Industrial field automatic monitoring method and system based on cloud edge cooperative detection

Abstract

The invention provides an industrial field automatic monitoring method and system based on cloud edge collaborative detection, and the method comprises the steps: obtaining video data through an edge calculation terminal, and carrying out the discrete image processing, so as to obtain first to-be-detected data; constructing an edge detection model, performing multi-scale static feature mining on the first to-be-detected data, and performing suspicious person detection; after a suspicious person is detected, sending the video data to a centralized control cloud computing center, and carrying out continuous image processing to obtain second to-be-detected data; a cloud detection model is constructed, space-time associated dynamic feature mining is performed on the second to-be-detected data, and abnormal behavior pattern detection is performed; and after an abnormal behavior is detected, storing the video data and generating visual alarm information. Aiming at the technical bottleneck of video monitoring of the industrial control system, the problem that the video traffic continuously occupies the bandwidth of a communication network is solved by utilizing a data processing mode of collaboration of the edge terminal and the cloud platform, and meanwhile, monitoring and diagnosis of abnormal behaviors such as illegal operation and deliberate destruction in an industrial production field can be automatically realized; and security assurance of physical facilities and information networks of the industrial control system is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of industrial control technology, and in particular to an industrial site automation monitoring method and system based on cloud-edge collaborative detection. Background Art

[0002] Industrial control systems face serious threats from misoperation by internal personnel and intrusion by illegal personnel. Misoperation by internal personnel may lead to equipment failure, production interruption or even safety accidents, while intrusion by illegal personnel is more likely to steal sensitive data, destroy system functions or perform malicious operations through physical destruction, network attacks and other means. These security threats not only endanger production safety, but may also cause damage to information security. For this reason, the application of video surveillance in industrial control systems is becoming more and more common. Video surveillance can not only improve security, but also provide real-time monitoring and fault diagnosis capabilities.

[0003] However, the widespread deployment of video surveillance in industrial control systems also brings significant challenges, especially the high demand for network bandwidth and the difficulty of automated monitoring and diagnosis. On the one hand, video surveillance in industrial control systems usually requires real-time transmission in order to detect and handle abnormal situations in a timely manner, which means that video data needs to continuously occupy network bandwidth. At the same time, many industrial scenarios require the use of high-resolution video to capture details, so the amount of video data is very large, and multi-channel video surveillance further increases the demand for network bandwidth; on the other hand, existing video surveillance systems usually rely on manual operators to view video images in real time for monitoring and decision-making. This not only increases labor costs, but is also easily affected by human factors. Traditional video surveillance systems lack efficient data processing and analysis capabilities, and it is difficult to extract valuable information from massive video data.

[0004] In order to solve the technical bottleneck of video surveillance in industrial control systems and address the important threat of abnormal behavior detection of on-site personnel, lightweight and automated video data processing, analysis and detection are required. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides an industrial site automation monitoring method and system based on cloud-edge collaborative detection.

[0006] In a first aspect, the present invention provides an industrial site automation monitoring method based on cloud-edge collaborative detection, the method comprising the following steps:

[0007] Acquire video data using an edge computing terminal, and perform discrete image processing to obtain first data to be inspected;

[0008] Constructing an edge detection model, performing multi-scale static feature mining on the first data to be inspected, and detecting suspicious persons;

[0009] After a suspicious person is detected, the video data is sent to the centralized control cloud computing center for continuous image processing to obtain the second data to be inspected;

[0010] Constructing a cloud detection model to mine the dynamic features of the second to-be-detected data in terms of time and space, and to detect abnormal behavior patterns;

[0011] When abnormal behavior is detected, the video data is saved and an alarm message is generated.

[0012] In some possible embodiments, the step of acquiring video data using an edge computing terminal and performing discrete image preprocessing to obtain first data to be inspected includes:

[0013] Use edge computing terminals to obtain video data streams collected by visual sensors;

[0014] Performing frame sampling on the video data stream according to a set frequency, extracting a discrete-time image, and normalizing pixel values ​​of the image;

[0015] A three-dimensional tensor of the image is generated according to the set size to obtain the first data to be inspected.

