Edge computing cloud-edge collaboration method, system and device

Abstract

The present invention relates to the field of cloud-edge collaboration technology, and in particular to an edge computing cloud-edge collaboration method, system and device. The present invention obtains real-time monitoring videos shot by each AI camera in a single target area, and divides them into valid video segments and invalid video segments. The valid video segments are divided into explicit clear frame groups and implicit clear frame groups according to the clarity, and the method for obtaining target information is determined according to the clear frame group category. The edge filtering module receives the target information of the explicit clear frame group and the clear enhanced frame group, determines all the target information of a single valid video segment according to the time relationship, and filters it to generate valid target information, and determines the traffic information of each time period according to the valid target information. The present invention reduces the dependence on cloud resources through cloud-edge collaboration and decentralization of the cloud, and realizes the optimal allocation of resources.

Description

Technical Field

[0001] The present invention relates to the field of cloud-edge collaboration technology, and in particular to an edge computing cloud-edge collaboration method, system and device. Background Art

[0002] Cloud-edge collaboration technology is a new computing model developed based on the integration of cloud computing and edge computing. It uses the central control capability of cloud computing and the local computing capability of edge computing to achieve efficient data processing and transmission through reasonable task allocation and resource scheduling. In the cloud-edge collaboration system, cloud computing, as the central control plane, is responsible for global management and scheduling; edge computing, as the computing platform, is responsible for local computing and storage; and terminal devices, as perception and execution units, are responsible for receiving instructions and performing corresponding operations.

[0003] Publication No. CN115766719A discloses an interaction method and device between an edge computing background system and a cloud-edge collaborative system, which can be applied to the field of cloud-edge collaborative technology. The invention establishes a connection relationship between the edge computing background system and the cloud-edge collaborative system through a TCP long link, thereby utilizing the characteristics of the TCP long link to improve the communication efficiency of the edge computing process, and performs data interaction between the edge computing background system and the cloud-edge collaborative system through the Json data format, thereby shortening the data length of the transmission process, and then updates the edge node information in the database based on the comparison results of all online edge node information and the edge node information in the database, and binds the user terminal, online edge node and application when the user terminal performs edge computing services, and performs operation authentication through the binding relationship when the edge computing background system interacts with the cloud-edge collaborative system, thereby effectively improving the security of edge communications. It can be seen that the invention has the following problems:

[0004] The invention does not consider how to reduce dependence on cloud resources in the target area through cloud-edge collaboration and decentralization of the cloud, so as to improve the understanding of the traffic health conditions in the target area by improving the accuracy of license plate recognition. Summary of the invention

[0005] To this end, the present invention provides an edge computing cloud-edge collaboration method, system and device to overcome the problem that the prior art does not consider reducing dependence on cloud resources in the target area through cloud-edge collaboration and decentralization of the cloud, so as to improve the understanding of the traffic health conditions in the target area by improving the accuracy of license plate recognition.

[0006] To achieve the above objectives, on the one hand, the present invention provides an edge computing cloud-edge collaboration method, comprising:

[0007] Obtain real-time surveillance video captured by each AI camera in a single target area;

[0008] According to the traffic light information in the monitoring direction of a single AI camera, the real-time monitoring video captured by it is divided into valid video segments and invalid video segments;

[0009] Determine the clarity of each frame of the effective video segment, divide the effective video segment into explicit clear frame groups and implicit clear frame groups according to the clarity, and determine a method for obtaining target information according to the clear frame group category, including:

[0010] Extracting target information of each explicit clear frame in a single explicit clear frame group;

[0011] Or, the implicit clear frame group is sent to a cloud computing module, each implicit clear frame of a single implicit clear frame group is processed into a clear enhancement frame to obtain a clear enhancement frame group, and target information of each clear enhancement frame in the single clear enhancement frame group is extracted;

[0012] receiving target information of explicit clear frame groups and clear enhanced frame groups;

[0013] Determine all target information of explicit clear frame groups and clear enhanced frame groups in a single effective video segment according to the temporal relationship;

[0014] Filter all target information of a single valid video segment to generate valid target information;

[0015] The flow information of each time period is determined according to the effective target information of each effective video segment.

[0016] Furthermore, according to the signal light information of the monitoring direction of a single AI camera, the real-time monitoring video captured by it is divided into valid video segments and invalid video segments, including:

[0017] If the monitoring direction of a single AI camera is green light or yellow light, the real-time monitoring video corresponding to the duration of the green light or yellow light is the valid video segment;

[0018] If the monitoring direction of a single AI camera is a red light, the real-time monitoring video corresponding to the red light duration is an invalid video segment.

