A lightweight transmission method for subway track monitoring video

Abstract

The present invention discloses a lightweight transmission method for subway track monitoring video, comprising: acquiring monitoring video data through an edge server, performing corresponding frame extraction according to the type of area where the monitoring video data is located, and obtaining first video monitoring data of different area types; decomposing the first monitoring video data through a primary server to obtain a plurality of frame images, performing image recognition according to the type of area where the data is located, and obtaining monitoring event data; dynamically adjusting the frame rate of the first monitoring data according to the type of area where the data is located and the monitoring event data, and obtaining second video monitoring data; storing the second video monitoring data in a queue according to different area types and the priority of the monitoring event data, and transmitting the second video monitoring data to a secondary server according to the priority of the video monitoring data and the situation of network resources; and dynamically compressing the second video monitoring data through the secondary server according to the duration, priority, and distribution ratio of different frame rates of the second video monitoring data, and obtaining third video monitoring data.

Description

Technical Field

[0001] The present invention relates to the technical field of lightweight video transmission, and in particular to a lightweight transmission method for subway track monitoring videos. Background Art

[0002] Under the environment of subway central monitoring and large-scale video surveillance at stations, activities in subway stations and carriages can be monitored in real time, helping to detect and prevent criminal activities and improve passenger safety; in the event of an unexpected incident (such as a fire or accident), video surveillance can help quickly locate the incident and assess the situation, thereby speeding up emergency response; through video surveillance, passenger flow conditions can be understood in real time, train and schedule arrangements can be optimized, congestion can be reduced, and transportation efficiency can be improved; in the event of an accident, video records can serve as important evidence to help investigate the cause of the accident and the division of responsibilities; equipment failures or areas that need inspection can be detected, thereby improving the operating efficiency and service quality of the subway system.

[0003] However, the unstructured storage format and content uncertainty of video surveillance data make it inconvenient to organize, manage and transmit data. The smoothness of the link channel, as an important part of the video transmission process, largely determines the efficiency of subsequent data transmission and server pressure. After acquiring the surveillance image, image recognition technology is required to process a large amount of video data, which places high demands on network bandwidth, storage space and device performance. If each frame is analyzed and transmitted, the amount of data will be too large, which will affect the performance and efficiency of the overall video surveillance.

[0004] The existing technology usually adopts video compression technology and frame rate reduction for lightweight transmission and storage. Lightweight transmission adopts a uniform frame rate reduction, but the subway video surveillance data needs to ensure the real-time data transmission. Low frame rate means that fewer pictures are taken per second, especially when it comes to security monitoring and instant response. Therefore, all details of fast-moving objects cannot be captured, which may cause key actions to be blurred, thereby affecting the accuracy of event recognition and causing certain data delays; at the same time, too high compression rate may make the encoding and decoding process more complicated, thereby increasing processing time and total delay. At the same time, high frame rate also requires high compression rate to support, but too high compression rate supports high frame rate at the cost of high computational cost, which may introduce processing delays and does not meet some timely response requirements of subway surveillance video data. Therefore, it is necessary to dynamically balance the relationship between compression rate, frame rate and delay.

[0005] The purpose of the present invention is to design a lightweight transmission method for subway track monitoring video in view of the above-mentioned problems in the prior art. Summary of the invention

[0006] In view of this, the purpose of the present invention is to propose a lightweight transmission method for subway track monitoring video, which can solve the above-mentioned problems.

[0007] The present invention provides a lightweight transmission method for subway track monitoring video, comprising:

[0008] Acquire surveillance video data through the edge server, extract frames accordingly according to the area type of the surveillance video data, and obtain first video surveillance data of different area types;

[0009] Decomposing the first video surveillance data by the primary server to obtain a plurality of frames of images, performing image recognition according to the type of the area in which the first video surveillance data is located to obtain surveillance event data, dynamically adjusting the frame rate of the first surveillance data according to the type of the area in which the first video surveillance data is located and the surveillance event data to obtain the second video surveillance data;

[0010] The second video surveillance data is queued and stored according to different area types and priorities of surveillance event data, and transmitted to the secondary server according to the priority of the video surveillance data and the situation of network resources;

[0011] The second video surveillance data is dynamically compressed by the secondary server according to the duration, priority, and distribution ratio of different frame rates of the second video surveillance data to obtain the third video surveillance data.