[0016] In some possible embodiments, the step of constructing an edge detection model, performing multi-scale static feature mining on the first to-be-detected data, and performing suspicious person detection includes:

[0017] Building an edge detection model based on a multi-channel convolutional neural network, including a first feature extractor and a first feature classifier;

[0018] Inputting the first to-be-tested data into the first feature extractor to obtain image static features;

[0019] The static features are input into the first feature classifier to detect whether there is a suspicious person in the image.

[0020] In some possible embodiments, the step of inputting the first to-be-detected data into the first feature extractor to obtain the image static features includes:

[0021] The first feature extractor uses a multi-channel convolution layer to perform parallel convolution calculation on the first to-be-detected data;

[0022] The calculation results are processed by ReLU nonlinearity and average pooling to obtain multi-scale features;

[0023] The multi-scale features are aligned and fused to form static features of the image.

[0024] In some possible embodiments, after a suspicious person is detected, the video data is sent to a centralized cloud computing center for continuous image preprocessing to obtain second data to be inspected, including:

[0025] After identifying a suspicious person, the visual sensor sends the collected video data stream to the centralized control computing center;

[0026] Reading out a continuous time image based on the video data stream, time-marking the image, and normalizing the pixel values ​​of the image;

[0027] A three-dimensional tensor set of the image is generated according to the set size to obtain the second data to be inspected.

[0028] In some possible embodiments, the step of constructing a cloud detection model, mining the dynamic features of the temporal and spatial correlation of the second to-be-detected data, and detecting abnormal behavior patterns includes:

[0029] A cloud detection model is constructed based on a convolutional long short-term memory network, including a second feature extractor and a second feature classifier;

[0030] The spatial attention module is used to assign weights to the second data to be inspected, thereby enhancing the ability to identify information in key areas of the image.

[0031] Inputting the second data to be inspected into the second feature extractor to obtain image dynamic features;

[0032] The dynamic features are input into the second feature classifier to detect whether there are abnormal behavior patterns of people in the video.

[0033] In some possible embodiments, the step of inputting the second to-be-detected data into the second feature extractor to obtain the image dynamic features includes:

[0034] The second feature extractor uses a stacked convolution layer to perform a cascade convolution calculation on the second to-be-detected data;

[0035] The calculation results are input into the input gate of the long short-term memory network unit in chronological order, and the feature sequence with contextual relevance is obtained from the output gate in sequence;

[0036] The feature sequences are combined to form dynamic features of the image.

[0037] In some possible embodiments, after detecting abnormal behavior, saving video data and generating visual alarm information include:

[0038] After identifying abnormal behavior of personnel, the centralized control cloud computing center sends the video data stream to the data server for storage and later audit work;

[0039] The computing center generates warning information including threat type, threat level, severity, etc., and performs human-computer interaction through a visualization platform.

[0040] In a second aspect, the present invention provides an industrial field automation monitoring system based on cloud-edge collaborative detection, the system comprising:

[0041] An edge data processing module, used to acquire video data using an edge computing terminal and perform discrete image processing to obtain first data to be inspected;

[0042] The edge data detection module is used to build an edge detection model, perform multi-scale static feature mining on the first data to be inspected, and detect suspicious persons; after detecting a suspicious person, the video data is sent to the centralized control cloud computing center for continuous image processing to obtain the second data to be inspected;

[0043] A cloud data detection module, used to build a cloud detection model, perform dynamic feature mining of temporal and spatial correlation on the second to-be-detected data, and perform abnormal behavior pattern detection;

[0044] The data storage module is used to save video data and generate alarm information after detecting abnormal behavior.

[0045] In a third aspect, the present invention provides an electronic device, comprising:

[0046] A memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the industrial site automation monitoring method based on cloud-edge collaborative detection as described above by executing the computer instructions.

[0047] In a fourth aspect, the present invention provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the industrial site automation monitoring method based on cloud-edge collaborative detection as described above.

