devices are used to calibrate mass flowmeters, which are limited by the measurement range of electronic scales and are mainly used to calibrate equipment with a small range; volume flow standard devices are mainly used to calibrate volume flowmeters, with a large range and high calibration accuracy, but they require manual receipts and records, and cannot objectively and impartially record the volume of the container. The consistency of calibration cannot be guaranteed, and it can also be measured by a liquid level meter, but the measurement accuracy of the liquid level meter is not high. Therefore, it is necessary to design an online detection system and method for volumetric measurement devices for instrument calibration to address challenges such as manual recording, calibration consistency, and measurement accuracy. Summary of the invention
[0004] The purpose of the present invention is to provide an online detection system and method for a volumetric measuring device for instrument calibration in view of the above-mentioned deficiencies in the prior art, so as to solve the problem of online detection of the volumetric calibration device.
[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:
[0006] In a first aspect, the present invention provides an online detection system for a volumetric measuring device for instrument calibration, the system comprising: a camera shooting module, a liquid level sensor acquisition module, a data processing module, an image recognition module, a human-computer interaction module, an alarm module, a data transmission module, a cloud service module, and a client;
[0007] The data processing module is respectively connected to the camera shooting module and the liquid level sensor acquisition module in communication, the image recognition module, the human-computer interaction module, the alarm module and the data transmission module are respectively connected to the data processing module in communication, the cloud service module is connected to the data transmission module in communication, and the client is connected to the cloud service module in communication;
[0008] The camera shooting module is fastened to the scale through a fixed bracket, and is used to take pictures of the scale at all times. The information of the scale picture includes the height of the water level and the display scale of the scale;
[0009] The liquid level sensor acquisition module is installed above the water pipe to collect the water level in real time;
[0010] The data processing module is used to store the pictures taken by the camera shooting module, send them to the image recognition module for recognition, and receive the returned recognition results; the data processing module is also used to receive the water level height collected by the liquid level sensor acquisition module, and compare it with the water level height result recognized by the image recognition module. When the comparison result is within the preset error range, the recognized water level height data is sent to the cloud service module through the data transmission module; when the comparison result exceeds the preset error range, the data processing module drives the alarm module to alarm;
[0011] The data processing module is also used to receive configuration information sent by the cloud service module via the data transmission module, and the configuration information includes shooting time, photo color, fill light brightness, and exposure gain;
[0012] The human-computer interaction module is used to exchange information with on-site inspection personnel. The human-computer interaction module is used to display the water level height data recognized by the image recognition module, the data collected by the liquid level sensor acquisition module, the data input by the on-site calibration of the inspection personnel, and the configuration information;
[0013] The cloud service module is used to receive and store, analyze and process data;
[0014] The client is used to directly call the test report template and generate a report based on the configuration information and water level data.
[0015] In a second aspect, the present invention further provides an online detection method for a volumetric measuring device for instrument calibration, the method being applied to the system according to the first aspect, the method comprising the following steps:
[0016] Step S1: Obtaining configuration parameters of the camera, the configuration parameters include shooting time parameters, photo color parameters, fill light brightness parameters and exposure gain parameters;
[0017] Step S2: Acquire image information and water level information. The image information is acquired by photographing with a camera, and the water level information is acquired by acquiring with a liquid level sensor acquisition module.
[0018] Step S3: performing image processing, which includes preprocessing, feature extraction and analysis of the image obtained in step S2 for subsequent recognition and analysis, and the feature values of feature extraction include color, shape, texture, edge and line;
[0019] Step S4: using a pre-trained model algorithm to analyze and identify the data obtained in step S3 to obtain a recognition result;
[0020] Step S5: comparing, calibrating, optimizing and storing the recognition result in step S4. The comparison is to compare with the water level data collected by the liquid level sensor acquisition module. The calibration is to calibrate by manually inputting the correct information. The optimization is to train and optimize the model parameters of the image recognition module so that it can converge to the values input by the detection personnel more quickly. The storage is to store the optimized model parameters formed after calibration in the register for subsequent improvement of recognition accuracy.