[0019] Further, the method of dividing the effective video segment into explicit clear frame groups and implicit clear frame groups according to the definition includes:

[0020] Determining a clear frame type of each frame according to the clarity of each frame in the effective video segment, wherein the clear frame type includes an explicit clear frame and an implicit clear frame;

[0021] dividing explicit clear frames into the explicit clear frame groups;

[0022] The implicit clear frames are grouped into the implicit clear frame groups.

[0023] Further, the clear frame type of each frame is determined according to the clarity of each frame in the valid video segment, including:

[0024] If the single frame definition is greater than or equal to the preset definition, the frame is an explicit definition frame;

[0025] If the definition of a single frame is less than the preset definition, the frame is a hidden clear frame.

[0026] Furthermore, target information is extracted from a single explicit clear frame and a single clear enhanced frame according to optical character recognition technology, and the target information includes license plate information.

[0027] Furthermore, the method of processing a single implicit clear frame into a clear enhanced frame includes:

[0028] Processing the implicit clear frame into a first clear enhanced frame according to a deblurring algorithm;

[0029] Processing the implicit clear frame into a second clear enhanced frame according to a super-resolution reconstruction algorithm;

[0030] The first clarity enhanced frame and the second clarity enhanced frame are fused into a clarity enhanced image through an image fusion technology.

[0031] Further, the method of filtering all target information of a single valid video segment to generate valid target information includes:

[0032] Determine duplicate target information and separate target information in all target information;

[0033] Filtering the repeated target information to generate filtered target information;

[0034] Determining effective target information according to the filtered target information and the separate target information;

[0035] The amount of the valid target information is the sum of the amount of the filtered target information and the amount of the separate target information.

[0036] Further, the flow information of each time period is determined according to the effective target information of each effective video segment;

[0037] Among them, the amount of valid target information in each time period is positively correlated with the flow information, and the flow information includes vehicle flow information.

[0038] The present invention also provides an edge computing cloud-edge collaboration system, comprising:

[0039] AI cameras to obtain real-time surveillance video of the target area;

[0040] An edge computing module, which is connected to the AI ​​camera and includes a plurality of edge computing units, wherein a single edge computing unit is used to divide the corresponding real-time monitoring video into a plurality of valid video segments and a plurality of invalid video segments, determine the clarity of any frame in a single valid video segment and divide the valid video segment into an explicit clear frame group and an implicit clear frame group according to the clarity, extract target information of each explicit clear frame in a single explicit clear frame group, and send the implicit clear frame group to the cloud computing module;

[0041] A cloud computing module connected to the edge computing module, configured to receive the implicit clear frame group, and to process each implicit clear frame of a single implicit clear frame group into a clear enhancement frame to obtain a clear enhancement frame group, and to extract target information of each clear enhancement frame in the single clear enhancement frame group;

[0042] The edge computing module and the cloud computing module both include optical character recognition technology to extract target information of explicit clear frames and clear enhanced frames;

[0043] an edge filtering module, which is connected to the edge computing module and the cloud computing module respectively, and is used to receive target information of explicit clear frame groups and clear enhanced frame groups, determine all target information of explicit clear frame groups and clear enhanced frame groups in a single valid video segment according to a temporal relationship, and filter all target information of a single valid video segment to generate valid target information;

[0044] A communication module, comprising a collaborative communication unit connecting the edge computing module and the cloud computing module, an edge communication unit connecting the edge computing module and the edge filtering module, and a cloud-edge communication unit connecting the cloud computing module and the edge filtering module;

[0045] The amount of valid target information in a single valid video segment is less than or equal to the total target information in the valid video segment.

[0046] The present invention also provides an edge computing cloud-edge collaboration device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, an edge computing cloud-edge collaboration method is implemented.

[0047] Compared with the prior art, the beneficial effect of the present invention is that the edge computing cloud-edge collaboration method provided by the present invention realizes efficient processing of the real-time monitoring video of the AI ​​camera in the target area by combining the advantages of the edge computing module and the cloud computing module. The edge computing module can instantly distinguish between valid and invalid video segments, and intelligently classify and process them according to the video clarity, which reduces the cloud computing pressure and ensures the rapid preliminary extraction of target information; the cloud computing module focuses on further clarifying the implicit clear frames, improving the accuracy of license plate recognition; the introduction of the edge filtering module not only integrates the target information from the edge and the cloud, but also ensures the integrity and accuracy of the information based on the time series relationship. The effective target information finally filtered and generated provides a reliable basis for accurately counting the traffic flow in each time period, reflecting the efficiency and intelligence level of cloud-centralized processing under cloud-edge collaboration.