[0048] An industrial site automation monitoring method and system based on cloud-edge collaborative detection in an embodiment of the present invention utilizes collaborative data processing between the edge side and the centralized control side to solve the problem of video traffic continuously occupying the communication network bandwidth. At the same time, it can automatically realize the monitoring and diagnosis of abnormal behaviors such as illegal operations and deliberate destruction in industrial production sites, and strengthen the security of physical facilities and information networks of industrial control systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 A flowchart of an industrial site automation monitoring method based on cloud-edge collaborative detection provided by an embodiment of the present invention;

[0051] Figure 2 A schematic diagram of a process for performing multi-scale static feature mining and suspicious person detection using an edge detection model provided in an embodiment of the present invention;

[0052] Figure 3 A schematic diagram of a process for mining spatiotemporal correlation dynamic features and detecting abnormal behaviors using a cloud-based detection model provided in an embodiment of the present invention;

[0053] Figure 4 A structural schematic diagram of an industrial field automation monitoring system based on cloud-edge collaborative detection provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0055] Unless otherwise specified, the technical terms or scientific terms used in the embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. "Including" or "comprising" used in the embodiments of the present invention neither limit the shapes, numbers, steps, actions, operations, components, originals and / or their groups mentioned, nor exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, originals and / or their groups, or the addition of these. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number and order of the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0056] Unless otherwise specifically stated, the relative arrangement of the components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be understood that for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship, and the techniques, methods and devices known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the techniques, methods and devices shown should be considered as part of the authorized specification. In all examples shown and discussed here, any specific other examples may have different values. It should be noted that similar symbols and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0057] In the description of the embodiments of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in the embodiments of the present invention and the features of the different embodiments or examples, without contradiction.

[0058] The embodiment of the present invention provides an industrial site automation monitoring method based on cloud-edge collaborative detection, such as Figure 1 As shown, the method specifically comprises the following steps:

[0059] Step S1: Use the edge computing terminal to obtain video data and perform discrete image processing to obtain first data to be inspected.

[0060] Specifically, the edge computing terminal is used to obtain the video data stream collected by the visual sensor. In this embodiment, the edge computing terminal is an embedded device deployed at the industrial production site, which can perform local lightweight processing and detection on the video data. The visual sensor is a monitoring device deployed at the industrial production site, which is mainly composed of an optical lens, an image sensor, a digital signal processing unit and an output interface. The video data stream is frame sampled according to the set frequency, a discrete time image is extracted, and the pixel value of the image is normalized. In this embodiment, the frame sampling frequency can be adjusted according to the work business situation of the industrial production site. Low-frequency sampling of the video stream can avoid the processing of a large number of repetitive images, thereby reducing the hardware and software requirements of the edge computing terminal. A three-dimensional tensor of the image is generated according to the set size to obtain the first data to be inspected. In this embodiment, after the image color space is converted into RGB space, a three-dimensional tensor of the image is generated according to three color channels.

[0061] Step S2: construct an edge detection model, perform multi-scale static feature mining on the first data to be inspected, and detect suspicious persons.

[0062] Specifically, in this step, refer to Figure 2 , an edge detection model is constructed based on a multi-channel convolutional neural network, including a first feature extractor and a first feature classifier. In this embodiment, the first feature extractor is composed of a multi-channel convolutional layer, and the first classifier is composed of a fully connected neural network; the first data to be inspected is input into the first feature extractor to obtain static features of the image. In this embodiment, the first feature extractor uses a multi-channel convolutional layer to perform parallel convolution calculations on the first data to be inspected, and the calculation results are processed by ReLU nonlinearity and average pooling to obtain multi-scale features, and the multi-scale features are aligned and fused to form static features of the image; the static features are input into the first feature classifier to detect whether there are suspicious persons in the image.

[0063] Step S3: After a suspicious person is detected, the video data is sent to the centralized control cloud computing center for continuous image processing to obtain the second data to be inspected.

