Ultra-large two-dimensional code positioning identification system and method in wind power scene

Through the combination of data encoding and QR code generation module, object detector model and analysis library, the problem of super large QR code recognition in wind power scenarios is solved, and rapid and accurate identification and correction are achieved, improving data transmission efficiency.

CN118968208APending Publication Date: 2024-11-15GUANGDONG MINGYANG WIND POWER IND GRP CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410784146.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-18
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing QR code positioning technology is difficult to identify large size, tilted or parallel super-large QR codes in wind power scenarios, resulting in unsatisfactory identification results and cannot meet the needs of data transmission.

Method used

Data encoding and QR code generation modules are used for data extraction, compression and encryption, and QR code pictures with four positioning icons are generated. The target detector model combined with opencv and pytorch is used for positioning and correction, and combined with pyzbar library for analysis and Splicing to realize automated identification and information restoration.

Benefits of technology

It realizes rapid and accurate identification and correction of super-large QR codes in wind power scenarios, reduces the workload of operation and maintenance personnel, improves the efficiency and accuracy of data transmission, and can handle identification tasks in complex scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968208A_ABST
    Figure CN118968208A_ABST
Patent Text Reader

Abstract

The invention discloses an ultra-large two-dimensional code positioning identification system and method in a wind power scene, and the system comprises a data coding and two-dimensional code generation module which is used for extracting data in an original file and carrying out the segmentation of the data, and generating a two-dimensional code; the user side module is used for photographing the two-dimensional code and transmitting the two-dimensional code to the remote server module; the remote server module is used for storing the two-dimensional code picture uploaded by the user side module and deploying a two-dimensional code decoding program; the two-dimensional code positioning module is used for identifying and positioning the two-dimensional code to obtain four-point coordinates of the two-dimensional code; the two-dimensional code correction module is used for correcting and enhancing the two-dimensional code picture; the two-dimensional code analyzing and splicing module is used for analyzing two-dimensional code information in the two-dimensional code picture and splicing the two-dimensional code information; according to the method, the defects of application in a wind power scene in the prior art can be overcome, large-batch data are processed, the two-dimensional code is accurately positioned and identified, and the purpose of offline data transmission is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of two-dimensional code positioning and recognition, and in particular to a large two-dimensional code positioning and recognition system and method in a wind power scenario. Background Art

[0002] The operation data of wind turbines is very important for analyzing and evaluating the operation status of each component of the wind turbines and the power generation efficiency. Since some wind turbines are located in remote areas with poor network signals, or due to certain restrictive reasons, the hardware interfaces of the workstation devices cannot perform online data transmission. At this time, it is necessary to transfer data offline, and generally, two-dimensional code technology is used to transfer data offline.

[0003] Regarding the two-dimensional code positioning technology, for example, in the Chinese invention patent application CN117787311A, a method for accurately positioning the two-dimensional code area is disclosed, which obtains the boundaries of the first code area and the second code area through two precise positionings, and preferably selects the target positioning boundary of the target two-dimensional code. However, the current two-dimensional code positioning technology has the following difficulties in this field: (1) Currently, the two-dimensional code positioning technology generally uses the "square within a square" positioning method to identify the position of the two-dimensional code. However, this "square within a square" only exists at three corners of a single two-dimensional code image. Since image correction requires information of four points, the existing positioning technology cannot meet the requirements; (2) The size of two-dimensional codes in general scenarios is relatively small, and the amount of information contained is also small, making them relatively easy to identify. However, in the wind power scenario, the amount of data is large. To compress and encrypt a large amount of data into a single two-dimensional code, the size of this two-dimensional code is relatively large, and there may be two two-dimensional codes juxtaposed in one picture. This is done to reduce the workload of maintenance personnel, and only a small number of photos need to be taken. Additionally, the two-dimensional codes in these photos are likely to be tilted. For such a situation where there are two large-sized tilted two-dimensional codes in one picture, the existing methods have unsatisfactory recognition effects. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and propose a large two-dimensional code positioning and recognition system and method in a wind power scenario, which can extract, compress, and encrypt the file information based on the file exported by the maintenance personnel, only need to take a picture of the two-dimensional code image and upload it, and then parse and merge the information of the uploaded two-dimensional code image to restore the original file, thereby achieving the effect of offline data transmission.

