Real-time image processing system based on edge calculation

Abstract

The invention provides a real-time image processing system based on edge computing, and relates to the technical field of edge computing. Environment image data are acquired through an image data acquisition unit; performing fuzzy masking on the environment image data through an environment feature extraction unit to obtain defuzzified image data, and further determining an environment feature highlight degree corresponding to each edge computing node; environment data is acquired through a node processing distribution unit, the task priority of each edge computing node is determined based on the corresponding environment data and the corresponding environment feature prominence degree, topological distribution is performed on each edge computing node through all the task priorities and an edge computing network, and an image processing topological graph is obtained; carrying out adaptive compression transmission by using the image processing topological graph through an image compression transmission unit; according to the technical scheme, the image processing results of the edge computing nodes can be compressed and transmitted in combination with the priorities of the image processing tasks, so that the real-time efficiency of an image processing system is improved.

Description

Technical Field

[0001] The present application relates to the field of edge computing technology, and more specifically, to a real-time image processing system based on edge computing. Background Art

[0002] The real-time image processing system based on edge computing is a distributed computing architecture that aims to transfer image processing tasks from the cloud or centralized servers to edge computing nodes (such as smart cameras, edge servers, etc.) that are closer to the data source for processing. The core goal of this system is to improve real-time performance, reduce network bandwidth consumption, enhance data security, and provide more flexible and efficient computing capabilities. It is particularly suitable for delay-sensitive application scenarios such as smart transportation, video surveillance, industrial inspection, medical imaging, etc. By distributing computing resources to the edge of the network, the bottlenecks encountered during data transmission can be reduced, and the computing burden of centralized servers can be reduced, so that the system can respond and process high-frequency, large-scale image data more quickly.

[0003] In an edge computing environment, edge computing nodes communicate through the network. However, the volatility of network conditions, bandwidth limitations, and latency issues may affect the performance of the image processing system. When the network bandwidth is limited, the transmission of high-resolution images or videos may cause network congestion, which in turn causes increased latency and affects the real-time performance of the system. In addition, in an edge computing system, the computing power and network connection of each edge computing node are different. Task scheduling not only needs to consider the computing resources, network bandwidth, latency, etc. of the node, but also needs to combine multiple factors such as the priority and complexity of the image processing task. If the scheduling is unreasonable, some nodes may be overloaded while other nodes are idle, reducing the overall efficiency of the system. Therefore, how to compress and transmit the image processing results of the edge computing nodes in combination with the priority of the image processing task under the premise of ensuring the quality of the transmitted image to improve the real-time efficiency of the image processing system is a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a real-time image processing system based on edge computing, which can compress and transmit the image processing results of the edge computing nodes in combination with the priority of the image processing task while ensuring the quality of the transmitted image, so as to improve the real-time and high efficiency of the image processing system.

[0005] The present application provides a real-time image processing system based on edge computing, the processing system comprising:

[0006] An image data acquisition unit, used to collect environmental image data corresponding to each edge computing node in the edge computing network in real time;

[0007] An environmental feature extraction unit is used to perform blur masking on environmental image data corresponding to each edge computing node in the edge computing network to obtain deblurred image data, determine the environmental feature prominence corresponding to the edge computing node according to the deblurred image data, and then obtain the environmental feature prominence corresponding to each edge computing node;

[0008] A node processing distribution unit is used to obtain the environmental data corresponding to each edge computing node, determine the task priority of each edge computing node based on the corresponding environmental data and the corresponding environmental feature prominence, and perform topological distribution on each edge computing node through all task priorities and the edge computing network to obtain an image processing topology map of all edge computing nodes;

[0009] The image compression transmission unit is used to adaptively compress and transmit the image processing results of each edge computing node using the image processing topology map.

[0010] In this embodiment, the environmental image data corresponding to each edge computing node in the edge computing network is collected in real time through the corresponding edge monitoring device.

[0011] In this embodiment, the environment image data corresponding to the edge computing node is blurred and masked, and the deblurred image data specifically includes:

[0012] Gray-scale the environment image data corresponding to the edge computing node to obtain gray-scale environment image data;

[0013] For each environmental grayscale image in the grayscale environmental image data, blurring and masking the environmental grayscale image to obtain a deblurred environmental grayscale image;

[0014] The deblurred environment grayscale image is converted into a deblurred environment image, thereby obtaining deblurred image data.