[0048] Furthermore, the real-time monitoring video captured by the AI ​​camera is intelligently divided into valid video segments and invalid video segments according to the traffic light information. This method significantly improves the video processing efficiency: in a complex traffic environment, by accurately identifying the green or yellow light period (i.e., the vehicle driving period) as a valid video segment, and the red light period (i.e., the vehicle stationary period) as an invalid video segment, it can effectively reduce unnecessary video data processing and avoid invalid analysis of images without vehicle traffic, thereby saving computing resources and accelerating the extraction process of important data such as target information. At the same time, it also improves the real-time response capability and intelligence level of the traffic monitoring system.

[0049] Furthermore, by carefully dividing the frames of the effective video segment into explicit clear frames and implicit clear frames, and classifying them into different frame groups accordingly, efficient classification and processing of video frames is achieved. Among them, explicit clear frames directly meet the high-definition standards, which is convenient for quickly and accurately extracting target information; while implicit clear frames are slightly lower than the preset clarity, they still have recognition potential; by setting the preset clarity as the minimum recognition threshold of the optical character recognition technology in the edge computing module and the cloud computing module, the scientificity and practicality of the classification are ensured, which not only optimizes resource allocation, but also improves the comprehensiveness and accuracy of target information extraction, and enhances the intelligent processing capability and efficiency of the traffic monitoring system.

[0050] Furthermore, a single-frame implicit clear frame is processed into a first clear enhanced frame and a second clear enhanced frame through a deblurring algorithm and a super-resolution reconstruction algorithm respectively, and then the two are merged into a clear enhanced image using image fusion technology. This method can significantly improve the image quality: this comprehensive processing strategy not only effectively reduces image blur, but also enhances the detail resolution of the image, making key information such as license plates more clear and discernible, thereby greatly improving the accuracy and reliability of license plate recognition, and providing a more accurate image foundation for traffic monitoring and data analysis.

[0051] Furthermore, by identifying and filtering repeated target information in a single valid video segment, only unique license plate records are retained, and at the same time, combined with individual target information, a concise and accurate set of valid target information is effectively generated; this filtering mechanism not only significantly reduces redundant data and reduces storage and processing costs, but also ensures the uniqueness and accuracy of target information, providing a more reliable data foundation for subsequent applications such as traffic flow analysis and vehicle tracking.

[0052] Furthermore, the traffic information of each time period is determined based on the effective target information of each effective video segment. This method not only realizes the accurate monitoring of traffic flow, but also can effectively evaluate the health of traffic conditions. By setting a preset number as the judgment standard for traffic congestion, when the number of effective target information is lower than the standard, the system can quickly issue a warning signal, and upgrade it to an alarm signal when congestion occurs in multiple consecutive time periods, and timely dispatch traffic police to conduct on-site command. This mechanism not only improves the intelligent level of traffic management, but also can take intervention measures at the early stage of traffic congestion, effectively alleviate traffic pressure, improve road traffic efficiency, and ensure traffic safety and smoothness.

[0053] Furthermore, the edge computing cloud-edge collaborative system of the present invention realizes efficient processing of real-time surveillance video of the target area through the close collaboration of AI cameras, edge computing modules, cloud computing modules, edge filtering modules and communication modules. This system makes full use of the rapid response capability of edge computing and the powerful processing capability of cloud computing to quickly extract the target information of explicit clear frames at the edge, and upload implicit clear frames to the cloud for clear processing and extraction of target information. This cloud-edge collaborative and decentralized processing method not only improves the accuracy and efficiency of license plate recognition, but also reduces dependence on cloud resources and realizes optimal resource allocation; at the same time, the filtering and processing of target information by the edge filtering module further ensures the accuracy and uniqueness of the target information, providing reliable data support for traffic management and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a step diagram of the edge computing cloud-edge collaboration method according to an embodiment of the present invention;

[0055] Figure 2 A diagram showing the steps of a method for processing a single implicit clear frame into a clear enhanced frame according to an embodiment of the present invention;

[0056] Figure 3 A step diagram of a method for filtering all target information of a single valid video segment to generate valid target information according to an embodiment of the present invention;

[0057] Figure 4 This is a connection diagram of the edge computing cloud-edge collaborative system of an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0059] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0060] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0061] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0062] See also Figure 1 As shown, it is a step diagram of the edge computing cloud-edge collaboration method of an embodiment of the present invention. An embodiment of the present invention provides an edge computing cloud-edge collaboration method, including:

[0063] Step S1, obtaining real-time monitoring videos captured by each AI camera in a single target area; it can be understood that the target area, taking an intersection as an example, refers to a standard intersection and a non-standard intersection (which may be a three-way intersection, a cross-shaped intersection, an X-shaped intersection, etc.);

[0064] Step S2: The edge computing module divides the real-time monitoring video captured by a single AI camera into valid video segments and invalid video segments according to the signal light information in the monitoring direction of the camera.