[0005] The purpose of the present invention is achieved through the following technical solutions: A large two-dimensional code positioning and recognition system in a wind power scenario, comprising:

[0006] A data encoding and two-dimensional code generation module, configured to extract data from the original file, segment the data according to a preset byte threshold to obtain encoded data, and finally generate a two-dimensional code based on the processed data;

[0007] The user-side module takes a picture of the QR code and transmits the obtained QR code picture to the remote server module;

[0008] The remote server module is used to store the QR code images uploaded by the user-side module and deploy the QR code positioning module, QR code correction module, and QR code parsing and splicing module;

[0009] The QR code positioning module identifies and locates the QR code according to the QR code image in the remote server module to obtain the four-point coordinates of the QR code;

[0010] The QR code correction module is used to correct and enhance the QR code image to highlight the QR code part;

[0011] The QR code parsing and splicing module is used to parse the QR code information in the QR code image and splice the parsed QR code information according to the original file name and serial number. If there is omission in the splicing process, the missing information will be recorded in the log.

[0012] Furthermore, the data encoding and QR code generation module includes:

[0013] Use Python program to read the data in the original file, and judge the number of bytes in the data in units of lines. When the number of bytes reaches the preset maximum carrying threshold of the QR code, compress this part of information using Brotli algorithm, encrypt the data using AES symmetric encryption algorithm, and then perform base64 encoding and store it in a list. Finally, take two adjacent elements in the list to generate two QR codes, and then splice the two QR codes together horizontally, and add small positioning icons to the four corners of the picture to generate a QR code picture. The QR code picture includes the original file name, segmentation sequence number and number of segments, which are used for information splicing and information omission reminder during subsequent decoding. At the same time, the data processing and QR code generation module is based on a visual interface and multi-threaded operation, which is convenient for on-site personnel to operate.

[0014] Furthermore, the two-dimensional code positioning module includes:

[0015] First, we use the open source QR code positioning program wechat_qrcode_WeChatQRCode integrated with opencv to directly identify the QR code position in the image;

[0016] If the QR code position is still not detected using the open source QR code positioning program, use several test files on the local computer to generate multiple corresponding QR code images, and annotate the four-corner positioning icons in these QR code images to construct training sets and validation sets. The proportion of the training set is 60%-80%, and the proportion of the validation set is 20%-40% accordingly;

[0017] Based on the pytorch framework, the target detector model uses yolox-tiny. The input of the target detector model training is a dynamic size with a size range of 960-1280. After the training is completed, it is converted into onnx format for deployment. The output of the target detector model is the coordinate bounding box of the four positioning icons. Then, these four bounding boxes are rearranged in a clockwise manner starting from the upper left point. The lower right coordinate of the first bounding box, the lower left coordinate of the second bounding box, the upper left coordinate of the third bounding box, and the upper right coordinate of the fourth bounding box are taken in turn. The coordinates of the four points are taken out in turn to complete the positioning of the QR code image.

[0018] Furthermore, the two-dimensional code correction module includes:

[0019] After obtaining the coordinates of the four points in the QR code image by relying on the QR code positioning module, a perspective transformation is performed on it, the QR code image is corrected, and data enhancement is performed on the corrected QR code image. The data enhancement methods include Gaussian blur and grayscale image to highlight the black part of the QR code image.

[0020] Furthermore, the two-dimensional code parsing and splicing module includes:

[0021] Use the pyzbar tool library to identify the QR code. First, directly identify the analyzed and corrected QR code image. If it is not recognized, perform data enhancement. The data enhancement includes grayscale conversion, Gaussian blur noise reduction and binarization. Then, analyze the enhanced QR code image and decrypt the information using the AES decoding algorithm. According to the file name and segmentation number added in the QR code before, perform information splicing and information omission reminder.

