CN114136882B - Fabric fiber component on-line detection system based on near infrared spectrum - Google Patents
Fabric fiber component on-line detection system based on near infrared spectrum Download PDFInfo
- Publication number
- CN114136882B CN114136882B CN202111298071.0A CN202111298071A CN114136882B CN 114136882 B CN114136882 B CN 114136882B CN 202111298071 A CN202111298071 A CN 202111298071A CN 114136882 B CN114136882 B CN 114136882B
- Authority
- CN
- China
- Prior art keywords
- data
- crowdsourcing
- component detection
- cloud
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 91
- 239000004744 fabric Substances 0.000 title claims abstract description 66
- 239000000835 fiber Substances 0.000 title claims abstract description 28
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 22
- 238000002372 labelling Methods 0.000 claims abstract description 19
- 238000000034 method Methods 0.000 claims description 48
- 230000004044 response Effects 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 14
- 238000007405 data analysis Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 12
- 238000003062 neural network model Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000013524 data verification Methods 0.000 claims description 8
- 238000001228 spectrum Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 5
- 230000000739 chaotic effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 14
- 238000013450 outlier detection Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 229920000742 Cotton Polymers 0.000 description 2
- 229920002334 Spandex Polymers 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013021 overheating Methods 0.000 description 2
- 238000004451 qualitative analysis Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 239000004759 spandex Substances 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004753 textile Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/01—Arrangements or apparatus for facilitating the optical investigation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Pathology (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Immunology (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Chemical & Material Sciences (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a near infrared spectrum-based fabric fiber component online detection system, which mainly comprises the following modules: the system comprises a cloud-based crowdsourcing platform, a data acquisition labeling module, a component detection module and an online display module; according to the invention, the cloud-based crowdsourcing platform is established, the data acquisition labeling module, the component detection module and the online display module are organically combined and unified into the cloud platform, so that the problems of scattered, chaotic, targeted and poor timeliness of the traditional data flow are solved, the data acquisition, labeling analysis, component detection and data display of the fabric fiber components are realized, and the full-link service has the characteristics of remarkable convenience, easiness in use, complete functions, easiness in expanding the service and the like.
Description
Technical Field
The invention relates to a near infrared spectrum-based online detection system for fabric fiber components, in particular to a cloud platform-based online detection system.
Background
The fabric production process is easily affected by various external factors, which leads to large quality fluctuation of finished products, and strict quality detection is required. The analysis of the fiber composition of the fabric is divided into two steps: qualitative analysis method and quantitative analysis method of components. The traditional qualitative analysis method and quantitative analysis method of the components have the advantages of more steps, long time consumption, high requirements on technicians, high cost and urgent need of technical innovation.
Recently, studies on analysis of fabric fiber components based on deep neural network methods such as LSTM and RNN have been carried out, but the fabric component analysis task is the same as the general classification qualitative task or regression quantitative task, and has the problems of a large total number of classes and uneven inter-class distribution. Further, the fabric component analysis task is different from the general classification regression task, multiple fabric component classification combinations exist, and the same multiple fabric component combination has the problem that different types of data are unevenly distributed, for example, the fabric production quantity of 90% cotton and 10% spandex is far less than that of 95% cotton and 5% spandex. This problem is more pronounced as the number of mixed component categories increases to triplets and even quaternions. Unbalanced data labels are distributed in a deep neural network model training stage to greatly misguide a model, so that a prediction result deviates towards a header class with a large number of samples, and the model prediction accuracy is reduced.
Disclosure of Invention
In view of the foregoing, embodiments of the present invention have been developed to provide an online detection system for fabric fiber components based on near infrared spectroscopy that overcomes or at least partially solves the foregoing problems.
According to the near infrared spectrum-based fabric fiber component online detection system, a plurality of originally scattered links from data acquisition to neural network model prediction are brought into a unified frame, data circulation logic is integrated through a cloud platform, data acquisition strategies are adjusted in real time according to existing data distribution of a database and prediction historical data of the neural network model, and data acquisition tasks are conveniently and rapidly issued in a crowdsourcing mode, so that the problem of unbalanced data distribution is solved.