[0065] Step S3, the edge computing module determines the clarity of each frame of the effective video segment, divides the effective video segment into an explicit clear frame group and an implicit clear frame group according to the clarity, and determines a method for obtaining target information according to the clear frame group category, including:

[0066] Step S31, extracting target information of each explicit clear frame in a single explicit clear frame group, packaging the target information and sending it to an edge filtering module, wherein the target information includes license plate information;

[0067] Or, step S32, sending the implicit clear frame group to a cloud computing module, the cloud computing module performs clear processing on each implicit clear frame of a single implicit clear frame group into a clear enhancement frame to obtain a clear enhancement frame group, extracts target information of each clear enhancement frame in the single clear enhancement frame group, and packages the target information and sends it to an edge filtering module;

[0068] Step S4, the edge filtering module receives target information of the explicit clear frame group and the clear enhanced frame group;

[0069] Step S5, the edge filtering module determines all target information of the explicit clear frame group and the clear enhancement frame group in a single valid video segment according to the temporal relationship; it can be understood that the first explicit clear frame and the last explicit clear frame in the explicit clear frame group after being arranged according to the temporal sequence are both in the single valid video segment, and the first implicit clear frame and the last implicit clear frame in the corresponding implicit clear frame group are also both in the single valid video segment; in implementation, the edge filtering module first determines the explicit clear frame group and the clear enhancement frame group of the single valid video segment, and then determines all target information of the valid video segment according to the determined two frame groups;

[0070] Step S6, filtering all target information of a single valid video segment to generate valid target information;

[0071] Step S7, determining the flow information of each time period according to the effective target information of each effective video segment.

[0072] It can be understood that the edge computing cloud-edge collaboration method provided by the present invention realizes efficient processing of the real-time monitoring video of the AI ​​camera in the target area by combining the advantages of the edge computing module and the cloud computing module. The edge computing module can instantly distinguish between valid and invalid video segments, and intelligently classify and process them according to the video clarity, which reduces the cloud computing pressure and ensures the rapid preliminary extraction of target information; the cloud computing module focuses on further clarifying the implicit clear frames, improving the accuracy of license plate recognition; the introduction of the edge filtering module not only integrates the target information from the edge and the cloud, but also ensures the integrity and accuracy of the information based on the time series relationship. The effective target information finally filtered and generated provides a reliable basis for accurately counting the traffic flow in each time period, reflecting the efficiency and intelligence level of cloud-centralized processing under cloud-edge collaboration.

[0073] Specifically, the real-time monitoring video captured by a single AI camera is divided into valid video segments and invalid video segments according to the signal light information in the monitoring direction, including:

[0074] If the monitoring direction of a single AI camera is green light or yellow light, the real-time monitoring video corresponding to the duration of the green light or yellow light is the valid video segment;

[0075] If the monitoring direction of a single AI camera is a red light, the real-time monitoring video corresponding to the red light duration is an invalid video segment.

[0076] During implementation, an X-shaped intersection has an AI camera for monitoring vehicle information traveling from the south to the northwest. The real-time monitoring video recorded by the AI ​​camera determines whether it is a valid video segment or an invalid video segment based on the color of the traffic light controlling that direction: when the traffic light is green or yellow, vehicles in that direction are allowed to travel and pass through the target area, and only then will vehicles pass through, so it is a valid video segment; when the traffic light is red, vehicles in that direction are not allowed to travel, and no vehicles pass through, so it is an invalid video segment.

[0077] It is understandable that the method of intelligently distinguishing the real-time monitoring video taken by the AI ​​camera into valid video segments and invalid video segments based on the traffic light information significantly improves the video processing efficiency: in a complex traffic environment, by accurately identifying the green or yellow light period (i.e., the vehicle driving period) as a valid video segment, and the red light period (i.e., the vehicle stationary period) as an invalid video segment, it can effectively reduce unnecessary video data processing and avoid invalid analysis of images without vehicle traffic, thereby saving computing resources and accelerating the extraction process of important data such as target information. At the same time, it also improves the real-time response capability and intelligence level of the traffic monitoring system.