[0022] Further, the remote server module includes:

[0023] The QR code positioning module, QR code correction module and QR code parsing and splicing module are deployed in the remote server module, leaving only one API interface. The backend only needs to call this API interface to call the entire decoding program, which is convenient for program updates and migration.

[0024] A positioning and identification method based on the super-large two-dimensional code positioning and identification system in the above wind power scenario includes the following steps:

[0025] S1. Input file data;

[0026] S2, encode according to the file data, generate a QR code image and take a photo of it;

[0027] S3, taking the QR code image obtained by taking the photo in step S2 as a training sample and marking it;

[0028] S4, preprocessing the training samples by selecting a dynamic input size;

[0029] S5. Input the training samples into the target detector model yolox-tiny based on the pytorch framework for training;

[0030] S6, deploying the trained target detector model to the remote server module;

[0031] S7. The target detector model in the remote server module parses the result of the QR code image and performs correction and splicing restoration operations.

[0032] Further, the step S2 comprises:

[0033] Use Python program to read the data in the original file, and judge the number of bytes in the data in units of lines. When the number of bytes reaches the preset maximum carrying threshold of the QR code, compress this part of information using Brotli algorithm, encrypt the data using AES symmetric encryption algorithm, and then perform base64 encoding and store it in a list. Finally, take two adjacent elements in the list to generate two QR codes, and then splice the two QR codes together horizontally, and add small positioning icons to the four corners of the picture to generate a QR code picture. The QR code picture includes the original file name, segmentation sequence number and number of segments, which are used for information splicing and information omission reminder during subsequent decoding. At the same time, the data processing and QR code generation module is based on a visual interface and multi-threaded operation, which is convenient for on-site personnel to operate.

[0034] Further, the step S5 comprises:

[0035] Based on the pytorch framework, the target detector model uses yolox-tiny. The input of the target detector model training is a dynamic size with a size range of 960-1280. After the training is completed, it is converted into onnx format for deployment. The output of the target detector model is the coordinate bounding box of the four positioning icons. Then, these four bounding boxes are rearranged in a clockwise manner starting from the upper left point. The lower right coordinate of the first bounding box, the lower left coordinate of the second bounding box, the upper left coordinate of the third bounding box, and the upper right coordinate of the fourth bounding box are taken in turn. The coordinates of the four points are taken out in turn to complete the positioning of the QR code image.

[0036] Further, the step S7 comprises:

[0037] Use the pyzbar tool library to identify the QR code. First, directly identify the analyzed and corrected QR code image. If it is not recognized, perform data enhancement. The data enhancement includes grayscale conversion, Gaussian blur noise reduction and binarization. Then, analyze the enhanced QR code image and decrypt the information using the AES decoding algorithm. According to the file name and segmentation number added in the QR code before, perform information splicing and information omission reminder.

[0038] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0039] 1. In the QR code generation stage, a visual interface is developed and multi-threaded processing is added, which can quickly and conveniently divide large quantities of data in sequence and generate QR codes, reducing the workload of on-site operation and maintenance personnel;

[0040] 2. Based on the positioning icons of the four corner points of the generated QR code image, train and optimize the target detection model and deploy it. According to the model output results, extract the four-point coordinates to achieve automatic extraction and correction in the scenario of multiple QR codes in parallel. In the later stage, the training model can be used to improve the recognition accuracy in complex and difficult business scenarios. When two QR codes are in parallel and there is background noise interference, the QR code area can be quickly and accurately located and corrected.

[0041] 3. Perform data enhancement on the corrected QR code image, then parse and splice the data. If any information is missing, it will be recorded in the system log for later tracing and supplementation;

[0042] 4. The QR code positioning module, QR code correction module, and QR code parsing and splicing module, i.e., the parsing program part, are deployed on the remote server module and called in API mode, and the model can be inferred only with the help of CPU devices, which is convenient for program use and migration. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of the visualization interface.

[0044] Figure 2 This is a schematic diagram of the generated QR code image.

[0045] Figure 3 This is the architecture diagram of the target detector model yolox-tiny.

[0046] Figure 4 This is a diagram showing the bounding box of the positioning icon output by reasoning and the coordinates of the four points to be retrieved.