The near infrared spectrum-based fabric fiber component online detection system mainly comprises the following modules: the system comprises a cloud-based crowdsourcing platform, a data acquisition labeling module, a component detection module and an online display module.
The cloud-based crowdsourcing platform is used for storing, checking, analyzing data, receiving, forwarding reasoning tasks and managing crowdsourcing tasks;
the data acquisition labeling module is used for acquiring and labeling near infrared spectrum data of fabric fiber components and receiving crowdsourcing tasks;
The component detection module is used for predicting the fiber components of the fabric and training a model on the new data line; the online display module is used for displaying data set information, crowdsourcing platform information, historical reasoning information and publishing crowdsourcing tasks.
Preferably, the cloud-based crowdsourcing platform further comprises a data acquisition verification method, a component detection module service registration method, a component detection request forwarding and load balancing method and a data analysis method.
Preferably, the data acquisition and verification method comprises a smooth comparison verification method and an outlier detection method.
According to the regular smooth comparison and verification method, aiming at the problems that the near infrared spectrum data of the fabric abnormally fluctuate due to error interference such as light leakage, instrument overheating, operation irregularity and the like in the acquisition process, the visible appearance is waveform saw tooth, the Savitzky-Golay smooth noise reduction algorithm is used for processing the spectrum data to obtain the smoothed near infrared spectrum of the fabric, and if the standard deviation is larger than a threshold value in comparison with the original data, the Savitzky-Golay smooth noise reduction algorithm has larger influence on the original data, namely the original data is not smooth enough and has more saw teeth, and the interference in the acquisition process needs to be acquired again.
The outlier detection method refers to the same cloth sample collected by a plurality of crowdsourcing collection staff, and if the standard deviation of the average value of data collected by a certain staff and the cloth uploading data is larger than a threshold value beta, the fact that outliers are likely to be collected is indicated, and errors need to be collected again. And (3) after the outlier data are removed, recalculating the data mean value, and repeating the process until the standard deviation of all the data and the mean value is smaller than beta. Waiting until the crowdsourcing staff required to re-collect and upload certain cloth sample data, and then executing the process again. If any outlier sample cannot be found in the process, all the cloth uploading samples are uniformly detected through outliers.
Preferably, the component detection module service registration method is a service for ensuring that the crowdsourcing platform accesses a plurality of component detection modules. Receiving a service registration request when a module is online, and enabling a cloud crowdsourcing platform user to select a specified component detection service; during this time, the cloud crowd-sourced platform always keeps a connection with each registered component detection module server in the form of heartbeat packets, and once the connection is lost, the component detection module logs off from the cloud platform, and the user cannot access the service. Preferably, the component detection module requests a forwarding and load balancing method, and the process is as follows: when the component detection module is registered for service, the cloud platform initializes the average response time of the module by using the average calculation time of the test submitted by the module. And the cloud platform automatically forwards the detection request to the component detection module with the minimum relative load according to the average response time and the latest request time of each registered component detection module corresponding to the component detection request without the specified module. Each successful request response updates the average response time of the corresponding component detection module of the cloud platform. Preferably, the data analysis method is characterized in that the cloud platform queries label distribution in the current database at regular time, records historical data of each component detection module, and obtains a data analysis result. Preferably, the data acquisition labeling module is characterized in that an acquisition labeling person acquires the near infrared spectrum data of the fabric by using a professional near infrared spectrometer according to the specified fabric fiber component data acquisition standard operation. The acquisition labeling module performs data verification on the data acquired by the near infrared spectrometer to eliminate the operation error of light leakage; simultaneously recording fabric sample components with labels, corresponding to the spectrum data, and submitting the fabric sample components to a cloud crowdsourcing platform after confirming the fabric sample components without errors; and receiving comparison results of spectrum data acquired by a plurality of crowdsourcing acquisition labeling staff of the same sample returned by the cloud crowdsourcing platform, and if the comparison results are failure, requesting the acquisition staff to acquire again.
