A home fire dual-direction early warning method based on internet of things

Abstract

The application discloses a kind of based on Internet of Things's family fire two-way early warning method, it is related to intelligent fire-fighting technical field, including, establish family early warning subarea and intelligent home equipment binding relationship, and call two-order exponential smoothing recurrent neural network time series extraction mode processing subarea fire signal, generate subarea state single;Evidence theory fire arbitration is carried out using subarea state single, and fire grade is divided in combination with personnel risk state, generates fire arbitration set;Subarea state single and fire arbitration set are encapsulated as two-way early warning message, and through single data frame, first data frame, continuous data frame and flow control frame are carried out sequence verification and frame loss retransmission, generate two-way early warning frame chain;Through two-way early warning frame chain carries out fuzzy logic fire signal mode identification and stacked converter encoder early warning arrangement, generates terminal early warning arrangement table.The application is identified by fuzzy logic fire signal mode, improves the accuracy of family fire two-way early warning, communication reliability and emergency disposal pertinence.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent fire protection, in particular to a two-way early warning method for household fire based on the Internet of Things. Background Art

[0002] With the development of Internet of Things communication, smart home control and household fire monitoring technologies, household fire alarm devices have gradually evolved from a single smoke alarm to the direction of multi-source perception, remote notification and linkage disposal. Existing household fire alarm schemes usually collect environmental abnormal information through smoke detectors, temperature sensors, carbon monoxide sensors, combustible gas detectors and other devices, and realize alarm prompt, remote push and basic linkage by means of home gateways, user terminals, community alarm receiving terminals or smart speakers. Some smart home scenarios also combine intelligent door locks, gas cut-off valves, intelligent circuit breakers, lighting devices, audible and visual alarms and other actuating devices to perform audible and visual prompting, gas shut-off, local power cut and escape assistance after a fire alarm, thereby forming a fire early warning and emergency disposal system based on the home Internet of Things for communication.

[0003] However, the existing household fire early warning technologies mostly rely on single-point detection signals or simple threshold triggering, and it is difficult to fully express the temporal relationship between smoke, temperature, gas and changes in adjacent rooms, which is easy to confuse cooking smoke, steam disturbance, short-term gas fluctuations with real fires. At the same time, existing alarm push usually focuses on one-way notification, and there is a lack of unified message organization and reliable return mechanism among alarm content, personnel position, door lock status and equipment disposal status. When frame loss, delay or terminal unacknowledgment occurs in the home Internet of Things link, user confirmation, alarm receiving confirmation and personnel retention review information are difficult to be consistent with the original fire evidence, which further affects the accuracy, traceability and linkage disposal reliability of household fire early warning. Summary of the Invention

[0004] The present invention is proposed in view of the above problems existing in the prior art.

[0005] Therefore, the present invention provides a two-way early warning method for household fire based on the Internet of Things to solve the problems of misjudgment of multi-source signals for household fire and insufficient reliability of two-way alarm confirmation.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides a two-way early warning method for household fire based on the Internet of Things, which comprises,

[0008] establishing a binding relationship between household early warning partitions and smart home devices, and calling a second-order exponential smoothing recurrent neural network time series extraction method to process partition fire signals to generate a partition status sheet;

[0009] Fire arbitration based on evidence theory is conducted using a zoned status sheet, and fire severity levels are classified in conjunction with personnel risk status to generate a fire arbitration set;

[0010] The partition status sheet and fire arbitration set are encapsulated into a two-way early warning message, and sequence verification and frame loss retransmission are performed through single data frames, first data frames, continuous data frames and flow control frames to generate a two-way early warning frame chain.

[0011] Fuzzy logic fire signal pattern recognition and stacked converter encoder early warning arrangement are performed through bidirectional early warning frame chain to generate terminal early warning arrangement table;

[0012] Based on the terminal warning scheduling table, user confirmation, alarm confirmation, and smart home response are triggered, and the status of personnel staying is verified by infrared and thermal imaging fusion human body detection, generating warning response results.

[0013] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the establishment of the binding relationship between the home early warning zone and smart home devices specifically involves,

[0014] The family warning zones are divided according to the functional areas of the family, and each family warning zone is assigned a zone code to generate a family zone relationship table.

[0015] Fire detection equipment, personnel detection equipment, door lock equipment, gas equipment, circuit breaker equipment, lighting equipment, and alarm notification equipment are bound to the corresponding zone codes to limit the source of fire signals and the objects to be dealt with, and an early warning zone binding table is generated.

[0016] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the generation of the partition status sheet specifically involves,

[0017] Smoke signals, temperature signals, carbon monoxide signals, and combustible gas signals are arranged in chronological order according to the warning zone of the same household to generate a zone fire signal sequence;

[0018] Trend extraction is performed on the fire signal sequence of the region, and it is divided into short-term disturbance segments, continuous rise segments, gas-preceding segments and heat-preceding segments to generate a set of time series segments of the region.

[0019] According to the household early warning zone, the time sequence fragment set of the zone, abnormal room, adjacent room, personnel detection status, door lock status, gas status, circuit breaker status, lighting status and executable device status are divided into zones and matched to form a zone status sheet.

[0020] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, the method of using zoned status sheets for evidence-based fire arbitration specifically involves...

[0021] Extract smoke evidence, heat evidence, gas evidence, and diffusion evidence from the zoning status sheet to generate a fire evidence group;

[0022] The fire evidence group is defined according to the time sequence fragment set of the partition, so that short-term disturbance fragments correspond to false alarm evidence, continuous rise fragments correspond to local fire evidence, gas-preceding fragments correspond to gas leak evidence, and heat-preceding fragments correspond to electrical overheating evidence, thus generating a time sequence evidence group.