[0021] Step S6: Upload the water level height data to the cloud service module for storage, analysis and processing;
[0022] Step S7: Produce a standard test report sample and give the calibration result.
[0023] Optionally, the model parameters of the image recognition module are trained and optimized by using the Adam optimization algorithm as an optimizer to update the recognition model parameters. The specific steps include:
[0024] S51: Define the loss function J(θ) to measure the gap between the model's prediction results and the actual labels;
[0025] S52: Pass in training data and corresponding labels, and call the Adam optimization algorithm to iteratively optimize model parameters so that the model can better fit the data.
[0026] Optionally, the training data and corresponding labels are passed in, and the Adam optimization algorithm is called to continuously iteratively optimize the model parameters so that the model can better fit the data. The specific steps include: S521: initializing the model parameters; S522: defining the Adam optimizer and setting the hyperparameters, which include the learning rate, momentum factor, and decay rate; S523: iteratively optimizing the model parameters,
[0027] Iterative optimization of model parameters, specifically including: inputting training data and corresponding labels into the model, calculating the prediction results of the model; back-propagating the model parameters according to the loss function J(θ) and calculating the gradient of the parameters; using the update formula of the Adam optimization algorithm to update the model parameters to reduce the value of the loss function; repeating the above steps until the preset number of iterations is reached or the loss function converges, so that the model can better fit the training data. The update formula of the Adam optimization algorithm is as follows:
[0028] m t =β1m t-1 +(1-β1)g t
[0029] v t =β2v t-1 +(1-β2)g t 2
[0030]
[0031] Among them, m t and v t denote the first-order moment estimate and the second-order moment estimate of the gradient, respectively. and They represent the first-order moment estimate and the second-order moment estimate after bias correction, θ t represents the updated model parameters, g t represents the gradient of the parameter, α represents the learning rate, β1 and β2 represent the momentum factor and decay rate respectively; ∈ is a preset constant used to avoid the case where the denominator is zero.
[0032] The beneficial effects of the present invention include:
[0033] The online detection system of the volumetric measuring device for instrument calibration provided by the present invention comprises: a camera shooting module, a liquid level sensor acquisition module, a data processing module connected to the camera shooting module and the liquid level sensor acquisition module, an image recognition module connected to the data processing module, a human-computer interaction module, an alarm module and a data transmission module, a cloud service module connected to the data transmission module, and a client connected to the cloud service module. Through the above method, the measurement data can be recorded online under the premise of high-precision calibration of the instrument, which provides a basis for the consistency and traceability of the detection process, and also gradually realizes digitalization of the management of the detection process, thereby improving the intelligent level of the instrument detection process. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art 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 work.
[0035] Figure 1 A structural principle block diagram of an online detection system for a volumetric measuring device for instrument calibration provided by an embodiment of the present invention is shown;
[0036] Figure 2 A schematic diagram of the installation structure of a camera and a scale of an online detection system of a volumetric measuring device for instrument calibration provided by an embodiment of the present invention is shown;
[0037] Figure 3A schematic diagram of the structure of a data processing module of an online detection system for a volumetric measuring device for instrument calibration provided by an embodiment of the present invention is shown;
[0038] Figure 4 A schematic diagram of the structure of a human-computer interaction module of an online detection system for a volumetric measuring device for instrument calibration provided by an embodiment of the present invention is shown;
[0039] Figure 5 A schematic flow chart of an online detection method of a volumetric measuring device for instrument calibration provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION
[0040] 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.