Preferably, the component detection module, wherein the method further comprises: the method for training the depth model on line comprises an automatic deployment method and a request task queue.
Preferably, the depth model on-line training method comprises the following steps: training a one-dimensional convolutional neural network based on Resnet, initializing a component detection module after deployment, using an automatic scheduling strategy to avoid a large request peak period occupied by resources, automatically downloading a new data set from a cloud platform to complete training, and updating on-line model parameters at regular time.
Preferably, the automatic deployment method comprises the following steps: the component detection module loads a neural network model when initializing an API server, and simulates 10 times of fabric fiber near infrared spectrum data input to obtain test average calculation time and test average single request resource occupancy rate; the component detection module actively sends a service registration request to the cloud crowdsourcing platform, establishes heartbeat packet establishment, and sends test average calculation time so as to initialize the cloud platform to establish average response time, namely load performance index, of the reasoning server, and load balancing is facilitated.
Preferably, the process of the request task queue is as follows: and calculating the maximum concurrent response quantity N by using the average single request resource occupancy rate and the unsolicited resource occupancy rate obtained when the component detection module initializes the API server, and establishing a request task queue. When the number of the received component detection requests is greater than N, the requests are placed in a request task queue until the requests are completed, and the requests are acquired from the head of the request task queue and responded after the resources are released.
Preferably, the online display module is configured to receive a data analysis result sent by the cloud crowd-sourced platform, and display the data analysis result to a user in a visual method. The user can issue crowdsourcing acquisition tasks of the fabrics with specified components according to the displayed database tag distribution diagram, and the data acquisition strategy is timely adjusted.
Compared with the prior art, the invention has the following advantages:
In an embodiment of the present invention, the fabric component on-line detection system includes: the system comprises a cloud-based crowdsourcing platform, a data acquisition labeling module, a component detection module and an online display module. According to the invention, the modules of data acquisition, data analysis, neural network model training, deployment, reasoning and the like from top to bottom are brought into a unified frame, the data circulation logic is integrated through the cloud platform, so that the loss of data circulation in each link is saved, the fabric component online detection system can accurately adjust the data acquisition strategy, flexibly send and receive the data acquisition task, rapidly predict the fabric fiber components, conveniently train the neural network model online, thereby solving or partially solving the problem of unbalanced data distribution, and finally improving the accuracy of detecting the fabric fiber components by the neural network model.
Drawings
The application will be described in further detail with reference to the drawings and the detailed description.
FIG. 1 is a schematic diagram of a fabric fiber composition detection system according to the present invention;
Fig. 2 is a waveform comparison chart of the same cloth multiple samplings provided by the invention.
Detailed Description
Specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. However, the present invention should be understood not to be limited to such an embodiment described below, and the technical idea of the present invention may be implemented in combination with other known technologies or other technologies having the same functions as those of the known technologies.
In the following description of the specific embodiments, for the sake of clarity in explaining the structure and operation of the present invention, description will be given by way of directional terms, but words of front, rear, left, right, outer, inner, outer, inner, axial, radial, etc. are words of convenience and are not to be construed as limiting terms.
The relevant terms are explained as follows:
and (3) data verification: the process of rechecking and verifying data aims to remove, correct errors that exist, and provide data consistency.
Service registration: an application may be composed of a group of services with a single responsibility, and after a single service instance is initialized successfully, the service instance information is submitted to a registry, and the connection is kept by a heartbeat packet, so that the elastic capacity expansion and independent services are realized.
Specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
Referring to fig. 1, the on-line detection system for fabric fiber components comprises the following steps S1 to S5:
1. Step S1: initializing cloud-based crowdsourcing platform
The cloud-based crowdsourcing platform is used for storing, checking, analyzing data, receiving, forwarding reasoning tasks and managing crowdsourcing tasks; comprising a number of
The system comprises a data acquisition and verification module, a component detection service registration module, a component detection request forwarding and load balancing module, a data analysis module and a crowdsourcing task management module.