[0047] Figure 5 This is a schematic diagram of the corrected QR code image.

[0048] Figure 6The figure is a flow chart of the method for positioning and identifying very large QR codes in wind power scenarios. DETAILED DESCRIPTION

[0049] The present invention will be further described below in conjunction with specific embodiments.

[0050] Example 1

[0051] The ultra-large QR code positioning and recognition system in the wind power scenario provided in this embodiment includes:

[0052] The data encoding and QR code generation module is used to extract the data in the original file, and segment the data according to the preset byte threshold to obtain the encoded data, and finally generate a QR code based on the processed data, including:

[0053] See also Figure 1 As shown in the figure, the data processing and QR code generation module is based on a visual interface and multi-threaded operation, which is convenient for on-site personnel to operate; the Python program is used to read the data in the original file, and the number of bytes in the data is determined in units of lines. When the number of bytes reaches the preset maximum carrying threshold of the QR code, the information is compressed using the Brotli algorithm, the AES symmetric encryption algorithm is used to encrypt the data, and then base64 encoding is performed and stored in a list. Finally, two adjacent elements in the list are taken to generate two QR codes, and then the two QR codes are horizontally spliced ​​together, and small positioning icons are added to the four corners of the picture, see Figure 2 As shown, a QR code image is generated, which includes the original file name, segmentation sequence number and number of segments, which is used for information splicing and information omission reminder during subsequent decoding.

[0054] The user-side module takes a picture of the QR code and transmits the obtained QR code picture to the remote server module;

[0055] The remote server module is used to store the QR code images uploaded by the user-side module and deploy the QR code positioning module, QR code correction module, and QR code parsing and splicing module, including:

[0056] The QR code positioning module, QR code correction module and QR code parsing and splicing module are deployed in the remote server module, leaving only one API interface. The backend only needs to call this API interface to call the entire decoding program, which is convenient for program updates and migration.

[0057] The QR code positioning module identifies and locates the QR code according to the QR code image in the remote server module, and obtains the four-point coordinates of the QR code, including:

[0058] First, we use the open source QR code positioning program wechat_qrcode_WeChatQRCode integrated with opencv to directly identify the QR code position in the image;

[0059] If the QR code position is still not detected using the open source QR code positioning program, use several test files on the local computer to generate multiple corresponding QR code images, and annotate the four-corner positioning icons in these QR code images to construct training sets and validation sets. The proportion of the training set is 60%-80%, and the proportion of the validation set is 20%-40% accordingly;

[0060] See also Figure 3 As shown in the figure, based on the pytorch framework, the target detector model uses yolox-tiny. The input of the target detector model training is a dynamic size, ranging from 960 to 1280. After the training is completed, it is converted into onnx format for deployment; see Figure 4 As shown in the figure, the output of the target detector model is the coordinate bounding boxes of the four positioning icons. The four bounding boxes are then rearranged in a clockwise rotation starting from the upper left point. The lower right coordinate of the first bounding box, the lower left coordinate of the second bounding box, the upper left coordinate of the third bounding box, and the upper right coordinate of the fourth bounding box are taken in turn to complete the positioning of the QR code image.

[0061] The QR code correction module is used to correct and enhance the QR code image to highlight the QR code part, including:

[0062] See also Figure 5 As shown, after obtaining the coordinates of the four points in the QR code image by relying on the QR code positioning module, a perspective transformation is performed on it, the QR code image is corrected, and data enhancement is performed on the corrected QR code image. The data enhancement methods include Gaussian blur and grayscale image to highlight the black part of the QR code image.

[0063] The QR code parsing and splicing module is used to parse the QR code information in the QR code image and splice the parsed QR code information according to the original file name and serial number. If there is any omission in the splicing process, the omission information will be recorded in the log, including:

[0064] Use the pyzbar tool library to identify the QR code. First, directly identify the analyzed and corrected QR code image. If it is not recognized, perform data enhancement. The data enhancement includes grayscale conversion, Gaussian blur noise reduction and binarization. Then, analyze the enhanced QR code image and decrypt the information using the AES decoding algorithm. According to the file name and segmentation number added in the QR code before, perform information splicing and information omission reminder.

