Edge gateway system for AI visual identification processing based on large model scene application

Abstract

The invention discloses an edge gateway system for AI visual identification processing based on large model scene application, which comprises a data acquisition module, a communication conversion module, an identification processing module and a model application optimization module, and relates to the technical field of artificial intelligence. According to the edge gateway system for AI visual identification processing based on large model scene application, data is preprocessed, original features are restored through visual identification, local feature analysis and extraction are carried out based on the original features, a result after identification operation is obtained and matched with data in a database, and finally target features are determined. Therefore, the AI visual identification processing is utilized to restore the abnormity, the analysis and restoration of the oblique features are utilized, and the matching processing of the restored features is realized, so that the accuracy and the identification efficiency in the data processing are effectively improved, and the equipment with the edge gateway system can better realize the application of a large model scene.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and specifically to an edge gateway system for AI visual recognition processing based on large model scene applications. Background Art

[0002] With the development of artificial intelligence technology, large models have been widely used in many fields. At the same time, AI visual recognition technology has become increasingly mature, playing an important role in security, intelligent transportation, industrial inspection and other scenarios.

[0003] The reference patent name is: Edge gateway system, data processing method of edge gateway system (patent publication number: CN118972433A, patent publication date: 2024-11-15). The system includes at least one edge gateway device, and the edge gateway device includes a multi-protocol access unit, an edge computing processing unit and an external interface unit; the external interface unit is used to connect to the terminal device, and the multi-protocol access unit is used to connect to the cloud device; the edge computing processing unit includes a central processing unit and an auxiliary computing processor for auxiliary computing of the original business data, and the auxiliary computing processor is connected to the central processing unit; the central processing unit is connected to the external interface unit to communicate with the terminal device through the external interface unit; the central processing unit is connected to the multi-protocol access unit to communicate with the cloud device through the multi-protocol access unit. The system helps to reduce the cloud load, and helps to improve the data processing efficiency of the edge gateway system and improve the business processing capability of the edge gateway device.

[0004] Based on the description in the above-mentioned document, in the existing edge gateway system at the toll road exit, when performing recognition and monitoring of the license plate of the driving vehicle, the recognized characters have directional deviations, which easily leads to the problem that the data after recognition and processing does not match the actual situation, or the processed data is still inaccurate, so that the final judgment result produces errors, and even the problem of failure to recognize the required data occurs. For this reason, the present invention provides an edge gateway system based on AI visual recognition processing for large model scene applications. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an edge gateway system based on AI visual recognition processing for large model scenario applications, which solves the problem that in the existing edge gateway system at the toll road exit, when performing recognition and monitoring of the driving vehicle license plate, the recognized characters have directional deviations, which easily leads to the problem that the recognized and processed data are inconsistent with the actual situation, or the processed data is still inaccurate, resulting in errors in the final judgment result, or even the problem of failure to recognize the required data.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an edge gateway system for AI visual recognition processing based on large model scene applications, comprising: The data acquisition module is equipped with sensors and cameras to realize image acquisition operations of physical objects in the scene; The communication conversion module performs format conversion and encoding processing on the collected data and establishes a transmission channel with the edge gateway device; The recognition processing module pre-processes the data, and then restores the original features through visual recognition, performs local feature analysis and extraction based on the original features, obtains the matching results after the recognition operation with the data in the database, and finally determines the target features; The model application optimization module summarizes the data of each recognition process and introduces it into the set model for training and optimization, so as to realize the practical application of the trained model.

[0007] Preferably, the operation steps of the communication conversion module are: A1. Identify the communication protocol type of the collected data; A2. According to the communication protocol requirements of the edge gateway device, the collected data is converted into a format and encoded; A3. Verify whether the converted data meets the protocol standard of the target system, and continue the conversion process until the protocol standard is the same, and then establish a transmission channel with the edge gateway device.