The steps further include:
Step S11: and initializing a data verification module.
The cloud-based crowdsourcing platform opens a designated port, receives acquired data of crowdsourcing staff, and uploads the data to the crowdsourcing staff
And performing data verification. The data acquisition and verification method comprises a smooth comparison verification method and an outlier detection method.
Step S12: and carrying out smooth comparison and verification on the data uploaded by the crowdsourcing staff.
Smooth comparison verification aims at near-red fabric caused by error interference such as light leakage, instrument overheating, irregular operation and the like in the acquisition process
The external spectrum data abnormally fluctuates, and the problem of waveform saw tooth is visually represented.
In the step, the Savitzky-Golay smoothing noise reduction algorithm can be used for processing the spectral data uploaded by the crowdsourcing staff to obtain a smoothed fabric near infrared spectrum, if the standard deviation of the comparison original data is larger than a threshold value, the comparison original data is set to be 0.01 in an experiment, the Savitzky-Golay smoothing noise reduction algorithm is proved to have a larger influence on the original data, namely the original data is not smooth enough and has more saw teeth, and interference in the acquisition process needs to be acquired again.
Step S12: and detecting outliers of the data uploaded by the crowdsourcing staff.
The detection result of the same piece of cloth sample may be affected by multiple factors such as temperature, humidity, irregular manual operation, etc., so that the acquired data may be different, as shown in fig. 2. When the variance is large, the data collected at this time is considered as outliers and should not be included in the database. The outlier detection method needs to wait until all crowdsourcing staff distributed by a certain cloth sample are uploaded, and is performed after smooth comparison and verification, namely the same cloth sample collected by a plurality of crowdsourcing collection staff, if the standard deviation of the data collected by a certain staff and the average value of the cloth uploading data is larger than a threshold value beta, the fact that the outlier is likely to be collected and errors are needed to be collected again is indicated. Because the existence of the outlier can affect the data mean, the process needs to be repeated after the employee is required to re-acquire the uploaded data, until the standard deviation of all the data and the mean is smaller than beta. And after waiting until all crowded staff upload the data acquisition, re-meeting the uploading condition of all crowded staff distributed by the cloth sample, and executing the process again. If any outlier sample cannot be found in the process, all the cloth uploading samples are uniformly detected through outliers.
Step S13: and initializing a component detection service registration module. The cloud-based crowdsourcing platform opens a designated port, receives service registration requests of a plurality of component detection modules, analyzes request contents, records information such as IP addresses and ports of the modules to be registered, tests average calculation time and the like, and sends heartbeat packets to each module at a frequency of once per second, so that connection is maintained. When the heartbeat packet loses the connection, retrying for 3 times, and logging off the module if the connection cannot be established, and informing the request forwarding and load balancing module to log off the module.
Step S14: and initializing a component detection request forwarding and load balancing module.
In this step, the module obtains registered component detection module information, such as IP address, port, and test average calculation time, from the service registration module. The module initializes its average response time using the test average computation time submitted by each component detection module.
The module opens a specified component detection request port, receives a component detection request of a user, corresponds to a component detection request of a non-specified module, and automatically forwards the detection request to a component detection module with the minimum relative load according to the average response time and the latest request time of each registered component detection module. And each successful request response can update the average response time of the corresponding component detection module of the cloud platform, and meanwhile, the detection result is recorded into a database.
Step S15: and initializing a data analysis module. The module initializes database connection, queries the current database data in the database at regular time, counts the data label distribution proportion and the sample number of each category and other information. And inquiring historical component detection request data from a database at regular time, taking the predicted components as pseudo tags, and counting the tag distribution proportion of the component detection request, the number of samples of each category and other information. And sending the statistical information to the data detection module at regular time.
Step S16: and initializing a crowdsourcing task management module. The open port of the module receives a crowdsourcing task dispatching request, calculates the load of the crowdsourcing staff according to the amount of the received tasks and the average acquisition time of the current online crowdsourcing staff, and dispatches the tasks to the crowdsourcing staff with smaller load.