[0065] Example 2

[0066] See also Figure 6 As shown, the positioning and identification method of the super-large two-dimensional code positioning and identification system in the wind power scenario provided by this embodiment based on the embodiment 1 includes the following steps:

[0067] S1. Input file data;

[0068] S2. Encode the file data, generate a QR code image and take a photo of it, including:

[0069] Use Python program to read the data in the original file, and judge the number of bytes in the data in units of lines. When the number of bytes reaches the preset maximum carrying threshold of the QR code, compress this part of information using Brotli algorithm, encrypt the data using AES symmetric encryption algorithm, and then perform base64 encoding and store it in a list. Finally, take two adjacent elements in the list to generate two QR codes, and then splice the two QR codes together horizontally, and add small positioning icons to the four corners of the picture to generate a QR code picture. The QR code picture includes the original file name, segmentation sequence number and number of segments, which are used for information splicing and information omission reminder during subsequent decoding. At the same time, the data processing and QR code generation module is based on a visual interface and multi-threaded operation, which is convenient for on-site personnel to operate.

[0070] S3, taking the QR code image obtained by taking the photo in step S2 as a training sample and marking it;

[0071] S4, preprocessing the training samples by selecting a dynamic input size;

[0072] S5. Input the training samples into the yolox-tiny target detector model based on the pytorch framework for training, including:

[0073] Based on the pytorch framework, the target detector model uses yolox-tiny. The input of the target detector model training is a dynamic size with a size range of 960-1280. After the training is completed, it is converted into onnx format for deployment. The output of the target detector model is the coordinate bounding box of the four positioning icons. Then, these four bounding boxes are rearranged in a clockwise manner starting from the upper left point. The lower right coordinate of the first bounding box, the lower left coordinate of the second bounding box, the upper left coordinate of the third bounding box, and the upper right coordinate of the fourth bounding box are taken in turn. The coordinates of the four points are taken out in turn to complete the positioning of the QR code image.

[0074] S6, deploying the trained target detector model to the remote server module;

[0075] S7, the target detector model in the remote server module analyzes the result of the QR code image and performs correction and splicing restoration operations, including:

[0076] Use the pyzbar tool library to identify the QR code. First, directly identify the analyzed and corrected QR code image. If it is not recognized, perform data enhancement. The data enhancement includes grayscale conversion, Gaussian blur noise reduction and binarization. Then, analyze the enhanced QR code image and decrypt the information using the AES decoding algorithm. According to the file name and segmentation number added in the QR code before, perform information splicing and information omission reminder.

[0077] Example 3

[0078] The method for using the super-large two-dimensional code positioning and recognition system in the wind power scenario provided in this embodiment based on the embodiment 1 includes:

[0079] 1) Coding part:

[0080] Taking the input .csv file as an example, the number of bytes is added line by line based on the behavior traversal benchmark to determine whether the number of bytes is greater than the maximum information carrying threshold of the QR code. The maximum information carrying threshold is preferably 28000-30000 bytes. When the maximum carrying threshold is reached, the previous line information is converted into a QR code and stored in a list. When the entire file is traversed, data is taken from the list in pairs and the two QR codes are spliced ​​into a picture to reduce the workload of on-site personnel taking pictures. After taking pictures, the on-site personnel upload them from the user-end module to the remote server module.

[0081] 2) Decoding part:

[0082] In order to remove background noise interference, we first train an object detector model for QR code position detection and prepare a data set. Since the detection model only needs to detect one category, the number of data sets does not need to be large. That is, 30 pictures with different lighting and shooting angles are collected, and the training set and validation set are divided according to the ratio of 8:2. Yolox-tiny is used as the object detection model, the image input size is set to 1200, batch_size is selected as 4, and pytorch is used as the training framework. After training, it is converted to onnx format for deployment. When using the output data, first arrange the data in clockwise order from the upper left corner to obtain the coordinates of the four corner points around the QR code.