[0008] Preferably, the identification processing module performs preprocessing operations on the data as follows: B1. After receiving the collected data, extract the image data according to the set frame period, and then screen the image data; B2. The license plate is located by extracting features from the image, and the license plate features are determined by the ratio of the square license plate in the historical data. If the license plate features are not determined in all images, the production direction adjustment instruction is transmitted to the adjustment unit to implement the camera adjustment operation, and the storage operation with the license plate features is completed; B3, and summarize the retained license plate feature image data and the parameter data currently extracted and collected to form an initial data set marked as P.

[0009] Preferably, the operation of restoring the original features by visual recognition by the recognition processing module is: C1. Extracting image data with license plate features, and performing a magnification operation on the license plate features in the image data; C2, and based on the edge limit of the license plate feature, the vertical and horizontal distance change values ​​are calculated, and after marking the character features in the feature, the character features are restored according to the change value to form image data with the license plate feature; C3. Perform local analysis and extraction operations on character features to confirm the actual situation of the features.

[0010] Preferably, the operation of enlarging the license plate feature in C1 is: c11, determining the horizontal boundary and the vertical boundary of the license plate feature in the image data, and the sizes of the horizontal boundary and the vertical boundary are L1 and L2 respectively; c12, determining the sizes of the horizontal boundary and the vertical boundary of the image data and marking them as H1 and H2, and determining the magnification ratio by calculating the ratio of the corresponding horizontal boundary and the vertical boundary; c13. Compare the ratio of H1 / L1 with the ratio of H2 / L2, select the smaller ratio as the magnification ratio, and transmit the magnification ratio to the adjustment unit to implement the adjustment operation of the camera.

[0011] Preferably, the adjustment group includes a direction adjustment group and a focal length adjustment group, and the direction adjustment group implements adjustment operations in the left and right or up and down directions after receiving the adjustment instruction until an image with license plate features appears, and the focal length adjustment group implements adjustment of the focal length after receiving the adjustment instruction to complete the magnification operation of the license plate features.

[0012] Preferably, the operation of restoring character features in C2 is: c21. Determine that the actual lateral distance of the license plate feature is L3, the actual vertical distance of the license plate feature is L4, the lateral distance of the license plate feature in the image data is H3, the vertical distance of the license plate feature in the image data is H4, and determine the lateral angle deviation α and the vertical angle deviation β; c22, then grayscale the license plate features, and then distinguish the character features according to the grayscale value to mark the character features; c23. Perform restoration operation on each character feature, determine the size of the character feature in the image data, and calculate the actual character feature restoration value based on the angle deviation to form image data that is directly opposite to the license plate feature.

[0013] Preferably, the angle deviation calculation operation in c21 is: Lateral angular deviation: Sinα=H3 / L3; Vertical angular deviation: Cosβ=H4 / L4.

[0014] Preferably, the operation of calculating the actual character feature restoration value according to the angle deviation in c23 is: D1. Select a character feature and determine the upper, lower, left and right extreme points of the current character feature. Connect the upper and lower extreme points and the left and right extreme points, and the intersection of the two lines is the center point of the character feature. D2. Establish an X / Y axis cross coordinate system with the center point as the origin, and set multiple lines at equal distances from the top to the bottom extreme point, and mark the intersection of the line and the boundary of the character feature as an expansion node. Based on the distance from the expansion node to the Y axis, expand according to the horizontal angle deviation value, and restore the expanded nodes according to the character feature. The distance from the expansion node to the Y axis is set to M1, and the corresponding actual horizontal distance is N1, and the calculation formula of the actual horizontal distance N1 is: N1=M1 / (H3 / L3); D3, then based on the connected character features in D2, multiple lines are set at equal distances from the left to the right extreme points, and the intersection of the lines and the boundaries of the character features is marked as an expansion node, and the expansion is performed according to the vertical angle deviation value based on the distance from the expansion node to the X-axis, and the expanded nodes are restored and connected according to the character features, so as to obtain the character features of the original size after restoration; The distance from the expansion node to the X-axis is set to M2, and the corresponding actual lateral distance is N2, and the calculation formula of the actual lateral distance N2 is: N2=M2 / (H4 / L4).

[0015] Preferably, the steps of performing local analysis and extraction of character features in C3 are: c31. After extracting the restored character features, extract various local features on the character features; c32. extracting local features corresponding to similar characters from the historical database, and matching and comparing the local features of the historical database with the local features of the character features; c33. Determine the result of the corresponding abnormal character feature based on the matching situation.