[0012] Further, the extracting frames corresponding to the area where the surveillance video data is located to obtain the first video surveillance data includes:

[0013] If the monitoring video data is site area video data, corresponding frame extraction is performed according to the event monitoring type of the site area;

[0014] If the monitoring video data is track area video data, corresponding frame extraction is performed according to the event monitoring type of the track area.

[0015] Further, if the monitoring video data is site area video data, performing corresponding frame extraction according to the event monitoring type of the site area includes:

[0016] If the event monitoring type of the station area is passenger statistics monitoring, a frame is extracted every 1 second within a preset time period;

[0017] If the event monitoring type of the site area is cleaning on-the-job monitoring, three frames are extracted every 40 minutes during the operating hours;

[0018] If the event monitoring type of the site area is security on-the-job monitoring, three frames are extracted every 10 minutes during the operating hours.

[0019] Furthermore, if the monitoring video data is track area video data, extracting frames according to the event monitoring type of the track area includes:

[0020] If the event monitoring type of the track area is vehicle entry monitoring, then when the train leaves the previous station, one frame is extracted per second until the train enters the station;

[0021] If the event monitoring type of the track area is obstacle monitoring, then when the train is about to enter the station, the next section will be inspected, with 3 frames drawn per second, until the train leaves the station.

[0022] Further, the image recognition is performed according to the type of the area to obtain the monitoring event data, including:

[0023] If the first video surveillance data is site area video data, human motion recognition is performed to obtain human surveillance event data;

[0024] If the first video monitoring data is track area video data, obstacle recognition is performed to obtain obstacle monitoring event data.

[0025] Further, dynamically adjusting the frame rate of the first monitoring data according to the type of area and the monitoring event data to obtain the second video monitoring data includes:

[0026] Input the area type and its corresponding monitoring event type into the pre-trained event prediction model to obtain the predicted event data of the current area and its occurrence time period;

[0027] If there are monitoring event data and predicted event data in the same time period in the area, the frame rate of the time period is adjusted to the highest frame rate;

[0028] If there is monitoring event data or predicted event data in the same time period of the area, the frame rate of the time period is adjusted to a medium frame rate;

[0029] If there is no monitoring event data and predicted event data in the same time period of the area, the frame rate of the time period is adjusted to the minimum frame rate.

[0030] Furthermore, the event prediction model is obtained through the following training steps:

[0031] Collect monitoring event types, historical monitoring event data and their occurrence time periods in each monitoring area to form training samples for event prediction models;

[0032] An event prediction model is constructed through a machine learning model. The region type and its monitored event type are used as the input of the event prediction model, and the historical monitored event data and its occurrence time period are used as the output. The event prediction model is trained to obtain the parameters of the event prediction model.

[0033] Further, the second video surveillance data is queued and stored according to different area types and priorities of surveillance event data, and transmitted to the secondary server according to the priority of the video surveillance data and the situation of network resources, including:

[0034] Assign priority tags to the second video surveillance data according to the urgency of the monitoring event data, construct queues of different priorities, store the second video surveillance data in queues of corresponding priorities according to the priority tags, and save the second video surveillance data in queues according to different area types;

[0035] According to the real-time network bandwidth and load of each secondary server, the secondary servers are dynamically sorted, and the transmission is allocated to the secondary servers in sequence according to the priority of the queue.

[0036] Further, the storing the second video surveillance data in a queue according to different area types includes:

[0037] If the second video surveillance data is site area video data, it is stored in a first-in, first-out queue;

[0038] If the second video surveillance data is track area video data, it is stored in a queue in a loop.

[0039] Further, the second video surveillance data is dynamically compressed by the secondary server according to the duration, priority, and distribution ratio of different frame rates of the second video surveillance data to obtain the third video surveillance data, including:

[0040] The duration, priority, and distribution ratio of different frame rates of historical video surveillance data are used as state variables, and the compression rate of historical video surveillance data is used as an action variable to build a compression reinforcement learning model.

[0041] When the distortion of the compressed second video surveillance data is less than the distortion threshold, it is marked as a positive reward signal, otherwise it is marked as a negative reward signal, and the compression reinforcement learning model is optimized according to the reward signal;

[0042] The duration, priority, and distribution ratio of different frame rates of the second video surveillance data are input into the compression reinforcement learning model to obtain a predicted compression rate, and the second video surveillance data is compressed according to the predicted compression rate to obtain third video surveillance data.