[0023] Evidence theory synthesis rules are used to synthesize chronological evidence groups, and the fire category is determined among false alarms, fires awaiting confirmation, local fires, spreading fires, and emergency fires, generating a basic set for fire arbitration.

[0024] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the generation of the fire arbitration set specifically involves,

[0025] Extract the personnel detection status and door lock status from the partition status sheet, determine the room where the personnel are located, the personnel's tendency to stay, and the door lock escape status, and generate the personnel risk status;

[0026] The risk status of personnel is used to correct the level of the fire arbitration base set. When the personnel are in the same room as the fire room and the door is locked, it corresponds to an emergency fire. When the personnel are in the same room as the adjacent affected room, it corresponds to a spreading fire, thus generating a fire level sequence.

[0027] The abnormal rooms corresponding to the fire type are identified as fire rooms, and the adjacent rooms corresponding to the spreading fire are identified as adjacent affected rooms, thus generating the fire room relationship;

[0028] The consistency of the fire severity sequence, the relationship between fire-affected rooms, and the corresponding time-series fragments of the fire evidence group is verified to generate a fire arbitration set.

[0029] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the encapsulation of the partition status sheet and the fire arbitration set into a two-way early warning message specifically involves,

[0030] Extract the fields of fire room, adjacent affected room, room where people are located, personnel detection status, door lock status, smart home device status and time segment from the partition status sheet and the relationship between fire rooms to generate a partition early warning field group;

[0031] Extract the fire severity sequence and fire evidence set from the fire arbitration set to generate an arbitration field set;

[0032] Based on the fire severity level sequence, determine the smart home handling field, user confirmation field, and alarm confirmation field, and determine the two-way early warning message number according to the fire room, the room where the person is located, the time segment field, and the fire severity level sequence to generate the confirmation handling field group;

[0033] The partitioned early warning field group, arbitration field group, and confirmation and handling field group are encapsulated into the same message payload to generate a two-way early warning message with a two-way early warning message number.

[0034] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the generation of the two-way early warning frame chain specifically comprises,

[0035] The payload length of the two-way early warning message is detected. When the payload length meets the single-frame carrying condition, the two-way early warning message is encapsulated into a single data frame. When the payload length exceeds the single-frame carrying condition, the initial payload of the two-way early warning message is encapsulated into the first data frame, and the remaining payload is divided into continuous data frames according to the sending order to generate the initial early warning frame group.

[0036] Write the frame sequence number to the consecutive data frames in the initial warning frame group, and store the frame sequence number, frame payload and bidirectional warning message number into the frame buffer to generate a warning frame group to be verified.

[0037] Verify the continuity of the frame sequence number of the early warning frame group to be verified. When the frame sequence number is continuous, splice the frame payload to generate a reconstructed early warning message. When there is a break in the frame sequence number, write the missing frame sequence number and the abnormal frame sequence number into the flow control frame to generate a retransmission request frame.

[0038] Based on the retransmission request frame, extract the corresponding consecutive data frames from the frame buffer, and limit the retransmission batch by the data block size in the flow control frame to generate a selective retransmission frame group.

[0039] The selective retransmission frame group is filled back into the early warning frame group to be verified, and the continuity of the reconstructed early warning message is verified again according to the frame sequence number to obtain the frame payload sequence that has passed the verification.

[0040] The bidirectional early warning message is reconstructed by combining the single data frame payload, the first data frame payload, and the consecutive data frame payloads in the frame payload sequence, and a bidirectional early warning frame chain is generated.

[0041] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the fuzzy logic fire signal pattern recognition through the two-way early warning frame chain specifically involves...

[0042] Fire signal identification data is generated by extracting the fire room, adjacent affected rooms, fire level, zonal time sequence fragment set and fire evidence group from the two-way early warning frame chain.

[0043] The fire signal identification data is classified into categories based on smoke persistence, temperature persistence, signal fluctuation, time segment type, and changes in adjacent rooms. The fire signal patterns are identified in categories such as cooking smoke, electrical overheating, gas leak, open flame spread, and people entrapment, and pattern recognition results are generated.

[0044] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the generation of the terminal early warning arrangement table specifically comprises,

[0045] The pattern recognition results, the room where the fire occurred, the adjacent affected rooms, the rooms where people were located, the door lock status, the smart home device status, and the set of time-series fragments of the partition were constructed into an early warning arrangement sequence according to the time order.

[0046] The stacked converter encoder is used to extract the room impact order, personnel risk order and equipment handling order in the early warning arrangement sequence, and to determine the next affected room, the priority notification terminal and the priority execution device;

[0047] A terminal warning scheduling table is generated based on the order of execution of notification terminals and smart home devices according to the next affected room, priority notification terminals, and priority execution devices.

[0048] As a preferred embodiment of the IoT-based two-way home fire early warning method of the present invention, wherein: the generation of early warning and response results specifically includes,

[0049] According to the terminal early warning scheduling table, user confirmation, alarm confirmation, alarm prompt, gas cut-off, partial power outage, door lock escape and lighting indication are triggered, and feedback messages and equipment handling records are generated.

[0050] When the pattern recognition result is a person staying, the user terminal does not return confirmation content, or the community alarm terminal requests verification of the person's location, the infrared human body contour information and thermal imaging heat source distribution information in the person detection state are extracted, and spatial registration is performed according to the family early warning zone to generate a person staying verification map.

[0051] In the personnel retention verification map, the overlapping areas of human body contour information and heat source distribution information are selected, and the room where the personnel are located, the number of personnel, and the location near the doors and windows are determined by combining the fire room, adjacent affected rooms, and door lock status, and the personnel retention verification results are generated.