[0016] The present invention provides an edge gateway system for AI visual recognition processing based on large model scene applications. Compared with the prior art, it has the following beneficial effects: 1. The edge gateway system based on AI visual recognition processing for large model scene applications pre-processes the data, then restores the original features through visual recognition, performs local feature analysis and extraction based on the original features, and obtains the matching results after the recognition operation with the data in the database, and finally determines the target features, thereby using AI visual recognition processing to restore the analysis and restoration of abnormal and oblique features, and realizes the matching processing of restored features, thereby effectively improving the accuracy and recognition efficiency in data processing, so that devices with edge gateway systems can better realize the application of large model scenes.

[0017] 2. The edge gateway system of AI visual recognition processing based on large model scene application extracts image data with license plate features, determines the horizontal and vertical boundaries of the image data and the actual license plate features, determines the magnification ratio by calculating the ratio of the corresponding horizontal and vertical boundaries, and selects a smaller ratio to amplify the license plate features in the image data, thereby maintaining the pixels while amplifying the collection of license plate features, improving the accuracy of data collection, and associating with subsequent processing algorithms to ensure the efficiency of data processing.

[0018] 3. The edge gateway system of AI visual recognition processing based on large model scene applications determines the size of the character features in the image data by restoring each character feature, and calculates the actual character feature restoration value based on the angle deviation to form image data facing the license plate feature. In this way, adaptive adjustments can be made according to the conditions of different vehicles to achieve the restoration of the character features facing the image data, thereby facilitating better subsequent matching and comparison of local features and ensuring the accuracy of recognition and processing operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a principle block diagram of the edge gateway system of the present invention; Figure 2 is an operation flow chart of the communication conversion module of the present invention; Figure 3 This is a flow chart of the pre-processing operation of the identification processing module of the present invention; Figure 4 This is a flow chart of the visual recognition and feature restoration operation of the recognition processing module of the present invention. DETAILED DESCRIPTION

[0020] 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.

[0021] See also Figure 1-Figure 4 , the present invention provides two technical solutions: Embodiment 1: An edge gateway system for AI visual recognition processing based on large model scene applications, comprising: The data acquisition module is equipped with sensors and cameras to realize image acquisition operations of physical objects in the scene; The communication conversion module performs format conversion and encoding processing on the collected data and establishes a transmission channel with the edge gateway device; The recognition processing module pre-processes the data, and then restores the original features through visual recognition, performs local feature analysis and extraction based on the original features, obtains the matching results after the recognition operation with the data in the database, and finally determines the target features; The model application optimization module summarizes the data of each recognition process and introduces it into the set model for training and optimization, so as to realize the practical application of the trained model.

[0022] Among them, the data is preprocessed, and then the original features are restored through visual recognition. Local feature analysis and extraction are performed based on the original features, and the results after the recognition operation are matched with the data in the database, and finally the target features are determined. The AI ​​visual recognition processing is used to restore the analysis and restoration of abnormal and oblique features, and the matching processing of the restored features is realized, so as to effectively improve the accuracy and recognition efficiency in data processing, so that devices with edge gateway systems can better realize the application of large model scenarios.

[0023] The edge gateway for physical object collection in the scene contains a camera and an audible and visual alarm. When the system fails to complete the recognition and processing operation, the audible and visual alarm can be used to inform personnel to handle it, thereby reducing the impact time after the failure occurs.

[0024] In the embodiment of the present invention, the operation steps of the communication conversion module are: A1. Identify the communication protocol type of the collected data; A2. According to the communication protocol requirements of the edge gateway device, the collected data is converted into a format and encoded; A3. Verify whether the converted data meets the protocol standard of the target system, and continue the conversion process until the protocol standard is the same, and then establish a transmission channel with the edge gateway device.

