A substation space video fusion and collaborative management method and system

Abstract

The application discloses a kind of based on substation space video fusion and collaborative management and control method and system, it is related to substation control technical field;Including: acquisition substation data, and execute pre-processing, establish based on longitude, latitude, elevation unified space-time reference system, based on the three-dimensional space analysis model of substation data after pre-processing is constructed, generates and contains the comprehensive database of all equipment parameters of substation, cloud table change matrix data, curved surface mapping data and space coding;Its technical key points are: the application is based on equipment collaborative screening mechanism, based on space coding, in combination with the mapping relationship between data, space unit, equipment and camera, equipment scheduling and target dynamic tracking are realized;Solve the problems of existing technology, such as equipment independent operation, artificial retrieval cumbersome, dynamic target is easy to lose, emergency response lag and low control efficiency, realize substation equipment collaborative scheduling and abnormal closed loop disposal, improve the real-time safety control and emergency disposal capability of substation.

Description

Technical Field

[0001] This invention relates to the field of substation control technology, specifically to a method and system for substation spatial video fusion and collaborative control. Background Technology

[0002] Intelligent substations are core hub nodes in new power systems. Based on an overall architecture of full-station information digitization, communication platform networking, and information sharing standardization, they integrate intelligent primary equipment and information-based secondary equipment. They can automatically complete basic functions such as power information collection, measurement, control, protection, and monitoring, while supporting advanced operation and maintenance functions such as intelligent grid regulation, online analysis, and collaborative interaction. Substations contain main transformers, GIS, high-voltage bushings, and other irregularly shaped curved surface power equipment, as well as various types of high-definition monitoring terminals. The equipment layout is highly complex, the operating conditions are dynamic, and the requirements for safety management precision are stringent, making it a core application scenario for intelligent operation and maintenance of the power grid.

[0003] Existing substation spatial video fusion technology has the following drawbacks: Existing technologies lack a unified spatial reference system for substations. Video acquisition equipment, power equipment, and substation building facilities belong to three independent spatial coordinate calculation systems. The calibration accuracy of camera intrinsic and extrinsic parameters (intrinsic parameters: focal length, pixel size; extrinsic parameters: position, angle) is insufficient, resulting in significant offsets in the projection of video images onto the 3D model, with an error rate generally exceeding 0.5%. This leads to distortion after the video is projected onto the 3D scene. Furthermore, substations contain numerous cylindrical devices, curved structures, and other complex surfaces, making it impossible for existing technologies to achieve distortion-free and accurate projection. In addition, when the pan-tilt camera adjusts its viewing angle, the mapping relationship cannot be updated synchronously in real time, resulting in "image drift" and loss of spatial correspondence.

[0004] Existing technologies only achieve surface window overlay of video and spatial scenes, which is a static mirroring mode without deep virtual-real fusion logic. Static 3D scenes cannot be updated by video in real time, and there is no deep data linkage between video and spatial scenes. At the same time, various monitoring, sensing, business, and alarm data of substations are isolated from video and spatial scene data, forming data silos. The utilization rate of multi-source data is low, and it is impossible to build a digital management and control foundation that can provide real-time feedback on the on-site status.

[0005] In summary, existing substation spatial video fusion technologies do not meet people's requirements. Therefore, we propose a method and system based on substation spatial video fusion and collaborative management. Summary of the Invention

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for spatial video fusion and collaborative management of substations includes: Data from substations is collected and preprocessed. Based on the preprocessed substation data, a three-dimensional spatial analysis model is constructed, generating a comprehensive database containing all equipment parameters of the substation, gimbal change matrix data, surface mapping data, and spatial coding. The national standard video stream decoding algorithm is used to perform unified transcoding on real-time acquired multi-format videos, and video pre-optimization is performed based on a comprehensive database. The optimized video is converted into dynamic textures and projected onto the corresponding spatial positions based on a region adaptation interpolation algorithm to obtain the video image base. The system accesses multi-source data collected from substations, binds multi-source data, spatial location, and camera footage based on spatial coding and video footage, constructs a data association system, and stores it in a comprehensive database. By collecting external environmental data and combining it with a data association system, the ambient light and shadow adaptation method is used to adjust the video parameters of the video frame base to obtain a virtual-real fusion scene. Obtain operation instructions, parse the corresponding spatial code from the virtual-real fusion scene, retrieve cameras covering the spatial code from the comprehensive database, filter cameras according to the set monitoring rules, and control the filtered cameras to perform focused monitoring; When the target moves, the inter-frame differential target recognition algorithm is used to extract the dynamic target, and the trajectory prediction algorithm is combined to predict the target position and perform focused monitoring. The system acquires substation environmental data in real time and compares it with preset standard rules. If the data meets the preset standard rules, no action is taken; otherwise, an anomaly alarm is triggered, and a maintenance work order is generated.

[0007] Preferably, the substation data includes substation laser point cloud data scanned by lidar, geodetic coordinate benchmark data measured on-site by national land benchmark points, coordinates of substation equipment and building structures queried from substation drawings, and camera installation location and altitude data measured by GPS equipment. The preprocessing includes removing noise points, outliers, and invalid acquisition points from the substation laser point cloud data using a combination of statistical and conditional filtering. Point cloud data is then stitched and registered using an iterative nearest-point algorithm to obtain a global point cloud dataset for the substation. Spatial data from the substation is converted into a unified national geodetic coordinate system and national elevation datum to obtain a unified coordinate dataset and a continuous point cloud surface. A three-dimensional mesh is then created on the continuous point cloud surface to obtain a three-dimensional mesh model. Partitioning is performed according to voltage and equipment function to obtain multi-level spatial units. A unique spatial code is assigned to each spatial unit, camera, and other device to obtain a spatial coding table.