[0015] In this embodiment, blurring and masking the environment grayscale image to obtain a deblurred environment grayscale image specifically includes:

[0016] Determine the pixel neighborhood of each pixel in the environmental grayscale image;

[0017] For each pixel point in the environmental grayscale image, deblurring is performed using a pixel neighborhood of the pixel point to obtain a deblurred pixel value of the pixel point, and further obtain a deblurred pixel value of each pixel point;

[0018] A deblurred environment grayscale image is determined based on the deblurred pixel values ​​of all pixels.

[0019] In this embodiment, deblurring is performed using the pixel neighborhood of a pixel point, and obtaining the pixel value of the pixel point after deblurring specifically includes:

[0020] Obtain all pixel points contained in the pixel neighborhood of the pixel point;

[0021] Determine a pixel variation coefficient corresponding to a pixel neighborhood of the pixel point;

[0022] Determine a pixel discrete coefficient corresponding to a pixel neighborhood of the pixel point;

[0023] Deblurring is performed according to all pixel points included in the pixel neighborhood, the pixel variation coefficient and the pixel discrete coefficient, thereby obtaining a pixel value of the pixel point after deblurring.

[0024] In this embodiment, determining the prominence of the environmental feature corresponding to the edge computing node according to the defuzzified image data specifically includes:

[0025] For each deblurred environment image in the deblurred image data, determining a first pixel extreme value and a second pixel extreme value of the deblurred environment image;

[0026] Determining the image prominence of the deblurred environment image by using the first pixel extreme value and the second pixel extreme value, thereby obtaining the image prominence of each deblurred environment image;

[0027] The environmental feature prominence corresponding to the edge computing node is determined according to all image prominences.

[0028] In this embodiment, the environmental data corresponding to each edge computing node is obtained through the sensor of the corresponding edge monitoring device.

[0029] In this embodiment, determining the task priority of each edge computing node based on the corresponding environmental data and the corresponding environmental feature prominence specifically includes:

[0030] For each edge computing node, extract all environmental features in the environmental data corresponding to the edge computing node;

[0031] The task priority of the edge computing node is determined by all environmental features and the prominence of the environmental features corresponding to the edge computing node, thereby obtaining the task priority of each edge computing node.

[0032] In this embodiment, the topological distribution of each edge computing node is performed through all task priorities and the edge computing network, and the image processing topological graph of all edge computing nodes is obtained, which specifically includes:

[0033] Extracting node connection data of all edge computing nodes in the edge computing network;

[0034] Converting the edge computing network into a node topology graph according to the node connection data;

[0035] Determine the node core coefficient of each edge computing node according to all task priorities and the node topology map;

[0036] The tasks of each edge computing node are distributed based on the core coefficients of all nodes, and then the image processing topology map of all edge computing nodes is obtained.

[0037] In this embodiment, using the image processing topology graph to adaptively compress and transmit the image processing results of each edge computing node specifically includes:

[0038] Determining the transmission performance of each transmission path in the image processing topology diagram;

[0039] For each edge computing node, selecting an image compression strategy for the image processing result of the edge computing node according to the task priority of the edge computing node and the transmission performance of the corresponding transmission path;

[0040] Then, the compressed image processing results of the edge computing nodes are transmitted in parallel.

[0041] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0042] The environmental image data corresponding to each edge computing node in the edge computing network is collected in real time by an image data collection unit; the environmental image data corresponding to the edge computing node is blurred and masked by an environmental feature extraction unit to obtain defuzzified image data, and the environmental feature prominence corresponding to the edge computing node is determined according to the defuzzified image data, thereby obtaining the environmental feature prominence corresponding to each edge computing node; the environmental data corresponding to each edge computing node is obtained by a node processing distribution unit, and the task priority of each edge computing node is determined based on the corresponding environmental data and the corresponding environmental feature prominence, and each edge computing node is topologically distributed through all task priorities and the edge computing network to obtain an image processing topology map of all edge computing nodes; the image processing results of each edge computing node are adaptively compressed and transmitted using the image processing topology map by an image compression and transmission unit.