[0041] Commonly used calibration methods include standard flowmeters, mass flow standard devices, and volume flow standard devices. Among them, standard flowmeters usually have a wide calibration range, but limited accuracy, so auxiliary calibration is usually required for high-precision calibration; mass flow standard devices are used to calibrate mass flowmeters, which are limited by the measurement range of electronic scales and are mainly used to calibrate equipment with a small range; volume flow standard devices are mainly used to calibrate volume flowmeters, with a large range and high calibration accuracy, but they require manual receipts and records, and cannot objectively and impartially record the volume of the container. The consistency of calibration cannot be guaranteed, and it can also be measured by a liquid level meter, but the measurement accuracy of the liquid level meter is not high. Therefore, it is necessary to design an online detection system and method for volumetric measurement devices for instrument calibration to address challenges such as manual recording, calibration consistency, and measurement accuracy.
[0042] In a first aspect, the present invention provides an online detection system for a volumetric measuring device for instrument calibration. Figure 1 FIG. 1 shows a block diagram of the structure principle of an online detection system for a volumetric measuring device for instrument calibration provided by an embodiment of the present invention. Figure 1 As shown, the system includes: a camera shooting module 1, a liquid level sensor acquisition module 2, a data processing module 3, an image recognition module 4, a human-computer interaction module 5, an alarm module 6, a data transmission module 7, a cloud service module 8, and a client 9.
[0043] The data processing module 3 is directly connected to the camera shooting module 1 and the liquid level sensor acquisition module 2 respectively, the image recognition module 4, the human-computer interaction module 5, the alarm module 6 and the data transmission module 7 are respectively connected to the data processing module 3, the cloud service module 8 is connected to the data transmission module 6, and the client 9 is connected to the cloud service module 8.
[0044] Figure 2 FIG. 1 shows a schematic diagram of the camera and scale installation structure of an online detection system of a volumetric measuring device for instrument calibration provided by an embodiment of the present invention, such as Figure 2 As shown, the camera shooting module 1 adopts a lightweight camera, and the camera shooting module 1 is fastened to the scale 103 through a fixed bracket 102, and is used to always take pictures of the scale 103 at a level view. In order to prevent the camera shooting module 1 and the scale 103 from tilting due to loose fastening during long-term use, the information of the taken picture of the scale 103 includes both the water level height and the display scale of the scale 103.
[0045] The liquid level sensor acquisition module 2 is installed above the water pipe and is used to collect the water level height in real time to assist in calibrating the water level height data recognized by the image recognition module 4 .
[0046] The data processing module 3 is used to store the pictures taken by the camera shooting module 1, and send them to the image recognition module 4 for recognition, and receive the returned recognition results; the data processing module 3 is also used to receive the water level height collected by the liquid level sensor acquisition module 2, and compare it with the water level height result recognized by the image recognition module 4. When the comparison result is within the preset error range, the recognized water level height data is sent to the cloud service module 8 through the data transmission module 7; when the comparison result exceeds the preset error range, the data processing module 3 drives the alarm module 6 to alarm. Specifically, if Figure 3As shown, the data processing module 3 includes functions such as data acquisition 301, data processing 302, data calculation 303, data storage 304, data control 305, data transmission 306 and data reception 307. Data acquisition 301 mainly collects the pictures taken by the camera shooting module 1 and the water level height collected by the liquid level sensor acquisition module 2; after the collected data is pre-processed by the data processing 302, the image acquisition data is sent to the image recognition module 4 for recognition, and the recognized result is returned to the data processing module 3, and then the water level height data collected by the liquid level sensor acquisition module 2 is compared by calling the data calculation 303. If the error is within the allowed range, the recognized data will be uploaded to the cloud server module 8 through the data transmission module 7. If it is not within the allowed range, the data control 304 of the data processing module 3 drives the alarm module 6 to alarm, and the on-site inspection personnel input the water level data actually read on the site through the human-computer interaction module 5 to realize the calibration of the water level height data, and the re-formed calibration model data is stored in the data storage 304 of the data processing module 3, which is convenient for improving the recognition accuracy of subsequent image data.