[0052] The consistency of the personnel stay verification results, feedback messages, equipment handling records, fire arbitration set and terminal early warning arrangement table is checked, and flow control frames are retransmitted for feedback messages and personnel stay verification results that have sequence breakpoints, and early warning handling results are generated.

[0053] The beneficial effects of this invention are as follows: By binding home early warning zones with smart home devices, the disconnect between fire location and response targets in traditional single-point alarms is avoided; by using a second-order exponential smoothing recurrent neural network for time-series extraction, short-term disturbance segments, continuously rising segments, gas-preceding segments, and heat-preceding segments are distinguished, and then combined with evidence theory fire arbitration and personnel risk status correction of fire level, differentiated judgments can be formed between situations such as cooking smoke, gas leaks, electrical overheating, open flame spread, and personnel entrapment. Compared with alarm methods triggered by a single threshold, the false alarm suppression capability and fire category identification capability are stronger; by using single data frames, first data frames, continuous data frames, and flow control frames for sequence verification and selective retransmission, the complete correspondence between alarm content, confirmation content, and response records can be maintained even when there are frame losses in the alarm communication link; by combining fuzzy logic fire signal pattern recognition, stacked converter encoder early warning orchestration, and infrared and thermal imaging fusion human body verification, the notification terminal, smart home devices, and personnel entrapment confirmation can be executed collaboratively according to the fire development sequence, improving the accuracy, communication reliability, and targeted emergency response of two-way home fire early warning. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of a two-way early warning method for household fires based on the Internet of Things.

[0056] Figure 2 This is a schematic diagram for the arbitration of fire evidence and the generation of a two-way early warning frame chain.

[0057] Figure 3 This diagram illustrates fire signal pattern recognition, early warning programming, and personnel stay verification.

[0058] Figure 4 A schematic diagram illustrating the binding of home early warning zone devices and the timing analysis of zone fire signals. Detailed Implementation

[0059] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0061] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0062] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a two-way early warning method for home fires based on the Internet of Things, including the following steps:

[0063] S1. Establish the binding relationship between the home early warning zone and smart home devices, and call the second-order exponential smooth recurrent neural network time series extraction method to process the zone fire signal and generate the zone status sheet.

[0064] S1.1. Divide the family into early warning zones according to family functional areas. Family functional areas include the kitchen, living room, bedroom, corridor, entrance hall, and electrical distribution box area. The family early warning zones are divided based on room boundaries, passage boundaries, and electricity and gas usage boundaries. The kitchen corresponds to the gas usage area, the living room and bedroom correspond to the areas where people stay, the corridor and entrance hall correspond to the evacuation passage areas, and the electrical distribution box area corresponds to the electricity risk area. Assign a unique zone code to each family early warning zone. Establish a correspondence between the zone code and the family functional area name, adjacent family early warning zones, and door lock locations to generate a family zone relationship table.

[0065] Fire detection equipment, personnel detection equipment, door lock equipment, gas equipment, circuit breaker equipment, lighting equipment, and alarm notification equipment are bound to their corresponding zone codes. Fire detection equipment includes smoke detectors, temperature sensors, carbon monoxide sensors, and combustible gas detectors; personnel detection equipment includes human presence sensors, infrared human body detectors, and thermal imaging human body detectors; door lock equipment is smart door locks; gas equipment is gas shut-off valves; circuit breaker equipment is smart circuit breakers; lighting equipment is entryway lighting equipment; and alarm notification equipment includes audible and visual alarms and smart speakers.

[0066] After the devices are bound together, a warning zone binding table is formed, which includes the zone code, device category, device location, fire signal source, response target, and neighboring household warning zones. This table limits the fire signals that each household warning zone can collect and the response targets that can be executed, preventing cross-room confusion between the fire signal source and the response targets of smart home devices.

[0067] S1.2. When any household warning zone generates a smoke signal, temperature signal, carbon monoxide signal, combustible gas signal, personnel detection status, door lock status, gas status, circuit breaker status, lighting status, and executable device status, first determine the corresponding zone code according to the warning zone binding table, and then arrange the smoke signal, temperature signal, carbon monoxide signal, and combustible gas signal according to the time sequence of the same household warning zone to generate a zone fire signal sequence.

[0068] In the fire zone signal sequence, the smoke signal is used to reflect changes in particulate fire conditions, the temperature signal is used to reflect changes in heat, and the carbon monoxide and combustible gas signals are used to reflect changes in gas risk.

[0069] A second-order exponential smoothing recurrent neural network (RNN) is used to extract trends from the fire signal sequences in different zones. The second-order exponential smoothing originates from time series smoothing prediction methods, while the recurrent neural network originates from time series feature extraction methods. In this embodiment, only the second-order exponential smoothing RNN is used to classify the trend state of continuous fire signals without altering the physical meaning of smoke, temperature, carbon monoxide, and combustible gas signals.

[0070] Specifically, the signal changes between adjacent moments in the fire signal sequence of the partition are smoothed while preserving the direction of change for consecutive moments; when the signal only increases suddenly for a short period of time and then falls back, it is classified as a short-term disturbance segment; when the signal continuously increases in time sequence without falling back, it is classified as a continuous rising segment; when the carbon monoxide signal and combustible gas signal appear before the smoke signal, it is classified as a gas-preceding segment; when the temperature signal appears before the smoke signal, it is classified as a heat-preceding segment, thus generating a partition time sequence segment set.

[0071] S1.3. According to the home early warning zone, the time sequence fragment set of the zone is matched with the abnormal room, adjacent room, personnel detection status, door lock status, gas status, circuit breaker status, lighting status and executable device status to form a zone status sheet.