[0064] Specifically, after identifying a suspicious person, the visual sensor sends the collected video data stream to the centralized control cloud computing center. In this embodiment, the centralized control cloud computing center is a computing service platform deployed on the centralized control side of the industrial plant, which can perform high-precision, high-complexity processing and detection of video data. After identifying a suspicious person, the edge computing terminal will modify the data upload address of the visual sensor through instructions to send the video data to the centralized control computing center; based on the video data stream, a continuous-time image is read out, the image is time-annotated, and the pixel values ​​of the image are normalized. In this embodiment, while reading the continuous-time image, a one-to-one mapping relationship table between the image and the frame sampling time is constructed to mark the time period information when the abnormal detection occurs; a three-dimensional tensor set of the image is generated according to the set size to obtain the second data to be inspected.

[0065] Step S4: construct a cloud detection model, perform dynamic feature mining of temporal and spatial correlation on the second data to be inspected, and perform abnormal behavior pattern detection.

[0066] Specifically, in this step, refer to Figure 3 , a cloud detection model is constructed based on a convolutional long short-term memory network, including a second feature extractor and a second feature classifier. In this embodiment, the second feature extractor is composed of a convolutional long short-term memory network, and the second feature classifier is composed of a fully connected neural network; a spatial attention module is used to assign weights to the second data to be inspected to enhance the ability to recognize information in key areas of the image. In this embodiment, the second data to be inspected is subjected to maximum pooling and average pooling processing, and the processing results are stacked and convolution calculation is performed to obtain attention weights, and then the weights are normalized by a sigmod function, and the attention weights are multiplied by the second data to be inspected; the second data to be inspected is input into the second feature extractor to obtain image dynamic features. In this embodiment, the second feature extractor uses a stacked convolution layer to perform cascade convolution calculations on the second data to be inspected, and the calculation results are input into the input gate of the long short-term memory network unit in chronological order, and a feature sequence with contextual relevance is obtained from the output gate in sequence, and the feature sequence is merged to form the dynamic features of the image; the dynamic features are input into the second feature classifier to detect whether there are abnormal behavior patterns of people in the video.

[0067] Step S5: After abnormal behavior is detected, the video data is saved and visual alarm information is generated.

[0068] Specifically, after identifying abnormal behavior of personnel, the centralized control cloud computing center sends the video data stream to the data server for storage for later auditing; the computing center generates alarm information including threat type, threat level, severity, etc., and conducts human-computer interaction through a visualization platform.

[0069] An embodiment of the present invention provides an automated industrial site monitoring method based on cloud-edge collaborative detection. It utilizes the collaborative data processing method of the edge side and the centralized control side to solve the problem of video traffic continuously occupying the communication network bandwidth. At the same time, it can automatically realize the monitoring and diagnosis of abnormal behaviors such as illegal operations and deliberate destruction in industrial production sites, and strengthen the security of physical facilities and information networks of industrial control systems.

[0070] Based on the same inventive concept, an embodiment of the present invention also provides an industrial field automation monitoring system based on cloud-edge collaborative detection. The system can be applied to the method recorded in the previous article. For details, please refer to the relevant records in the previous article and will not be repeated here.

[0071] like Figure 4 As shown, the system includes: an edge data processing module 401, an edge data detection module 402, a cloud data detection module 403 and a data storage module 404.

[0072] Specifically, the edge data processing module 401 is used to acquire video data using an edge computing terminal, and perform discrete image processing to obtain first data to be inspected. The edge data detection module 402 is used to construct an edge detection model, perform multi-scale static feature mining on the first data to be inspected, and detect suspicious persons; after detecting a suspicious person, the video data is sent to the centralized control cloud computing center for continuous image processing to obtain second data to be inspected. The cloud data detection module 403 is used to construct a cloud detection model, perform spatiotemporal-correlated dynamic feature mining on the second data to be inspected, and detect abnormal behavior patterns. The data storage module 404 is used to save video data and generate alarm information after abnormal behavior is detected.

[0073] An embodiment of the present invention provides an industrial site automation monitoring system based on cloud-edge collaborative detection. It utilizes the collaborative data processing method of the edge side and the centralized control side to solve the problem of video traffic continuously occupying the communication network bandwidth. At the same time, it can automatically realize the monitoring and diagnosis of abnormal behaviors such as illegal operations and deliberate destruction in industrial production sites, and strengthen the security of physical facilities and information networks of industrial control systems.