[0078] It is understandable that the real-time surveillance video is always composed of a valid video segment, an invalid video segment, a valid video segment, and an invalid video segment. A single valid video segment can be divided into an explicit clear frame group and a hidden clear frame group. A single explicit clear frame group includes several explicit clear frames, and a single hidden clear frame group includes several hidden clear frames.

[0079] In implementation, a single valid video segment is composed of several frames, among which the frames with a clarity greater than or equal to a preset clarity are explicit clear frames, and the frames with a clarity less than a preset clarity are implicit clear frames. The explicit clear frames and the implicit clear frames should be staggered to form a single valid video segment; therefore, the explicit clear frames within a single explicit clear frame group should be discontinuous, and the implicit clear frames within a single implicit clear frame group should also be discontinuous.

[0080] Specifically, in step S3, the method of dividing the effective video segment into explicit clear frame groups and implicit clear frame groups according to the definition includes:

[0081] In step S301, the clear frame type of each frame is determined according to the clarity of each frame in the valid video segment, and the clear frame type includes an explicit clear frame and an implicit clear frame;

[0082] In step S302, the explicit clear frames are divided into the explicit clear frame groups;

[0083] In step S303, the implicit clear frames are grouped into the implicit clear frame groups.

[0084] Specifically, in step S301, the clear frame type of each frame is determined according to the clarity of each frame in the valid video segment, including:

[0085] If the single frame definition is greater than or equal to the preset definition, the frame is an explicit definition frame;

[0086] If the definition of a single frame is less than the preset definition, the frame is a hidden clear frame.

[0087] In implementation, the preset clarity is the minimum clarity that can be recognized by the optical character recognition technology included in the edge computing module and the cloud computing module.

[0088] It can be understood that this method achieves efficient classification and processing of video frames by carefully dividing the frames of the effective video segment into explicit clear frames and implicit clear frames, and classifying them into different frame groups accordingly. Among them, explicit clear frames directly meet the high-definition standard, which is convenient for quickly and accurately extracting target information; while implicit clear frames are slightly lower than the preset clarity, but still have recognition potential; by setting the preset clarity as the minimum recognition threshold of the optical character recognition technology in the edge computing module and the cloud computing module, the scientificity and practicality of the classification are ensured, which not only optimizes resource allocation, but also improves the comprehensiveness and accuracy of target information extraction, and enhances the intelligent processing capability and efficiency of the traffic monitoring system.

[0089] Specifically, in step S5, target information is extracted from a single explicit clear frame and a single clear enhanced frame according to optical character recognition technology, and the target information includes license plate information.

[0090] It is understandable that the optical character recognition technology is updated regularly (the update frequency is generally once every half a month to a month, preferably once every half a month), so the preset clarity is updated along with the update of the optical character recognition technology.

[0091] It is understandable that the use of optical character recognition technology to extract target information in a single explicit clear frame and clear enhanced frame, combined with the regular update frequency of the technology, dynamically adjusts the preset clarity threshold, which significantly improves the accuracy and adaptability of license plate recognition. With the continuous advancement of optical character recognition technology, the timely update of the preset clarity ensures that the system can maintain efficient target information extraction capabilities even in the face of complex and changing traffic monitoring environments, thereby enhancing the intelligence level and response speed of the traffic monitoring system, and providing more reliable data support for traffic management and analysis.

[0092] See also Figure 2 As shown, it is a step diagram of a method for processing a single implicit clear frame into a clear enhanced frame according to an embodiment of the present invention. Specifically, in step S32, the method for processing a single implicit clear frame into a clear enhanced frame includes:

[0093] Step S321, processing the implicit clear frame into a first clear enhanced frame according to a deblurring algorithm;

[0094] Step S322, processing the implicit clear frame into a second clear enhanced frame according to a super-resolution reconstruction algorithm;

[0095] Step S323: fusing the first clarity enhanced frame and the second clarity enhanced frame into a clarity enhanced image through an image fusion technology.

[0096] It can be understood that the deblurring algorithms include Wiener filtering that uses the statistical characteristics of the signal for deblurring and blind deblurring that restores the image without knowing the blur kernel; the super-resolution reconstruction algorithms include interpolation methods that increase image resolution to enhance details (such as bilinear interpolation, cubic interpolation) and deep learning methods that use convolutional neural networks (CNN) for super-resolution processing of images.