[0072] An abnormal room is a family warning zone where a short-term disturbance segment, a continuously rising segment, a gas-first segment, or a heat-first segment appears in the fire signal sequence. Adjacent rooms are family warning zones that are adjacent to the abnormal room in the family zoning relationship table. Personnel detection status is obtained from human presence sensors, infrared human detectors, and thermal imaging human detectors. Door lock status is obtained from smart door locks. Gas status is obtained from gas shut-off valves. Circuit break status is obtained from smart circuit breakers. Lighting status is obtained from entryway lighting equipment. Executable device status is obtained from gas shut-off valves, smart circuit breakers, entryway lighting equipment, audible and visual alarms, and smart speakers. A zoning status sheet consists of a zoning code, a zoning time sequence segment set, an abnormal room, adjacent rooms, personnel detection status, door lock status, gas status, circuit break status, lighting status, and executable device status.

[0073] By using a zone status sheet to confine fire signals, personnel status, and smart home device status within the same household early warning zone chain, it can directly serve as the basis for subsequent fire arbitration and two-way early warning message encapsulation.

[0074] S2. Fire arbitration based on evidence theory is conducted using a zoned status sheet, and fire severity is classified in conjunction with personnel risk status to generate a fire arbitration set.

[0075] S2.1. Extract the partition code, partition time sequence fragment set, abnormal room, adjacent room, personnel detection status, door lock status, gas status, circuit breaker status, lighting status, and executable device status from the partition status sheet, and determine the home early warning partitions that require fire arbitration according to the partition code. Smoke signals correspond to smoke evidence, temperature signals correspond to heat evidence, carbon monoxide signals and combustible gas signals correspond to gas evidence, and the following changes in smoke and temperature signals in adjacent rooms correspond to diffusion evidence. Smoke evidence, heat evidence, gas evidence, and diffusion evidence are combined to generate a fire evidence group. By limiting the alarm content of different fire detection devices to the home early warning partition through the fire evidence group, compared with the method of triggering an alarm based on only a single smoke detector, it can provide verifiable multiple sources of evidence for two-way home fire early warning.

[0076] The fire evidence group is constrained according to the time-series fragment set of the partitions. Short-term disturbance fragments correspond to false alarm evidence, indicating that smoke, temperature, carbon monoxide, and combustible gas signals suddenly increase and then decrease within a short period of time. Continuously rising fragments correspond to localized fire evidence, indicating that smoke and temperature signals continuously increase chronologically within the abnormal room. Gas-preceding fragments correspond to gas leak evidence, indicating that carbon monoxide and combustible gas signals appear before smoke signals. Heat-preceding fragments correspond to electrical overheating evidence, indicating that temperature signals appear before smoke signals. After constraining the evidence in the partitioned time-series fragment set, the fire evidence group is converted into a time-series evidence group, so that fire arbitration is not only based on whether a fire signal appears, but also on the sequence and change pattern of the fire signal.

[0077] S2.2. Evidence theory synthesis rules are used to synthesize the time-series evidence group. The evidence theory synthesis rules adopt the classic rules of multi-evidence synthesis in Dempster-Shafer evidence theory, which are used to combine the support results of smoke evidence, heat evidence, gas evidence and diffusion evidence for different fire categories.

[0078] Specifically, when smoke evidence corresponds to a short-term disturbance segment and the heat, gas, and diffusion evidence do not support a local fire, gas leak, electrical overheating, or diffuse fire, the fire type is determined to be a fire pending confirmation; when smoke and heat evidence correspond to a continuously rising segment, the fire type is determined to be a local fire; when gas evidence corresponds to a gas-preceding segment, the fire type is determined to be a local fire related to a gas leak; when heat evidence corresponds to a heat-preceding segment, the fire type is determined to be a local fire related to electrical overheating; when diffusion evidence corresponds to a continuously rising segment in an adjacent room, the fire type is determined to be a diffuse fire; when the fire evidence set only corresponds to a short-term disturbance segment and does not form a continuously rising segment, gas-preceding segment, or heat-preceding segment, the fire type is determined to be a false alarm.

[0079] Once the fire type is determined, a fire arbitration base set is generated. The fire arbitration base set includes a fire evidence group, a chronological evidence group, and a fire type. By distinguishing between fires to be confirmed, localized fires, and spreading fires through the fire arbitration base set, the problem of mixed triggering of cooking fumes, steam, and actual fires in ordinary household alarms can be reduced.

[0080] S2.3. Extract the personnel detection status and door lock status from the partition status sheet, determine the room where the person is located, the person's tendency to linger, and the door lock escape status, and generate a personnel risk status. The personnel detection status is obtained from the human presence sensor, infrared human detector, and thermal imaging human detector; the room where the person is located is the home warning partition corresponding to the personnel detection status; the person's tendency to linger is the personnel risk content formed when the room where the person is located overlaps with an abnormal room or an adjacent room; the door lock escape status is obtained from the open and closed status of the smart door lock.

[0081] After the personnel risk status is generated, the fire arbitration no longer only describes the fire itself, but also reflects whether the personnel are in the fire room, adjacent affected rooms, and door lock restricted environments, providing a source for the user confirmation field and alarm confirmation field in the subsequent two-way early warning message.

[0082] The fire arbitration base set is corrected by using the risk status of personnel. The situation where the personnel are in the same room as the fire room and the door is locked corresponds to an emergency fire. The situation where the personnel are in the same room as the adjacent affected room corresponds to a spreading fire. The situation where the personnel are in a room that is not in the fire room or the adjacent affected room and the door is locked maintains the fire category in the fire arbitration base set.

[0083] After the level correction is completed, a fire level sequence is generated. The fire level sequence is arranged according to false alarm, fire to be confirmed, local fire, spreading fire and emergency fire, and retains the partition code, fire evidence group and partition time sequence fragment set corresponding to each fire level.