[0074] In some possible embodiments, Figure 4 As shown, the edge data processing module 401 is specifically used to: use the edge computing terminal to obtain the video data stream collected by the visual sensor; perform frame sampling on the video data stream according to the set frequency, extract the discrete time image, and normalize the pixel value of the image; generate a three-dimensional tensor of the image according to the set size to obtain the first data to be inspected.

[0075] In some possible embodiments, Figure 4As shown, the edge data detection module 402 is specifically used to: construct an edge detection model based on a multi-channel convolutional neural network, including a first feature extractor and a first feature classifier; input the first data to be inspected into the first feature extractor to obtain image static features; input the static features into the first feature classifier to detect whether there are suspicious persons in the image.

[0076] In some possible embodiments, Figure 4 As shown, the edge data detection module 402 is specifically used for: the first feature extractor uses a multi-channel convolution layer to perform parallel convolution calculations on the first data to be detected; the calculation results are subjected to ReLU nonlinear processing and average pooling processing to obtain multi-scale features; the multi-scale features are aligning and fusing to form static features of the image.

[0077] In some possible embodiments, Figure 4 As shown, the edge data detection module 402 is specifically used for: after identifying a suspicious person, the visual sensor sends the collected video data stream to the centralized control computing center; based on the video data stream, a continuous time image is read out, the image is time-annotated, and the pixel values ​​of the image are normalized; a three-dimensional tensor set of the image is generated according to a set size to obtain the second data to be inspected.

[0078] In some possible embodiments, Figure 4 As shown, the cloud data detection module 403 is specifically used to: construct a cloud detection model based on a convolutional long short-term memory network, including a second feature extractor and a second feature classifier; use a spatial attention module to assign weights to the second data to be inspected to enhance the ability to recognize information in key areas of the image; input the second data to be inspected into the second feature extractor to obtain dynamic features of the image; input the dynamic features into the second feature classifier to detect whether there are abnormal behavior patterns of people in the video.

[0079] In some possible embodiments, Figure 4 As shown, the cloud data detection module 403 is specifically used for: the second feature extractor uses a stacked convolution layer to perform cascade convolution calculations on the second data to be detected; the calculation results are input into the input gate of the long short-term memory network unit in chronological order, and feature sequences with contextual relevance are obtained from the output gate in sequence; the feature sequences are merged to form dynamic features of the image.

[0080] In some possible embodiments, Figure 4As shown, the data storage module 404 is also specifically used for: after identifying abnormal behavior of personnel, the centralized control cloud computing center sends the video data stream to the data server for storage for later audit work; the computing center generates alarm information including threat type, threat level, severity, etc., and conducts human-computer interaction through a visualization platform.

[0081] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. An industrial site automation monitoring method based on cloud-edge collaborative detection, characterized in that: The method comprises the following steps: Acquire video data using an edge computing terminal, and perform discrete image processing to obtain first data to be inspected; Constructing an edge detection model, performing multi-scale static feature mining on the first data to be inspected, and detecting suspicious persons; After a suspicious person is detected, the video data is sent to the centralized control cloud computing center for continuous image processing to obtain the second data to be inspected; Constructing a cloud detection model to mine the dynamic features of the second to-be-detected data in terms of time and space, and to detect abnormal behavior patterns; When abnormal behavior is detected, the video data is saved and an alarm message is generated.

2. According to claim 1, an industrial site automation monitoring method based on cloud-edge collaborative detection is characterized in that: The method of acquiring video data by using an edge computing terminal and performing discrete image preprocessing to obtain first data to be inspected includes: Use edge computing terminals to obtain video data streams collected by visual sensors; Performing frame sampling on the video data stream according to a set frequency, extracting a discrete-time image, and normalizing pixel values ​​of the image; A three-dimensional tensor of the image is generated according to the set size to obtain the first data to be inspected.