[0097] It can be understood that image fusion technology includes: (1) weighted averaging method, which directly performs weighted averaging of two images to obtain a fused image; selecting appropriate weights according to their respective clarity and detail levels (the weight of the first clear enhanced frame is the ratio of the clarity of the first clear enhanced frame to the clarity of the second clear enhanced frame, and the weight of the second clear enhanced frame is the ratio of the clarity of the second clear enhanced frame to the clarity of the first clear enhanced frame); (2) Laplacian pyramid fusion, which constructs a Laplacian pyramid for the two images, and in each pyramid layer, selects a fusion strategy according to clarity and detail (for example, selects an image with more details), and reconstructs the fused pyramid into the final image; (3) multi-scale fusion, which performs wavelet transform on the two images to obtain detail information of different frequencies, selects appropriate coefficients in each wavelet coefficient layer for fusion, and uses inverse wavelet transform to reconstruct the fusion result into an image; (4) deep learning fusion method, which trains a convolutional neural network, with two images as input and a fused image as output, and optimizes the fusion effect by learning features.

[0098] It can be understood that the method of processing a single-frame implicit clear frame into a first clear enhanced frame and a second clear enhanced frame through a deblurring algorithm and a super-resolution reconstruction algorithm respectively, and then fusing the two into a clear enhanced image using image fusion technology can significantly improve the image quality: this comprehensive processing strategy not only effectively reduces image blur, but also enhances the detail resolution of the image, making key information such as license plates more clear and discernible, thereby greatly improving the accuracy and reliability of license plate recognition, and providing a more accurate image basis for traffic monitoring and data analysis.

[0099] See also Figure 3 As shown, it is a step diagram of a method for filtering all target information of a single valid video segment to generate valid target information according to an embodiment of the present invention. Specifically, in step S5, the method for filtering all target information of a single valid video segment to generate valid target information includes:

[0100] Step S51, determining repeated target information and single target information in all target information;

[0101] Step S52, filtering the repeated target information to generate filtered target information; it can be understood that the repeated target information includes a plurality of identical target information, and only one of the target information remains after filtering; in one implementation, the repeated target information includes n1 a target information, n2 b target information, n3 c target information, ... ni i target information; the filtered filtered target information includes 1 a target information, 1 b target information, 1 c target information, ... 1 i target information;

[0102] Step S53, determining valid target information according to the filtered target information and the separate target information;

[0103] The amount of the valid target information is the sum of the amount of the filtered target information and the amount of the separate target information.

[0104] It is understandable that when a single vehicle passes by the target area, the target information should appear in more than one video frame, so there is a lot of repeated target information, so the number of valid target information should be less than the total target information.

[0105] It can be understood that this method effectively generates a concise and accurate set of valid target information by identifying and filtering duplicate target information in a single valid video segment, retaining only unique license plate records, and combining individual target information. This filtering mechanism not only significantly reduces redundant data and reduces storage and processing costs, but also ensures the uniqueness and accuracy of target information, providing a more reliable data foundation for subsequent applications such as traffic flow analysis and vehicle tracking.

[0106] In implementation, any filtering algorithm in the prior art may be used to filter duplicate target information, including a non-maximum suppression (NMS) algorithm, a dictionary filtering algorithm, a hash algorithm, and a K-Means clustering algorithm.

[0107] Specifically, in step S7, the flow information of each time period is determined according to the effective target information of each effective video segment;

[0108] Among them, the amount of valid target information in each time period is positively correlated with the flow information, and the flow information includes vehicle flow information.

[0109] It is understandable that a greater number of valid target information in a single valid video segment indicates a greater traffic volume in the time period corresponding to the valid video segment, that is, a healthier (less congested) traffic condition in the time period.

[0110] During implementation, if the number of valid target information in the time period corresponding to a certain valid video segment is less than the preset number, it is determined that traffic congestion has occurred in the time period, and a warning signal needs to be issued. If the time periods corresponding to multiple consecutive (generally 3) valid video segments are determined to have traffic congestion, an alarm signal needs to be issued and the nearest traffic police will be dispatched to the congested section to direct traffic.

[0111] It can be understood that the preset number should be determined based on the number of lanes in that direction of the intersection and the green light duration. The preset number of single-row lanes corresponding to a green light duration of 30 seconds is generally 15 to 20 vehicles, and the preset number of double-row lanes corresponding to a green light duration of 30 seconds is twice that of a single-row lane. The longer the green light time, the more the preset number should be.