[0084] S2.4. Identify the abnormal rooms corresponding to the fire type as fire rooms, and identify the adjacent rooms corresponding to the spreading fire as adjacent affected rooms, thus generating fire room relationships. Fire room relationships include the fire room, adjacent affected rooms, the zone code corresponding to the fire room, the zone code corresponding to the adjacent affected rooms, and the door lock location. After the fire room relationships are generated, local fires, spreading fires, and emergency fires in the fire level sequence all have clearly defined spatial objects, avoiding the problem that subsequent two-way early warning messages only contain the alarm level but lack room locations.

[0085] The consistency of the fire severity sequence, the relationship between fire-affected rooms, and the corresponding temporal fragment set of the fire evidence set is verified. Localized fires need to be consistent with the continuously rising, gas-preceding, or heat-preceding fragments within the fire-affected room; spreading fires need to be consistent with changes in smoke or temperature signals in adjacent affected rooms; emergency fires need to be consistent with the room where personnel are located, the door lock escape status, and the relationship between the fire-affected rooms; and fires awaiting confirmation need to be consistent with short-term disturbance fragments and thermal, gas, and spreading evidence that does not support localized fires. After the consistency verification is passed, the fire severity sequence, the relationship between fire-affected rooms, the fire evidence set, the temporal fragment set of the partition, and the personnel risk status together form the fire arbitration set.

[0086] S3. Encapsulate the partition status sheet and fire arbitration set into a two-way early warning message, and perform sequence verification and frame loss retransmission through single data frames, first data frames, continuous data frames and flow control frames to generate a two-way early warning frame chain.

[0087] S3.1. Extract the fields of fire room, adjacent affected room, room where personnel are located, personnel detection status, door lock status, smart home device status and time segment from the partition status sheet and the relationship between fire rooms, and generate a partition warning field group.

[0088] Among them, the fire room is used to indicate the home warning zone corresponding to the type of fire that occurred, the adjacent affected room is used to indicate the adjacent home warning zone corresponding to the spread of the fire, the room where the person is located is used to indicate the home warning zone corresponding to the person detection status, the door lock status is used to indicate the open and closed status of the smart door lock, the smart home device status is used to indicate the executable status of gas equipment, circuit breaker equipment, lighting equipment and alarm prompting equipment, and the time sequence segment field is used to retain short-term disturbance segments, continuous rise segments, gas-first segments and heat-first segments.

[0089] Through the partitioned early warning field group, the two-way early warning message can simultaneously carry the fire location, personnel location, and status of executable equipment. Compared with the home alarm method that only sends the alarm level, the alarm communication content has verifiable spatial objects and objects to be dealt with.

[0090] Fire severity levels and evidence sets are extracted from the fire arbitration database to generate arbitration field sets. The fire severity levels include the zone codes corresponding to false alarms, pending confirmation fires, localized fires, spreading fires, and emergency fires. The fire evidence sets include smoke evidence, heat evidence, gas evidence, and diffusion evidence. The arbitration field sets and the zone warning field sets share the same zone code, ensuring that each fire severity level corresponds to a specific fire-affected room and adjacent affected rooms, avoiding inconsistencies between fire severity levels and home warning zones in two-way warning messages.

[0091] S3.2. Determine the smart home response field, user confirmation field, and alarm confirmation field based on the fire severity level sequence. The user confirmation field corresponds to a fire pending confirmation; the smart home response field corresponds to a localized fire; the user confirmation field and the alarm confirmation field correspond to a spreading fire; and the smart home response field and the alarm confirmation field correspond to an emergency fire. The smart home response field records the execution content corresponding to gas cut-off, partial power outage, door lock escape, lighting indication, and alarm prompt; the user confirmation field records the confirmation content that the user terminal needs to return; and the alarm confirmation field records the confirmation content that the community alarm receiving terminal needs to return. The smart home response field, user confirmation field, and alarm confirmation field form a confirmation response field group, which enables two-way alarm messages to have a two-way communication structure for sending alarm content and receiving confirmation content.

[0092] The two-way early warning message number is determined based on the fire room, the room where the personnel are located, the time segment field, and the fire level sequence. The two-way early warning message number is formed by sequentially arranging the partition code corresponding to the fire room, the partition code corresponding to the room where the personnel are located, the segment type corresponding to the time segment field, and the fire category corresponding to the fire level sequence. The two-way early warning message number and the confirmation and handling field group are included in the same message payload, ensuring that subsequent single data frames, first data frames, consecutive data frames, and flow control frames can all correspond to the same two-way early warning message, avoiding the problem of feedback messages and retransmission frame groups not matching the original alarm content.

[0093] The zonal early warning field group, arbitration field group, and confirmation and handling field group are encapsulated into a single message payload to generate a two-way early warning message with a two-way early warning message number. The message payload is arranged in the order of two-way early warning message number, zonal early warning field group, arbitration field group, and confirmation and handling field group, so that the two-way early warning message first indicates the message's attribution, and then indicates the fire spatial object, the basis for the fire level, and the terminal confirmation content.

[0094] S3.3. Detect the payload length of the two-way warning message and compare it with the maximum payload of the data frame used in the alarm communication link. The maximum payload of the data frame is the message payload length that a single data frame can carry after removing the frame type field and the check field. The maximum payload of the data frame is determined by the data frame format of the alarm communication link. When the payload length of the two-way warning message is less than or equal to the maximum payload of the data frame, the two-way warning message is encapsulated into a single data frame. When the payload length of the two-way warning message is greater than the maximum payload of the data frame, the initial payload of the two-way warning message is encapsulated into the first data frame, and the remaining payload is divided into consecutive data frames according to the transmission order to generate the initial warning frame group.