[0008] Preferably, the three-dimensional spatial analysis model includes: an input layer, a feature point extraction and database construction layer, a camera intrinsic and extrinsic parameter calibration and solution layer, a gimbal dynamic mapping matrix iteration layer, a complex surface projection parameter calculation layer, and an output layer; The input layer takes into account the substation's global point cloud dataset, 3D mesh model, unified coordinate dataset, and spatial coding table. The feature point extraction and database construction layer extracts the mesh vertices of the 3D mesh model, calculates the Gaussian curvature and average curvature of each mesh vertex, selects curvature extreme points, and constructs a curvature candidate feature point set. A secondary selection is performed on the curvature candidate feature points according to preset feature selection rules to obtain the power feature point set. After correcting the coordinates of the power feature points, a unique feature code is assigned to each corrected power feature point, and the 3D spatial coordinates, corresponding spatial units, and equipment information are associated to obtain the standard point set. The feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, and the gimbal dynamic mapping matrix iteration layer sequentially perform feature extraction analysis, attitude calculation, and surface analysis processing. The output layer integrates the computational results parameters from the feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, the gimbal dynamic mapping matrix iteration layer, and the complex surface projection parameter calculation layer to form the original parameter set. Based on the spatial encoding table, spatial encoding is performed, and the binding of device information and parameter labels is realized. The binding data is validated, and data that fails the validation is removed. After format unification, a comprehensive database is constructed.

[0009] Preferably, the feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration solution layer, and the gimbal dynamic mapping matrix iteration layer sequentially perform feature extraction analysis, attitude calculation, and surface analysis processing, including: The camera intrinsic and extrinsic parameter calibration solution layer extracts two-dimensional feature points from the captured standard image to obtain a two-dimensional image feature set; it then performs matching between the three-dimensional feature points in the standard point set and the two-dimensional feature points in the two-dimensional image feature set to obtain a feature matching set; the matching results in the feature matching set are iteratively filtered using a random sampling consensus algorithm to obtain a precise pairing set; based on the standard pinhole camera imaging model and using the data in the precise pairing set as a basis, the camera intrinsic and extrinsic parameter attitude parameters are solved, and then iterative optimization is performed using the bundle adjustment method to obtain the parameter set of all cameras in the substation. The parameter set includes the intrinsic parameter matrix, distortion coefficient set, extrinsic parameter rotation matrix, extrinsic parameter translation, and camera calibration reprojection error. The gimbal dynamic mapping matrix iteration layer performs three-dimensional attitude calculation on the gimbal's horizontal and pitch rotation angles, generating horizontal rotation attitude matrices and pitch rotation attitude matrices. After superposition, the attitude increment rotation matrix is ​​obtained. The extrinsic parameter rotation matrix and extrinsic parameter translation are used as static references to update the extrinsic parameters, obtaining the real-time extrinsic parameter rotation matrix and real-time extrinsic parameter translation after attitude changes. The equivalent focal length parameter of the camera is corrected based on the gimbal's optical zoom, generating a corrected intrinsic parameter matrix. When the gimbal attitude data changes, a real-time spatial change matrix is ​​generated, and the dynamic correction focal length and mapping consistency verification results are obtained. The complex surface projection parameter calculation layer calls the 3D mesh model and the 3D mesh point and surface structure data of the equipment to generate the surface equipment mesh coordinate set; it calls the parameter set and real-time spatial change matrix of all cameras in the substation to construct the spatial relationship between the camera view and the surface equipment; it performs distortion-free plane unfolding on the 3D surface of irregular equipment, establishes and corrects the bidirectional binding mapping relationship between the 3D surface spatial coordinates and the 2D video texture pixel coordinates, and obtains the texture interpolation coefficients, sampling step size and correction coefficient set of each surface. After integration, it generates the surface texture mapping parameter table, correction coefficient set, surface correspondence matrix and distortion-free projection rules.

[0010] Preferably, the pre-processing optimization of the video includes: performing frame segmentation on the video; retrieving the distortion coefficient set; correcting the pixel coordinates of each image in the video based on the distortion coefficient set to obtain the corrected video frame; identifying the device information and spatial unit corresponding to the video based on the spatial coding table; and determining whether the spatial unit contains a curved device after analyzing the device information. If no curved device exists, no additional processing is performed; if a curved device exists, the device information is used to determine whether to call the curved texture mapping parameter table or the correction coefficient set to perform correction and obtain the video image base.

[0011] Preferably, the optimized video is converted into a dynamic texture and projected onto the corresponding spatial position based on a region-adaptive interpolation algorithm to obtain the video image base. This includes: converting the video image base into a dynamic texture pixel resource adapted to a three-dimensional spatial surface, and unifying the texture pixel format and sampling precision; classifying each region within the spatial unit into types, including planar regions and curved regions, and using a bilinear interpolation texture mapping algorithm to perform uniform sampling and interpolation filling on the video texture pixels in the planar regions; and using a curved interpolation mapping algorithm to perform texture adaptation projection on the curved regions based on the curved texture mapping parameter table and correction coefficient set.

[0012] Preferably, the multi-source data collected by the substation includes equipment status data, environmental monitoring data, safety alarm data, and operation and maintenance data. The multi-source data collected by the substation is then subjected to formatting and time-series alignment processing. Based on spatial coding and video footage, the binding of multi-source data, spatial location, and camera footage is performed, including: using a spatial coding table and the unique code of the equipment as the primary key, matching the time-series aligned multi-source data with equipment, spatial units, and cameras, clarifying the spatial units corresponding to the multi-source data and the camera footage, and constructing a mapping relationship between data, spatial units, equipment, and cameras; binding control locations with video footage, binding equipment data with 3D spatial analysis models, and binding alarm data with spatiotemporal data to obtain a data association system.

[0013] Preferably, the external environmental data includes measured data from ambient light sensors at the substation site, meteorological data, and time-series data. Combined with an ambient light and shadow adaptation method, video parameter adjustment is performed, including extracting real-time light intensity, day / night time, sunny / cloudy weather, and ambient color temperature from the external environmental data; matching corresponding brightness, contrast, saturation, and shadow compensation correction parameters based on a preset light and shadow threshold rule library; performing adaptive parameter adjustment; performing frame-by-frame pixel-level light and shadow adaptation correction on the dynamic video texture projected onto the surface of the 3D spatial analysis model to unify the brightness levels, color tones, and light and shadow transitions between the video image and the 3D spatial analysis model; calling an edge feathering transition algorithm to perform smooth feathering processing on the splicing boundary between the video texture and the 3D spatial analysis model substrate; and monitoring ambient light and shadow fluctuations frame by frame and dynamically adjusting parameters to obtain a virtual-real fusion scene.