[0043] It can be seen that in the present application, the image processing results of the edge computing nodes can be compressed and transmitted in combination with the priority of the image processing tasks while ensuring the quality of the transmitted image; wherein, the noise and unnecessary details in the image can be removed by fuzzy masking processing, so that the key features are more prominent, and the redundancy of the image data is reduced, which can help reduce the burden of storage and transmission, improve the processing efficiency of the image processing system, and determine the prominence of environmental features to quantify the information content of the monitoring environment corresponding to the edge computing node; then, the task priority is determined based on the environmental data and environmental feature prominence of the edge computing node, which can ensure that the key tasks are first assigned to the most suitable edge computing node, thereby improving the response speed and processing efficiency of the overall system; further, through topological distribution, the tasks can be effectively and evenly distributed among the edge computing nodes to avoid overloading of some edge computing nodes while other edge computing nodes are idle, which can reduce unnecessary data transmission, thereby reducing network bandwidth consumption and energy consumption, and through the construction of the image processing topology map, the image processing system can process and respond to environmental changes in a timely manner; finally, the image processing topology map is used to adaptively compress and transmit the image processing results of each edge computing node to improve the image data transmission speed.

[0044] To sum up, the technical solution adopted in this application can compress and transmit the image processing results of the edge computing node in combination with the priority of the image processing task while ensuring the quality of the transmitted image, so as to improve the real-time and high efficiency of the image processing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 It is a module structure diagram of a real-time image processing system based on edge computing provided by the present application;

[0047] Figure 2 is a schematic diagram of a process for determining deblurred image data provided by the present application;

[0048] Figure 3 It is a flowchart of determining task priority according to the application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0050] The embodiment of the present application provides a real-time image processing system based on edge computing, the core of which is to collect the environmental image data corresponding to each edge computing node in the edge computing network in real time through an image data acquisition unit; to blur and mask the environmental image data corresponding to the edge computing node through an environmental feature extraction unit to obtain defuzzified image data, to determine the environmental feature prominence corresponding to the edge computing node according to the defuzzified image data, and then to obtain the environmental feature prominence corresponding to each edge computing node; to obtain the environmental data corresponding to each edge computing node through a node processing distribution unit, to determine the task priority of each edge computing node based on the corresponding environmental data and the corresponding environmental feature prominence, to topologically distribute each edge computing node through all task priorities and the edge computing network, to obtain the image processing topology map of all edge computing nodes; to adaptively compress and transmit the image processing results of each edge computing node using the image processing topology map through an image compression transmission unit. The present application adopts the above technical solution to compress and transmit the image processing results of the edge computing node in combination with the priority of the image processing task under the premise of ensuring the quality of the transmitted image, so as to improve the real-time efficiency of the image processing system.

[0051] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown, the figure is a module structure diagram of a real-time image processing system based on edge computing according to the embodiment of the present application. The processing system includes: an image data acquisition unit 100, an environmental feature extraction unit 200, a node processing distribution unit 300, and an image compression transmission unit 400, which are described as follows:

[0052] The image data acquisition unit 100 is used to acquire environmental image data corresponding to each edge computing node in the edge computing network in real time.

[0053] In specific implementation, the environmental image data corresponding to each edge computing node in the edge computing network can be collected in real time through the corresponding edge monitoring equipment. Appropriate edge monitoring equipment needs to be deployed near each edge computing node. The edge monitoring equipment usually includes a smart camera, a sensor, and a microprocessor, which can collect environmental image data in real time. The environmental image data is the image data of the monitoring environment collected by the edge monitoring device corresponding to the edge computing node. In order to ensure the accuracy and real-time nature of the environmental image data, an effective data collection strategy needs to be formulated, which will not be repeated here.

[0054] The environmental feature extraction unit 200 is used to blur and mask the environmental image data corresponding to each edge computing node in the edge computing network to obtain defuzzified image data, determine the environmental feature prominence corresponding to the edge computing node based on the defuzzified image data, and then obtain the environmental feature prominence corresponding to each edge computing node.