[0047] The data processing module 3 is also used to receive the configuration information sent by the cloud service module 8 via the data transmission module 7, and store it in the data storage 304. The configuration information includes shooting parameters such as shooting time, photo color, fill light brightness, exposure gain, etc., to facilitate clearer and more controllable shooting of water level scale information on site.
[0048] The human-computer interaction module 5 is used to exchange information with the on-site inspection personnel. The human-computer interaction module 5 is used to display the water level data recognized by the image recognition module 4, the data collected by the liquid level sensor acquisition module 2, the data input by the on-site inspection personnel for calibration, and the configuration information. Specifically, Figure 4 As shown, the human-computer interaction module 5 includes a data collection display 501, a recognition result display 502, an actual data input 503, a fitting result display 504, a configuration parameter display 505, a configuration parameter input 506, and an equipment operation display 507. On-site inspection personnel can view the data collected by the liquid level sensor acquisition module 2 through the data collection display 501, and view the results of image recognition by the camera shooting module 1 through the recognition result display 502. At the same time, when the alarm module 6 alarms, the recognition result calibration is performed by viewing the actual data on site and inputting it in the actual data input 503, and the fitting result display 504 is the calibrated data. For the convenience of on-site debugging, the configuration parameter display 505 will display the configuration parameters related to the camera shooting module 1. When the photo appears dim light, exposure, etc., the parameter value is changed through the configuration parameter input 506. The equipment operation display 507 mainly displays the operation status of the measuring device, so that when there is a problem with the equipment, it can be located faster and more accurately.
[0049] The cloud service module 8 is used to receive, store, analyze and process data, including water level scale information, relevant configuration information and optimized model parameters.
[0050] The client 9 is used to directly call the test report template and generate a report based on the configuration information and water level height data. Specifically, the client 9 generates a standard test report by calling the water level scale information and related configuration information stored in the cloud service module 8 for submission to customers and archiving.
[0051] Through the above method, the measurement data can be recorded online under the premise of high-precision calibration of the instrument, which provides a basis for the consistency and traceability of the detection process, and also enables the management of the detection process to be gradually digitized, thereby improving the intelligence level of the instrument detection process.
[0052] In a second aspect, the present invention further provides an online detection method for a volumetric measuring device for instrument calibration, the method being applied to the system according to the first aspect above, Figure 5 A schematic flow chart of an online detection method of a volumetric measuring device for instrument calibration provided by an embodiment of the present invention is shown.
[0053] Reference Figure 5 , the method comprises the following steps:
[0054] Step S1: Obtaining configuration parameters of the camera, the configuration parameters include shooting time parameters, photo color parameters, fill light brightness parameters and exposure gain parameters;
[0055] Step S2: Acquire image information and water level information. The image information is acquired by photographing with a camera, and the water level information is acquired by acquiring with a liquid level sensor acquisition module.
[0056] Step S3: performing image processing, which includes preprocessing, feature extraction and analysis of the image obtained in step S2 for subsequent recognition and analysis, and the feature values of feature extraction include information such as color, shape, texture, edge, line, etc.;
[0057] Step S4: using a pre-trained model algorithm to analyze and identify the data obtained in step S3 to obtain a recognition result;
[0058] Step S5: Compare, calibrate, optimize and store the recognition result in step S4. Comparison is to compare with the water level data collected by the liquid level sensor acquisition module. Calibration is to calibrate by manually inputting correct information. Optimization is to train the model parameters of the image recognition module and optimize the parameters so that it can converge to the numerical values input by the detection personnel more quickly. Storage is to store the optimized model parameters after calibration in the register for subsequent recognition accuracy improvement. Compare the image recognition result to see if it is within the error range with the water level data collected by the liquid level sensor acquisition module. If not, drive the alarm module and the detection personnel to view the actual data on site and input it into the corresponding display interface.
[0059] Step S6: Upload the water level data to the cloud service module for storage, analysis and processing.
[0060] Step S7: Produce a standard test report sample and give the calibration result.