[0014] Preferably, the method of retrieving cameras covering spatial codes from the comprehensive database is as follows: Based on spatial codes, the corresponding camera set number is extracted from the mapping relationship between data, spatial units, equipment and camera devices, and then the optimal scheduling camera is selected according to the preset priority filtering rules; combined with the camera parameter set, as well as the spatial distance and three-dimensional size of the target corresponding to the instruction, the camera parameters are adjusted to achieve target focused monitoring.

[0015] A substation spatial video fusion and collaborative management system includes a model building module, a video processing module, a data processing module, a real-time adjustment module, and a real-time management module. Model building module: Collects substation data and performs preprocessing. Based on the preprocessed substation data, it builds a three-dimensional spatial analysis model and generates a comprehensive database containing all equipment parameters of the substation, PTZ change matrix data, surface mapping data and spatial coding. Video processing module: It adopts the national standard video stream decoding algorithm to perform unified transcoding processing on real-time acquired multi-format videos, and performs video pre-optimization based on a comprehensive database. The optimized video is converted into dynamic textures and projected to the corresponding spatial position based on the region adaptation interpolation algorithm to obtain the video image base. Data processing module: It receives multi-source data collected by the substation, performs binding of multi-source data, spatial location and camera footage based on spatial coding and video image base, and stores it in the comprehensive database; Real-time adjustment module: Collects external environmental data, combines it with a data association system, and uses an ambient light and shadow adaptation method to adjust the video parameters of the video frame base to obtain a virtual-real fusion scene; Real-time management module: It acquires operation commands, parses the corresponding spatial codes from the virtual-real fusion scene, retrieves cameras covering the spatial codes from the comprehensive database, filters cameras according to the set monitoring rules, and controls the filtered cameras to perform focused monitoring; when the target moves, it uses the inter-frame differential target recognition algorithm to extract the dynamic target, combines the trajectory prediction algorithm to predict the target position, and performs focused monitoring; it acquires substation environmental data in real time and compares it with preset standard rules. If it meets the preset standard rules, no action is taken; otherwise, an abnormal alarm is triggered, and a maintenance work order is generated.

[0016] This invention provides a method and system for spatial video fusion and collaborative management of substations, which has the following beneficial effects: This invention constructs a three-dimensional spatial analysis model of a substation based on the national geodetic coordinate system and elevation datum, realizing the logical unification of spatial coordinates of equipment, video, and scene; it solves the problems of multiple coordinate systems being chaotic and insufficient accuracy of manual calibration in the spatial video fusion process of existing technologies, and realizes the virtual-real fusion and alignment of planar and curved surface equipment in substation videos, meeting the accuracy requirements of substation monitoring.

[0017] This invention is based on spatial coding technology, using the unique code of the device as the primary key to construct a mapping relationship between data, spatial units, devices and cameras. Based on a three-dimensional spatial analysis model, it analyzes the fixed parameters to realize the mounting and updating of video textures and data. It solves the defects of existing technologies that only realize the surface overlay of video, static solidification of three-dimensional scenes, isolation of multiple types of data, and extremely low data utilization. It realizes real-time synchronization of scene and device status, and provides data support for operation and maintenance and emergency response.

[0018] This invention utilizes a collaborative screening mechanism based on spatial coding, combining the mapping relationship between data, spatial units, equipment, and cameras to achieve dynamic tracking of equipment and targets. It addresses the shortcomings of existing technologies, such as independent equipment operation, cumbersome manual retrieval, easy loss of dynamic targets, delayed emergency response, and low management efficiency. It improves the flexibility of collaborative scheduling of substation equipment, enhances the real-time safety management and emergency response capabilities of substations, and has shown good application results with broad prospects for promotion and application. Attached Figure Description

[0019] Figure 1 This is a flowchart of a substation spatial video fusion and collaborative management method according to the present invention; Figure 2 This is a structural block diagram of a substation spatial video fusion and collaborative management system according to the present invention; Figure 3 This is a trajectory display diagram of the target monitoring during the operation of this invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Please see Figure 1 This embodiment provides a method for spatial video fusion and collaborative management of substations, including: S1. Collect substation data and perform preprocessing. Based on the preprocessed substation data, construct a three-dimensional spatial analysis model and generate a comprehensive database containing all equipment parameters of the substation, PTZ change matrix data, surface mapping data and spatial coding. Substation data is mainly used to build models. The more comprehensive the data, the more accurate the model. Substation data includes substation laser point cloud data scanned by LiDAR, geodetic coordinate benchmark data measured on-site by national land benchmark points, coordinates of substation equipment and building structures queried from substation drawings, and camera installation location and altitude data measured by GPS equipment. Preprocessing includes removing noise, outliers, and invalid acquisition points from the substation laser point cloud data using a combination of statistical and conditional filtering methods, resulting in noise-free discrete point cloud data. Then, an iterative nearest-point algorithm is used to perform point cloud data stitching and registration, accurately aligning and fusing the segmented point clouds from multiple regions to obtain the substation's global point cloud dataset. This data processing is a common procedure and will not be described in detail here. Finally, the spatial data from the substation data is converted to a unified national geodetic coordinate system and national elevation datum, and spatial data alignment and normalization are performed to obtain unified coordinate data. The process involves: 1) Gathering and integrating continuous point cloud surfaces; 2) Performing spatial surface interpolation fitting to fill in spatial gaps between discrete points, generating a normalized continuous point cloud surface; 3) Performing 3D meshing on the continuous point cloud surface to generate uniformly distributed mesh vertices and mesh patches, thus obtaining a 3D mesh model; 4) Performing zoning according to voltage and equipment function, i.e., dividing the substation into multi-level spatial units based on the substation voltage level and equipment function attributes, forming different voltage regions and equipment function zones; 5) Obtaining multi-level spatial units, assigning unique spatial codes to spatial units, cameras, and other equipment, resulting in a spatial code table, where other equipment refers to physical devices excluding cameras.