[0055] Preferably, in this embodiment, the environment image data corresponding to the edge computing node is blurred and masked to obtain deblurred image data, referring to Figure 2 The figure is a schematic diagram of a process for determining deblurred image data in some embodiments of the present application. In this embodiment, determining the deblurred image data can be implemented by the following steps:

[0056] First, in step S21, the environment image data corresponding to the edge computing node is grayed to obtain grayed environment image data;

[0057] Then, in step S22, for each environmental grayscale image in the grayscale environmental image data, the environmental grayscale image is blurred and masked to obtain a deblurred environmental grayscale image;

[0058] Finally, in step S23, the deblurred environment grayscale image is converted into a deblurred environment image, thereby obtaining deblurred image data.

[0059] In the specific implementation, first, the environmental image data corresponding to the edge computing node can be grayscaled to obtain grayscale environmental image data. Grayscale is the process of converting a color image into a grayscale image. The purpose is to reduce the complexity of image processing while retaining important feature information. An appropriate image processing library (such as OpenCV) can be used to convert each environmental image in the environmental image data corresponding to the edge computing node into an environmental grayscale image, thereby obtaining grayscale environmental image data; then, for each environmental grayscale image in the grayscale environmental image data, the environmental grayscale image can be blurred and masked to obtain a deblurred environmental grayscale image; finally, the deblurred environmental grayscale image is converted into a deblurred environmental image to obtain deblurred image data. Pseudo-colorization technology can be applied. For example, color mapping can be used to map grayscale values ​​to color images, thereby saving the processed image as the final deblurred environmental image, and using data composed of all deblurred environmental images as deblurred image data.

[0060] In this embodiment, the following method can be used to perform blur masking on the environment grayscale image to obtain a deblurred environment grayscale image, namely:

[0061] Determine the pixel neighborhood of each pixel in the environmental grayscale image;

[0062] For each pixel point in the environmental grayscale image, deblurring is performed using a pixel neighborhood of the pixel point to obtain a deblurred pixel value of the pixel point, and further obtain a deblurred pixel value of each pixel point;

[0063] A deblurred environment grayscale image is determined based on the deblurred pixel values ​​of all pixels.

[0064] In the specific implementation, first, it is necessary to determine the length and width of the pixel neighborhood. The length and width of the pixel neighborhood determine the size of each pixel neighborhood. Usually, the pixel neighborhood is square or rectangular. The determined pixel neighborhood must ensure that the environmental grayscale image is just divided. Therefore, the length and width of the pixel neighborhood need to be adjusted according to the size of the environmental grayscale image. For each pixel point in the environmental grayscale image, the current pixel point is taken as the center point of the pixel neighborhood. According to the determined pixel neighborhood length and width, a subset of pixels around the center point is determined in the environmental grayscale image. This subset is the pixel neighborhood corresponding to the current pixel point; then, for each pixel point in the environmental grayscale image, the pixel neighborhood of the pixel point is used to deblur the pixel point to obtain the pixel value after deblurring the pixel point. The deblurred pixel value of each pixel point can be obtained by the above method; finally, the deblurred environmental grayscale image is determined based on the deblurred pixel values ​​of all pixels, that is, the deblurred pixel values ​​of all pixels are combined, so that the deblurred environmental grayscale image can be obtained.

[0065] In this embodiment, the pixel neighborhood of a pixel point is used for deblurring, and the pixel value of the pixel point after deblurring can be obtained in the following manner, namely:

[0066] Obtain all pixel points contained in the pixel neighborhood of the pixel point;

[0067] Determine a pixel variation coefficient corresponding to a pixel neighborhood of the pixel point;

[0068] Determine a pixel discrete coefficient corresponding to a pixel neighborhood of the pixel point;

[0069] Deblurring is performed according to all pixel points included in the pixel neighborhood, the pixel variation coefficient and the pixel discrete coefficient, thereby obtaining a pixel value of the pixel point after deblurring.

[0070] It should be noted that in the present application, the deblurred pixel value represents the pixel value obtained after the deblurring processing is performed on the pixel point; the pixel variation coefficient represents the degree of difference between all the pixel points in the pixel field. The smaller the pixel variation coefficient, the more similar all the pixel points in the pixel field are; the pixel discrete coefficient represents the degree of discreteness between all the pixel points in the pixel field. The larger the pixel discrete coefficient, the more obvious the difference between all the pixel points in the pixel field.