2. Step S2: initializing component detection module
The method comprises the steps of automatically deploying and initializing a request task queue and training on a depth model line.
The automatic deployment comprises the steps that an API server is initialized by the module, a neural network model is loaded, 10 times of fabric fiber near infrared spectrum data input is simulated, and the test average calculation time and the test average single request resource occupancy rate are obtained; the component detection module actively sends a service registration request to the cloud crowdsourcing platform, establishes heartbeat packet connection, and sends test average calculation time so as to initialize the cloud platform to establish average response time of the reasoning server, namely load performance index, and load balancing is facilitated.
The process of initializing the request task queue is as follows: and calculating the maximum concurrent response quantity N by using the average single request resource occupancy rate and the unsolicited resource occupancy rate obtained when the component detection module initializes the API server, and establishing a request task queue. When the number of the received component detection requests is greater than N, the requests are placed in a request task queue until the requests are completed, and the requests are acquired from the head of the request task queue and responded after the resources are released.
The depth model on-line training comprises the following steps: the model adopts a one-dimensional convolutional neural network based on Resnet, uses an automatic scheduling strategy, avoids the large request peak period of resource occupation, automatically downloads a new data set from a cloud platform to complete training, and updates on-line model parameters at regular time.
3. Step S3: initializing data display module
In the step, the module initializes an API server, establishes WebSocket connection with the cloud platform, receives a data analysis result of the cloud platform, and visually displays a data distribution diagram, a request distribution diagram and the like to a user. The user can send data acquisition tasks to the cloud platform crowdsourcing task management module according to the data distribution situation in the database, the data distribution situation of the actual request, the user's own needs and the like.
4. Step S4: acquisition of near infrared spectrum data of fabric fibers
The crowdsourcing staff logs in a data acquisition labeling module to get an acquisition task, follow the data acquisition standard, go deep into textile clothing factories, cloth manufacturers and the like, acquire data such as reflectivity, absorptivity, strength and the like of cloth in a near infrared band by adopting a hyperspectral near infrared instrument, clean the data and only adopt spectral sequence data with the wavelength of 900nm-1700 nm. And (3) labeling the acquired data with component labels, uploading the labeled data to a cloud crowdsourcing platform, checking to complete the acquisition task, and if errors exist, re-acquiring near infrared spectrum sequence data of the designated cloth sample.
5. Step S5: fabric fiber composition detection
The user logs in the data acquisition labeling module and labels empty marks after acquiring the near infrared spectrum data of the fabric fiber according to the acquisition standard
And if the label is a component detection mode instead of a data acquisition mode, only smooth comparison verification is carried out after the label is uploaded to the cloud crowdsourcing platform. And after the cloud crowdsourcing platform data verification module confirms that the data is valid, forwarding the request to the corresponding component detection module according to the load condition and timing. And when the timing is greater than the maximum response time T, the component detection module is busy, and the cloud platform forwards the request to other component detection modules.
After receiving the request, the component reasoning module firstly checks whether the number of the current processing requests is smaller than the maximum concurrent request of the module
If the number N is smaller than the number N, inputting the data into the model, waiting for outputting a result, and returning the result to the cloud platform after encoding; if the number of the current processing requests is greater than N, the requests are added to a request task queue.
The foregoing is merely a preferred embodiment of the present application, and it should be noted that modifications and substitutions can be made by those skilled in the art without departing from the technical principles of the present application, and these modifications and substitutions should also be considered as being within the scope of the present application.
