Remote control method, device and equipment of security and protection system and storage medium

Abstract

The invention provides a remote control method and device for a security system, equipment and a storage medium, and the method comprises the steps: carrying out the feature extraction of security data of a fire alarm host, and obtaining a fire alarm scene vector; a preset linkage strategy library is matched according to the fire alarm scene vector, and a PTZ positioning instruction is obtained; controlling a camera to collect an alarm image based on the PTZ positioning instruction; based on a preset fire alarm analysis model, generating an alarm confirmation report according to the alarm condition image and the fire alarm scene vector; and notifying a plurality of preset user terminals to perform remote interaction according to the alarm confirmation report, and performing start-stop control on the fire-fighting equipment.

Description

Technical Field

[0001] The present application relates to the field of remote control technology, and in particular to a remote control method, device, equipment and storage medium for a security system. Background Art

[0002] Currently, security systems still face significant challenges in remote control capabilities and intelligence when responding to emergencies like fires. Existing remote control methods often focus on simple startup and shutdown or status monitoring of individual devices, lacking a unified, efficient coordination mechanism. Firefighting equipment control commands are often disconnected from interactive notifications from multiple terminals (such as security, firefighting, and management personnel). This results in delayed information transmission and decentralized decision-making, making it difficult to achieve rapid, coordinated, and coordinated control across multiple remote roles. This fragmented response model severely restricts the effectiveness of remote control of security systems at critical moments. Summary of the Invention

[0003] The present application provides a remote control method, apparatus, device and storage medium for a security system, which are used to improve the remote control efficiency of the security system.

[0004] In a first aspect, an embodiment of the present application provides a remote control method for a security system, the method comprising: Extract features from the security data of the fire alarm host to obtain the fire alarm scene vector; According to the fire alarm scene vector, a preset linkage strategy library is matched to obtain a PTZ positioning instruction; Controlling the camera to collect alarm images based on the PTZ positioning instructions; Based on a preset fire alarm analysis model, generating an alarm confirmation report according to the alarm image and the fire alarm scene vector; According to the confirmed alarm report, multiple preset user terminals are notified to perform remote interaction and start and stop control of fire-fighting equipment.

[0005] In a second aspect, an embodiment of the present application provides a remote control device for a security system, the remote control device for the security system comprising: Feature extraction module, used to extract features from the security data of the fire alarm host and obtain the fire alarm scene vector; An instruction generation module is used to match a preset linkage strategy library according to the fire alarm scene vector to obtain a PTZ positioning instruction; An image collection module is used to control the camera to collect alarm images based on the PTZ positioning instructions; A report generation module, configured to generate an alarm confirmation report based on the alarm image and the fire alarm scene vector based on a preset fire alarm analysis model; The remote control module is used to notify multiple preset user terminals to perform remote interaction according to the confirmed alarm report and to control the start and stop of the fire-fighting equipment.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device including a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the remote control method of the security system as described in any one of the embodiments of the present application when executing the computer program.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor implements a remote control method for a security system as described in any one of the embodiments of the present application.

[0008] The embodiment of the present application provides a remote control method for a security system, the method comprising: extracting features from the security data of a fire alarm host to obtain a fire alarm scene vector; matching a preset linkage strategy library according to the fire alarm scene vector to obtain a PTZ positioning instruction; controlling a camera to collect alarm images based on the PTZ positioning instruction; generating an alarm confirmation report according to the alarm image and the fire alarm scene vector based on a preset fire alarm analysis model; notifying multiple preset user terminals to remotely interact and start and stop firefighting equipment based on the alarm confirmation report. In the above method, by extracting the security data of the fire alarm host as a scene vector and intelligently matching the strategy library, accurate PTZ positioning instructions are generated, the initial response time is shortened, the real-time alarm image and the fire alarm scene vector are integrated for intelligent alarm confirmation analysis, and the remote interaction of multiple user terminals and the accurate start and stop control of firefighting equipment are synchronously triggered based on the automatically generated alarm confirmation report, thereby achieving rapid and orderly linkage across systems and multiple roles. In this way, not only the critical time window from alarm to disposal is compressed, but also resource allocation and coordination efficiency are optimized through data-driven decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic flow chart of a remote control method for a security system provided in an embodiment of the present application; Figure 2 A schematic block diagram of a remote control device for a security system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0011] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0012] The flowcharts shown in the drawings are only exemplary and do not necessarily include all contents and operations / steps, nor are they necessarily executed in the described order. For example, some operations / steps can be further decomposed, combined or partially merged, so that the actual execution order can be changed according to actual situations.

[0013] It should also be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application and the appended claims, unless otherwise clear from the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0014] It should be further understood that the term "and / or" used in the present application and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0015] Please refer to Figure 1 , Figure 1 is a schematic flowchart of a remote control method of a security system provided by an embodiment of the present application. As shown in Figure 1 , the specific steps of the remote control method of the security system include: S101-S105.

[0016] S101, feature extraction is performed on security data of a fire alarm host to obtain a fire alarm scene vector.

[0017] Exemplarily, multi-wire disc signals of the fire alarm host are read through an RS485 interface. Bus disc signals are collected using a CAN bus interface. Loop bus device signals are polled through a loop interface board, a query command is sent and a device response is received, and a device type and a current state field are analyzed.