[0071] In specific implementation, first, all the pixels contained in the pixel neighborhood of the pixel point can be traversed; then, the Euclidean distance between all the pixels in the pixel neighborhood is determined, and then the standard deviation of the Euclidean distance between all the pixels is calculated, and the standard deviation is used as the pixel discrete coefficient corresponding to the pixel neighborhood of the pixel point; secondly, in the pixel neighborhood corresponding to the pixel point, the pixel mean of all the pixels is calculated, and the difference between the pixel value corresponding to each pixel point and the pixel mean is determined, that is, the pixel value corresponding to each pixel point is subtracted from the pixel mean, and the square of all the differences is determined and averaged, and then the square root is taken, and the result is used as the pixel variation coefficient corresponding to the pixel neighborhood of the pixel point; finally, deblurring can be performed based on all the pixels contained in the pixel neighborhood, the pixel variation coefficient and the pixel discrete coefficient, so as to obtain the pixel value of the pixel after deblurring. In actual implementation, the deblurred pixel value can be determined according to the following formula:

[0072]

[0073] Wherein, F(x, y) represents the deblurred pixel value corresponding to the pixel point with coordinates (x, y) in the environmental grayscale image, and f(x, y) represents the pixel value of the pixel point with coordinates (x, y). represents the pixel area corresponding to the pixel point with coordinates (x, y), f(i, j) represents the pixel value corresponding to the pixel point with coordinates (i, j), a represents the intermediate variable, Represents the pixel discrete coefficient corresponding to the pixel neighborhood, Represents the pixel variation coefficient corresponding to the pixel neighborhood.

[0074] In this embodiment, the following method may be used to determine the prominence of the environmental features corresponding to the edge computing node according to the defuzzified image data, namely:

[0075] For each deblurred environment image in the deblurred image data, determining a first pixel extreme value and a second pixel extreme value of the deblurred environment image;

[0076] Determining the image prominence of the deblurred environment image by using the first pixel extreme value and the second pixel extreme value, thereby obtaining the image prominence of each deblurred environment image;

[0077] The environmental feature prominence corresponding to the edge computing node is determined according to all image prominences.

[0078] It should be noted that in the present application, the first pixel extreme value represents the maximum pixel value in the deblurred environmental image; the second pixel extreme value represents the minimum pixel value in the deblurred environmental image; the image prominence represents the clarity of the image features in the deblurred environmental image; the environmental feature prominence represents the feature prominence of the environment monitored by the edge monitoring device corresponding to the edge computing node.

[0079] In a specific implementation, first, for each deblurred environment image in the deblurred image data, the deblurred environment image can be traversed, so that the maximum pixel value in the deblurred environment image is used as the first pixel extreme value, and the minimum pixel value in the deblurred environment image is used as the second pixel extreme value; then, the image prominence of the deblurred environment image can be determined by the first pixel extreme value and the second pixel extreme value. In actual implementation, the image prominence of the deblurred environment image can be determined by the following formula:

[0080]

[0081] Among them, CLE represents the image prominence of the defuzzified environment image, λ1 represents the first pixel extreme value of the defuzzified environment image, λ2 represents the second pixel extreme value of the defuzzified environment image, N represents the total number of rows of the defuzzified environment image, M represents the total number of columns of the defuzzified environment image, and F(i, j) represents the value of the pixel point in the i-th row and i-th column of the defuzzified environment image. The image prominence of each defuzzified environment image can be obtained in the above manner; finally, the prominence of the environmental feature corresponding to the edge computing node can be determined according to the prominence of all images, that is, the standard deviation of the prominence of all images can be used as the prominence of the environmental feature corresponding to the edge computing node.

[0082] In specific implementation, for each edge computing node in the edge computing network, the environmental feature prominence corresponding to each edge computing node can be obtained through the above method.

[0083] It should be noted that fuzzy masking processing can remove noise and unnecessary details in the image, making key features more prominent and reducing the redundancy of image data, which can help reduce the burden of storage and transmission, improve the processing efficiency of the image processing system, and determine the prominence of environmental features to quantify the information content of the monitoring environment corresponding to the edge computing node.