[0061] In practical applications, the system of the present invention is used to perform online detection on the volumetric measuring device. The data processing module 3 obtains the configuration parameters and configures the parameters of the camera through the cloud service module 8. The camera shooting module 1 takes pictures, and the liquid level sensor acquisition module 2 collects the liquid level height data. The data processing module 3 obtains the image data and the liquid level height data. First, the image data is preprocessed to reduce noise and enhance image features to facilitate subsequent recognition and analysis; then feature extraction and analysis are performed to extract representative features from the image, including color, shape, texture, edge, line and other information and analyze them to provide support and reference for subsequent instrument calibration. Then, the image recognition module 4 is called for recognition, and the result is compared with the liquid level height data. If it is within the error range, it is uploaded to the cloud service module 8. If not, the inspection personnel input the actual numerical value for calibration. S51: The data processing module 3 first defines the loss function J(θ) to measure the gap between the model's prediction results and the actual labels. S52: Then the training data and the corresponding labels are passed in, and the Adam optimization algorithm is called to continuously iterate and optimize the model parameters so that the model can better fit the input data. The specific steps include: S521: Initialize the model parameters. Before using the Adam optimization algorithm, the model parameters need to be initialized; S522 : Define the Adam optimizer and set some hyperparameters, including learning rate, momentum factor, and decay rate; S523: Iteratively optimize model parameters, iteratively optimize model parameters, specifically including: input training data and corresponding labels into the model, and calculate the prediction results of the model; back-propagate the parameters of the model according to the loss function J(θ) and calculate the gradient of the parameters; use the update formula of the Adam optimization algorithm to update the parameters of the model to reduce the value of the loss function; repeat the above steps until the preset number of iterations is reached or the loss function converges, so that the model can better fit the training data. The update formula of the Adam optimization algorithm is as follows:
[0062] m t =β1m t-1 +(1-β1)g t
[0063] v t =β2v t-1 +(1-β2)g t 2
[0064]
[0065] Among them, m t and v t denote the first-order moment estimate and the second-order moment estimate of the gradient, respectively. and They represent the first-order moment estimate and the second-order moment estimate after bias correction, θ trepresents the updated model parameters, g t represents the gradient of the parameter, α represents the learning rate, β1 and β2 represent the momentum factor and decay rate respectively; ∈ is a preset constant used to avoid the case where the denominator is zero.
[0066] After Adam optimization, the new model parameters are stored in the register of the data processing module 3, which is convenient for improving the subsequent accuracy. At the same time, the client calls the online detection data through the cloud service module 8 to generate a standard detection report, issue it to the customer and archive it.
[0067] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it. They cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.
Claims
1. An online detection system for a volumetric measuring device for instrument calibration, characterized in that: The system includes: a camera shooting module, a liquid level sensor acquisition module, a data processing module, an image recognition module, a human-computer interaction module, an alarm module, a data transmission module, a cloud service module, and a client; The data processing module is respectively connected to the camera shooting module and the liquid level sensor acquisition module in communication, the image recognition module, the human-computer interaction module, the alarm module and the data transmission module are respectively connected to the data processing module in communication, the cloud service module is connected to the data transmission module in communication, and the client is connected to the cloud service module in communication; The camera shooting module is fastened to the scale through a fixed bracket, and is used to take pictures of the scale at all times, wherein the information of the scale picture includes the height of the water level and the display scale of the scale; The liquid level sensor acquisition module is installed above the water pipe and is used to collect the water level in real time; The data processing module is used to store the pictures taken by the camera shooting module, send them to the image recognition module for recognition, and receive the returned recognition results; the data processing module is also used to receive the water level height collected by the liquid level sensor acquisition module, and compare it with the water level height result recognized by the image recognition module. When the comparison result is within the preset error range, the recognized water level height data is sent to the cloud service module through the data transmission module; when the comparison result exceeds the preset error range, the data processing module drives the alarm module to alarm; The data processing module is also used to receive configuration information sent by the cloud service module via the data transmission module, wherein the configuration information includes shooting time, photo color, fill light brightness, and exposure gain; The human-computer interaction module is used to exchange information with on-site inspection personnel, and the human-computer interaction module is used to display the water level height data recognized by the image recognition module, the data collected by the liquid level sensor acquisition module, the data input by the inspection personnel for on-site calibration, and the configuration information; The cloud service module is used to receive and store, analyze and process data; The client is used to directly call the test report template and generate a report based on the configuration information and the water level height data.