Claims (2)
1. A detection method of a near infrared spectrum-based fabric fiber component online detection system, which is characterized in that the detection system comprises:
the cloud-based crowdsourcing platform is used for storing, checking, analyzing data, receiving, forwarding reasoning tasks and managing crowdsourcing tasks;
the data acquisition labeling module is used for acquiring and labeling near infrared spectrum data of fabric fiber components and receiving crowdsourcing tasks;
the component detection module is used for predicting the fiber components of the fabric;
The online display module is used for displaying the data set information, the crowdsourcing platform information, the historical reasoning information and the issuing crowdsourcing task,
The detection method comprises the following steps:
S11: the data acquisition and verification method comprises the steps that a crowdsourcing platform distributes a single data set acquisition task to three crowdsourcing data collectors, and submitted data of three crowdsourcing staff of the same acquisition task are detected; if the standard deviation between the fabric near infrared spectrum data acquired by a crowd-sourced employee and the task submitted data mean value is greater than a threshold value, the acquisition of outliers is indicated, and the acquisition needs to be repeated due to errors; processing the spectrum data by using a Savitzky-Golay smoothing noise reduction algorithm, and comparing the original data if the standard deviation is larger than a threshold value, so as to indicate that interference exists in acquisition and re-acquisition is needed;
S12: the component detection module service registration method includes that a crowdsourcing platform accesses a plurality of component detection modules, receives a service registration request when the modules are online, and a cloud crowdsourcing platform user can select a specified component detection service; during the period, the cloud crowdsourcing platform is always connected with each registered component detection module server in the form of heartbeat packets, once the connection is lost, the component detection module is logged off from the cloud platform, and a user cannot access the service;
S13: the cloud platform automatically forwards the detection request to the component detection module with the minimum relative load according to the average response time and the latest request time of each registered component detection module;
S14: according to the data analysis method, a cloud platform queries label distribution in a current database at regular time, records historical data of each component detection module to obtain a data analysis result, wherein fabric data of which labels are fewer, component detection precision of which labels are relatively poor, reasoning precision difference among the component detection modules is used for researchers and developers to analyze data, a data acquisition strategy is timely adjusted, and neural network model performance is estimated and improved, and S21: collecting and labeling personnel use a professional near infrared spectrometer to collect near infrared spectrum data of the fabric according to the standard operation of collecting and labeling fiber component data of the appointed fabric; the acquisition labeling module performs data verification on the data acquired by the near infrared spectrometer to eliminate the operation error of light leakage; simultaneously recording fabric sample components with labels, corresponding to the spectrum data, and submitting the fabric sample components to a cloud crowdsourcing platform after confirming the fabric sample components without errors; receiving a data verification result returned by the cloud crowdsourcing platform, if the data verification fails, requesting an acquirer to acquire again,
S31: the on-line training method of the depth model comprises the steps that the depth model is based on a Resnet one-dimensional convolutional neural network, a new data set is automatically downloaded from a cloud platform by using an automatic scheduling strategy to complete training, and on-line model parameters are updated regularly;
S32: according to the automatic deployment method, a neural network model is loaded when the component detection module initializes an API server, 10 times of fabric fiber near infrared spectrum data input are simulated, and the test average calculation time and the test average single request resource occupancy rate are obtained; the component detection module actively sends a service registration request to the cloud crowdsourcing platform, establishes heartbeat packet establishment, and sends test average calculation time so as to initialize the cloud platform to establish average response time of the server, namely load performance index, and load balancing is facilitated;
s33: the method comprises the steps of requesting a task queue, calculating the maximum concurrent response quantity N by testing average single request resource occupancy rate and unsolicited resource occupancy rate obtained when an API server is initialized by a component detection module, and establishing the request task queue; when the number of the received component detection requests is greater than N, the requests are placed in a request task queue until the requests are completed, and the requests are acquired from the head of the request task queue and responded after the resources are released.