[0084] The node processing distribution unit 300 is used to obtain the environmental data corresponding to each edge computing node, determine the task priority of each edge computing node based on the corresponding environmental data and the corresponding environmental feature prominence, and topologically distribute each edge computing node through all task priorities and the edge computing network to obtain an image processing topology map of all edge computing nodes.

[0085] In specific implementation, the environmental data corresponding to each edge computing node can be obtained through the sensors of the corresponding edge monitoring device; it should be noted that in this application, the environmental data includes temperature data and light data of the monitored environment. In actual implementation, the temperature sensors and light sensors in the edge monitoring device can be used to obtain the environmental data corresponding to the edge computing node.

[0086] Preferably, in this embodiment, the task priority of each edge computing node is determined based on the corresponding environmental data and the corresponding environmental feature prominence, referring to Figure 3 As described above, the figure is a schematic diagram of the process of determining the priority of tasks in some embodiments of the present application. In this embodiment, the determination of the priority of tasks can be achieved by the following steps:

[0087] First, in step S31, for each edge computing node, all environmental features in the environmental data corresponding to the edge computing node are extracted;

[0088] Then, in step S32, the task priority of the edge computing node is determined by all environmental features and the prominence of the environmental features corresponding to the edge computing node, thereby obtaining the task priority of each edge computing node.

[0089] In the specific implementation, first, the environmental data collected by each edge computing node is preprocessed, including denoising, normalization, smoothing, etc., to improve the accuracy of feature extraction, so as to extract all relevant environmental features from the processed environmental data, that is, the temperature characteristics and light characteristics of the monitoring environment; then, the task priority of the edge computing node is determined by all environmental features and the environmental feature prominence corresponding to the edge computing node. In actual implementation, machine learning methods (such as weighted regression, decision trees, etc.) can be used to determine the function between environmental features and environmental feature prominence, so as to calculate the task priority of the edge computing node through the function. The task priority indicates the urgency of image processing at the edge computing node. The task priority of each edge computing node can be obtained in the above manner.

[0090] In this embodiment, the topological distribution of each edge computing node is performed through all task priorities and the edge computing network, and the image processing topological graph of all edge computing nodes is obtained in the following manner, namely:

[0091] Extracting node connection data of all edge computing nodes in the edge computing network;

[0092] Converting the edge computing network into a node topology graph according to the node connection data;

[0093] Determine the node core coefficient of each edge computing node according to all task priorities and the node topology map;

[0094] The tasks of each edge computing node are distributed based on the core coefficients of all nodes, and then the image processing topology map of all edge computing nodes is obtained.

[0095] In the specific implementation, first, the node connection data of all edge computing nodes can be extracted in the edge computing network, that is, the connection information between each edge computing node in the edge computing network is collected, including the ID of the edge computing node, the connection type (such as wired or wireless), the bandwidth and delay, which can be obtained through a network monitoring tool or a protocol analysis tool; then, the edge computing network is converted into a node topology graph according to the node connection data, each edge computing node can be regarded as a node in the graph, the connection between the nodes can be regarded as the edge in the graph, and the bandwidth and delay of the edge computing node can be regarded as the value of the edge, so that the node topology graph can be obtained; then, the node core coefficient of each edge computing node is determined according to all task priorities and the node topology graph. The node core coefficient is used to indicate the core importance of the edge computing node, which can be calculated based on the degree of connection of the node (i.e. The node core coefficient of each edge computing node is calculated based on the node core coefficient (i.e., the number of edges) and the task priority. The node core coefficient can be expressed as the node's degree of connection (i.e., the number of edges) multiplied by the task priority. Finally, tasks are distributed to each edge computing node based on all the node core coefficients, that is, each edge computing node is sorted according to the node core coefficient, so as to assign tasks to each edge computing node. For example, tasks are preferentially assigned to edge computing nodes with high node core coefficients to improve processing efficiency and response speed, and the current load of the edge computing node is considered to avoid overloading of a certain edge computing node as much as possible to ensure the stability of the image processing system, so as to generate the final image processing topology diagram according to the task distribution and the connection relationship between the edge computing nodes, which clearly shows the task allocation of each edge computing node and the direction of image data flow.