3. The method for automated monitoring of industrial sites based on cloud-edge collaborative detection according to claim 1 is characterized in that: The step of constructing an edge detection model, performing multi-scale static feature mining on the first to-be-detected data, and detecting suspicious persons includes: Building an edge detection model based on a multi-channel convolutional neural network, including a first feature extractor and a first feature classifier; Inputting the first to-be-tested data into the first feature extractor to obtain image static features; The static features are input into the first feature classifier to detect whether there is a suspicious person in the image.

4. The method for automated monitoring of industrial sites based on cloud-edge collaborative detection according to claim 3 is characterized in that: The step of inputting the first to-be-detected data into the first feature extractor to obtain the image static features comprises: The first feature extractor uses a multi-channel convolution layer to perform parallel convolution calculation on the first to-be-detected data; The calculation results are processed by ReLU nonlinearity and average pooling to obtain multi-scale features; The multi-scale features are aligned and fused to form static features of the image.

5. The method for automated monitoring of industrial sites based on cloud-edge collaborative detection according to claim 1, characterized in that: After the suspicious person is detected, the video data is sent to the centralized cloud computing center for continuous image preprocessing to obtain the second data to be inspected, including: After identifying a suspicious person, the visual sensor sends the collected video data stream to the centralized control computing center; Reading out a continuous time image based on the video data stream, time-marking the image, and normalizing the pixel values ​​of the image; A three-dimensional tensor set of the image is generated according to the set size to obtain the second data to be inspected.

6. The method for automated monitoring of industrial sites based on cloud-edge collaborative detection according to claim 1, characterized in that: The step of constructing a cloud detection model, mining the dynamic features of the temporal and spatial correlation of the second to-be-detected data, and detecting abnormal behavior patterns includes: A cloud detection model is constructed based on a convolutional long short-term memory network, including a second feature extractor and a second feature classifier; The spatial attention module is used to assign weights to the second data to be inspected, thereby enhancing the ability to identify information in key areas of the image. Inputting the second data to be inspected into the second feature extractor to obtain image dynamic features; The dynamic features are input into the second feature classifier to detect whether there are abnormal behavior patterns of people in the video.

7. The method for automated monitoring of industrial sites based on cloud-edge collaborative detection according to claim 6 is characterized in that: The step of inputting the second to-be-detected data into the second feature extractor to obtain the image dynamic features comprises: The second feature extractor uses a stacked convolution layer to perform a cascade convolution calculation on the second to-be-detected data; The calculation results are input into the input gate of the long short-term memory network unit in chronological order, and the feature sequence with contextual relevance is obtained from the output gate in sequence; The feature sequences are combined to form dynamic features of the image.

8. The method for automated monitoring of industrial sites based on cloud-edge collaborative detection according to claim 1, characterized in that: After the abnormal behavior is detected, the video data is saved and visual alarm information is generated, including: After identifying abnormal behavior of personnel, the centralized control cloud computing center sends the video data stream to the data server for storage and later audit work; The computing center generates warning information including threat type, threat level, severity, etc., and performs human-computer interaction through a visualization platform.

9. An industrial site automation monitoring system based on cloud-edge collaborative detection, characterized in that: The system comprises: An edge data processing module, used to acquire video data using an edge computing terminal and perform discrete image processing to obtain first data to be inspected; The edge data detection module is used to build an edge detection model, perform multi-scale static feature mining on the first data to be inspected, and detect suspicious persons; after detecting a suspicious person, the video data is sent to the centralized control cloud computing center for continuous image processing to obtain the second data to be inspected; A cloud data detection module, used to build a cloud detection model, perform dynamic feature mining of temporal and spatial correlation on the second to-be-detected data, and perform abnormal behavior pattern detection; The data storage module is used to save video data and generate alarm information after detecting abnormal behavior.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the industrial site automation monitoring method based on cloud-edge collaborative detection as described in any one of claims 1 to 8 by executing the computer instructions.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the industrial site automation monitoring method based on cloud-edge collaborative detection as described in any one of claims 1 to 8.

Images