[0025] In the embodiment of the present invention, the identification processing module performs preprocessing operations on the data as follows: B1. After receiving the collected data, extract the image data according to the set frame period, and then screen the image data; B2. The license plate is located by extracting features from the image, and the license plate features are determined by the ratio of the square license plate in the historical data. If the license plate features are not determined in all images, the production direction adjustment instruction is transmitted to the adjustment unit to implement the camera adjustment operation, and the storage operation with the license plate features is completed; B3, and summarize the retained license plate feature image data and the parameter data currently extracted and collected to form an initial data set marked as P.

[0026] In the embodiment of the present invention, the operation of restoring the original features by the recognition processing module through visual recognition is: C1. Extracting image data with license plate features, and performing a magnification operation on the license plate features in the image data; C2, and based on the edge limit of the license plate feature, the vertical and horizontal distance change values ​​are calculated, and after marking the character features in the feature, the character features are restored according to the change value to form image data with the license plate feature; C3. Perform local analysis and extraction operations on character features to confirm the actual situation of the features.

[0027] In the embodiment of the present invention, the operation of enlarging the license plate feature in C1 is: c11, determining the horizontal boundary and the vertical boundary of the license plate feature in the image data, and the sizes of the horizontal boundary and the vertical boundary are L1 and L2 respectively; c12, determining the sizes of the horizontal boundary and the vertical boundary of the image data and marking them as H1 and H2, and determining the magnification ratio by calculating the ratio of the corresponding horizontal boundary and the vertical boundary; c13. Compare the ratio of H1 / L1 with the ratio of H2 / L2, select the smaller ratio as the magnification ratio, and transmit the magnification ratio to the adjustment unit to implement the adjustment operation of the camera.

[0028] Among them, by extracting image data with license plate features, determining the horizontal and vertical boundaries of the image data and the actual license plate features, determining the magnification ratio by calculating the ratio of the corresponding horizontal and vertical boundaries, and selecting a smaller ratio to perform a magnification operation on the license plate features in the image data, thereby maintaining the pixels while enlarging the acquisition of the license plate features, improving the accuracy of data acquisition, and linking with subsequent processing algorithms to ensure the efficiency of data processing.

[0029] In an embodiment of the present invention, the adjustment group includes a direction adjustment group and a focal length adjustment group, and the direction adjustment group implements left-right or up-down adjustment operations after receiving an adjustment instruction until an image with license plate features appears, and the focal length adjustment group implements focal length adjustment after receiving an adjustment instruction to complete the magnification operation of the license plate features.

[0030] The direction adjustment group and focal length adjustment group are mature existing devices, which are used to realize the adjustment and control operations after being combined with the edge gateway system. Any device that can complete the direction adjustment and focal length adjustment can be used, and this article will not elaborate on them in detail.

[0031] In the embodiment of the present invention, the operation of restoring the character features in C2 is: c21. Determine that the actual lateral distance of the license plate feature is L3, the actual vertical distance of the license plate feature is L4, the lateral distance of the license plate feature in the image data is H3, the vertical distance of the license plate feature in the image data is H4, and determine the lateral angle deviation α and the vertical angle deviation β; c22, then grayscale the license plate features, and then distinguish the character features according to the grayscale value to mark the character features; c23. Perform restoration operation on each character feature, determine the size of the character feature in the image data, and calculate the actual character feature restoration value based on the angle deviation to form image data that is directly opposite to the license plate feature.

[0032] Among them, after grayscale conversion, the grayscale values ​​corresponding to different features are also different. Multiple detection points are set equidistantly on the image data, and the position of the character feature is determined according to the change of the grayscale value of the detection point. Then, the first value change detection point located in the outer circle of the character feature is extracted, and connected accordingly in sequence to form the outer boundary of the character feature.

[0033] In the embodiment of the present invention, the angle deviation calculation operation in c21 is: Lateral angular deviation: Sinα=H3 / L3; Vertical angular deviation: Cosβ=H4 / L4.