[0043] Beneficial effects of the present invention:

[0044] First, we selectively extract frames according to different regions and monitoring event types to generate monitoring data focused on key events, reducing storage requirements and data redundancy. Real-time processing of full-frame rate video requires a lot of bandwidth and computing resources. By extracting frames on the edge server of the video monitoring system, we can reduce unnecessary subsequent video frame transmissions. According to different regional monitoring requirements, we can strategically increase the frame extraction frequency during the key time period when the event occurs to ensure the timeliness and integrity of the data, thereby improving the accuracy and response speed of event detection.

[0045] The second is to use the prediction model to predict the predicted event data and the time period of occurrence in the current area based on the image recognition of human motion recognition and obstacle detection through the first-level server. According to the prediction results, the frame rate in this area and time period is increased in advance. After the frame rate is increased, the frame rate is further adjusted in combination with the real-time event data to achieve dynamic adjustment of the frame rate. This can effectively avoid uniformly reducing the frame rate for lightweight transmission.

[0046] Third, queues of different priorities are used to organize and store these video data, so that they can be processed and transmitted in an orderly manner. Video data is differentiated by region, and then the appropriate queue entry method is selected according to the characteristics of the regional video data. Each secondary server is sorted in real time according to its current network bandwidth and load, and video data is distributed and transmitted to these servers in turn according to the priority of the queue to improve processing efficiency and timeliness of data transmission.

[0047] Fourth, the video compression rate is automatically determined by the duration, priority, and distribution ratio of different frame rates of the second video surveillance data, ensuring that the compressed video maximizes compression efficiency while maintaining low distortion. By dynamically adjusting the compression rate, video data can be processed without significantly reducing video quality or increasing processing delays. The compression reinforcement learning model can dynamically balance the relationship between compression rate, frame rate, and delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 It is a flow chart of the method of embodiment 1.

[0050] Figure 2 It is a transmission schematic diagram of the first embodiment. DETAILED DESCRIPTION

[0051] To facilitate understanding by those skilled in the art, the structure of the present invention is now further described in detail with reference to the embodiments and the accompanying drawings. It should be understood that the steps mentioned in the present embodiment, unless otherwise specified in their order, can be adjusted in order according to actual needs, and can even be executed simultaneously or partially simultaneously.

[0052] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a lightweight transmission method for subway track monitoring video, comprising:

[0053] S1 obtains surveillance video data through the edge server, extracts frames according to the area type of the surveillance video data, and obtains first video surveillance data of different area types;

[0054] S101: if the monitoring video data is site area video data, extract corresponding frames according to the event monitoring type of the site area;

[0055] S1011 If the event monitoring type of the station area is passenger statistics monitoring, extract a frame every 1 second within a preset time period;

[0056] S1012: If the event monitoring type of the site area is cleaning on-the-job monitoring, three frames are extracted every 40 minutes during the operating hours;

[0057] S1013 If the event monitoring type of the site area is on-the-job security monitoring, three frames are extracted every 10 minutes during the operating hours.

[0058] In this step, the types of event monitoring in the station area mainly include: passenger statistics, cleaning staff on duty, security on duty and other scenarios. In the passenger statistics scenario, according to the set time (8:00 am peak time, 19:00 pm peak time), all monitoring devices at the platform and security checkpoints of the same station automatically extract a frame every 1s. In the security on duty scenario, during the operating hours, all monitoring devices at the escalators, toilets, and public areas of the same station continuously extract 3 frames every 40 minutes to detect whether there are cleaning staff entering and exiting during this time period, which is used as the basis for their on-duty status. In the security on-duty scenario, during the operating hours, all monitoring devices at the security checkpoints of the same station continuously extract 3 frames every 10 minutes to detect the number of security personnel on duty at the security checkpoints, which is used as the basis for their on-duty status and whether there are too many or too few security personnel.

[0059] S102: If the monitoring video data is track area video data, corresponding frame extraction is performed according to the event monitoring type of the track area.