[0022] The three-dimensional spatial analysis model includes: an input layer, a feature point extraction and database construction layer, a camera intrinsic and extrinsic parameter calibration and solution layer, a gimbal dynamic mapping matrix iteration layer, a complex surface projection parameter calculation layer, and an output layer; The input layer takes into account the substation's global point cloud dataset, 3D mesh model, unified coordinate dataset, and spatial encoding table. The feature point extraction and database construction layer extracts the mesh vertices of the 3D mesh model, calculates the Gaussian curvature and average curvature of each vertex, and filters out curvature extrema. These curvature extrema correspond to key geometric locations such as corners, edges, and inflection points. Based on these key geometric locations, all candidate points can be filtered to construct a curvature candidate feature point set. A secondary filtering process is then performed on the curvature candidate feature points according to preset feature filtering rules, primarily to eliminate points with low stability, such as those found on lawns, ground surfaces, and equipment flat surfaces. A set of power feature points is obtained, which consists of power feature points with high stability, high precision, and easy identification characteristics, including the top of the equipment flange, the vertical corner of the building, the four corners of the equipment identification plate, the top of the lightning rod, the end point of the frame beam, and the edge of the equipment foundation. The coordinates of the power feature points are corrected by a sub-pixel coordinate refinement algorithm to obtain sub-pixel level high-precision feature coordinate data, i.e., high-precision power feature points. A unique feature code is assigned to each corrected power feature point, and the three-dimensional spatial coordinates, corresponding spatial units, and equipment information are associated to obtain a standard point set. The equipment information is the ID of the equipment. The feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, and the gimbal dynamic mapping matrix iteration layer sequentially perform feature extraction analysis, attitude calculation, and surface analysis processing. The feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, and the gimbal dynamic mapping matrix iteration layer sequentially perform feature extraction analysis, attitude calculation, and surface analysis processing, including: The camera intrinsic and extrinsic parameter calibration solution layer extracts two-dimensional feature points from the captured standard image to obtain a two-dimensional image feature set. It then performs matching between the three-dimensional feature points in the standard point set and the two-dimensional feature points in the two-dimensional image feature set to obtain a feature matching set. The matching results in the feature matching set are iteratively filtered using a random sampling consensus algorithm to remove all mismatched point pairs, resulting in a precise matching set. Based on the standard pinhole camera imaging model and using the data in the precise matching set, the basic optical parameters and spatial attitude parameters of the camera are initially solved using the PnP (perspective n-point problem) attitude solution algorithm. The camera's intrinsic and extrinsic parameter attitude parameters are then solved, and iterative optimization is performed using the bundle adjustment method to fit and obtain the radial and tangential distortion coefficients of the lens. This completes the precise correction of lens optical distortion, resulting in a parameter set for all cameras in the substation. The parameter set includes an intrinsic parameter matrix, a distortion coefficient set, an extrinsic parameter rotation matrix, extrinsic parameter translation, and camera calibration reprojection error. The precise formula for correcting lens optical distortion is: ; In the formula: The original horizontal coordinates of the image pixels under lens distortion conditions. The original vertical coordinates of the image pixels under lens distortion conditions; These are the precise lateral coordinates of the pixels after radial and tangential distortion correction. These are the precise vertical coordinates of the pixels after radial and tangential distortion correction; The horizontal pixel coordinates of the principal point of the image represent the horizontal position of the intersection of the lens optical axis and the imaging plane; The vertical pixel coordinates of the principal point of the image represent the vertical position of the intersection of the lens optical axis and the imaging plane; This represents the Euclidean radial distance between the distorted pixel and the principal point of the image. ; These are the first, second, and third order radial distortion coefficients of the lens. These are the first and second order tangential distortion coefficients of the lens; A standard image refers to a high-definition video captured by a camera that meets preset requirements. When performing two-dimensional feature point extraction, the ORB (Oriented Rotation Binary Feature) feature extraction and descriptor generation algorithm is used to perform two-dimensional feature point extraction on each frame of the video image, and at the same time generate a unique feature descriptor with rotation invariance and scale invariance, thus obtaining a two-dimensional image feature set.

[0023] The gimbal dynamic mapping matrix iteration layer performs 3D attitude calculation on the gimbal's horizontal and pitch rotation angles, generating horizontal rotation attitude matrices and pitch rotation attitude matrices. The gimbal's horizontal angle, pitch rotation angle, and optical zoom matrices are acquired in real time, resulting in high data accuracy. The two matrices are superimposed to obtain the attitude increment rotation matrix. The extrinsic parameter rotation matrix and extrinsic parameter translation are used as static references to update the extrinsic parameter, obtaining the real-time extrinsic parameter rotation matrix and real-time extrinsic parameter translation after attitude changes. The equivalent focal length parameter of the camera is corrected based on the gimbal's optical zoom matrices, generating a corrected intrinsic parameter matrix. When the gimbal attitude data changes, a real-time spatial change matrix is ​​generated, and the dynamic correction focal length and mapping consistency verification results are obtained simultaneously.

[0024] The complex surface projection parameter calculation layer calls the 3D mesh model and the 3D mesh point and surface structure data of the equipment. The surface equipment includes cylindrical and arc-shaped surface equipment such as transformers and combined electrical appliances, generating a set of surface equipment mesh coordinates. It calls the parameter set and real-time spatial change matrix of all cameras in the substation to build the spatial relationship between the camera view and the surface equipment. It performs distortion-free plane unfolding on the 3D surface of irregular equipment, establishes and corrects the bidirectional binding mapping relationship between the 3D surface spatial coordinates and the 2D video texture pixel coordinates, and obtains the texture interpolation coefficients, sampling step size and correction coefficient set of each surface. After integration, it generates a surface texture mapping parameter table, correction coefficient set, surface correspondence matrix and distortion-free projection rules.

[0025] Differentiated algorithms are used to perform distortion-free planar unfolding on the 3D surfaces of different types of equipment: a conformal parametric surface unfolding algorithm is used for the 3D surfaces of irregular equipment, a fixed cylindrical surface ratio correction algorithm is used for regular cylindrical equipment, and an adaptive sampling interpolation correction algorithm is used for irregular arc-shaped equipment. The texture interpolation coefficients, sampling step size and correction coefficients of each surface region are obtained by solving.

[0026] The output layer integrates the computational results parameters from the feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, the gimbal dynamic mapping matrix iteration layer, and the complex surface projection parameter calculation layer to form the original parameter set. Spatial coding is performed based on the spatial coding table, and the binding of device information and parameter labels is realized. The binding data is validated, and data that fails the validation is removed. After the data is formatted uniformly, a comprehensive database is constructed. The spatial coding adopts existing conventional coding rules.