[0083] The decoding program first uses the WeChat open source QR code detection module integrated with OpenCV. If it is not detected, it uses the trained yolox-tiny target detector model to get the four-point coordinates. Then, based on these four-point coordinates, it performs image correction, so that the tilted QR code image can be converted into a regular rectangular area.

[0084] Use pyzbar to parse the corrected QR code image. If no recognition result is obtained, perform data enhancement, convert to grayscale, Gaussian blur and binarization to increase the prominence of the black area of ​​the QR code and improve the recognition accuracy. Then use the AES decryption algorithm to decrypt the content, get the order information of the first line, and restore it in the correct order. If data enhancement is performed and the decoding information is still not obtained, record the name of the picture in the system log, indicating that the picture quality is too poor and needs to be re-shot.

[0085] The above decoding program is deployed on the server side in the form of an API, constantly listening for incoming image signals. If an incoming image signal is received, the decoding program is enabled and the parsed and restored file is returned.

[0086] The embodiments described above are only preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. Therefore, all changes made according to the shape and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A super-large QR code positioning and recognition system in wind power scenarios, characterized in that: include: The data encoding and QR code generation module is used to extract the data in the original file, segment the data according to the preset byte threshold, obtain the encoded data, and finally generate a QR code based on the processed data; The user-side module takes a picture of the QR code and transmits the obtained QR code picture to the remote server module; The remote server module is used to store the QR code images uploaded by the user-side module and deploy the QR code positioning module, QR code correction module, and QR code parsing and splicing module; The QR code positioning module identifies and locates the QR code according to the QR code image in the remote server module to obtain the four-point coordinates of the QR code; The QR code correction module is used to correct and enhance the QR code image to highlight the QR code part; The QR code parsing and splicing module is used to parse the QR code information in the QR code image and splice the parsed QR code information according to the original file name and serial number. If there is omission in the splicing process, the missing information will be recorded in the log.

2. According to claim 1, a super-large two-dimensional code positioning and recognition system in a wind power scenario is characterized in that: The data encoding and QR code generation module includes: Use Python program to read the data in the original file, and judge the number of bytes in the data in units of lines. When the number of bytes reaches the preset maximum carrying threshold of the QR code, compress this part of information using Brotli algorithm, encrypt the data using AES symmetric encryption algorithm, and then perform base64 encoding and store it in a list. Finally, take two adjacent elements in the list to generate two QR codes, and then splice the two QR codes together horizontally, and add small positioning icons to the four corners of the picture to generate a QR code picture. The QR code picture includes the original file name, segmentation sequence number and number of segments, which are used for information splicing and information omission reminder during subsequent decoding. At the same time, the data processing and QR code generation module is based on a visual interface and multi-threaded operation, which is convenient for on-site personnel to operate.

3. The ultra-large two-dimensional code positioning and recognition system in a wind power scenario according to claim 1 is characterized in that: The two-dimensional code positioning module includes: First, we use the open source QR code positioning program wechat_qrcode_WeChatQRCode integrated with opencv to directly identify the QR code position in the image; If the QR code position is still not detected using the open source QR code positioning program, use several test files on the local computer to generate multiple corresponding QR code images, and annotate the four-corner positioning icons in these QR code images to construct training sets and validation sets. The proportion of the training set is 60%-80%, and the proportion of the validation set is 20%-40% accordingly; Based on the pytorch framework, the target detector model uses yolox-tiny. The input of the target detector model training is a dynamic size with a size range of 960-1280. After the training is completed, it is converted into onnx format for deployment. The output of the target detector model is the coordinate bounding box of the four positioning icons. Then, these four bounding boxes are rearranged in a clockwise manner starting from the upper left point. The lower right coordinate of the first bounding box, the lower left coordinate of the second bounding box, the upper left coordinate of the third bounding box, and the upper right coordinate of the fourth bounding box are taken in turn. The coordinates of the four points are taken out in turn to complete the positioning of the QR code image.

4. The ultra-large two-dimensional code positioning and recognition system in a wind power scenario according to claim 1 is characterized in that: The two-dimensional code correction module includes: After obtaining the coordinates of the four points in the QR code image by relying on the QR code positioning module, a perspective transformation is performed on it, the QR code image is corrected, and data enhancement is performed on the corrected QR code image. The data enhancement methods include Gaussian blur and grayscale image to highlight the black part of the QR code image.