[0060] S1021 If the event monitoring type of the track area is vehicle entry monitoring, then when the train leaves the previous station, one frame is extracted per second until the train enters the station;

[0061] S1022 If the event monitoring type of the track area is obstacle monitoring, then when the train is about to enter the station, the next section will be inspected, 3 frames will be taken per second, and this will continue until the train leaves the station.

[0062] In this step, the types of event monitoring in the track area include: vehicle entry, obstacle detection and other scenarios. In the vehicle entry scenario, the time when the train leaves the previous station is obtained based on monitoring, and the station's frame extraction plan is started. Frame extraction starts when the train leaves the previous station, and one frame is extracted per second until the train is detected entering the station. In the obstacle detection scenario, the system starts the obstacle inspection frame extraction in the track area based on the monitored train position. When the train enters the station, the system automatically starts inspection of the next section and extracts frames at 3 frames per second until the train leaves the station.

[0063] Long-term and high-frame-rate surveillance videos generate a large amount of data, which is difficult to store and manage efficiently. Selectively extract frames according to different areas and types of surveillance events to generate surveillance data focused on key events, reducing storage requirements and data redundancy. Real-time processing of full-frame rate videos requires a lot of bandwidth and computing resources. By extracting frames on the edge servers that come with video surveillance, unnecessary subsequent video frame transmissions can be reduced. According to different regional monitoring needs, the frame extraction frequency is strategically increased during the key time period when the event occurs to ensure the timeliness and integrity of the data, thereby improving the accuracy and response speed of event detection.

[0064] S2 decomposes the first video surveillance data through the primary server to obtain a number of frames of images, performs image recognition according to the type of area, obtains monitoring event data, dynamically adjusts the frame rate of the first monitoring data according to the type of area and the monitoring event data, and obtains second video surveillance data;

[0065] S201: if the first video surveillance data is site area video data, human motion recognition is performed to obtain human surveillance event data;

[0066] S202: If the first video monitoring data is track area video data, obstacle recognition is performed to obtain obstacle monitoring event data.

[0067] S203: Input the area type and its corresponding monitoring event type into a pre-trained event prediction model to obtain the predicted event data of the current area and its occurrence time period;

[0068] S204: if there are monitoring event data and predicted event data in the same time period of the area, adjust the frame rate of the time period to the highest frame rate;

[0069] S205: if there is monitoring event data or predicted event data in the same time period of the area, adjust the frame rate of the time period to a medium frame rate;

[0070] S206: If there is no monitoring event data and prediction event data in the same time period of the area, the frame rate of the time period is adjusted to the lowest frame rate.

[0071] Furthermore, the event prediction model is obtained through the following training steps:

[0072] Collect monitoring event types, historical monitoring event data and their occurrence time periods in each monitoring area to form training samples for event prediction models;

[0073] An event prediction model is constructed through a machine learning model. The region type and its monitored event type are used as the input of the event prediction model, and the historical monitored event data and its occurrence time period are used as the output. The event prediction model is trained to obtain the parameters of the event prediction model.

[0074] In this step, the first-level server performs image recognition for human motion recognition and obstacle detection. The image recognition method can be a trained machine learning model, etc. The image recognition method is a prior art and will not be described in detail. By using historical monitoring data to train the prediction model, it is possible to predict in advance what may happen in the area and in the time period, and increase the frame rate in advance in the area and in the time period according to the prediction results. After the frame rate is increased, the frame rate is further adjusted in combination with real-time event data to achieve dynamic adjustment of the frame rate. This can effectively avoid uniformly reducing the frame rate for lightweight transmission. Since the subway video surveillance data needs to ensure the real-time data transmission, the low frame rate means that fewer pictures are taken per second, so it is impossible to capture all the details of fast-moving objects, which may cause key actions to be blurred, thereby affecting the accuracy and timeliness of event recognition, especially fast or short events may occur between frames and are not recorded, resulting in the omission of key evidence or event information, and also causing certain data delays.

[0075] S3 queues and stores the second video surveillance data according to different area types and priorities of surveillance event data, and transmits it to the secondary server according to the priority of the video surveillance data and the situation of network resources;

[0076] S301 assigns a priority tag to the second video surveillance data according to the urgency of the monitoring event data, constructs queues of different priorities, stores the second video surveillance data in queues of corresponding priorities according to the priority tags, and stores the second video surveillance data in queues according to different area types;

[0077] S3011 If the second video surveillance data is site area video data, save it in a first-in, first-out queue;

[0078] S3012 If the second video surveillance data is track area video data, it is stored in a queue in a loop.