[0027] S2. The national standard video stream decoding algorithm is used to perform unified transcoding on the real-time acquired multi-format video, and video pre-optimization is performed based on the comprehensive database. The optimized video is converted into dynamic texture, and projected onto the corresponding spatial position based on the region adaptation interpolation algorithm to obtain the video image base. The pre-processing optimization of the video includes frame segmentation, retrieving the distortion coefficient set, correcting the pixel coordinates of each image in the video based on the distortion coefficient set, and obtaining the corrected video frame. Based on the spatial coding table, the device information and spatial unit corresponding to the video are identified. After analyzing the device information, it is determined whether the spatial unit contains a curved device. If no curved device exists, no additional processing is performed. If a curved device exists, the device information is used to determine whether to call the curved texture mapping parameter table or the correction coefficient set to perform pre-interpolation correction, so as to pre-compensate the stretching distortion problem caused by the mapping of the three-dimensional curved surface to the two-dimensional plane, and obtain the video image base.

[0028] The optimized video is converted into a dynamic texture and projected onto the corresponding spatial location based on a region-adaptive interpolation algorithm. This process results in a video image base consisting of: converting the video image base into dynamic texture pixel resources adapted to a 3D curved surface, unifying the texture pixel format and sampling precision to eliminate projection misalignment caused by differences in image format and pixel specifications; classifying each region within the spatial unit into planar and curved regions, and using a bilinear interpolation texture mapping algorithm to perform uniform sampling and interpolation filling on the video texture pixels in planar regions; and using a surface interpolation mapping algorithm based on a surface texture mapping parameter table and correction coefficient set to perform texture adaptation projection on curved regions. Specifically, for planar regions, the bilinear interpolation texture mapping algorithm is called to perform uniform sampling and interpolation filling on the video texture pixels, ensuring a smooth texture without jagged edges or stretching deformation. For curved regions, the surface texture mapping parameter table and correction coefficient set are called, and texture adaptation projection is completed by combining the device's surface curvature and radius changes, offsetting the geometric deformation generated during the 2D unfolding of the 3D curved surface.

[0029] S3. Access multi-source data collected by the substation, and based on spatial coding and video footage, bind multi-source data, spatial location, and camera footage to build a data association system and store it in the comprehensive database; The substation collects multi-source data including equipment status data, environmental monitoring data, safety alarm data, and operation and maintenance data. Equipment status data includes equipment temperature sensing, partial discharge monitoring, and equipment operating conditions; environmental monitoring data includes temperature, humidity, smoke detection, wind speed, and rainfall; safety alarm data includes restricted area intrusion, equipment anomalies, and fire alarms; and operation and maintenance data includes inspection work orders, equipment ledgers, and personnel trajectories. The multi-source data collected from the substation is then formatted and time-series aligned sequentially. Based on spatial coding and video footage, the binding of multi-source data, spatial location, and camera footage is performed, including: using a spatial coding table with the equipment's unique code as the primary key, and aligning the time sequence... After alignment processing, multi-source data is matched with devices, spatial units, and cameras to clarify the spatial units and camera images corresponding to the multi-source data, and to construct the mapping relationship between data, spatial units, devices, and cameras. The binding of control positions with video images, device data with 3D spatial analysis models, and alarm data with spatiotemporal data is performed to obtain a data association system. That is, changes in data in a specified area can automatically trigger the retrieval of the corresponding area's monitoring video, and real-time operating data of devices can be mounted to the corresponding 3D device entities in real time to achieve dynamic updates of the model status. Alarm events can be automatically associated with the location of the event, the corresponding monitoring video, device information, and timestamp to complete the coupling of all elements in the spatiotemporal dimension.

[0030] S4. Collect external environmental data, combine it with the data association system, and use the ambient light and shadow adaptation method to perform partitioned video parameter adjustment on the video screen base to obtain a virtual-real fusion scene; External environmental data includes measured data from ambient light sensors at the substation, meteorological data, and time-series data. Combined with an environmental lighting and shadow adaptation method, video parameter adjustments are performed. This includes extracting real-time light intensity, day / night time, sunny / cloudy weather, and ambient color temperature from the external environmental data. Based on a preset lighting and shadow threshold rule library, corresponding brightness, contrast, saturation, and shadow compensation correction parameters are matched, and adaptive parameter adjustments are performed. The lighting and shadow threshold rule library adjusts global basic lighting and shadow parameters for day / night switching, fine-tunes color and color temperature parameters according to sunny / cloudy weather, and adapts local brightness and darkness compensation parameters for strong light, backlight, and low light scenes. This applies to the projection onto the surface of the 3D spatial analysis model. The system performs frame-by-frame pixel-level light and shadow adaptation correction on the dynamic video texture, unifying the brightness levels, color tones, and light and shadow transitions between the video image and the 3D spatial analysis model. It also calls an edge feathering transition algorithm to smoothly feather the splicing boundary between the video texture and the 3D spatial analysis model base, eliminating hard boundaries between virtual and real elements, light and shadow breaks, and abrupt color changes, achieving a uniform global light and shadow transition. Furthermore, it monitors ambient light and shadow fluctuations frame-by-frame, dynamically adjusting parameters to obtain a virtual-real fusion scene. Strict constraints are placed on the amplitude of light and shadow changes between frames to avoid screen flickering and color jumps, ensuring the continuity and stability of light and shadow timing during dynamic video refresh.

[0031] S5. Obtain operation instructions and parse the corresponding spatial code from the virtual-real fusion scene. Retrieve cameras covering the spatial code from the comprehensive database, filter cameras according to the set monitoring rules, and control the filtered cameras to perform focused monitoring. When the target moves, use the inter-frame differential target recognition algorithm to extract the dynamic target, combine it with the trajectory prediction algorithm to predict the target position, and perform focused monitoring. Obtain substation environmental data in real time and compare it with the preset standard rules. If it meets the preset standard rules, no action is taken. Otherwise, execute the abnormal alarm and generate an operation and maintenance work order. The preset standard rules are set according to historical data and consist of several data intervals. The corresponding data intervals are different in different seasons and regions, such as temperature intervals, humidity intervals, etc.

[0032] The system retrieves cameras covered by spatial coding from the comprehensive database as follows: Based on spatial coding, the corresponding camera set number is extracted from the mapping relationship between data, spatial units, equipment and camera devices, and then the optimal scheduling camera is selected according to the preset priority filtering rules; combined with the camera parameter set, as well as the spatial distance and three-dimensional size of the target corresponding to the instruction, the camera parameters are adjusted to achieve target-focused monitoring.