[0034] In the embodiment of the present invention, the operation of calculating the actual character feature restoration value according to the angle deviation in c23 is: D1. Select a character feature and determine the upper, lower, left and right extreme points of the current character feature. Connect the upper and lower extreme points and the left and right extreme points, and the intersection of the two lines is the center point of the character feature. D2. Establish an X / Y axis cross coordinate system with the center point as the origin, and set multiple lines at equal distances from the top to the bottom extreme point, and mark the intersection of the line and the boundary of the character feature as an expansion node. Based on the distance from the expansion node to the Y axis, expand according to the horizontal angle deviation value, and restore the expanded nodes according to the character feature. The distance from the expansion node to the Y axis is set to M1, and the corresponding actual horizontal distance is N1, and the calculation formula of the actual horizontal distance N1 is: N1=M1 / (H3 / L3); D3, then based on the connected character features in D2, multiple lines are set at equal distances from the left to the right extreme points, and the intersection of the lines and the boundaries of the character features is marked as an expansion node, and the expansion is performed according to the vertical angle deviation value based on the distance from the expansion node to the X-axis, and the expanded nodes are restored and connected according to the character features, so as to obtain the character features of the original size after restoration; The distance from the expansion node to the X-axis is set to M2, and the corresponding actual lateral distance is N2, and the calculation formula of the actual lateral distance N2 is: N2=M2 / (H4 / L4).

[0035] Among them, by restoring each character feature, the size of the character feature in the image data is determined, and the actual character feature restoration value is calculated based on the angle deviation to form image data facing the license plate feature. In this way, adaptive adjustments are made according to the conditions of different vehicles to achieve restoration of the character feature facing the image data, thereby facilitating better subsequent matching and comparison of local features and ensuring the accuracy of recognition and processing operations.

[0036] In the embodiment of the present invention, the steps of performing local analysis and extraction of character features in C3 are as follows: c31. After extracting the restored character features, extract various local features on the character features; c32. extracting local features corresponding to similar characters from the historical database, and matching and comparing the local features of the historical database with the local features of the character features; c33. Determine the result of the corresponding abnormal character feature based on the matching situation.

[0037] Embodiment 2 is different from Embodiment 1 in that, in the local analysis operation of character features, for example, when distinguishing and identifying the character features "2" and "Z", through the recognition of local features, the upper half of the character feature "2" is in the shape of an arc, while the upper half of the character feature "Z" is in the shape of a combination of a straight line and a corner, so the actual character feature is confirmed after matching and comparing the local features of the corresponding restored character features as the distinguishing features; When distinguishing and identifying the character features "D" and "O", the left half of the character feature "D" is vertical through the recognition of local features, while the left half of the character feature "O" is arc-shaped. Therefore, the actual character features are confirmed after matching and comparing the local features with the corresponding restored character features as the distinguishing features.

[0038] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0039] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0040] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An edge gateway system for AI visual recognition processing based on large model scene applications, characterized by: include: The data acquisition module is equipped with sensors and cameras to realize image acquisition operations of physical objects in the scene; The communication conversion module performs format conversion and encoding processing on the collected data and establishes a transmission channel with the edge gateway device; The recognition processing module pre-processes the data, and then restores the original features through visual recognition, performs local feature analysis and extraction based on the original features, obtains the matching results after the recognition operation with the data in the database, and finally determines the target features; The model application optimization module summarizes the data of each recognition process and introduces it into the set model for training and optimization, so as to realize the practical application of the trained model.

2. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 1 is characterized in that: The operation steps of the communication conversion module are: A1. Identify the communication protocol type of the collected data; A2. According to the communication protocol requirements of the edge gateway device, the collected data is converted into a format and encoded; A3. Verify whether the converted data meets the protocol standard of the target system, and continue the conversion process until the protocol standard is the same, and then establish a transmission channel with the edge gateway device.

3. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 1 is characterized in that: The identification processing module performs preprocessing operations on the data as follows: B1. After receiving the collected data, extract the image data according to the set frame period, and then screen the image data; B2. The license plate is located by extracting features from the image, and the license plate features are determined by the ratio of the square license plate in the historical data. If the license plate features are not determined in all images, the production direction adjustment instruction is transmitted to the adjustment unit to implement the camera adjustment operation, and the storage operation with the license plate features is completed; B3, and summarize the retained license plate feature image data and the parameter data currently extracted and collected to form an initial data set marked as P.

4. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 3 is characterized in that: The operation of restoring the original features by visual recognition in the recognition processing module is as follows: C1. Extracting image data with license plate features, and performing a magnification operation on the license plate features in the image data; C2, and based on the edge limit of the license plate feature, the vertical and horizontal distance change values ​​are calculated, and after marking the character features in the feature, the character features are restored according to the change value to form image data with the license plate feature; C3. Perform local analysis and extraction operations on character features to confirm the actual situation of the features.

5. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 4 is characterized in that: The operation of enlarging the license plate feature in C1 is: c11, determining the horizontal boundary and the vertical boundary of the license plate feature in the image data, and the sizes of the horizontal boundary and the vertical boundary are L1 and L2 respectively; c12, determining the sizes of the horizontal boundary and the vertical boundary of the image data and marking them as H1 and H2, and determining the magnification ratio by calculating the ratio of the corresponding horizontal boundary and the vertical boundary; c13. Compare the ratio of H1 / L1 with the ratio of H2 / L2, select the smaller ratio as the magnification ratio, and transmit the magnification ratio to the adjustment unit to implement the adjustment operation of the camera.

6. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 5 is characterized in that: The adjustment group includes a direction adjustment group and a focal length adjustment group, and the direction adjustment group implements left-right or up-down adjustment operations after receiving an adjustment instruction until an image with license plate features appears, and the focal length adjustment group implements focal length adjustment to complete the magnification operation of the license plate features after receiving an adjustment instruction.

7. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 4 is characterized in that: The operation of restoring character features in C2 is: c21. Determine that the actual lateral distance of the license plate feature is L3, the actual vertical distance of the license plate feature is L4, the lateral distance of the license plate feature in the image data is H3, the vertical distance of the license plate feature in the image data is H4, and determine the lateral angle deviation α and the vertical angle deviation β; c22, then grayscale the license plate features, and then distinguish the character features according to the grayscale value to mark the character features; c23. Perform restoration operation on each character feature, determine the size of the character feature in the image data, and calculate the actual character feature restoration value based on the angle deviation to form image data that is directly opposite to the license plate feature.

8. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 7 is characterized in that: The angle deviation calculation operation in c21 is: Lateral angular deviation: Sinα=H3 / L3; Vertical angular deviation: Cosβ=H4 / L4.

9. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 8 is characterized in that: The operation of calculating the actual character feature restoration value based on the angle deviation in c23 is: D1. Select a character feature and determine the upper, lower, left and right extreme points of the current character feature. Connect the upper and lower extreme points and the left and right extreme points, and the intersection of the two lines is the center point of the character feature. D2. Establish an X / Y axis cross coordinate system with the center point as the origin, and set multiple lines at equal distances from the top to the bottom extreme point, and mark the intersection of the line and the boundary of the character feature as an expansion node. Based on the distance from the expansion node to the Y axis, expand according to the horizontal angle deviation value, and restore the expanded nodes according to the character feature. The distance from the expansion node to the Y axis is set to M1, and the corresponding actual horizontal distance is N1, and the calculation formula of the actual horizontal distance N1 is: N1=M1 / (H3 / L3); D3, then based on the connected character features in D2, multiple lines are set at equal distances from the left to the right extreme points, and the intersection of the lines and the boundaries of the character features is marked as an expansion node, and the expansion is performed according to the vertical angle deviation value based on the distance from the expansion node to the X-axis, and the expanded nodes are restored and connected according to the character features, so as to obtain the character features of the original size after restoration; The distance from the expansion node to the X-axis is set to M2, and the corresponding actual lateral distance is N2, and the calculation formula of the actual lateral distance N2 is: N2=M2 / (H4 / L4).

10. The edge gateway system for AI visual recognition processing based on large model scene application according to claim 4 is characterized in that: The steps of performing local analysis and extraction of character features in C3 are as follows: c31. After extracting the restored character features, extract various local features on the character features; c32. extracting local features corresponding to similar characters from the historical database, and matching and comparing the local features of the historical database with the local features of the character features; c33. Determine the result of the corresponding abnormal character feature based on the matching situation.

Images