[0018] Data cleaning is performed on the collected security data to remove duplicate records and outliers. A two-dimensional matrix of floors and fire zones is created based on the cleaned security data. The spatial distribution of alarms is calculated, and the number of alarms at the corresponding locations is recorded in the corresponding matrix elements. Alarm trends are analyzed based on the cleaned security data. A 30-second sliding window is used to calculate the growth rate of alarm devices and changes in alarm levels. Device associations are then constructed, with nodes representing alarm devices and edges representing the physical distance between devices. Edge weights are set based on historical linkage records.

[0019] The spatial distribution matrix, time series data, and device associations are converted into numerical features. Spatial features include the coordinates of the alarm center, diffusion radius, and device density; temporal features include the start time, duration, and rate of change; and association features include the number of associated devices and average association strength. All features are normalized to generate a 128-dimensional fire alarm scene vector, where each dimension corresponds to a specific scene feature attribute.

[0020] S102: Match the preset linkage strategy library according to the fire alarm scene vector to obtain a PTZ positioning instruction.

[0021] For example, load the linkage policy library file in XML format, use the DOM parser to traverse the policy nodes, and extract the policy ID, trigger conditions, and linkage device list. Match the fire alarm scene vector with the trigger conditions of each policy. The trigger conditions include three types of rules: threshold judgment, range check, and pattern matching. Calculate the cosine similarity between the scene vector and the predefined pattern. When the similarity exceeds 0.8, the pattern is considered to be matched successfully. Perform multi-rule combination judgment on the successfully matched pattern, build a judgment tree connected by logical operators, and recursively calculate the Boolean result of each node. Filter all policies with true trigger conditions, sort them by priority, and select the highest priority policy. Extract the camera configuration information from the selected policy, including the IP address, port number, and preset position number.

[0022] Query the PTZ parameter mapping table and obtain the corresponding horizontal angle, vertical angle, and zoom factor based on the preset position number. Set the PTZ movement speed, in-position wait time, and image acquisition parameters to generate a complete PTZ positioning command. The command specifies a capture resolution of 1920×1080, a frame rate of 25fps, and an H.264 encoding format.

[0023] S103: Control the camera to collect alarm images based on the PTZ positioning instruction.

[0024] For example, use the RTSP protocol to connect to the target camera and send a DESCRIBE request to obtain media description information. Establish an RTP transmission channel in TCP mode to avoid UDP packet loss. Send a GetStatus request to query the current PTZ position and calculate the difference from the target position. Execute the AbsoluteMove command to control the camera's movement, checking the movement status every 500 milliseconds.

[0025] Once the camera is in place, it sends a PLAY request to receive the video stream, parses the RTP packets, and sorts them by sequence number. It then identifies the NALU type in the H.264 stream and assembles the complete frame data for decoding. It also calculates image clarity metrics, including Laplace variance and edge density, and adjusts focus and exposure parameters if they do not meet the requirements.

[0026] Use frame difference analysis to identify dynamically changing areas within the image, and perform morphological processing to extract the outlines of moving targets. Alarm images are stored in chronological order, with file names containing timestamps, camera numbers, and alarm event IDs. Image metadata is saved, recording shooting parameters and associated alarm device information, providing complete image data for subsequent analysis.

[0027] S104: Based on a preset fire alarm analysis model, generate an alarm confirmation report according to the alarm situation image and the fire alarm scene vector.

[0028] For example, the alarm image is resized to 224×224 pixels and normalized. Image features are extracted using a convolutional layer. The convolutional layer uses a 3×3 convolution kernel to extract visual features such as edges, texture, and smoke patterns, outputting a 256-dimensional image feature vector. The image feature vector and the fire scene vector are fed into a fully connected layer to calculate feature similarity and anomaly deviation. A similarity matrix and a deviation matrix are constructed, and an attention mechanism is used to weight the association of important features. The fused features represent a comprehensive score encompassing similarity and deviation, generating a confidence score between 0 and 1. The fused features are fed into an LSTM recurrent network to analyze temporal changes and identify key time nodes. Fire spread characteristics, including spread speed, direction, and impact range, are extracted to construct a five-level risk situation prediction matrix. Fire equipment topology data is read, and graph convolutional layers are used to process equipment nodes and connections. The risk prediction is mapped to a topological hypergraph, and the activation priority of each device is calculated. Considering resource constraints such as water pressure and power, integer programming is used to determine the device call configuration framework. The encoder vectorizes the configuration framework, and the decoder generates equipment start and stop sequences and evacuation routes. Output alarm confirmation report in JSON format, including alarm situation summary, risk assessment, equipment plan and evacuation plan.

[0029] S105: Notify multiple preset user terminals to remotely interact based on the confirmed alarm report and start and stop the firefighting equipment.

[0030] For example, the system analyzes evacuation path information from confirmed alarm reports and determines the primary and secondary terminals based on the importance of the paths and the number of people covered. It analyzes the device start / stop sequence, identifies critical paths, and allocates control authority time windows. Each time window defines start and end times, permission levels, and a list of controllable devices. A virtual token object is created to manage control permissions. The token contains an ID, holder, and expiration date. Interaction commands are pushed to the primary and secondary terminals, respectively, with the interface layout adapted to the terminal type. After executing an operation, the terminal generates an interaction confirmation message containing a digital signature. A conflict detection matrix is ​​constructed to analyze operation commands from multiple terminals. The matrix records each terminal's operations within a 100-millisecond time slice. Timestamps, terminal permission levels, and command priority are extracted to identify conflicting operations targeting the same device. Conflicts are resolved using arbitration rules that prioritize primary control, timestamps, and device safety. The valid command set is sorted according to the device start / stop sequence, and redundant operations are merged to create a unified control sequence. The device protocol mapping table is queried to convert the control sequence into Modbus or BACnet protocol format. Protocol headers such as the device address and checksum are added, and the start / stop commands are sent through the corresponding communication interface. Collect equipment operating parameters to verify execution results and ensure that fire-fighting equipment completes start and stop control as expected.

[0031] The embodiment of the present application provides a remote control method for a security system, the method comprising: extracting features from the security data of a fire alarm host to obtain a fire alarm scene vector; matching a preset linkage strategy library according to the fire alarm scene vector to obtain a PTZ positioning instruction; controlling a camera to collect alarm images based on the PTZ positioning instruction; generating an alarm confirmation report according to the alarm image and the fire alarm scene vector based on a preset fire alarm analysis model; notifying multiple preset user terminals to remotely interact and start and stop firefighting equipment based on the alarm confirmation report. In the above method, by extracting the security data of the fire alarm host as a scene vector and intelligently matching the strategy library, accurate PTZ positioning instructions are generated, the initial response time is shortened, the real-time alarm image and the fire alarm scene vector are integrated for intelligent alarm confirmation analysis, and the remote interaction of multiple user terminals and the accurate start and stop control of firefighting equipment are synchronously triggered based on the automatically generated alarm confirmation report, thereby achieving rapid and orderly linkage across systems and multiple roles. In this way, not only the critical time window from alarm to disposal is compressed, but also resource allocation and coordination efficiency are optimized through data-driven decision-making.

[0032] In order to more clearly introduce the technical solution of the present application, the technical solution of the present application will be introduced through specific embodiments below. It should be noted that the specific embodiments are used to expand the technical solution of the present application, but are not intended to limit the present application.

[0033] In some embodiments, a preset fire alarm analysis model includes: a convolutional layer, a fully connected layer, a recurrent network, a graph convolutional layer, an encoder and a decoder. Based on the preset fire alarm analysis model, a confirmation report is generated according to the alarm image and the fire scene vector, including: extracting the image feature vector of the alarm image based on the convolutional layer; calculating the feature similarity matrix and the abnormal deviation matrix of the image feature vector and the fire scene vector based on the fully connected layer, generating a fused feature representation based on the feature similarity matrix and the abnormal deviation matrix, and the fused feature representation including: a confidence score; inputting the fused feature representation into the recurrent network to obtain key time nodes and fire spread characteristics, and determining a risk situation prediction matrix based on the key time nodes and the fire spread characteristics; obtaining topological relationship data of fire-fighting equipment, constructing a topological hypergraph corresponding to the risk situation prediction matrix and the topological relationship data through the graph convolutional layer, and determining the call configuration framework corresponding to the fire-fighting equipment based on the topological hypergraph and preset resource constraints; encoding the call configuration framework through the encoder to obtain a resource configuration encoding vector, decoding the resource configuration encoding vector through the decoder, generating equipment start and stop timing and personnel evacuation path, and outputting a confirmation alarm report based on the equipment start and stop timing and personnel evacuation path.

[0034] For example, a police image is input into a convolutional layer for feature extraction. The convolutional layer uses multiple 3×3 convolution kernels to scan the image. As each convolution kernel slides across the image, it calculates the product of the pixel value at the corresponding location and the kernel weight, generating an element of the feature map. The first convolution layer uses 64 convolution kernels to extract low-level features such as edges and corners. The second convolution layer uses 128 convolution kernels to extract mid-level features such as texture and local patterns based on the features of the previous layer. The third convolution layer uses 256 convolution kernels to extract high-level semantic features such as smoke outlines and flame shapes. After each convolution operation, the ReLU activation function is applied to zero negative values ​​and retain positive values, enhancing the nonlinear representation of the features. A max pooling operation is used to reduce the spatial size of the feature map, taking the maximum value within a 2×2 region as the output, reducing computational effort while preserving key features. The output of the final convolution layer is flattened into a one-dimensional vector to form the image feature vector.

[0035] The fully connected layer receives the image feature vector and the fire scene vector as input and computes the interaction between the two vectors across different dimensions through matrix multiplication. When constructing the feature similarity matrix, the inner product of each dimension of the image feature and each dimension of the scene vector is calculated, resulting in matrix elements representing the strength of feature association. The number of rows in the similarity matrix is ​​equal to the image feature dimension, and the number of columns is equal to the scene vector dimension. Larger values ​​in the matrix indicate stronger correlation between the corresponding features. The anomaly deviation matrix is ​​constructed by calculating the square of the feature differences, reflecting the degree to which the image features deviate from a normal fire scene. Each element of the similarity matrix is ​​multiplied by the inverse of the corresponding position deviation matrix element to obtain a fused value that comprehensively considers both similarity and anomaly. A weighted average of all fused values ​​is calculated, with weights preset based on feature importance, to generate a confidence score between 0 and 1. A confidence score close to 1 indicates a high degree of consistency between the image content and the fire scene, while a confidence score close to 0 indicates a possible false alarm. The fused feature representation not only contains confidence scores but also preserves the structural information of the feature matrix, providing input for subsequent time series analysis.

[0036] The recurrent network uses an LSTM structure to process the time series represented by the fused features. The fused features at the current moment and the hidden state at the previous moment are input into the LSTM unit. The forget gate determines how much historical information is retained, the input gate controls how much new information is accepted, and the output gate determines the output hidden state. The hidden state of the LSTM encodes the temporal pattern of the fire's development. Key time nodes are identified by analyzing the rate of change of the hidden state. When the rate of change exceeds a preset threshold, the moment is marked as a key node, such as the fire outbreak point or a point of rapid spread. Fire spread characteristics are obtained by calculating the difference between the fused features at adjacent moments. The direction vector of the feature change indicates the spread direction, the magnitude of the change indicates the spread speed, and the number of affected feature dimensions indicates the spread range. Based on the identified key time nodes, the future time is divided into multiple stages. The risk level of each stage is predicted based on the fire spread characteristics. The rows of the risk situation prediction matrix represent different spatial regions, and the columns represent future time points. The matrix elements are integers from 1 to 5, representing the risk level at the corresponding spatiotemporal location.

[0037] The topology relationship data of the fire-fighting equipment is read from the database, including the unique identification of each equipment, the equipment type, the installation position coordinates, the connection relationship with other equipment, etc. The equipment is abstracted as a node of a graph structure, and the node attributes include the basic information and the current state of the equipment. The physical connection or logical control relationship between the equipment is represented as an edge of the graph, and the weight of the edge reflects the closeness of the connection. The graph convolution layer propagates the feature of the topology graph, and each node aggregates the feature information of its neighbor nodes to update its own feature representation. The information of the risk situation prediction matrix is mapped to the graph node, and each equipment node is attached with the risk level attribute of the area where it is located. When constructing the topology hypergraph, multiple equipment related to the function of the hyperedge are created, such as all the sprinkler heads in the same fire zone connected by a hyperedge. The graph convolution operation propagates the risk information on the hypergraph, and the activation priority score of each equipment is calculated. The resource constraint conditions include the fire pool capacity, the power upper limit of the power supply system, the number limit of concurrent equipment operation, etc., which are represented as linear inequalities. The integer programming algorithm is used to maximize the equipment coverage range and response speed under the constraint conditions, and the optimal equipment calling combination is solved. The calling configuration framework includes the selected equipment list, the start parameter of each equipment, the start order dependency relationship between the equipment, etc.

[0038] The encoder converts the calling configuration framework into a fixed-length vector representation. The device ID in the equipment list is embedded and coded, mapping the discrete ID to a continuous vector space. The position encoding is used to preserve the order information of the equipment in the list, and the position encoding is added to the equipment embedding. The combined vector is nonlinearly transformed by a multilayer perceptron to extract the deep features of the configuration information. After concatenating the encoding vectors of all equipment, a dimension reduction operation is performed to generate a fixed-dimensional resource configuration encoding vector. The decoder uses a sequence generation method to convert the encoding vector into a specific control scheme. Each decoding step outputs a control instruction for one equipment, including the equipment ID, the start time, the running parameter, etc. The attention mechanism is used to make the decoder focus on different parts of the encoding vector when generating each instruction, ensuring that important information is not missed. The start and stop timing of the equipment is determined by a sorting algorithm, considering the preheating time of the equipment, the dependency relationship, the resource competition, etc. The personnel evacuation path is planned on the building plan by a graph search algorithm, with the starting point being the current position of the personnel and the ending point being the safety exit, avoiding high-risk areas during the search process. All the generated information is organized into a structured alarm report in JSON format for easy parsing and transmission.

[0039] In some embodiments, multiple preset user terminals are notified to conduct remote interaction based on the confirmed alarm report to control the start and stop of fire-fighting equipment, including: determining the main control terminal and the auxiliary terminal based on the personnel evacuation path, and determining the control authority time window based on the equipment start and stop timing; establishing a token transfer communication mechanism between the main control terminal and the auxiliary terminal based on the control authority time window; pushing interactive instructions of the control interface to the main control terminal and the auxiliary terminal respectively based on the token transfer communication mechanism; receiving interactive confirmation of the interactive instructions from the main control terminal and the auxiliary terminal, and arbitrating the interactive confirmation in real time based on the preset instruction conflict detection algorithm and the equipment start and stop timing to generate a unified control sequence; converting the unified control sequence into a start and stop instruction that complies with the communication protocol of the fire-fighting equipment, and controlling the corresponding fire-fighting equipment to be turned on or off through the start and stop instructions.

[0040] For example, the evacuation routes from confirmed alarm reports are parsed, and the starting point, end point, nodes along the way, and estimated evacuation time for each route are extracted. The personnel positioning system is queried to obtain the real-time number of people in each area, and the number of people covered by each route is calculated. A path importance score is calculated based on the number of people covered, path length, and the number of hazardous areas. The management terminal corresponding to the highest-scoring path is designated as the master control terminal, typically the main console in the fire control room. Other related terminals are designated as auxiliary terminals. The master control terminal has the highest control authority and can override the operational instructions of other terminals.

[0041] Analyze device start and stop sequence data to extract the start time, operating duration, and stop time of each fire protection device. This identifies dependencies between devices and distinguishes between sequential and parallel start groups. Based on the device timing distribution and terminal management scope, each terminal is assigned a control permission time window. This time window includes start and end timestamps, a list of controllable devices, and the level of operational permission. A buffer period is set between adjacent windows to facilitate permission switching.

[0042] Create a token object to manage control permissions, including a unique identifier, expiration date, the holding terminal ID, and a list of controllable resources. The initial token is assigned to the master terminal, and token transfer between terminals is achieved through a message queue. Each transfer is confirmed by a three-way handshake, and a timeout mechanism is set to prevent token loss. This token transfer communication mechanism ensures that each device is controlled by only one terminal at a time.

[0043] The control interface configuration is generated based on the terminal type, with large virtual buttons used on touch screens and a full menu interface on PCs. Interaction commands are formatted in JSON and include information such as the interface element type, location, bound device ID, and permitted operations. Full-duplex communication is established via WebSocket, with an independent session space allocated to each terminal. When pushing commands, metadata such as message type, priority, and timestamp is added to implement a message confirmation mechanism.

[0044] After executing an operation, the terminal generates an interactive confirmation message containing the operation type, target device, operation parameters, user identity, and a timestamp. The message is digitally signed using the private key, and the server verifies the signature using the public key. A conflict detection matrix is ​​constructed to record the operation history, with rows representing terminals and columns representing 100-millisecond time slices. The matrix is ​​scanned to identify conflicting operations on the same device within the same time slice.

[0045] Arbitration rules are applied to resolve conflicts: commands from the master terminal are executed first, commands with earlier timestamps are selected for identical permissions, and safe alternatives are selected for dangerous operations. Valid commands are sorted by device start / stop sequence, consolidating consecutive operations and eliminating redundant queries. The device protocol mapping table is queried to convert commands to Modbus or BACnet format. Fields such as the device address, transaction identifier, and checksum are added, and start / stop commands are sent through the corresponding interface. A timeout retry mechanism is implemented to ensure command delivery, with up to three retries.

[0046] In some embodiments, interactive confirmations are arbitrated in real time based on a preset instruction conflict detection algorithm and device start-stop timing to generate a unified control sequence, including: parsing the interactive confirmations to construct a conflict detection matrix, the conflict detection matrix including: timestamp, terminal authority level and instruction priority; based on the conflict detection matrix and device start-stop timing, identifying device operation instructions with time overlap in the interactive confirmations, and generating a list of conflicting instructions; sorting and screening the list of conflicting instructions according to the arbitration rules of master terminal priority, timestamp priority and device security priority to obtain a valid instruction set; sorting and logically reorganizing the valid instruction set according to the device start-stop timing to generate a unified control sequence.

[0047] For example, the interactive confirmation message is parsed to extract the timestamp, terminal permission level, and command priority. Timestamps are accurate to the millisecond level, and the NTP protocol is used to synchronize the clocks of all terminals to eliminate skew. Terminal permissions are categorized into three levels: master control, operation, and monitoring. Command priority is assigned based on the operation type, with emergency stop having the highest priority, parameter adjustment having a medium priority, and status query having the lowest priority. Operation instructions are stored by constructing a conflict detection matrix, with rows corresponding to terminal IDs and columns corresponding to 100-millisecond time slices. Timestamps are mapped to time slice indices, and detailed instruction information is stored in the corresponding locations. Matrix elements use linked lists to organize multiple operations within the same time slice, and data that expires within 5 minutes is regularly cleared. Each column in the conflict detection matrix is ​​traversed to examine all instructions within the time slice, grouping them by target device ID. When a device group contains multiple instructions, the conflict type is identified: different operation types are considered state conflicts, and different parameters are considered parameter conflicts. Device start and stop timing constraints are checked, and violations are marked as timing conflicts. Conflicting instructions and their types are recorded in a conflict instruction list.

[0048] Conflicts are handled using a three-level arbitration system: The terminal permissions of the command source are checked, and commands from the master terminal are automatically selected as valid. Timestamps are compared for commands with the same permission level, and the earliest command is selected. The impact of commands on the device is evaluated, and potentially damaging operations are rejected. Conflicts that cannot be automatically arbitrated are reported and require manual resolution. The arbitration process is recorded in an audit log.

[0049] Read the execution order of the device start and stop sequence and use quick sort to sort valid instructions. The primary key is the device sequence number, and the secondary key is the original timestamp to ensure that dependencies are met. Execution logic reorganization: merge continuous operations of the same device and delete redundant intermediate states. Check time intervals to avoid frequent starts and stops, and insert delays when necessary. Analyze resource usage to prevent overload and adjust the sequence to stagger execution. The generated unified control sequence contains an optimized instruction list, and each instruction has a clear execution time and conditions. Assign a version number and checksum to the sequence to ensure integrity. The sequence uses a linked list structure to facilitate dynamic adjustment, support the insertion of emergency instructions, and support the cancellation of pending instructions.

[0050] In some embodiments, feature extraction is performed on the security data of the fire alarm host to obtain a fire alarm scene vector, including: obtaining the multi-line disk signal, bus disk signal and loop bus device signal of the fire alarm host; parsing the multi-line disk signal, bus disk signal and loop bus device signal according to a preset signal protocol to obtain security data; extracting the spatial distribution of alarm situations, alarm situation change trends and device correlation of the security data to construct a fire alarm scene vector.

[0051] For example, the multi-disk output port of the fire alarm host is connected via the RS485 interface, and the serial port communication parameters are configured as 9600 baud rate, 8 data bits, 1 stop bit, and no parity bit. The multi-disk signal is transmitted in 16-byte packets, and the reading program receives the data packets in a loop and stores them in the buffer. The bus disk signal is collected using the CAN bus interface, the baud rate of the CAN controller is set to 250Kbps, and the receiving filter is configured to only receive fire alarm-related message IDs. The bus disk signal uses an extended frame format, and the 29-bit identifier contains the device type and address information. The loop bus is connected through a dedicated loop interface board. Each loop supports up to 127 device nodes, and the device status is queried one by one in a polling manner.

[0052] The preset signal protocol includes data frame format definitions, field meanings, and a verification algorithm. When parsing multi-line disk signals, the device address (bytes 1-2), signal type (bytes 3-4), detector number (bytes 5-8), and alarm value (bytes 9-12) are extracted from the 16-byte data packet. Bus disk signal parsing requires identifying the identifier field of the CAN message to extract the message priority, device type code, and specific address. Parsing loop bus device signals includes identifying the response frame's function code, device type, current status, and analog value. The parsed raw data is validated, and abnormal data that fails the CRC check and falls outside the normal range is eliminated.

[0053] The spatial distribution of alarms is extracted by constructing a two-dimensional matrix of floors and fire zones, where each matrix element records the number of alarm devices and the number of alarms at the corresponding location. All alarm records in the security data are traversed, and the values ​​of the corresponding matrix elements are updated based on the device installation location information. The center coordinates and diffusion radius of the alarm-intensive area are calculated to identify the spatial clustering characteristics of fire alarms. Alarm trends are analyzed through time series analysis, using a 30-second sliding window to calculate the number of alarm devices, alarm level distribution, and growth rate. The difference between adjacent time windows is calculated to determine whether the alarm is rising, stable, or declining. Device association is calculated by constructing a device relationship graph, where nodes represent alarm devices and edge weights are determined by historical linkage records and physical distance. Graph algorithms are used to calculate the degree centrality and betweenness centrality of each device node to identify key device nodes.

[0054] The extracted spatial distribution features, temporal trend features, and device association features are numerically encoded. Spatial features are encoded as 32-dimensional vectors, including parameters such as the coordinates of the alarm center, diffusion radius, device density, and distribution uniformity. Temporal features are encoded as 64-dimensional vectors, including parameters such as start time, peak time, duration, rate of change, and periodicity. Association features are encoded as 32-dimensional vectors, including parameters such as average association degree, maximum connected subgraph size, and number of key nodes. The three types of feature vectors are concatenated to form a 128-dimensional fire alarm scene vector. Each dimension is normalized to ensure that the value range is between 0 and 1.

[0055] In some embodiments, a preset linkage strategy library is matched according to the fire alarm scene vector to obtain a PTZ positioning instruction, including: based on the spatial features and alarm levels in the fire alarm scene vector, a matching linkage strategy template is retrieved from the preset linkage strategy library to filter out a candidate strategy set; according to the camera configuration information in the candidate strategy set and the preset PTZ parameter mapping table, the device identification and the corresponding preset position number of the target camera are determined, and the camera call parameters are generated; based on the camera call parameters, a PTZ positioning instruction is constructed, and the PTZ positioning instruction includes: camera identification, PTZ coordinates, focal length parameters and image acquisition mode.

[0056] For example, spatial feature dimensions are extracted from the fire alarm scene vector, including the center coordinates of the alarm area, coverage, and device density. Alarm level information is parsed, mapping continuous alarm intensity values ​​to discrete level labels (levels 1 to 5). Spatial features and alarm levels are combined to form a search key, which is used to find matching policy templates in the linkage policy library. The linkage policy library is organized using an index structure, supporting fast searches based on spatial range and alarm level. Each policy template contains fields such as trigger conditions, applicable scenarios, linkage device list, and execution priority. A range query is performed to identify all policies that spatially cover the alarm area, and then filtered based on alarm level. The similarity between the fire alarm scene vector and the policy template feature vector is calculated, using cosine similarity to measure the degree of match. Policies with a similarity exceeding a threshold (0.7) are added to the candidate policy set and sorted in descending order of similarity. The candidate policy set is traversed to extract the camera configuration information defined in each policy. This configuration information includes the camera IP address, port number, user authentication credentials, and a list of associated preset positions. Query the PTZ parameter mapping table, which uses preset position numbers as keys and stores corresponding horizontal angles, vertical angles, and zoom factors. Select the optimal preset position number based on the geometric relationship between the alarm location and the camera's coverage area. Calculate the azimuth and pitch angles from the alarm center to the camera to verify the rationality of the preset parameters. Determine the device identifier of the target camera and generate camera call parameters containing the IP address, preset position number, and PTZ parameters. Construct a PTZ positioning command compliant with the ONVIF protocol based on the camera call parameters. Create a SOAP message structure with a message header containing the camera address, message ID, and timestamp. The message body encapsulates the AbsoluteMove command, specifying the ProfileToken parameter to identify the video stream profile. Set the Position parameter, with the PanTilt field filled in with the horizontal and vertical angle values ​​obtained from the mapping table. Set the focal length parameter in the Zoom field, dynamically adjusting the zoom factor based on the size of the alarm area. Add the Speed ​​parameter to control the PTZ movement speed, setting high speed for emergency situations and medium speed for normal situations. The image acquisition mode parameters specify the resolution (1920×1080), frame rate (25fps), encoding format (H.264), and bit rate (4Mbps). The constructed PTZ positioning instructions are serialized into XML format and prepared to be sent to the camera.

[0057] In some embodiments, after notifying multiple preset user terminals to conduct remote interaction based on the confirmed alarm report and controlling the start and stop of the fire-fighting equipment, the method also includes: monitoring the execution effect of the fire-fighting equipment at the scene, comparing the execution effect with the expected effect corresponding to the confirmed alarm report, and generating an effectiveness report; sending the effectiveness report to the main control terminal and the auxiliary terminal respectively.

[0058] For example, after firefighting equipment starts and stops, various sensors are deployed to monitor on-site performance. For firefighting water pumps, current transformers are used to collect operating current, pressure sensors measure outlet pressure, and vibration sensors detect equipment vibration amplitude. Sprinkler system monitoring includes pipe network pressure sensors reading pressure values ​​at each node, flow meters calculating water volume, and level sensors monitoring water tank levels. Smoke exhaust fans use speed sensors to obtain actual speed, wind pressure sensors to measure duct pressure, and temperature sensors to monitor motor temperature. Fire doors and fire shutters use position sensors to confirm their closed position, and current sensors to determine whether the motors are functioning properly. Expected performance indicators defined in the alarm confirmation report are read, including target operating parameters, response time requirements, and coverage for each device. Actual monitoring data is compared against expected indicators item by item, and deviation rates and compliance status are calculated. For quantitative indicators such as pressure and flow, the relative percentage error between the actual value and the target value is calculated. For qualitative indicators such as device startup success or failure, the success rate and reasons for failure are recorded. Execution sequence compliance is analyzed to verify that devices are started in the scheduled order and time interval. Evaluate the linkage effect and verify whether the collaborative work of multiple devices can achieve the expected protection effect.

[0059] Generate a structured effectiveness report based on the comparative analysis results. The report header contains basic information such as the event ID, report generation time, and evaluation time range. The execution overview section counts the number of devices that successfully executed, a list of failed devices, and the overall compliance rate. The detailed evaluation section lists the expected parameters, actual parameters, deviation analysis, and problem diagnosis for each device. The timing analysis section displays a Gantt chart of device startup, annotating the time points of delays and anomalies. The risk assessment section identifies weak links in the execution process and provides improvement suggestions. The report is organized in JSON format to facilitate program parsing and human reading.

[0060] Effectiveness reports are pushed to the master control terminal via the WebSocket channel, ensuring the fire control room is kept informed of execution status. Upon receiving the report, the master control terminal displays key metrics and abnormality alerts on a large screen. Reports are also distributed to each auxiliary terminal, ensuring personnel in each position understand the effectiveness of execution within their jurisdiction. Report views are customized for different terminals, with the master terminal displaying global statistics and decision-making recommendations, while auxiliary terminals focus on specific equipment details. A report push confirmation mechanism is implemented, recording the receipt time and viewing status of each terminal. A report update mechanism is set up to automatically generate incremental updates and push them to relevant terminals when the on-site situation changes.

[0061] See also Figure 2 , Figure 21 is a schematic block diagram of a remote control device for a security system provided in an embodiment of the present application, wherein the remote control device 200 for the security system is used to execute the aforementioned remote control method for the security system. The remote control device 200 for the security system may be configured in a server.

[0062] Among them, the server can be an independent server, a server cluster, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0063] like Figure 2 As shown, the remote control device 200 of the security system includes: a feature extraction module 201 , an instruction generation module 202 , an image collection module 203 , a report generation module 204 and a remote control module 205 .

[0064] The feature extraction module 201 is used to extract features from the security data of the fire alarm host to obtain a fire alarm scene vector.

[0065] The instruction generation module 202 is used to match the preset linkage strategy library according to the fire alarm scene vector to obtain the PTZ positioning instruction.

[0066] The image collection module 203 is used to control the camera to collect alarm images based on the PTZ positioning instructions.

[0067] The report generation module 204 is used to generate an alarm confirmation report based on the preset fire alarm analysis model, the alarm situation image and the fire alarm scene vector.

[0068] The remote control module 205 is used to notify multiple preset user terminals to perform remote interaction according to the confirmed alarm report and to control the start and stop of the fire-fighting equipment.

[0069] An embodiment of the present application provides an electronic device, which includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement a remote control method for a security system such as any one of the embodiments of the present application when executing the computer program.

[0070] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor implements a remote control method for a security system as described in any one of the embodiments of the present application.

[0071] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present application, and such modifications or substitutions should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A remote control method for a security system, characterized in that: The method comprises: Extract features from the security data of the fire alarm host to obtain the fire alarm scene vector; According to the fire alarm scene vector, a preset linkage strategy library is matched to obtain a PTZ positioning instruction; Controlling the camera to collect alarm images based on the PTZ positioning instructions; Based on a preset fire alarm analysis model, generating an alarm confirmation report according to the alarm image and the fire alarm scene vector; According to the confirmed alarm report, multiple preset user terminals are notified to perform remote interaction and start and stop control of fire-fighting equipment.

2. The remote control method for a security system according to claim 1, wherein: The preset fire alarm analysis model includes: a convolutional layer, a fully connected layer, a recurrent network, a graph convolutional layer, an encoder, and a decoder. The preset fire alarm analysis model is used to generate a confirmed alarm report according to the alarm image and the fire alarm scene vector, including: Extracting an image feature vector of the police image based on the convolutional layer; Based on the fully connected layer, a feature similarity matrix and an abnormal deviation matrix of the image feature vector and the fire scene vector are calculated, and a fused feature representation is generated according to the feature similarity matrix and the abnormal deviation matrix, wherein the fused feature representation includes: a confidence score; Inputting the fused feature representation into the recurrent network to obtain key time nodes and fire spread characteristics, and determining a risk situation prediction matrix based on the key time nodes and the fire spread characteristics; Obtaining the topological relationship data of the fire-fighting equipment, constructing a topological hypergraph corresponding to the risk situation prediction matrix and the topological relationship data through the graph convolution layer, and determining a call configuration framework corresponding to the fire-fighting equipment based on the topological hypergraph and preset resource constraints; The call configuration framework is encoded by the encoder to obtain a resource configuration coding vector, and the resource configuration coding vector is decoded by the decoder to generate an equipment start-stop sequence and a personnel evacuation path, and an alarm confirmation report is output according to the equipment start-stop sequence and the personnel evacuation path.

3. The remote control method for a security system according to claim 2, wherein: The method of notifying multiple preset user terminals to remotely interact with the alarm report and start and stop the firefighting equipment includes: Determine the main control terminal and auxiliary terminal according to the personnel evacuation path, and determine the control authority time window according to the start and stop sequence of the equipment; Based on the control authority time window, establishing a token passing communication mechanism between the master terminal and the auxiliary terminal; Based on the token passing communication mechanism, push interactive instructions of the control interface to the main control terminal and the auxiliary terminal respectively; receiving interactive confirmations of the interactive instructions from the master terminal and the auxiliary terminal, performing real-time arbitration on the interactive confirmations based on a preset instruction conflict detection algorithm and the device start / stop sequence, and generating a unified control sequence; The unified control sequence is converted into a start / stop instruction that complies with the communication protocol of the fire-fighting equipment, and the corresponding fire-fighting equipment is controlled to be turned on or off through the start / stop instruction.

4. The remote control method for a security system according to claim 3, wherein: The method of performing real-time arbitration on the interactive confirmation based on a preset instruction conflict detection algorithm and the device start-stop timing to generate a unified control sequence includes: Parsing the interaction confirmation to construct a conflict detection matrix, wherein the conflict detection matrix includes: a timestamp, a terminal authority level, and a command priority; Based on the conflict detection matrix and the device start and stop sequence, identifying device operation instructions with time overlap in the interactive confirmation, and generating a conflict instruction list; Sorting and screening the conflicting instruction list according to arbitration rules of master terminal priority, timestamp priority, and device security priority to obtain a valid instruction set; The valid instruction set is sorted and logically reorganized according to the device start and stop timing to generate a unified control sequence.

5. The remote control method for a security system according to claim 1, wherein: The feature extraction of the security data of the fire alarm host to obtain the fire alarm scene vector includes: Obtaining the multi-line disk signal, bus disk signal and loop bus device signal of the fire alarm host; Analyze the multi-line disk signal, the bus disk signal and the loop bus device signal according to a preset signal protocol to obtain security data; The spatial distribution of alarms, alarm change trends, and device correlations of the security data are extracted to construct a fire alarm scene vector.

6. The remote control method for a security system according to claim 1, wherein: The method of matching a preset linkage strategy library according to the fire alarm scene vector to obtain a PTZ positioning instruction includes: Based on the spatial features and alarm levels in the fire alarm scene vector, searching for matching linkage strategy templates in the preset linkage strategy library to screen out a candidate strategy set; Determine the device identifier of the target camera and the corresponding preset position number according to the camera configuration information in the candidate strategy set and the preset PTZ parameter mapping table, and generate camera call parameters; Based on the camera call parameters, a PTZ positioning instruction is constructed, where the PTZ positioning instruction includes: camera identification, PTZ coordinates, focal length parameters, and image acquisition mode.

7. The remote control method for a security system according to claim 3, wherein: After notifying multiple preset user terminals to remotely interact with the fire-fighting equipment according to the confirmed alarm report and controlling the start and stop of the fire-fighting equipment, the method further includes: Monitor the performance of the fire-fighting equipment at the scene, compare the performance with the expected performance corresponding to the alarm confirmation report, and generate an effectiveness report; The validity report is sent to the main control terminal and the auxiliary terminal respectively.

8. A remote control device for a security system, characterized in that: The remote control device of the security system is used to execute the remote control method of the security system according to any one of claims 1 to 7, and the remote control device of the security system includes: Feature extraction module, used to extract features from the security data of the fire alarm host and obtain the fire alarm scene vector; An instruction generation module is used to match a preset linkage strategy library according to the fire alarm scene vector to obtain a PTZ positioning instruction; An image collection module is used to control the camera to collect alarm images based on the PTZ positioning instructions; A report generation module, configured to generate an alarm confirmation report based on the alarm image and the fire alarm scene vector based on a preset fire alarm analysis model; The remote control module is used to notify multiple preset user terminals to perform remote interaction according to the confirmed alarm report and to control the start and stop of the fire-fighting equipment.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and implement the remote control method for the security system according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor enables the processor to implement the remote control method of the security system according to any one of claims 1 to 7.

Images