[0096] It should be noted that determining the task priority based on the environmental data and environmental feature prominence of the edge computing nodes can ensure that critical tasks are first assigned to the most suitable edge computing nodes, thereby improving the response speed and processing efficiency of the overall system. Through topological distribution, tasks can be effectively and evenly distributed among the edge computing nodes to avoid overloading of some edge computing nodes while others are idle, which can reduce unnecessary data transmission and thus reduce network bandwidth consumption and energy consumption. By constructing an image processing topology map, the image processing system can process and respond to environmental changes in a timely manner.

[0097] The image compression transmission unit 400 is used to adaptively compress and transmit the image processing results of each edge computing node using the image processing topology map.

[0098] In this embodiment, the image processing topology diagram is used to adaptively compress and transmit the image processing results of each edge computing node in the following manner, namely:

[0099] Determining the transmission performance of each transmission path in the image processing topology diagram;

[0100] For each edge computing node, selecting an image compression strategy for the image processing result of the edge computing node according to the task priority of the edge computing node and the transmission performance of the corresponding transmission path;

[0101] Then, the compressed image processing results of the edge computing nodes are transmitted in parallel.

[0102] In the specific implementation, first, the performance of each transmission path in the image processing topology diagram is evaluated. The bandwidth and delay of the transmission path can be obtained through the network monitoring tool, and then the comprehensive function of bandwidth and delay is constructed through the machine learning method, that is, different weights are assigned to the bandwidth and delay respectively, and the sum of the corresponding weight products is used as the transmission performance of the transmission path; then, for each edge computing node, according to the transmission performance and task priority of the edge computing node corresponding to the transmission path, select the appropriate image compression strategy. For example, for the image processing results of edge computing nodes with high task priority or high transmission performance, a lossless compression algorithm (such as PNG, GIF) can be used to ensure that the image quality is not damaged; finally, the compressed image processing results of the edge computing nodes are transmitted in parallel, that is, the compressed image processing results are sent in batches, and multiple transmission paths are used for parallel transmission to improve the data transmission speed, thereby completing the processing of real-time images by the edge monitoring device.

[0103] It can be seen that in the present application, the image processing results of the edge computing nodes can be compressed and transmitted in combination with the priority of the image processing tasks while ensuring the quality of the transmitted image; wherein, the noise and unnecessary details in the image can be removed by fuzzy masking processing, so that the key features are more prominent, and the redundancy of the image data is reduced, which can help reduce the burden of storage and transmission, improve the processing efficiency of the image processing system, and determine the prominence of environmental features to quantify the information content of the monitoring environment corresponding to the edge computing node; then, the task priority is determined based on the environmental data and environmental feature prominence of the edge computing node, which can ensure that the key tasks are first assigned to the most suitable edge computing node, thereby improving the response speed and processing efficiency of the overall system; further, through topological distribution, the tasks can be effectively and evenly distributed among the edge computing nodes to avoid overloading of some edge computing nodes while other edge computing nodes are idle, which can reduce unnecessary data transmission, thereby reducing network bandwidth consumption and energy consumption, and through the construction of the image processing topology map, the image processing system can process and respond to environmental changes in a timely manner; finally, the image processing topology map is used to adaptively compress and transmit the image processing results of each edge computing node to improve the image data transmission speed.

[0104] To sum up, the technical solution adopted in this application can compress and transmit the image processing results of the edge computing node in combination with the priority of the image processing task while ensuring the quality of the transmitted image, so as to improve the real-time and high efficiency of the image processing system.

[0105] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0106] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0107] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A real-time image processing system based on edge computing, characterized in that: The processing system comprises: An image data acquisition unit, used to collect environmental image data corresponding to each edge computing node in the edge computing network in real time; An environmental feature extraction unit is used to perform blur masking on environmental image data corresponding to each edge computing node in the edge computing network to obtain deblurred image data, determine the environmental feature prominence corresponding to the edge computing node according to the deblurred image data, and then obtain the environmental feature prominence corresponding to each edge computing node; A node processing distribution unit is used to obtain the environmental data corresponding to each edge computing node, determine the task priority of each edge computing node based on the corresponding environmental data and the corresponding environmental feature prominence, and perform topological distribution on each edge computing node through all task priorities and the edge computing network to obtain an image processing topology map of all edge computing nodes; The image compression transmission unit is used to adaptively compress and transmit the image processing results of each edge computing node using the image processing topology map.