[0079] S302 dynamically sorts the secondary servers according to the real-time network bandwidth and load of each secondary server, and sequentially allocates the transmission to the secondary servers according to the priority of the queues.

[0080] In this step, a priority label is assigned to each video surveillance data. This label is based on the urgency of the data, so as to clarify which part of the data needs to be processed first during the transmission process. Queues of different priorities are used to organize and store these video data, so as to process and transmit them in an orderly manner. Site video data usually involves monitoring of the flow of people and the order of events. The use of first-in, first-out queues can simplify the logic of event detection and management. Each video clip is processed in the order of generation, which can effectively distinguish and record events. The track area generally requires continuous monitoring to ensure safety, especially for continuous monitoring of tracks, vehicles, etc. The circular queue method is suitable for this continuous and almost uninterrupted data flow management. The monitoring data in the track area is updated quickly, and the latest data needs to be processed in time. The circular queue can ensure that the most current data fragments are always processed without accumulating too much outdated information. Each secondary server is sorted in real time according to its current network bandwidth and load, and the video data is distributed and transmitted to these servers in turn according to the priority of the queue to improve processing efficiency and timeliness of data transmission.

[0081] S4 dynamically compresses the second video surveillance data according to the duration, priority, and distribution ratio of different frame rates of the second video surveillance data through the secondary server to obtain third video surveillance data.

[0082] S401 uses the duration, priority, and distribution ratio of different frame rates of historical video surveillance data as state variables and the compression rate of historical video surveillance data as an action variable to construct a compression reinforcement learning model;

[0083] S402: when the distortion of the compressed second video surveillance data is less than the distortion threshold, it is marked as a positive reward signal; otherwise, it is marked as a negative reward signal, and the compression reinforcement learning model is optimized according to the reward signal;

[0084] S403 inputs the duration, priority, and distribution ratio of different frame rates of the second video surveillance data into the compression reinforcement learning model to obtain a predicted compression rate, and compresses the second video surveillance data according to the predicted compression rate to obtain third video surveillance data.

[0085] In this step, a reinforcement learning model is used to automatically determine the video compression rate to ensure that the compressed video maximizes the compression efficiency while maintaining a low distortion. The duration, priority, and distribution ratio of different frame rates of the historical video are used as input states. These variables represent the characteristics of the video, and the compression rate is used as an action variable. In this way, the model can adjust the compression level of different video clips according to the duration, priority, and frame rate to optimize the overall performance. The video length may involve the proportion of complex content or irrelevant information, which affects the choice of compression strategy; the data transmission quality of high-priority videos is given priority, and the image quality requirements after compression are higher, which affects the choice of compression rate strategy; different monitoring event types and time periods may require different frame rates to ensure the effectiveness of monitoring, and the compression strategy must flexibly adapt to these differences. Through positive or negative reward feedback, the model is guided and optimized to learn how to determine the optimal compression rate. When the distortion of the compressed video is lower than the preset threshold, it is recorded as a positive reward, otherwise it is a negative reward. The positive reward will strengthen the model's current model decision, while the negative reward will prompt the model to adjust the model decision. The application of the trained and optimized compression reinforcement learning model can obtain the best compression rate of the video, rather than a fixed high compression rate for lightweight, thereby increasing processing delay. By dynamically adjusting the compression rate, video data can be processed without significantly reducing video quality or increasing processing delay. The compression reinforcement learning model can dynamically balance the relationship between compression rate, frame rate, and delay.

[0086] Furthermore, the compressed video may be directly stored in the secondary server, or the compressed third video surveillance data may be transmitted to the cloud server according to the transmission method of step S3.

[0087] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0089] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0090] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0091] It should be noted that in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claim. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0092] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0093] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0094] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like 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 an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0095] In the description of this specification, 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 this specification, the schematic representation of the above terms should not be understood as necessarily referring to the same embodiment or example. 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 this specification and the features of the different embodiments or examples, unless they are contradictory.

Claims

1. A lightweight transmission method for subway track monitoring video, characterized in that: include: The monitoring video data is obtained through the edge server, and corresponding frames are extracted according to the area type of the monitoring video data to obtain the first video monitoring data of different area types, specifically: If the monitoring video data is site area video data, corresponding frame extraction is performed according to the event monitoring type of the site area, specifically: If the event monitoring type of the station area is passenger statistics monitoring, a frame is extracted every 1 second within a preset time period; If the event monitoring type of the site area is cleaning on-the-job monitoring, three frames are extracted every 40 minutes during the operating hours; If the event monitoring type of the site area is security on-the-job monitoring, three frames are extracted every 10 minutes during the operating hours; If the monitoring video data is track area video data, corresponding frame extraction is performed according to the event monitoring type of the track area, specifically: If the event monitoring type of the track area is vehicle entry monitoring, then when the train leaves the previous station, one frame is extracted per second until the train enters the station; If the event monitoring type of the track area is obstacle monitoring, then when the train is about to enter the station, the next section will be inspected, and 3 frames will be drawn per second until the train leaves the station; The first video surveillance data is decomposed by the primary server to obtain a plurality of frames of images, and image recognition is performed according to the type of the area to obtain surveillance event data. The frame rate of the first surveillance data is dynamically adjusted according to the type of the area and the surveillance event data to obtain the second video surveillance data. Specifically: If the first video surveillance data is site area video data, human motion recognition is performed to obtain human surveillance event data; If the first video monitoring data is track area video data, obstacle identification is performed to obtain obstacle monitoring event data; Input the area type and its corresponding monitoring event type into the pre-trained event prediction model to obtain the predicted event data of the current area and its occurrence time period; If there are monitoring event data and predicted event data in the same time period in the area, the frame rate of the time period is adjusted to the highest frame rate; If there is monitoring event data or predicted event data in the same time period of the area, the frame rate of the time period is adjusted to a medium frame rate; If there is no monitoring event data and predicted event data in the same time period of the area, the frame rate of the time period is adjusted to the lowest frame rate; The second video surveillance data is queued and stored according to different area types and priorities of surveillance event data, and transmitted to the secondary server according to the priority of the video surveillance data and the situation of network resources; The second video surveillance data is dynamically compressed by the secondary server according to the duration, priority, and distribution ratio of different frame rates of the second video surveillance data to obtain the third video surveillance data, specifically: The duration, priority, and distribution ratio of different frame rates of historical video surveillance data are used as state variables, and the compression rate of historical video surveillance data is used as an action variable to build a compression reinforcement learning model. When the distortion of the compressed second video surveillance data is less than the distortion threshold, it is marked as a positive reward signal, otherwise it is marked as a negative reward signal, and the compression reinforcement learning model is optimized according to the reward signal; The duration, priority, and distribution ratio of different frame rates of the second video surveillance data are input into the compression reinforcement learning model to obtain a predicted compression rate, and the second video surveillance data is compressed according to the predicted compression rate to obtain third video surveillance data.

2. A lightweight transmission method for subway track monitoring video according to claim 1, characterized in that: The event prediction model is obtained through the following training steps: Collect monitoring event types, historical monitoring event data and their occurrence time periods in each monitoring area to form training samples for event prediction models; An event prediction model is constructed through a machine learning model. The region type and its monitored event type are used as the input of the event prediction model, and the historical monitored event data and its occurrence time period are used as the output. The event prediction model is trained to obtain the parameters of the event prediction model.

3. A lightweight transmission method for subway track monitoring video according to claim 2, characterized in that: The second video surveillance data is queued and stored according to different area types and priorities of surveillance event data, and transmitted to the secondary server according to the priorities of the video surveillance data and the network resources, including: Assign priority tags to the second video surveillance data according to the urgency of the monitoring event data, construct queues of different priorities, store the second video surveillance data in queues of corresponding priorities according to the priority tags, and save the second video surveillance data in queues according to different area types; According to the real-time network bandwidth and load of each secondary server, the secondary servers are dynamically sorted, and the transmission is allocated to the secondary servers in sequence according to the priority of the queue.

4. A lightweight transmission method for subway track monitoring video according to claim 3, characterized in that: The step of storing the second video surveillance data in a queue according to different area types includes: If the second video surveillance data is site area video data, it is stored in a first-in, first-out queue; If the second video surveillance data is track area video data, it is stored in a queue in a loop.

Images