[0033] The priority filtering rules are as follows: first, invalid devices that are offline or faulty, have obstructed views, or have restricted permissions are removed. The specific priority ranking rules are: devices with unobstructed views have the highest priority, followed by devices with the best monitoring view and target alignment, then devices with the closest spatial distance to the target, and fixed camera devices have a higher priority than PTZ devices. Through multi-level rule weighted comparison, the optimal scheduling monitoring device is output to avoid the problems of redundant retrieval of multiple devices and chaotic images.

[0034] When the regulatory focus is on the target, a pop-up window displays real-time high-definition monitoring footage of the target area. The synchronous linkage model highlights the target device and its associated spatial unit, achieving linkage positioning between virtual and real scenarios. Information such as the target code, called device, parameter adjustment data, and scheduling time of this scheduling is recorded, and the scheduling log is archived to provide data support for subsequent operation and maintenance traceability, data statistics, and intelligent review.

[0035] When the target moves, the inter-frame differential target recognition algorithm is used to extract the dynamic target, and the trajectory prediction algorithm is combined to predict the target position. Focused monitoring is then performed, including: real-time identification of moving targets on site using the inter-frame differential method; calculation of the target's moving speed and direction based on the target coordinate data of continuous frames, and solving the target's spatial position at the next moment using the trajectory prediction formula; and after predicting the target's field of view switching area, triggering adjacent cameras in advance to complete the adaptation adjustment of the viewing angle and focal length, so as to achieve seamless relay tracking of the target and avoid target loss.

[0036] The inter-frame difference target recognition algorithm compares the differences in video pixels frame by frame and only performs re-rendering for dynamic areas where the pixel change exceeds a fixed threshold; for static areas where the pixels do not change, historical texture data is directly reused, thereby reducing the consumption of unnecessary computing power.

[0037] The trajectory prediction formula is: In the formula, and These are the two-dimensional spatial coordinates of the dynamic target at the current moment. and These are the instantaneous lateral and longitudinal movement velocities of the dynamic target, respectively. is the time interval between video frames, and is a fixed sampling period parameter; and These are the predicted spatial coordinates of the dynamic target at the next moment.

[0038] When in use, the programmatic resolution adaptation is performed based on the line-of-sight distance of the operation and maintenance: when the line-of-sight distance is less than 5m, the 4K high-definition mode is enabled; when the line-of-sight distance is between 5m and 20m, the 1080P standard mode is enabled; when the line-of-sight distance is greater than 20m, the 720P lightweight mode is enabled. This solution can stably support the synchronous fusion processing of 64 channels of high-definition video with an end-to-end latency of no more than 100ms.

[0039] like Figure 3 As shown, Figure 3 This is a trajectory display diagram for target monitoring during runtime. You can click on the camera in the diagram to view the video taken at the corresponding time. In addition, the black dot in the middle of the broken line in the diagram indicates the change of the target's direction, making it easy to understand the time point of the target's direction change.

[0040] In practical applications, the system monitors manual clicks on the 3D interface of operation and maintenance management in real time, capturing the coordinates of the target device or area selected by the user. Based on spatial coding, it performs point coordinate matching, parsing to obtain the unique device ID and spatial unit code corresponding to the target. After accurately locking the scheduling object, it performs monitoring device coverage matching retrieval: using the target spatial code as the retrieval key, it batch filters out all fixed cameras and PTZ cameras that can cover the target area, have no view obstruction, and are in normal online operation, generating a candidate monitoring device set. Invalid devices that are offline, have view obstruction, or have restricted permissions are removed to ensure the effectiveness and availability of the scheduling devices. After accurate sorting and optimization based on preset priority filtering rules, it outputs the optimal scheduling for a single device. Monitoring equipment: For the selected optimal PTZ monitoring equipment, based on the target spatial distance and the equipment's 3D dimensions, the optimal horizontal rotation angle, pitch angle, and optical zoom are matched. The PTZ viewpoint rotation and focal length adaptive adjustment are completed in milliseconds, accurately focusing on the core monitoring area of ​​the target equipment. No manual fine-tuning of the viewpoint and focal length is required, eliminating problems such as image offset, target too small a proportion, or image overflow. After the viewpoint and focal length are adapted, a pop-up window displays the real-time high-definition monitoring image of the target area, and the target equipment and its spatial area are highlighted in the 3D model simultaneously, realizing virtual and real scene linkage positioning. The system automatically records information such as the target code, called equipment, parameter adjustment data, and scheduling time for this scheduling, completing the scheduling log archiving and providing data support for subsequent operation and maintenance traceability, data statistics, and review.

[0041] In practical applications, the local edge server undertakes computational tasks that have strict requirements for low latency, such as video decoding, texture projection, PTZ control, and real-time alarm analysis, while the cloud data center undertakes offline computational tasks such as historical storage, data statistics, parameter archiving, and retrospective analysis.

[0042] After completing equipment modification or site changes, import the remeasured point cloud data of the changed area; compare the old and new point cloud data in a programmatic way to achieve accurate positioning of the changed range; only the spatial grid, spatial coding and local mapping parameters need to be updated for the changed area, without the need to carry out full-site model reconstruction and full recalibration.

[0043] Example 2: Please see Figure 2 A substation spatial video fusion and collaborative management system includes a model building module, a video processing module, a data processing module, a real-time adjustment module, and a real-time management module. Model building module: Collects substation data and performs preprocessing. Based on the preprocessed substation data, it builds a three-dimensional spatial analysis model and generates a comprehensive database containing all equipment parameters of the substation, PTZ change matrix data, surface mapping data and spatial coding. Video processing module: It adopts the national standard video stream decoding algorithm to perform unified transcoding processing on real-time acquired multi-format videos, and performs video pre-optimization based on a comprehensive database. The optimized video is converted into dynamic textures and projected to the corresponding spatial position based on the region adaptation interpolation algorithm to obtain the video image base. Data processing module: It receives multi-source data collected by the substation, performs binding of multi-source data, spatial location and camera footage based on spatial coding and video image base, and stores it in the integrated database; Real-time adjustment module: Collects external environmental data, combines it with a data association system, and uses an ambient light and shadow adaptation method to adjust the video parameters of the video frame base to obtain a virtual-real fusion scene; Real-time management module: It acquires operation commands, parses the corresponding spatial codes from the virtual-real fusion scene, retrieves cameras covering the spatial codes from the comprehensive database, filters cameras according to the set monitoring rules, and controls the filtered cameras to perform focused monitoring; when the target moves, it uses the inter-frame differential target recognition algorithm to extract the dynamic target, combines the trajectory prediction algorithm to predict the target position, and performs focused monitoring; it acquires substation environmental data in real time and compares it with preset standard rules. If it meets the preset standard rules, no action is taken; otherwise, an abnormal alarm is triggered, and a maintenance work order is generated.