5. The ultra-large two-dimensional code positioning and recognition system in a wind power scenario according to claim 1 is characterized in that: The two-dimensional code parsing and splicing module includes: Use the pyzbar tool library to identify the QR code. First, directly identify the analyzed and corrected QR code image. If it is not recognized, perform data enhancement. The data enhancement includes grayscale conversion, Gaussian blur noise reduction and binarization. Then, analyze the enhanced QR code image and decrypt the information using the AES decoding algorithm. According to the file name and segmentation number added in the QR code before, perform information splicing and information omission reminder.

6. The ultra-large two-dimensional code positioning and recognition system in a wind power scenario according to claim 1 is characterized in that: The remote server module comprises: The QR code positioning module, QR code correction module and QR code parsing and splicing module are deployed in the remote server module, leaving only one API interface. The backend only needs to call this API interface to call the entire decoding program, which is convenient for program updates and migration.

7. A positioning and identification method based on the ultra-large two-dimensional code positioning and identification system in a wind power scenario according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1. Input file data; S2, encode according to the file data, generate a QR code image and take a photo of it; S3, taking the QR code image obtained by taking the photo in step S2 as a training sample and marking it; S4, preprocessing the training samples by selecting a dynamic input size; S5. Input the training samples into the target detector model yolox-tiny based on the pytorch framework for training; S6, deploying the trained target detector model to the remote server module; S7. The target detector model in the remote server module parses the result of the QR code image and performs correction and splicing restoration operations.

8. The method for positioning and identifying a super-large two-dimensional code in a wind power scenario according to claim 7, characterized in that: The step S2 comprises: Use Python program to read the data in the original file, and judge the number of bytes in the data in units of lines. When the number of bytes reaches the preset maximum carrying threshold of the QR code, compress this part of information using Brotli algorithm, encrypt the data using AES symmetric encryption algorithm, and then perform base64 encoding and store it in a list. Finally, take two adjacent elements in the list to generate two QR codes, and then splice the two QR codes together horizontally, and add small positioning icons to the four corners of the picture to generate a QR code picture. The QR code picture includes the original file name, segmentation sequence number and number of segments, which are used for information splicing and information omission reminder during subsequent decoding. At the same time, the data processing and QR code generation module is based on a visual interface and multi-threaded operation, which is convenient for on-site personnel to operate.

9. The method for positioning and identifying a super-large two-dimensional code in a wind power scenario according to claim 7, characterized in that: The step S5 comprises: Based on the pytorch framework, the target detector model uses yolox-tiny. The input of the target detector model training is a dynamic size with a size range of 960-1280. After the training is completed, it is converted into onnx format for deployment. The output of the target detector model is the coordinate bounding box of the four positioning icons. Then, these four bounding boxes are rearranged in a clockwise manner starting from the upper left point. The lower right coordinate of the first bounding box, the lower left coordinate of the second bounding box, the upper left coordinate of the third bounding box, and the upper right coordinate of the fourth bounding box are taken in turn. The coordinates of the four points are taken out in turn to complete the positioning of the QR code image.

10. The method for positioning and identifying a super-large two-dimensional code in a wind power scenario according to claim 7, characterized in that: The step S7 comprises: Use the pyzbar tool library to identify the QR code. First, directly identify the analyzed and corrected QR code image. If it is not recognized, perform data enhancement. The data enhancement includes grayscale conversion, Gaussian blur noise reduction and binarization. Then, analyze the enhanced QR code image and decrypt the information using the AES decoding algorithm. According to the file name and segmentation number added in the QR code before, perform information splicing and information omission reminder.

Citation Information

Cited By

  • Two-dimensional code recognition unlocking method suitable for entrance machine and related equipment

    CN119251939A

  • Downloading method of traceability tamper-proof file, storage medium and equipment

    CN119814762A

  • A method for downloading a traceable tamper-proof file, a storage medium and an equipment

    CN119814762B