[0095] It should be noted that the payload length comparison originates from the frame length determination rules in the data communication field. The payload length of a two-way pre-alarm message is the byte length after encapsulating the partitioned early warning field group, arbitration field group, and confirmation and handling field group. A single data frame is used to carry two-way pre-alarm messages with a payload length less than or equal to the maximum payload of the data frame. The first data frame is used to carry the initial payload of two-way pre-alarm messages with a payload length greater than the maximum payload of the data frame. Continuous data frames are used to carry the remaining payload after the initial payload. By distinguishing between single data frames, the first data frame, and continuous data frames, short messages for pending fire confirmation and long messages for emergency fires can be covered, preventing long messages from being truncated in home IoT alarm communication.

[0096] Frame sequence numbers are written to consecutive data frames in the initial warning frame group, and the frame sequence numbers, frame payloads, and bidirectional warning message numbers are stored in the frame buffer to generate a warning frame group to be verified. The frame sequence numbers are written sequentially according to the transmission order of the consecutive data frames, the frame payload is the message segment carried by the consecutive data frames, and the bidirectional warning message number indicates the bidirectional warning message to which the consecutive data frames belong. The frame buffer stores the correspondence between the frame sequence numbers, frame payloads, and bidirectional warning message numbers, enabling the location of consecutive data frames that need to be retransmitted when frame loss occurs, without needing to retransmit the complete bidirectional warning message.

[0097] S3.4. Verify the continuity of the frame sequence number of the early warning frame group to be verified. Collect consecutive data frames belonging to the same two-way early warning message according to the two-way early warning message number, and read the frame sequence number of the consecutive data frames in the order of arrival. When the frame sequence number of the first consecutive data frame is consistent with the starting sequence number after the first data frame, and the frame sequence number of the next consecutive data frame is equal to the frame sequence number of the previous consecutive data frame plus one, the frame sequence number is determined to be continuous. When the frame sequence number of the next consecutive data frame is greater than the frame sequence number of the previous consecutive data frame plus one, the frame sequence number is determined to have a breakpoint.

[0098] When frame sequence numbers are consecutive, the frame payloads in the alarm frame group to be verified are concatenated in ascending order of frame sequence number to generate a reconstructed alarm message. When there is a break in the frame sequence number, the frame sequence numbers between the previous consecutive data frame's sequence number plus one and the next consecutive data frame's sequence number minus one are identified as missing frame sequence numbers, and the frame sequence numbers of consecutive data frames arriving after the break point are identified as abnormal frame sequence numbers. The missing frame sequence number and the abnormal frame sequence number are written into the flow control frame to generate a retransmission request frame. The missing frame sequence number is used to indicate consecutive data frames that have not arrived, and the abnormal frame sequence number is used to indicate consecutive data frames that have arrived but are not in the consecutive order. The flow control frame does not carry new fire information, but rather carries retransmission request information, enabling the alarm communication link to provide feedback for missing frames.

[0099] Based on the retransmission request frame, the corresponding consecutive data frames are extracted from the frame buffer, and the retransmission batch is limited by the data block size in the flow control frame to generate a selective retransmission frame group. The data block size is the number of consecutive data frames allowed to be retransmitted in a single instance. The data block size is derived from the transmission control field of the flow control frame to prevent consecutive retransmission data frames from filling the home IoT alarm communication link at once. The selective retransmission frame group only contains consecutive data frames corresponding to missing frame sequence numbers and abnormal frame sequence numbers. Compared with the whole packet retransmission method, it can reduce the repeated transmission of two-way early warning messages in high-concurrency fire notification scenarios.

[0100] S3.5. The selective retransmission frame group is backfilled into the pre-warning frame group to be verified, and the continuity of the reconstructed pre-warning messages is verified again according to the frame sequence number to obtain a verified frame payload sequence. The frame payload sequence includes single data frame payload, first data frame payload, and continuous data frame payload; the single data frame payload directly forms a verified frame payload sequence when the single frame carrying condition is met, while the first data frame payload and continuous data frame payload form a verified frame payload sequence after retransmission backfilling and continuity verification are completed. This re-verification ensures that the user confirmation field, alarm confirmation field, and smart home handling field are not missing due to frame loss.

[0101] The bidirectional early warning message is reconstructed by combining the single data frame payload, the first data frame payload, and the consecutive data frame payloads in the frame payload sequence to generate a bidirectional early warning frame chain. The bidirectional early warning frame chain retains the correspondence between the bidirectional early warning message number, frame sequence number, frame payload, zone early warning field group, arbitration field group, and confirmation and handling field group. The bidirectional early warning frame chain serves as input for subsequent fire signal pattern recognition, early warning transmission sequence determination, and smart home linkage orchestration, enabling subsequent steps to continue based on the complete fire-affected room, adjacent affected rooms, personnel location rooms, fire severity sequence, user confirmation field, and alarm receipt confirmation field.

[0102] S4. Perform fuzzy logic fire signal pattern recognition and stacked converter encoder early warning arrangement through bidirectional early warning frame chain to generate terminal early warning arrangement table.

[0103] S4.1. Extract the fire-affected room, adjacent affected rooms, fire level sequence, zonal time sequence fragment set, and fire evidence group from the two-way early warning frame chain to generate fire signal identification data. The fire-affected room is used to determine the home early warning zone where the fire signal pattern occurs; adjacent affected rooms are used to determine whether the fire signal has spread to adjacent home early warning zones; the fire level sequence is used to determine the level and location of false alarms, pending confirmation fires, localized fires, spreading fires, and emergency fires; the zonal time sequence fragment set is used to determine short-term disturbance fragments, continuously rising fragments, gas-preceding fragments, and heat-preceding fragments; and the fire evidence group is used to provide smoke evidence, heat evidence, gas evidence, and diffusion evidence. By simultaneously incorporating fire level, fire location, and signal timing into the fire signal identification process, compared to pushing alarm content only based on fire level, subsequent alarm communication content has a clearer source of the fire signal pattern.