[0112] It is understandable that the method of determining the flow information of each time period based on the effective target information of each effective video segment not only realizes the accurate monitoring of traffic flow, but also can effectively evaluate the health of traffic conditions. By setting a preset number as the judgment standard for traffic congestion, when the number of effective target information is lower than the standard, the system can quickly issue a warning signal, and upgrade it to an alarm signal when congestion occurs for multiple consecutive time periods, and promptly dispatch traffic police to conduct on-site command. This mechanism not only improves the level of intelligence of traffic management, but also can take intervention measures at the early stage of traffic congestion, effectively alleviate traffic pressure, improve road traffic efficiency, and ensure traffic safety and smoothness.

[0113] See also Figure 4 As shown, it is a connection diagram of the edge computing cloud-edge collaboration system of an embodiment of the present invention. The embodiment of the present invention also provides an edge computing cloud-edge collaboration system, including:

[0114] AI cameras are used to obtain real-time surveillance video of the target area. It is understandable that an AI camera is equipped with at least one edge computing module and one edge filtering module.

[0115] An edge computing module, which is connected to the AI ​​camera and includes a plurality of edge computing units, wherein a single edge computing unit is used to divide the corresponding real-time monitoring video into a plurality of valid video segments and a plurality of invalid video segments, determine the clarity of any frame in a single valid video segment and divide the valid video segment into an explicit clear frame group and an implicit clear frame group according to the clarity, extract target information of each explicit clear frame in the single explicit clear frame group, and send the implicit clear frame group to the cloud computing module;

[0116] A cloud computing module connected to the edge computing module, for receiving the implicit clear frame group, and processing each implicit clear frame of a single implicit clear frame group into a clear enhanced frame to obtain a clear enhanced frame group, and extracting target information of each clear enhanced frame in the single clear enhanced frame group; it can be understood that the cloud computing module stores a deblurring algorithm, a super-resolution reconstruction algorithm and an image fusion technology, for processing a single implicit clear frame into a clear enhanced frame;

[0117] The edge computing module and the cloud computing module both include optical character recognition technology to extract target information of explicit clear frames and clear enhanced frames;

[0118] an edge filtering module, which is connected to the edge computing module and the cloud computing module respectively, and is used to receive target information of explicit clear frame groups and clear enhanced frame groups, determine all target information of explicit clear frame groups and clear enhanced frame groups in a single valid video segment according to a temporal relationship, and filter all target information of a single valid video segment to generate valid target information;

[0119] A communication module, comprising a collaborative communication unit connecting the edge computing module and the cloud computing module, an edge communication unit connecting the edge computing module and the edge filtering module, and a cloud-edge communication unit connecting the cloud computing module and the edge filtering module;

[0120] The amount of valid target information in a single valid video segment is less than or equal to the total target information in the valid video segment.

[0121] It can be understood that the edge computing cloud-edge collaborative system of the present invention realizes efficient processing of real-time surveillance video of the target area through the close collaboration of AI cameras, edge computing modules, cloud computing modules, edge filtering modules and communication modules. This system makes full use of the rapid response capability of edge computing and the powerful processing capability of cloud computing to quickly extract the target information of explicit clear frames at the edge, and upload implicit clear frames to the cloud for clear processing and extraction of target information. This cloud-edge collaborative and decentralized processing method not only improves the accuracy and efficiency of license plate recognition, but also reduces dependence on cloud resources and realizes optimal resource allocation; at the same time, the filtering and processing of target information by the edge filtering module further ensures the accuracy and uniqueness of the target information, providing reliable data support for traffic management and analysis.

[0122] An embodiment of the present invention also provides an edge computing cloud-edge collaboration device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the edge computing cloud-edge collaboration method of the present invention is implemented.

[0123] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0124] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An edge computing cloud-edge collaboration method, characterized in that: include: Obtain real-time surveillance video captured by each AI camera in a single target area; According to the traffic light information in the monitoring direction of a single AI camera, the real-time monitoring video captured by it is divided into valid video segments and invalid video segments; Determine the clarity of each frame of the effective video segment, divide the effective video segment into explicit clear frame groups and implicit clear frame groups according to the clarity, and determine a method for obtaining target information according to the clear frame group category, including: Extracting target information of each explicit clear frame in a single explicit clear frame group; Or, the implicit clear frame group is sent to a cloud computing module, each implicit clear frame of a single implicit clear frame group is processed into a clear enhancement frame to obtain a clear enhancement frame group, and target information of each clear enhancement frame in the single clear enhancement frame group is extracted; receiving target information of explicit clear frame groups and clear enhanced frame groups; Determine all target information of explicit clear frame groups and clear enhanced frame groups in a single effective video segment according to the temporal relationship; Filter all target information of a single valid video segment to generate valid target information; The flow information of each time period is determined according to the effective target information of each effective video segment.