[0044] During system operation, the calibration process will be automatically started on a fixed monthly cycle. The calibration process uses a lightweight approach to call the feature point extraction and library construction layer and the camera intrinsic and extrinsic parameter calibration solution layer. It calculates parameter offset errors by comparing real-time image features with the global feature library. Parameters that exceed the threshold range will be automatically corrected to compensate for accuracy loss caused by equipment loosening, lens aging, and positional offset, and can maintain sub-millimeter level fusion accuracy for a long time.

[0045] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware; whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0046] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for spatial video fusion and collaborative management of substations, characterized in that, include: Data from substations is collected and preprocessed. Based on the preprocessed substation data, a three-dimensional spatial analysis model is constructed, generating a comprehensive database containing all equipment parameters of the substation, gimbal change matrix data, surface mapping data, and spatial coding. The national standard video stream decoding algorithm is used to perform unified transcoding on real-time acquired multi-format videos, and video pre-optimization is performed based on a comprehensive database. The optimized video is converted into dynamic textures and projected onto the corresponding spatial positions based on a region adaptation interpolation algorithm to obtain the video image base. The system accesses multi-source data collected from substations, binds multi-source data, spatial location, and camera footage based on spatial coding and video footage, constructs a data association system, and stores it in a comprehensive database. By collecting external environmental data and combining it with a data association system, the ambient light and shadow adaptation method is used to adjust the video parameters of the video frame base to obtain a virtual-real fusion scene. The system acquires operation commands and parses the corresponding spatial codes from the virtual-real fusion scene. It then retrieves cameras covering the spatial codes from the comprehensive database, filters cameras according to the set monitoring rules, and controls the filtered cameras to perform focused monitoring. When a target moves, it uses an inter-frame differential target recognition algorithm to extract the dynamic target, combines it with a trajectory prediction algorithm to predict the target position, and performs focused monitoring. It acquires substation environmental data in real time and compares it with preset standard rules. If the data meets the preset standard rules, no action is taken; otherwise, an anomaly alarm is triggered, and a maintenance work order is generated.

2. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: The substation data includes laser point cloud data of the substation scanned by LiDAR, geodetic coordinate benchmark data measured on-site by national land benchmark points, coordinates of substation equipment and building structures obtained by querying substation drawings, and camera installation location and altitude data measured by GPS equipment. The preprocessing includes removing noise points, outliers, and invalid acquisition points from the substation laser point cloud data using a combination of statistical and conditional filtering. Point cloud data is then stitched and registered using an iterative nearest-point algorithm to obtain a global point cloud dataset for the substation. Spatial data from the substation is converted to a unified national geodetic coordinate system and national elevation datum to obtain a unified coordinate dataset and a continuous point cloud surface. A three-dimensional mesh is then created on the continuous point cloud surface to obtain a three-dimensional mesh model. Finally, the data is partitioned according to voltage and equipment function to obtain multi-level spatial units. A unique spatial code is assigned to each spatial unit, camera, and other device to obtain a spatial coding table.

3. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: The three-dimensional spatial analysis model includes: an input layer, a feature point extraction and database construction layer, a camera intrinsic and extrinsic parameter calibration and solution layer, a gimbal dynamic mapping matrix iteration layer, a complex surface projection parameter calculation layer, and an output layer; The input layer takes into account the substation's global point cloud dataset, 3D mesh model, unified coordinate dataset, and spatial coding table. The feature point extraction and database construction layer extracts the mesh vertices of the 3D mesh model, calculates the Gaussian curvature and average curvature of each mesh vertex, selects curvature extreme points, and constructs a curvature candidate feature point set. A secondary selection is performed on the curvature candidate feature points according to preset feature selection rules to obtain the power feature point set. After correcting the coordinates of the power feature points, a unique feature code is assigned to each corrected power feature point, and the 3D spatial coordinates, corresponding spatial units, and equipment information are associated to obtain the standard point set. The feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, and the gimbal dynamic mapping matrix iteration layer sequentially perform feature extraction analysis, attitude calculation, and surface analysis processing. The output layer integrates the computational results parameters from the feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, the gimbal dynamic mapping matrix iteration layer, and the complex surface projection parameter solution layer to form the original parameter set. Based on the spatial encoding table, spatial encoding is performed, and the binding of device information and parameter labels is realized. The binding data is validated, and data that fails the validation is removed. After the data is processed in a unified format, a comprehensive database is constructed.

4. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: The feature point extraction and database construction layer, the camera intrinsic and extrinsic parameter calibration and solution layer, and the gimbal dynamic mapping matrix iteration layer sequentially perform feature extraction analysis, attitude calculation, and surface analysis processing, including: The camera intrinsic and extrinsic parameter calibration solution layer extracts two-dimensional feature points from the captured standard image to obtain a two-dimensional image feature set; it then matches the three-dimensional feature points in the standard point set with the two-dimensional feature points in the two-dimensional image feature set to obtain a feature matching set; the matching results in the feature matching set are iteratively filtered using a random sampling consensus algorithm to obtain a precise pairing set; based on the standard pinhole camera imaging model and using the data in the precise pairing set as a basis, the camera intrinsic and extrinsic parameter attitude parameters are solved, and then iteratively optimized using the bundle adjustment method to obtain the parameter set of all cameras in the substation. The parameter set includes the intrinsic parameter matrix, distortion coefficient set, extrinsic parameter rotation matrix, extrinsic parameter translation, and camera calibration reprojection error. The gimbal dynamic mapping matrix iteration layer performs three-dimensional attitude calculation on the gimbal's horizontal and pitch rotation angles, generating horizontal rotation attitude matrices and pitch rotation attitude matrices. After superposition, the attitude increment rotation matrix is ​​obtained. The extrinsic parameter rotation matrix and extrinsic parameter translation are used as static references to update the extrinsic parameters, obtaining the real-time extrinsic parameter rotation matrix and real-time extrinsic parameter translation after attitude changes. The equivalent focal length parameter of the camera is corrected based on the gimbal's optical zoom, generating a corrected intrinsic parameter matrix. When the gimbal attitude data changes, a real-time spatial change matrix is ​​generated, and the dynamic correction focal length and mapping consistency verification results are obtained. The complex surface projection parameter calculation layer calls the 3D mesh model and the 3D mesh point and surface structure data of the equipment to generate the surface equipment mesh coordinate set; it calls the parameter set and real-time spatial change matrix of all cameras in the substation to construct the spatial relationship between the camera view and the surface equipment; it performs distortion-free plane unfolding on the 3D surface of irregular equipment, establishes and corrects the bidirectional binding mapping relationship between the 3D surface spatial coordinates and the 2D video texture pixel coordinates, and obtains the texture interpolation coefficients, sampling step size and correction coefficient set of each surface. After integration, it generates the surface texture mapping parameter table, correction coefficient set, surface correspondence matrix and distortion-free projection rules.

5. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: Performing video pre-optimization includes segmenting the video into frames, retrieving the distortion coefficient set, and correcting the pixel coordinates of each image in the video based on the distortion coefficient set to obtain the corrected video frames. Based on the spatial coding table, the device information and spatial units corresponding to the video are identified. After analyzing the device information, it is determined whether there is a curved device in the spatial unit. If there is no curved device, no additional processing is performed. If there is a curved device, the device information is combined to determine whether to call the curved texture mapping parameter table or the correction coefficient set to perform correction and obtain the video image base.

6. The method for substation spatial video fusion and collaborative management according to claim 5, characterized in that: The optimized video is converted into a dynamic texture and projected onto the corresponding spatial location based on a region-adaptive interpolation algorithm. The resulting video image base includes: converting the video image base into a dynamic texture pixel resource adapted to a 3D spatial curved surface, and unifying the texture pixel format and sampling precision; classifying each region within the spatial unit into types, including planar regions and curved regions, and using a bilinear interpolation texture mapping algorithm to perform uniform sampling and interpolation filling on the video texture pixels in the planar regions; and using a curved interpolation mapping algorithm to perform texture adaptation projection on the curved regions based on the curved texture mapping parameter table and correction coefficient set.

7. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: The multi-source data collected by the substation includes equipment status data, environmental monitoring data, safety alarm data, and operation and maintenance data; and the multi-source data collected by the substation is sequentially formatted and time-series aligned. Based on spatial coding and video footage, the binding of multi-source data, spatial location, and camera footage includes: using a spatial coding table and the device's unique code as the primary key, matching the time-aligned multi-source data with the device, spatial unit, and camera; clarifying the spatial unit and camera footage corresponding to the multi-source data; and constructing a mapping relationship between data, spatial unit, device, and camera. The binding of control location with video footage, device data with a 3D spatial analysis model, and alarm data with spatiotemporal context is performed to obtain a data association system.

8. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: External environmental data includes measured data from ambient light sensors at the substation site, meteorological data, and time-series data. Combined with the ambient light and shadow adaptation method, video parameter adjustment is performed by extracting real-time light intensity, day and night time, sunny and cloudy weather, and ambient color temperature from the external environmental data. Based on the preset light and shadow threshold rule library, corresponding brightness, contrast, saturation, and shadow compensation and correction parameters are matched, and adaptive parameter adjustment is performed. For the dynamic video texture projected onto the surface of the 3D spatial analysis model, perform frame-by-frame pixel-level light and shadow adaptation correction to unify the brightness levels, color tones, and light and shadow transitions between the video image and the 3D spatial analysis model; call the edge feathering transition algorithm to perform smooth feathering processing on the splicing boundary between the video texture and the 3D spatial analysis model base; monitor ambient light and shadow fluctuations frame by frame and perform dynamic parameter adjustments to obtain a virtual-real fusion scene.

9. The method for substation spatial video fusion and collaborative management according to claim 1, characterized in that: The system retrieves cameras covered by spatial coding from the comprehensive database as follows: Based on spatial coding, the corresponding camera set number is extracted from the mapping relationship between data, spatial units, equipment and camera devices, and then the optimal scheduling camera is selected according to the preset priority filtering rules; combined with the camera parameter set, as well as the spatial distance and three-dimensional size of the target corresponding to the instruction, the camera parameters are adjusted to achieve target-focused monitoring.

10. A substation spatial video fusion and collaborative management system, characterized in that: include: Model building module: Collects substation data and performs preprocessing. Based on the preprocessed substation data, it builds a three-dimensional spatial analysis model and generates a comprehensive database containing all equipment parameters of the substation, PTZ change matrix data, surface mapping data and spatial coding. Video processing module: It adopts the national standard video stream decoding algorithm to perform unified transcoding processing on real-time acquired multi-format videos, and performs video pre-optimization based on a comprehensive database. The optimized video is converted into dynamic textures and projected to the corresponding spatial position based on the region adaptation interpolation algorithm to obtain the video image base. Data processing module: It receives multi-source data collected by the substation, performs binding of multi-source data, spatial location and camera footage based on spatial coding and video image base, and stores it in the comprehensive database; Real-time adjustment module: Collects external environmental data, combines it with a data association system, and uses an ambient light and shadow adaptation method to adjust the video parameters of the video frame base to obtain a virtual-real fusion scene; Real-time management module: It acquires operation commands, parses the corresponding spatial codes from the virtual-real fusion scene, retrieves cameras covering the spatial codes from the comprehensive database, filters cameras according to the set monitoring rules, and controls the filtered cameras to perform focused monitoring; when the target moves, it uses the inter-frame differential target recognition algorithm to extract the dynamic target, combines the trajectory prediction algorithm to predict the target position, and performs focused monitoring; it acquires substation environmental data in real time and compares it with preset standard rules. If it meets the preset standard rules, no action is taken; otherwise, an abnormal alarm is triggered, and a maintenance work order is generated.

Images