2. A real-time image processing system based on edge computing as claimed in claim 1, characterized in that: The corresponding edge monitoring devices are used to collect the environmental image data corresponding to each edge computing node in the edge computing network in real time.

3. A real-time image processing system based on edge computing as claimed in claim 1, characterized in that: The environment image data corresponding to the edge computing node is blurred and masked, and the defuzzified image data specifically includes: Gray-scale the environment image data corresponding to the edge computing node to obtain gray-scale environment image data; For each environmental grayscale image in the grayscale environmental image data, blurring and masking the environmental grayscale image to obtain a deblurred environmental grayscale image; The deblurred environment grayscale image is converted into a deblurred environment image, thereby obtaining deblurred image data.

4. A real-time image processing system based on edge computing as claimed in claim 3, characterized in that: The environmental grayscale image is blurred and masked to obtain a deblurred environmental grayscale image, which specifically includes: Determine the pixel neighborhood of each pixel in the environmental grayscale image; For each pixel point in the environmental grayscale image, deblurring is performed using a pixel neighborhood of the pixel point to obtain a deblurred pixel value of the pixel point, and further obtain a deblurred pixel value of each pixel point; A deblurred environment grayscale image is determined based on the deblurred pixel values ​​of all pixels.

5. A real-time image processing system based on edge computing as claimed in claim 4, characterized in that: Deblurring using the pixel neighborhood of a pixel point to obtain the pixel value of the pixel point after deblurring specifically includes: Obtain all pixel points contained in the pixel neighborhood of the pixel point; Determine a pixel variation coefficient corresponding to a pixel neighborhood of the pixel point; Determine a pixel discrete coefficient corresponding to a pixel neighborhood of the pixel point; Deblurring is performed according to all pixel points included in the pixel neighborhood, the pixel variation coefficient and the pixel discrete coefficient, thereby obtaining a pixel value of the pixel point after deblurring.

6. A real-time image processing system based on edge computing as claimed in claim 1, characterized in that: Determining the prominence of the environmental features corresponding to the edge computing node according to the defuzzified image data specifically includes: For each deblurred environment image in the deblurred image data, determining a first pixel extreme value and a second pixel extreme value of the deblurred environment image; Determining the image prominence of the deblurred environment image by using the first pixel extreme value and the second pixel extreme value, thereby obtaining the image prominence of each deblurred environment image; The environmental feature prominence corresponding to the edge computing node is determined according to all image prominences.

7. A real-time image processing system based on edge computing as claimed in claim 1, characterized in that: The environmental data corresponding to each edge computing node is obtained through the sensors of the corresponding edge monitoring devices.

8. The real-time image processing system based on edge computing according to claim 1, characterized in that: The task priority of each edge computing node is determined based on the corresponding environmental data and the prominence of the corresponding environmental features, including: For each edge computing node, extract all environmental features in the environmental data corresponding to the edge computing node; The task priority of the edge computing node is determined by all environmental features and the prominence of the environmental features corresponding to the edge computing node, thereby obtaining the task priority of each edge computing node.

9. The real-time image processing system based on edge computing according to claim 1, characterized in that: The topological distribution of each edge computing node is performed through all task priorities and the edge computing network, and the image processing topology diagram of all edge computing nodes is obtained, which specifically includes: Extracting node connection data of all edge computing nodes in the edge computing network; Converting the edge computing network into a node topology graph according to the node connection data; Determine the node core coefficient of each edge computing node according to all task priorities and the node topology map; The tasks of each edge computing node are distributed based on the core coefficients of all nodes, and then the image processing topology map of all edge computing nodes is obtained.

10. A real-time image processing system based on edge computing as claimed in claim 1, characterized in that: Adaptively compressing and transmitting the image processing results of each edge computing node using the image processing topology graph specifically includes: Determining the transmission performance of each transmission path in the image processing topology diagram; For each edge computing node, selecting an image compression strategy for the image processing result of the edge computing node according to the task priority of the edge computing node and the transmission performance of the corresponding transmission path; Then, the compressed image processing results of the edge computing nodes are transmitted in parallel.

Images