2. The edge computing cloud-edge collaboration method according to claim 1, characterized in that: According to the signal light information of the monitoring direction of a single AI camera, the real-time monitoring video captured by it is divided into valid video segments and invalid video segments, including: If the monitoring direction of a single AI camera is green light or yellow light, the real-time monitoring video corresponding to the duration of the green light or yellow light is the valid video segment; If the monitoring direction of a single AI camera is a red light, the real-time monitoring video corresponding to the red light duration is an invalid video segment.

3. The edge computing cloud-edge collaboration method according to claim 2 is characterized in that: The method of dividing the effective video segment into explicit clear frame groups and implicit clear frame groups according to the definition includes: Determining a clear frame type of each frame according to the clarity of each frame in the effective video segment, wherein the clear frame type includes an explicit clear frame and an implicit clear frame; dividing explicit clear frames into the explicit clear frame groups; The implicit clear frames are grouped into the implicit clear frame groups.

4. The edge computing cloud-edge collaboration method according to claim 3 is characterized in that: Determine the clear frame type of each frame according to the clarity of each frame in the valid video segment. include, If the single frame definition is greater than or equal to the preset definition, the frame is an explicit definition frame; If the definition of a single frame is less than the preset definition, the frame is a hidden clear frame.

5. The edge computing cloud-edge collaboration method according to claim 1, characterized in that: Target information is extracted from a single explicit clear frame and a single clear enhanced frame according to optical character recognition technology, and the target information includes license plate information.

6. The edge computing cloud-edge collaboration method according to claim 5, characterized in that: The method for processing a single implicit clear frame into a clear enhanced frame includes: Processing the implicit clear frame into a first clear enhanced frame according to a deblurring algorithm; Processing the implicit clear frame into a second clear enhanced frame according to a super-resolution reconstruction algorithm; The first clarity enhanced frame and the second clarity enhanced frame are fused into a clarity enhanced image through an image fusion technology.

7. The edge computing cloud-edge collaboration method according to claim 1, characterized in that: The method of filtering all target information of a single valid video segment to generate valid target information includes: Determine duplicate target information and separate target information in all target information; Filtering the repeated target information to generate filtered target information; Determining effective target information according to the filtered target information and the separate target information; The amount of the valid target information is the sum of the amount of the filtered target information and the amount of the separate target information.

8. The edge computing cloud-edge collaboration method according to claim 1, characterized in that: Determine the flow information of each time period according to the effective target information of each effective video segment; Among them, the amount of valid target information in each time period is positively correlated with the flow information, and the flow information includes vehicle flow information.

9. An edge computing cloud-edge collaboration system using the edge computing cloud-edge collaboration method according to any one of claims 1 to 8, characterized in that: include, AI cameras to obtain real-time surveillance video of the target area; An edge computing module, which is connected to the AI ​​camera and includes a plurality of edge computing units, wherein a single edge computing unit is used to divide the corresponding real-time monitoring video into a plurality of valid video segments and a plurality of invalid video segments, determine the clarity of any frame in a single valid video segment and divide the valid video segment into an explicit clear frame group and an implicit clear frame group according to the clarity, extract target information of each explicit clear frame in a single explicit clear frame group, and send the implicit clear frame group to the cloud computing module; A cloud computing module connected to the edge computing module, configured to receive the implicit clear frame group, and to process each implicit clear frame of a single implicit clear frame group into a clear enhancement frame to obtain a clear enhancement frame group, and to extract target information of each clear enhancement frame in the single clear enhancement frame group; The edge computing module and the cloud computing module both include optical character recognition technology to extract target information of explicit clear frames and clear enhanced frames; an edge filtering module, which is connected to the edge computing module and the cloud computing module respectively, and is used to receive target information of explicit clear frame groups and clear enhanced frame groups, determine all target information of explicit clear frame groups and clear enhanced frame groups in a single valid video segment according to a temporal relationship, and filter all target information of a single valid video segment to generate valid target information; A communication module, comprising a collaborative communication unit connecting the edge computing module and the cloud computing module, an edge communication unit connecting the edge computing module and the edge filtering module, and a cloud-edge communication unit connecting the cloud computing module and the edge filtering module; The amount of valid target information in a single valid video segment is less than or equal to the total target information in the valid video segment.

10. An edge computing cloud-edge collaboration device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the edge computing cloud-edge collaboration method as described in any one of claims 1-8 is implemented.

Images