[0104] A fuzzy logic-based fire signal pattern recognition method is employed to classify fire signal identification data into membership categories. This method utilizes existing fuzzy logic techniques for membership classification and rule matching. Smoke persistence is determined by the continuously rising segments corresponding to smoke evidence; temperature persistence by the continuously rising segments corresponding to heat evidence; signal fluctuations by short-term disturbance segments; time-series segment types by the partitioned time-series segment set; and changes in adjacent rooms are determined by diffusion evidence and adjacent affected rooms. Membership classification does not alter the physical meaning of smoke, temperature, carbon monoxide, and combustible gas signals; it simply assigns fire signal identification data to corresponding fire signal patterns, avoiding the problem of directly determining fire signal patterns based on the anomaly of a single detection device.

[0105] S4.2. Determine the fire signal pattern based on the affiliation classification results of the fire signal identification data, and generate pattern recognition results. If the fire room belongs to the kitchen, the fire level sequence corresponds to a fire to be confirmed, the fire evidence group mainly consists of smoke evidence, and the partitioned time sequence fragment set corresponds to a short-term disturbance fragment, then the fire signal pattern is determined to be cooking smoke type. If the fire room belongs to the electrical distribution box area, the fire level sequence corresponds to a local fire, and the heat evidence corresponds to a heat-preceding fragment, then the fire signal pattern is determined to be electrical overheating type. If the fire room belongs to the kitchen, and the gas evidence corresponds to a gas-preceding fragment, then the fire signal pattern is determined to be gas leak type. If the fire level sequence corresponds to a spreading fire, and the spreading evidence is consistent with the persistence of smoke or temperature in adjacent affected rooms, then the fire signal pattern is determined to be open flame spreading type. If the fire level sequence corresponds to an emergency fire, and the room where the personnel are located overlaps with the fire room or adjacent affected rooms, then the fire signal pattern is determined to be personnel entrapment type.

[0106] The pattern recognition results include the fire signal pattern, the room on fire, the adjacent affected rooms, and the corresponding fire level sequence, enabling subsequent terminal warning programming to be based on the specific fire signal pattern, instead of using a uniform alarm prompt content.

[0107] S4.4. Construct an early warning orchestration sequence by arranging the pattern recognition results, the room experiencing the fire, adjacent affected rooms, the room where people are located, door lock status, smart home device status, and the set of time-series segments of each zone in chronological order. The early warning orchestration sequence is arranged according to the time when the fire signal appears in the room experiencing the fire, the time when adjacent affected rooms show a corresponding change, the time when the room where people are located generates a personnel detection status, the time when the door lock status changes, and the time when the smart home device status becomes executable. By placing the fire signal pattern, the room's impact process, the personnel's risk location, and the executable device status into the same time chain through the early warning orchestration sequence, the alarm notification and the smart home device's response sequence are prevented from becoming disconnected.

[0108] A stacked converter encoder is used to extract the room impact order, personnel risk order, and equipment handling order from the early warning orchestration sequence to determine the next affected room, the priority notification terminal, and the priority execution device. The stacked converter encoder uses existing methods of position encoding and self-attention calculation in the sequence data. Position encoding is used to maintain the sequential position in the early warning orchestration sequence, and self-attention calculation is used to identify the temporal dependencies between the fire room, adjacent affected rooms, rooms where personnel are located, door lock status, and smart home device status.

[0109] Specifically, the records arranged in chronological order in the early warning sequence are converted into a sequence vector matrix. Each record includes at least the pattern recognition result, the room where the fire occurred, adjacent affected rooms, the room where the person is located, the door lock status, the smart home device status, and the encoded content corresponding to the partitioned time sequence segment set; the stacked converter encoder processes the sequence vector matrix. Perform a linear mapping to obtain the query matrix. Key matrix Sum matrix The attention output matrix is ​​obtained by using the scaled dot product self-attention method from the classic Transformer encoder. The expression is:

[0110] ;

[0111] in, The attention output matrix represents the sequence representation after the association calculation of each record in the early warning arrangement sequence. This represents the query matrix, used to indicate the matching requests of the current record for other records; The key matrix represents the features that allow other records to be matched. The value matrix represents the output content corresponding to the matched record. Indicates transpose; Key matrix The dimension of each record vector in the vector; This represents the normalized exponential function, used to... Convert to an attention matrix.

[0112] Based on the attention output matrix The warning sequence is read as follows: After the record corresponding to the room with the fire, the association strength order of the records corresponding to adjacent affected rooms is read to obtain the room impact order; after the record corresponding to the room where the person is located, the association strength order of the records corresponding to the fire room, adjacent affected rooms, and door lock status is read to obtain the personnel risk order; after the record corresponding to the pattern recognition result, the association strength order of the records corresponding to the smart home device status is read to obtain the device handling order. The room impact order is determined by the change order from the fire room to the adjacent affected rooms; the personnel risk order is determined by the correspondence between the room where the person is located, the fire room, adjacent affected rooms, and door lock status; and the device handling order is determined by the correspondence between the smart home device status and the fire signal pattern.

[0113] The next affected room is the adjacent affected room immediately following the room with the fire in the room impact sequence. The priority notification terminals are the user terminals, family member terminals, and community alarm receiving terminals corresponding to the personnel risk sequence. The priority execution devices are the gas equipment, circuit breaker equipment, lighting equipment, door lock equipment, and alarm notification equipment that need to be executed first in the equipment handling sequence.