2. An online detection method for a volumetric measuring device for instrument calibration, characterized in that: The method is applied to the system according to claim 1, and the method comprises the following steps: Step S1: obtaining configuration parameters of the camera, wherein the configuration parameters include shooting time parameters, photo color parameters, fill light brightness parameters and exposure gain parameters; Step S2: acquiring image information and water level information, wherein the image information is acquired by photographing with a camera, and the water level information is acquired by acquiring with a liquid level sensor acquisition module; Step S3: performing image processing, wherein the image processing includes preprocessing, feature extraction and analysis of the image acquired in step S2 for subsequent recognition and analysis, wherein the feature values of the feature extraction include color, shape, texture, edge and line; Step S4: using a pre-trained model algorithm to analyze and identify the data obtained in step S3 to obtain a recognition result; Step S5: comparing, calibrating, optimizing and storing the recognition result in step S4, wherein the comparison is to compare with the water level data collected by the liquid level sensor acquisition module, the calibration is to calibrate by manually inputting the correct information, the optimization is to train the model parameters of the image recognition module and optimize the parameters so that it can converge to the numerical values input by the detection personnel more quickly, and the storage is to store the optimized model parameters formed after calibration in the register for subsequent improvement of recognition accuracy; Step S6: Upload the water level height data to the cloud service module for storage, analysis and processing; Step S7: Produce a standard test report sample and give the calibration result.
3. The online detection method of a volumetric measuring device for instrument calibration according to claim 2, characterized in that: The training and optimization of the model parameters of the image recognition module is to use the Adam optimization algorithm as an optimizer to update the recognition model parameters. The specific steps include: S51: Define the loss function J(θ) to measure the gap between the model's prediction results and the actual labels; S52: Pass in training data and corresponding labels, and call the Adam optimization algorithm to iteratively optimize model parameters so that the model can better fit the data.
4. The online detection method of a volumetric measuring device for instrument calibration according to claim 3, characterized in that: The training data and corresponding labels are passed in, and the Adam optimization algorithm is called to continuously iteratively optimize the model parameters so that the model can better fit the data. The specific steps include: S521: initializing the model parameters; S522: defining the Adam optimizer and setting the hyperparameters, wherein the hyperparameters include the learning rate, the momentum factor, and the decay rate; S523: iteratively optimizing the model parameters, The iterative optimization of model parameters specifically includes: inputting training data and corresponding labels into the model, calculating the prediction results of the model; back-propagating the parameters of the model according to the loss function J(θ) and calculating the gradient of the parameters; updating the parameters of the model using the update formula of the Adam optimization algorithm to reduce the value of the loss function; repeating the above steps until a preset number of iterations is reached or the loss function converges, so that the model can better fit the training data. The update formula of the Adam optimization algorithm is as follows: m t =β1m t-1 +(1-β1)g t v t =β2v t-1 +(1-β2)g t 2 Among them, m t and v t denote the first-order moment estimate and the second-order moment estimate of the gradient, respectively. and They represent the first-order moment estimate and the second-order moment estimate after bias correction, θ t represents the updated model parameters, g t represents the gradient of the parameter, α represents the learning rate, β1 and β2 represent the momentum factor and decay rate respectively; ∈ is a preset constant used to avoid the case where the denominator is zero.
Citation Information
Cited By
Measuring cylinder liquid level intelligent identification and analysis system based on computer vision and method thereof
CN121917016A