2. The detection method of the near infrared spectrum-based fabric fiber component online detection system according to claim 1, wherein an online display module of the detection system receives data analysis results sent by a cloud crowdsourcing platform and displays the data analysis results to a user in a visual way; users can issue crowdsourcing acquisition tasks of the specified component fabrics according to the displayed database tag distribution diagram.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111298071.0A CN114136882B (en) | 2021-11-04 | 2021-11-04 | Fabric fiber component on-line detection system based on near infrared spectrum |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111298071.0A CN114136882B (en) | 2021-11-04 | 2021-11-04 | Fabric fiber component on-line detection system based on near infrared spectrum |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114136882A CN114136882A (en) | 2022-03-04 |
| CN114136882B true CN114136882B (en) | 2024-04-19 |
Family
ID=80392484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111298071.0A Active CN114136882B (en) | 2021-11-04 | 2021-11-04 | Fabric fiber component on-line detection system based on near infrared spectrum |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114136882B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117973439A (en) * | 2024-04-01 | 2024-05-03 | 长春理工大学 | Water turbidity prediction method, turbidity prediction system and computer readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109525665A (en) * | 2018-11-16 | 2019-03-26 | 济南浪潮高新科技投资发展有限公司 | A kind of wound visitor's cloud center configuration recommended method based on crowdsourcing |
| CN110110545A (en) * | 2019-03-25 | 2019-08-09 | 湖州师范学院 | The space crowdsourcing Quality Control Model detected based on location privacy protection and tricker |
| CN110297990A (en) * | 2019-05-23 | 2019-10-01 | 东南大学 | The associated detecting method and system of crowdsourcing marketing microblogging and waterborne troops |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9297748B2 (en) * | 2013-03-13 | 2016-03-29 | Aclima Inc. | Distributed sensor system with remote sensor nodes and centralized data processing |
-
2021
- 2021-11-04 CN CN202111298071.0A patent/CN114136882B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109525665A (en) * | 2018-11-16 | 2019-03-26 | 济南浪潮高新科技投资发展有限公司 | A kind of wound visitor's cloud center configuration recommended method based on crowdsourcing |
| CN110110545A (en) * | 2019-03-25 | 2019-08-09 | 湖州师范学院 | The space crowdsourcing Quality Control Model detected based on location privacy protection and tricker |
| CN110297990A (en) * | 2019-05-23 | 2019-10-01 | 东南大学 | The associated detecting method and system of crowdsourcing marketing microblogging and waterborne troops |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114136882A (en) | 2022-03-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI742980B (en) | Computerized system for application programming interface migration and computer-implemented method for application programming interface migration | |
| US6856942B2 (en) | System, method and model for autonomic management of enterprise applications | |
| CN109582301A (en) | Method for processing business, device, equipment and the medium of task based access control scheduling system | |
| US8180762B2 (en) | Database tuning methods | |
| US7970755B2 (en) | Test execution of user SQL in database server code | |
| CN103095475B (en) | The method for inspecting and system of multimode communication device | |
| CN114136882B (en) | Fabric fiber component on-line detection system based on near infrared spectrum | |
| KR102254653B1 (en) | Data labeling system using legacy service and method thereof | |
| CN106598823B (en) | A kind of the residual quantity calculation method and system of network behavior feature | |
| CN104573977A (en) | Quality data management system and method | |
| US20120303625A1 (en) | Managing heterogeneous data | |
| CN112183990A (en) | Self-adaptive inspection monitoring management platform and method based on big data machine learning | |
| CN106980572A (en) | The on-line debugging method and system of distributed system | |
| CN109063433A (en) | Recognition methods, device and the readable storage medium storing program for executing of fictitious users | |
| CN110321531A (en) | Laboratory Instruments information management system | |
| CN106850316A (en) | A kind of optical fiber into end graphically managing method and system | |
| CN110895506A (en) | Construction method and construction system of test data | |
| CN113360353B (en) | Test server and cloud platform | |
| CN107040434A (en) | Terminal computer actual bandwidth remote centralized acquisition system and method | |
| CN116777297B (en) | Machine room evaluation index configuration method and system based on IDC equipment monitoring data | |
| CN110764909B (en) | Method and system for load balancing of storage device of cloud computing platform | |
| CN104268680A (en) | Case detection process real-time synchronizing system | |
| CN112988868B (en) | Steel laboratory data processing method and system | |
| KR101039874B1 (en) | System for integration platform of information communication | |
| CN114154668A (en) | IT system capacity expansion prediction method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |