Analysis and alarm video monitoring platform for wind power generation equipment

Abstract

The application provides an analysis alarm video monitoring platform for wind power generation equipment, relates to the technical field of video monitoring, and comprises a real-time picture acquisition module, a historical picture acquisition module, a grid division module, an analysis module and an alarm generation and management module. The video monitoring platform is also built with a three-level architecture, which comprises a group headquarters unit, a provincial unit and a basic unit. After triggering the alarm through grid analysis, the current / historical picture data is automatically associated, the alarm event with levels is generated by means of the event playback function, the false alarm caused by environmental interference is greatly reduced, and the alarm accuracy is improved. The current picture transmitted by the video acquisition module is acquired through the real-time preview mode, and is compared with the specified historical time picture retrieved from the hierarchical storage system through the intelligent index. The historical picture is used as a reference to determine whether there is an abnormality and whether alarm processing is needed.

Description

Technical Field

[0001] This invention relates to the field of video surveillance technology, specifically to an analysis and alarm video surveillance platform for wind power generation equipment. Background Technology

[0002] The analysis and alarm video monitoring platform for wind power generation equipment monitors the operating status and external environmental changes of wind turbines (including blades, nacelles, towers, foundations, and surrounding environment) in real time, remotely, and automatically. It intelligently identifies abnormalities or potential faults such as personnel loitering, safety helmet detection, smoking detection, remote equipment monitoring, oil leak detection, and breather detection, and issues accurate alarms immediately. The platform's significance lies in greatly improving the safety protection level and operation and maintenance response efficiency of wind farms, transforming traditional manual passive inspections into proactive predictive maintenance, and effectively avoiding the risk of major safety accidents.

[0003] In the prior art, CN 117201737A discloses a method and system for cascading video stream transmission based on upper and lower level video surveillance platforms. This technology includes: announcing the public IP address and port of the upper-level video surveillance platform to the lowest-level video surveillance platform, which is the source of the video stream; after the lowest-level video surveillance platform defines the required internal IP address and port, it sends a packet to a specific port of the upper-level monitoring platform. In this way, the upper-level platform can obtain the corresponding public IP address and port required by the lower-level monitoring platform. Then, both platforms modify the IP address and port contained in the application layer protocol unit to the public IP address and port, and can then perform signaling interaction and direct transmission of video streams through these public IP addresses and ports.

[0004] However, existing technologies rely solely on single-point-of-time image analysis, primarily focusing on real-time monitoring footage, and lack a mechanism for comparison with historical baseline images. This isolation-based analysis of single real-time frames fails to establish a historical baseline image library and lacks the ability to capture spatiotemporally continuous dynamic features. For example, it cannot verify the authenticity of object movement trajectories through multiple consecutive frames. Consequently, in dynamic scenarios like wind farms, high-frequency environmental noise such as periodic vegetation swaying, gradual changes in equipment projection shape, weather disturbances, and low-light noise interference at night are misidentified as real threats. Furthermore, by neglecting the correlation between equipment operating status and the environment—for example, the morphological changes in wind turbine projection at different speeds are not modeled—the system's filtering capability for non-threatening dynamic targets is severely inadequate, ultimately resulting in a high false alarm rate and significantly reduced monitoring effectiveness.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of the present application is to provide an analysis alarm video monitoring platform for wind power generation equipment to solve the problems raised in the background art. After triggering an alarm through grid analysis, the present application automatically associates current / historical picture data, generates an alarm event with levels with the help of event playback function, and pushes it to the management interface, greatly reducing false positives caused by environmental interference and improving alarm accuracy. By real-time preview mode, the current picture transmitted by the video acquisition module is obtained and compared with the specified historical time picture retrieved from the hierarchical storage system through intelligent indexing. The historical picture is used as a reference to determine whether there is an anomaly and whether it needs to be alarmed, which can filter transient interference and improve accuracy.

[0007] To achieve the above purpose, the present application provides the following technical solutions:

[0008] The analysis alarm video monitoring platform for wind power generation equipment comprises the following functional modules:

[0009] Real-time picture acquisition module: real-time acquisition of current picture data captured by the camera, the system transmits the video stream to the storage and analysis unit based on real-time preview mode through encoding; historical video files are quickly located and picture data of a specified time point before the current time point is obtained through the historical picture acquisition module, and through indexing and association, a hierarchical storage system triggered by events is combined for subsequent grid comparison and analysis;

[0010] Grid division module: the division module is used to generate a grid, the grid is used to divide the picture into multiple independent units, the same position grids in the current picture and the historical picture are pixel-level matched to eliminate the coordinate offset caused by camera shaking or light change, a corresponding detection algorithm is bound for each grid, and the sensitivity is adjusted by the user, the non-key area is isolated, and only the specified grid is focused for change comparison, and the user manually selects the partial area after grid division through the front-end interactive interface;

[0011] Analysis module: used for detecting anomalies in each grid area, the anomalies include personnel behavior, equipment state and environmental risk, the running process is that the analysis engine loads the grid data of the current and historical pictures, the absolute difference value of the current frame grid and the historical reference frame grid is calculated by comparing the picture changes of the two time points, whether there is an anomaly in the current monitoring picture is judged by the absolute difference value, and if an anomaly is detected, an alarm signal is triggered;

[0012] Alarm generation and management module: the module comprises an alarm server and a front-end management interface, and is used for automatically generating alarm information when detecting an anomaly, issuing an alarm through an alarm management function, and the running process is that after the alarm is triggered by the anomaly signal, the system stores the video before and after the alarm in combination with the traceable playback technology, and realizes fast response and processing.

[0013] Furthermore, the real-time video acquisition module captures the raw data stream in real time, performs frame-level compression and encapsulation using the H.264 / H.265 encoding standard, establishes multiple transmission channels, and distributes the video stream, simultaneously embedding timestamps, device IDs, and GPS positioning metadata; for the historical video acquisition module, the video surveillance platform constructs a dual indexing system of time and address to generate a key frame fingerprint map;

[0014] The real-time image acquisition module also includes a dynamic bitrate adjustment mechanism, which adaptively adjusts the resolution and frame rate according to network bandwidth fluctuations. In case of abnormalities, it triggers a frame loss compensation algorithm and starts data verification and retransmission to ensure image integrity.

[0015] Furthermore, the gridding module acquires the resolution and object scale features of the real-time monitoring image and historical images, and uses an adaptive grid algorithm to divide the real-time video frame into several independent unit grids. Each grid has a unique coordinate identifier. The grayscale value / feature point comparison is performed on the grids at the same position in the current frame and historical frames. The offset vector is calculated through a feature matching algorithm. If the overall displacement of the grid exceeds a threshold, the homography transformation matrix is ​​triggered to automatically correct camera shake. For changes in illumination, histogram equalization is used to eliminate interference. The threshold for the overall displacement is 5 pixels.

[0016] The grid division module also provides sensitivity adjustment and area focusing functions. Users can set the change detection sensitivity by dragging the slider. The sensitivity adjustment range is 1%-100%. Low sensitivity can only detect the movement of large objects, while high sensitivity can capture subtle changes.

[0017] In the region focusing function, key grids are manually selected on the front-end interface, and the system automatically isolates non-selected areas, performing inter-frame difference or background subtraction algorithms only on the target grids to reduce computing power consumption.

[0018] Furthermore, the analysis module is equipped with a time and address comparison engine, and the anomaly detection mechanism of the time and address comparison engine includes the following steps:

[0019] The analysis module calls the preprocessed data from the meshing module: the current frame mesh and the historical reference frame mesh, which are strictly matched by coordinates. The mesh position deviation caused by the camera offset is automatically corrected by affine transformation to ensure the consistency of the comparison space.

[0020] First, pixel-level spatiotemporal alignment is performed, and the results are corrected using the coordinate offset of the meshing module. Then, data within each mesh cell is collected, including the mean absolute difference of grayscale values ​​of all pixels in the current frame mesh and the historical reference frame mesh. The absolute difference between the current frame mesh and the historical reference frame mesh is calculated using the following formula.

[0021] ,

[0022] in:

[0023] C(i,j) is the gray value of the pixel in the i-th row and j-th column of the current frame image grid;

[0024] S(i,j) is the gray value of the pixel in the i-th row and j-th column of the current frame image grid;

[0025] This represents the absolute difference in grayscale between the current frame C and the historical frame S.

[0026] W is the image grid width;

[0027] H is the image grid height;

[0028] The calculated results Δ1 and Δ2 are then compared, where Δ2 is the threshold for judging anomalies in the change area. If:

[0029] Δ1<Δ2

[0030] If no abnormality has occurred, no alarm signal will be triggered.

[0031] Δ1≥Δ2

[0032] If an anomaly is detected, an alarm signal will be sent from the station to the provincial unit.

[0033] Furthermore, the alarm generation and management module implements end-to-end alarm handling through an intelligent pipeline: when the grid analysis module detects an abnormal signal, the alarm server immediately triggers a five-step closed-loop process:

[0034] The system automatically associates abnormal grid coordinates, device type, and abnormality type to generate structured alarms. It utilizes traceable playback technology to store video clips of 60 seconds before and after the alarm trigger, marks the location of the abnormal grid, and distributes alarms according to preset rules. Maintenance personnel can view alarm videos and mark processing status through an interactive interface. If processing times out, the system automatically escalates to supervision. After processing is completed, the system automatically archives alarm data to the database and associates it with equipment maintenance records to provide samples for algorithm optimization.

[0035] Furthermore, the alarm distribution is based on alarm type, risk level and management responsibility to achieve precise routing and distribution: when the alarm server generates an event, the scope of impact is divided and the push level is determined according to the scope of impact.

[0036] Further, the video monitoring platform is also built with a three-level architecture, the three-level architecture includes a group headquarters unit, a province unit and a basic unit, the group headquarters unit is a top-level decision center of the monitoring platform; the province unit is used as a regional hub to collect video streams and equipment data from the basic unit in real time, to perform preliminary screening and compressed storage, and to support real-time monitoring, historical playback and event alarm, so as to facilitate the quick response and coordination of provincial managers, and the basic unit is directly deployed in the photovoltaic power station and the land wind power station to provide video monitoring capability.

[0037] Further, in the three-level architecture of the video monitoring platform, the alarm information follows a top-down transmission process: after the front-end equipment detects an abnormality, local alarm is triggered and real-time pictures are captured, preliminary filtering and formatting are performed through the station-end system, and the video monitoring platform is uploaded to the integrated control level master station of the province unit; the integrated control level master station integrates multi-station alarm information, performs risk research and judgment and linkage analysis, and if cross-domain coordination or major event disposal is required, further reports are made to the master station of the group headquarters unit; the master station of the group headquarters unit starts an emergency plan based on global data, displays the alarm situation through a large-screen visual system, and coordinates resource scheduling.

[0038] Further, in the three-level architecture of the video monitoring platform, the control instruction is transmitted from top to bottom: the master station of the group headquarters unit issues an emergency command strategy or system parameter configuration to the integrated control level master station of the province unit through a management server, the integrated control level master station executes the instruction and distributes it to the station-end system, and finally the station-end system controls the front-end equipment.

[0039] Compared with the prior art, the video monitoring platform has the following beneficial effects:

[0040] After grid analysis triggers an alarm, it automatically associates current / historical video data (using event playback functionality) to generate graded alarm events and pushes them to the management interface. Users can quickly process alarms through multi-dimensional filtering (time / status / type), achieving full-process traceability. The system divides the video feed into customizable detection units, combining multiple detection algorithms (such as personnel loitering detection, safety helmet detection, smoking detection, off-site equipment detection, equipment oil leak detection, respirator detection, open flame detection, ground water accumulation detection, cabinet door detection, personnel climbing over fences detection, human boundary crossing recognition, warehouse exit reminder, smoke detection, tooling detection, oil level detection, and fire exit blockage detection), focusing only on key areas for video comparison. This design significantly reduces false alarms caused by environmental interference and improves alarm accuracy. The system acquires the current video feed transmitted by the video capture module through real-time preview mode and compares it with specified historical footage retrieved from the hierarchical storage system via intelligent indexing. Using historical footage as a reference, it determines whether anomalies exist and whether alarm processing is necessary, filtering out momentary interference and improving accuracy. Attached Figure Description

[0041] Fig. 1 This is a flowchart illustrating the operation of the analysis and alarm video monitoring platform for wind power generation equipment according to the present invention.

[0042] Fig. 2 This is a three-level architecture diagram of the video surveillance platform in an embodiment of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0044] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly. Example

[0045] Please see Figs. 1-2 The present invention provides the following technical solution:

[0046] The analysis alarm video monitoring platform for wind power generation equipment includes the following functional modules:

[0047] Real-time picture acquisition module: real-time acquisition of current picture data captured by the camera, the system transmits the video stream to the storage and analysis unit through encoding based on real-time preview mode; through the historical picture acquisition module, the historical video file is quickly located and the picture data of the specified time point before the current time point is acquired, through indexing and association, combined with the event-triggered hierarchical storage system, for subsequent grid comparison analysis;

[0048] Grid division module: the division module is used to generate a grid, the picture is divided into multiple independent units through the grid, the same position grids in the current picture and the historical picture are pixel-level matched, the coordinate offset caused by camera shaking or light change is eliminated, each grid is bound with a corresponding detection algorithm, and the sensitivity is adjusted by the user, through isolation of non-key areas, only the specified grid is focused for change comparison, and the user manually selects the partial area after grid division through the front-end interactive interface;

[0049] Analysis module: used for detecting abnormalities in each grid area, the abnormalities include personnel behavior, equipment state and environmental risk, the running process is that the analysis engine loads the grid data of the current and historical pictures, the absolute difference value of the current frame grid and the historical reference frame grid is calculated by comparing the picture changes of the two time points, whether the current monitoring picture exists abnormalities is judged through the absolute difference value, if the abnormalities are detected, an alarm signal is triggered;

[0050] Alarm generation and management module: the module includes an alarm server and a front-end management interface, used for automatically generating alarm information when detecting abnormalities, issuing alarms through the alarm management function, the running process is that after the alarm is triggered by the abnormal signal, the system stores the video before and after the alarm in combination with the traceable playback technology, realizing fast response and processing.

[0051] In the embodiment, the video monitoring platform accesses various monitoring terminals, the monitoring terminals include a gun camera, a ball camera and a unmanned aerial vehicle device, the gun camera is a fixed monitoring: deployed at key points, using a high-precision target detection algorithm to identify personnel, vehicles or abnormal objects intrusion in real time, combining with a behavior analysis algorithm to automatically trigger regional intrusion alarm for photovoltaic panel theft, wind power equipment damage and the like, and the ball camera is linked to track;

[0052] In this embodiment, the bolt as a fixed monitoring terminal, through the high-resolution CMOS / CCD sensor captures the key point of continuous picture, using H.265 encoding real-time compression video stream, via Ethernet or optical fiber network transmission to the platform center server, in the preprocessing stage automatically execute gray correction and Gaussian filter denoising, while embedding timestamp and geographic location metadata, real-time identification of personnel, vehicles or abnormal object intrusion behavior, and specific action (such as photovoltaic panel theft, wind power equipment damage) pattern matching, once the threat to automatically trigger regional intrusion alarm mechanism, linkage ball machine for dynamic tracking, ensure the whole process delay is less than 300ms, for the photovoltaic panel disassembly, wind power equipment damage, illegal stay and other behaviors modeling, automatically trigger regional intrusion alarm and snapshot evidence;

[0053] The ball machine relies on panoramic stitching algorithm to realize large-scale dynamic monitoring, and intelligently analyzes the fire risk around the wind tower and the photovoltaic area through deep learning behavior recognition; the personnel density is counted through the heat map algorithm to assist the grassroots unit in scheduling security resources, and the abnormal behavior is immediately pop-up alarmed to the provincial platform. The ball machine in this embodiment provides horizontal 360°+vertical 90° PTZ rotation, eliminates parallax ghosting through local homography grid transformation (APAP algorithm), generates seamless panorama, and adapts to changes in light, and can control the automatic rotation of the ball machine view angle in any area of the panorama;

[0054] The unmanned aerial vehicle provides mobile inspection, carries sensor fusion algorithm, autonomously patrols the photovoltaic array and fan group through the path planning algorithm, locates photovoltaic panel faults through the infrared hot spot detection algorithm, and identifies fan blade cracks or icing through the blade deformation analysis algorithm; combined with SLAM technology, a three-dimensional real scene model is generated, and abnormal data is returned to the algorithm platform of the group headquarters in real time to generate equipment health statistical report. The core sensor configuration of the unmanned aerial vehicle in this embodiment includes a hyperspectral camera, which analyzes abnormal conditions such as dust accumulation degree, bird droplet corrosion and blade coating aging degree on the surface of the photovoltaic panel.

[0055] In this embodiment, the real-time picture acquisition module captures raw data stream in real time, performs frame-level compression and packaging using H.264 / H.265 encoding standard, establishes a multi-channel transmission channel for video stream distribution, and synchronously embeds timestamp, device ID and GPS positioning metadata; for the historical picture acquisition module, the video monitoring platform constructs a dual index system of time and address to generate a key frame fingerprint map.

[0056] A dynamic code rate adjustment mechanism is also established in the real-time picture acquisition module, which is used to adaptively adjust the resolution and frame rate according to network bandwidth fluctuations, trigger the frame loss compensation algorithm and start data verification and retransmission in abnormal conditions, and ensure the integrity of the picture.

[0057] The grid division module also provides sensitivity adjustment function and area focusing function. A user sets change detection sensitivity by dragging a slider. The sensitivity adjustment range is 1%-100%. Low sensitivity only identifies large object movement, and high sensitivity captures subtle changes.

[0058] In the area focusing function, a key grid is manually framed in the front-end interface. The system automatically isolates non-selected areas and only performs inter-frame difference method or background subtraction algorithm on the target grid to reduce computing power consumption.

[0059] In this embodiment, the analysis module is used to detect abnormalities in each grid area in the real-time monitoring picture and the contrast historical picture. The running process is that the analysis engine loads the grid data of the current and historical pictures. By comparing the picture changes at two time points, the absolute difference value of the current frame grid and the historical reference frame grid is calculated. Whether the current monitoring picture has an abnormality is determined by the absolute difference value. If an abnormality is detected, an alarm signal is triggered.

[0060] In this embodiment, the analysis module is used to detect abnormalities in each grid area in the real-time monitoring picture and the contrast historical picture. The running process is that the analysis engine loads the grid data of the current and historical pictures. By comparing the picture changes at two time points, the absolute difference value of the current frame grid and the historical reference frame grid is calculated. Whether the current monitoring picture has an abnormality is determined by the absolute difference value. If an abnormality is detected, an alarm signal is triggered.

[0061] The analysis module calls the data preprocessed by the grid module: the current frame grid and the historical reference frame grid. The coordinates are strictly matched. The grid position deviation caused by lens offset is automatically corrected through affine transformation to ensure the consistency of the comparison space.

[0062] First, pixel-level space-time alignment is performed. The coordinate offset correction result of the grid module is used. Then, data in each grid unit is collected. The collection range includes the absolute difference mean of the gray values of all pixels of the current frame grid and the historical reference frame grid. The absolute difference value of the current frame grid and the historical reference frame grid is calculated by the following formula:

[0063] ,

[0064] Wherein:

[0065] C(i,j) is the gray value of the pixel in the i-th row and j-th column of the current frame image grid, and the theoretical range of the gray value is 0-255.

[0066] S(i,j) is the gray value of the pixel in the i-th row and j-th column of the current frame image grid, and the theoretical range of the gray value is 0-255.

[0067] is the absolute difference value of the gray value of the current frame C and the historical frame S. Each group (i,j) represents a fixed pixel point. In the image collected by the unified monitoring device in this formula, each pixel point is compared and calculated with the pixel point in the same position of the historical image.

[0068] W is the image grid width, that is, the number of width pixels in the grid within the range;

[0069] H is the image grid height, that is, the number of height pixels in the grid within the range;

[0070] When the value of Δ1 in the calculation result is larger, it indicates that the degree of change of the picture within the current frame grid and the historical reference frame grid is stronger.

[0071] The calculation result Δ1 is compared with Δ2, which is a change area abnormality judgment threshold value, and if:

[0072] Δ1 < Δ2

[0073] it is determined that no abnormality is generated, and no alarm signal is triggered, and if:

[0074] Δ1 ≥ Δ2

[0075] it is determined that an abnormality is generated, and an alarm signal is sent from the station to the provincial unit.

[0076] In the embodiment, an average absolute difference calculation scheme is also provided, that is:

[0077] ,

[0078] where Δ m is the average absolute difference, and the theoretical range thereof is also 0-255. When Δ m < 5, it represents a slight change, for example, a light fluctuation factor can cause the change; if 5 ≤ Δ m ≤ 30, it is determined that an abnormality is generated, and if Δ m ≥ 30, it is determined that a significant change is generated, for example, object movement or scene change can cause the change. The formula is not sensitive to abnormal pixels (such as transient noise), and can further improve the reliability of the final detection result.

[0079] In the embodiment, the alarm generation and management module implements full-link alarm handling through an intelligent pipeline: when the grid analysis module detects an abnormal signal, the alarm server immediately triggers a five-step closed-loop process:

[0080] an abnormal grid coordinate, a device type, and an abnormal type are automatically associated to generate a structured alarm, a traceable playback technology is called, a video clip of 60 seconds before and 60 seconds after the alarm triggering is stored, and the abnormal grid position is marked, the alarm is distributed according to a preset rule; an operation and maintenance personnel views the alarm video and a processing state through an interactive interface, and the processing is automatically upgraded and supervised when it is overdue; after the processing is completed, the system automatically archives the alarm data to a database, associates a device maintenance record, and provides a sample for algorithm optimization.

[0081] The distribution alarm in this embodiment is precisely routed and distributed according to alarm types, risk levels and management responsibilities: when the alarm server generates an event, the influence range is divided, and the push level is determined according to the influence range:

[0082] The alarm information is pushed to the basic unit: trigger the monitoring large screen pop-up window and sound and light alarm, synchronously link the ball machine to automatically turn and lock the target, and push the disposal guide;

[0083] The alarm information is pushed to the provincial unit: generate a structured operation and maintenance work order, including video clips, grid coordinates, device ID, assign to the nearest maintenance team mobile terminal, and automatically upgrade to the on-duty supervisor if not responded within 10 minutes;

[0084] The alarm information is pushed to the group headquarters unit: aggregate alarms according to device types / geographical areas, generate health degree heat map and prediction report, and drive algorithm model iteration.

[0085] In this embodiment, the video monitoring platform also has a three-level architecture, which includes the group headquarters unit, the provincial unit and the basic unit. The group headquarters unit is the top decision-making center of the monitoring platform. The provincial unit serves as a regional hub for collecting video streams and device data from the basic unit in real time, performing preliminary screening and compressed storage, and supporting real-time monitoring, historical playback and event alarm, which facilitates the rapid response and coordination of provincial management personnel. The basic unit is directly deployed in photovoltaic power stations and land-based wind power stations to provide video monitoring capabilities.

[0086] In the three-level architecture of the video monitoring platform, the alarm information follows a top-down transmission process: after the front-end device detects an anomaly, it triggers a local alarm and captures real-time images, which are preliminarily filtered and formatted by the station-end system and uploaded to the master control station of the provincial unit. The master control station integrates multi-site alarm information, conducts risk research and linkage analysis, and if cross-domain coordination or major event disposal is required, it further reports to the master station of the group headquarters unit. The master station of the group headquarters unit starts the emergency plan based on global data, displays the alarm situation through the large screen visualization system, and coordinates resource scheduling. Conversely, control instructions are transmitted from top to bottom: the master station of the group headquarters unit issues emergency command strategies or system parameter configurations to the master control station of the provincial unit through the management server, the master control station executes the instructions and distributes them to the station-end system, and finally the station-end system controls the front-end device.

[0087] In the three-level architecture of the video monitoring platform, control instructions are transmitted from top to bottom: the master station of the group headquarters unit issues emergency command strategies or system parameter configurations to the master control station of the provincial unit through the management server, the master control station executes the instructions and distributes them to the station-end system, and finally the station-end system controls the front-end device.

[0088] The above formulas are all dimensionless values calculated, the formula is obtained by collecting a large amount of data to simulate the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0089] The above embodiments can be implemented wholly or partially by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented wholly or partially in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solutions.

[0090] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, which can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiments according to actual needs.

[0091] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. An analytic alarm video surveillance platform for wind power plants, characterized in that The video monitoring platform comprises the following functional modules: A real-time picture acquisition module: real-time acquisition of current picture data captured by a camera, the system transmits the video stream to the storage and analysis unit through encoding based on the real-time preview mode; the historical picture acquisition module is used for searching and quickly positioning a historical video file, and picture data of a specified time point before a current time point is acquired, and through indexing and association, a hierarchical storage system triggered by an event is combined, and is used for subsequent grid comparison and analysis; A grid division module: the division module is used for generating a grid, the picture is divided into a plurality of independent units through the grid, pixel-level matching is performed on the grids at the same position in the current picture and the historical picture, coordinate offset caused by camera shaking or light change is eliminated, a corresponding detection algorithm is bound for each grid, and sensitivity is supported to be adjusted by a user, non-key areas are isolated, and only specified grids are focused for change comparison, and the user manually selects a partial area after grid division through a front-end interactive interface; An analysis module: used for detecting an exception in each grid area, the exception includes personnel behavior, equipment state and environmental risk, a running process is that an analysis engine loads grid data of the current and historical pictures, changes of the two time points are compared, an absolute difference value of the current frame grid and the historical reference frame grid is calculated, whether the current monitoring picture exists an exception is judged through the absolute difference value, and if the exception is detected, an alarm signal is triggered; An alarm generation and management module: the module includes an alarm server and a front-end management interface, and is used for automatically generating alarm information when the exception is detected, issuing an alarm through an alarm management function, and realizing quick response and processing by storing video before and after the alarm in combination with traceable playback technology after the alarm is triggered.

2. The wind power plant oriented analytical alarm video surveillance platform according to claim 1, characterized in that: The real-time picture acquisition module captures a raw data stream in real time, performs frame-level compression and packaging through H.264 / H.265 coding standards, performs video stream distribution after establishing a plurality of transmission channels, and synchronously embeds timestamp, device ID and GPS positioning metadata; for the historical picture acquisition module, the video monitoring platform constructs a dual-index system of time and address, and generates a key frame fingerprint atlas; A dynamic code rate adjustment mechanism is further established in the real-time picture acquisition module, used for adaptively adjusting resolution and frame rate according to network bandwidth fluctuation, triggering a frame loss compensation algorithm and starting data retransmission in an abnormal condition, and ensuring picture integrity.

3. The wind power plant oriented analytical alarm video surveillance platform according to claim 2, characterized in that: The grid division module acquires resolution and object scale features of real-time monitoring pictures and historical pictures, divides the real-time video frame into a plurality of independent unit grids through an adaptive grid algorithm, each grid has a unique coordinate identifier, the same position grids of the current frame and the historical frame are compared in terms of gray value / feature point, an offset vector is calculated through a feature matching algorithm, if the overall displacement of the grid exceeds a threshold value, a homographic transformation matrix is triggered to automatically correct camera shaking; for light change, histogram equalization is used to eliminate interference, and the threshold value of the overall displacement is 5 pixel points.

4. The wind power plant oriented analytical alarm video surveillance platform according to claim 3, characterized in that: The grid division module also provides sensitivity adjustment function and area focusing function. The user sets the change detection sensitivity by dragging the slider. The sensitivity adjustment range is 1%-100%. Low sensitivity only identifies large object movement, and high sensitivity captures subtle changes. In the area focusing function, the key grid is manually selected in the front-end interface. The system automatically isolates the non-selected area and only performs inter-frame difference method or background subtraction algorithm on the target grid to reduce the computing power consumption.

5. The wind power plant oriented analytical alarm video surveillance platform according to claim 4, characterized in that: The analysis module is equipped with a time and address comparison engine. The abnormal detection mechanism of the time and address comparison engine includes the following contents: The analysis module calls the preprocessed data of the grid module: the current frame grid and the historical reference frame grid. The coordinates are strictly matched. The grid position deviation caused by lens offset is automatically corrected through affine transformation to ensure the consistency of the comparison space. First, the pixel-level space-time alignment is performed. The coordinate offset correction result of the grid module is used to collect data in each grid unit. The collection range includes the absolute difference mean of the gray value of all pixels of the current frame grid and the historical reference frame grid. The absolute difference value of the current frame grid and the historical reference frame grid is calculated by the following formula: ， Where: C(i,j) is the gray value of the i-th row and j-th column pixel of the current frame image grid. S(i,j) is the gray value of the i-th row and j-th column pixel of the historical frame image grid. Gabs is the absolute difference of gray scale between the current frame C and the history frame S; W is the width of the image grid. H is the height of the image grid. The calculated result Δ1 is compared with Δ2, which is the change area abnormality judgment threshold. If: Δ1 < Δ2 It is determined that no abnormality has occurred, and no alarm signal is triggered. If: Δ1 ≥ Δ2 It is determined that an abnormality has occurred, and an alarm signal is sent from the station to the provincial unit.

6. The wind power plant oriented analytical alarm video surveillance platform of claim 1, wherein: The alarm generation and management module realizes full-link alarm handling through intelligent pipeline: when the grid analysis module detects an abnormal signal, the alarm server immediately triggers a five-step closed-loop process: Automatically associate abnormal grid coordinates, device type, and abnormal type to generate structured alarms. Call traceable playback technology to store 60 seconds of video clips before and after the alarm trigger, and mark the abnormal grid position. According to the preset rules, distribute the alarm; operation and maintenance personnel view the alarm video and mark the processing status through the interactive interface. If the processing time is exceeded, it will be automatically upgraded and supervised; after processing is completed, the system automatically archives the alarm data to the database and associates the device maintenance record to provide samples for algorithm optimization.

7. The wind power plant oriented analytical alarm video surveillance platform according to claim 6, characterized in that: The distributed alarm is accurately routed and distributed according to the alarm type, risk level, and management responsibilities: after the alarm server generates an event, the impact range is determined, and the push level is determined according to the impact range.

8. The wind power plant oriented analytical alarm video surveillance platform of claim 1, wherein: The video monitoring platform is also built with a three-level architecture, which includes a group headquarters unit, a provincial unit and a basic unit, the group headquarters unit is the top decision-making center of the monitoring platform; the provincial unit is used as a regional hub to collect video streams and equipment data from the basic unit in real time, to perform preliminary screening and compressed storage, and to support real-time monitoring, historical playback and event alarm, so as to facilitate the quick response and coordination of provincial managers, and the basic unit is directly deployed in photovoltaic power stations and land wind power stations to provide video monitoring capabilities.

9. The wind power plant oriented analytical alarm video surveillance platform according to claim 8, characterized in that: In the three-level architecture of the video monitoring platform, the alarm information follows a top-down transmission process: after the front-end equipment detects an anomaly, local alarm is triggered and real-time pictures are captured, preliminary filtering and formatting are performed through the station-end system, and the alarm information is uploaded to the integrated control level master station of the provincial unit; The integrated control level master station integrates multi-site alarm information, conducts risk research and linkage analysis, and if cross-domain coordination or major event disposal is required, it further reports to the master station of the group headquarters unit; The master station of the group headquarters unit starts an emergency plan based on global data, displays the alarm situation through a large-screen visual system, and coordinates resource scheduling; on the contrary, the control command is transmitted from top to bottom: the master station of the group headquarters unit issues emergency command strategies or system parameter configurations to the integrated control level master station of the provincial unit through the management server, the integrated control level master station executes the command and distributes it to the station-end system, and finally the station-end system controls the front-end equipment.

10. The wind power plant oriented analytical alarm video surveillance platform according to claim 9, characterized in that: In the three-level architecture of the video monitoring platform, the control command is transmitted from top to bottom: the master station of the group headquarters unit issues emergency command strategies or system parameter configurations to the integrated control level master station of the provincial unit through the management server, the integrated control level master station executes the command and distributes it to the station-end system, and finally the station-end system controls the front-end equipment.

Images