[0114] S4.5. Generate a terminal warning arrangement table according to the execution order of the notification terminals and smart home devices based on the next affected room, priority notification terminals, and priority execution devices.

[0115] When the pattern recognition result is cooking smoke and the fire level sequence corresponds to the fire to be confirmed, the terminal warning scheduling table first arranges the user terminal and the smart speaker, and the user terminal displays the room with the fire and the confirmation content, while the smart speaker prompts confirmation of cooking smoke; when the pattern recognition result is electrical overheating and the fire level sequence corresponds to a local fire, the terminal warning scheduling table first arranges the circuit breaker and the user terminal, and the circuit breaker performs a partial power outage, while the user terminal displays a prompt message indicating that the area is away from the distribution box; when the pattern recognition result is gas leak, the terminal warning scheduling table first arranges the gas equipment and the alarm prompt device. The system will trigger gas shut-off for gas appliances and prompt alarm devices to prohibit switching on or off electrical appliances. When the pattern recognition result indicates an open flame spread, the terminal warning scheduling table will first arrange lighting devices, door lock devices, and alarm devices, then arrange family member terminals and community alarm receiving terminals, displaying the room where the fire is occurring, the next affected room, and evacuation instructions on both terminals. When the pattern recognition result indicates a person remaining in an area, the terminal warning scheduling table will first arrange community alarm receiving terminals, then arrange user terminals and alarm devices, displaying the room where the person is located, door lock status, and accessible locations on the community alarm receiving terminal. The terminal warning scheduling table includes the pattern recognition result, the next affected room, priority notification terminals, priority execution devices, and execution order. This table serves as the basis for subsequent triggering of user confirmation, alarm confirmation, and smart home intervention.

[0116] S5. Based on the terminal early warning scheduling table, trigger user confirmation, alarm confirmation and smart home handling, and verify the personnel's lingering status through infrared and thermal imaging fusion human body detection, and generate early warning handling results.

[0117] S5.1. The home fire alarm controller triggers user confirmation, alarm confirmation, alarm prompt, gas cut-off, partial power outage, door lock escape and lighting indication according to the terminal warning arrangement table, and generates feedback messages and equipment handling records.

[0118] The system includes several steps: User confirmation, returned by the user terminal, indicates feedback from the user terminal regarding the room where the fire occurred, the fire signal mode, and the confirmed content; Alarm confirmation, returned by the community alarm receiving terminal, indicates feedback from the community alarm receiving terminal regarding the room where the fire occurred, adjacent affected rooms, the rooms where people are located, and the door lock status; Alarm prompts are executed by audible and visual alarms and smart speakers, used to issue audible, visual, and voice prompts to people within the home's warning zone; Gas shut-off is executed by the gas shut-off valve; Partial power outage is executed by the smart circuit breaker; Door lock escape is executed by the smart door lock; and Lighting indication is executed by the entryway lighting equipment. Feedback messages carry the two-way warning message number, user confirmation content, alarm receiving confirmation content, and feedback time. Equipment handling records carry the two-way warning message number, zone code, equipment type, executed content, and executed status.

[0119] S5.2. When any of the following situations occur: the pattern recognition result is "personnel lingering," the user terminal does not return confirmation content, or the community alarm receiving terminal requests verification of the personnel location, the home fire alarm controller extracts infrared human body contour information and thermal imaging heat source distribution information from the personnel detection status, and performs spatial registration according to the home warning zone to generate a personnel lingering verification map. The infrared human body contour information is output by the infrared human body detector and is used to represent the boundary of the person's shape; the thermal imaging heat source distribution information is output by the thermal imaging human body detector and is used to represent the location of the human body heat source; spatial registration uses the room boundary, door and window positions, and personnel detection device positions of the home warning zone as the same spatial reference, and transforms the contour coordinates corresponding to the infrared human body contour information and the heat source coordinates corresponding to the thermal imaging heat source distribution information into the same home warning zone coordinates.

[0120] It should be noted that the personnel stagnation verification map is formed by superimposing infrared human body contour information and thermal imaging heat source distribution information within the same household warning zone coordinates. This can avoid the problem of being unable to confirm the specific location of personnel based solely on the presence of human body sensors.

[0121] S5.3. In the personnel retention verification map, filter the overlapping areas of human body contour information and heat source distribution information, and combine the fire room, adjacent affected rooms and door lock status to determine the room where the personnel are located, the number of personnel and their location near doors and windows, and generate personnel retention verification results.

[0122] The overlapping area between human body contour information and heat source distribution information indicates that the area corresponding to the infrared human body contour information and the area corresponding to the thermal imaging heat source distribution information overlap within the same home warning zone coordinates; each unconnected overlapping area corresponds to one person target, and the number of unconnected overlapping areas is determined as the number of people; the home warning zone where the overlapping area is located is determined as the room where the person is located; the door or window position with the smallest distance between the center of the overlapping area and the door or window position is determined as the position closest to the door or window.

[0123] If the door lock is closed and the person is in the same room as the room where the fire is occurring, the personnel stay verification result records the person as being in the room where the fire is occurring. If the door lock is closed and the person is in the same room as an adjacent affected room, the personnel stay verification result records the person as being in an adjacent affected room. The personnel stay verification result includes the room the person is in, the number of people, their location near doors and windows, and the door lock status, so that subsequent alarm confirmation can obtain the accessible location, rather than just obtaining a personnel stay indication.

[0124] S5.4. The home fire alarm controller performs consistency verification on the personnel stay verification results, feedback messages, equipment handling records, fire arbitration set, and terminal early warning arrangement table. If the user confirmation content in the feedback message is a false alarm, and the smoke evidence, heat evidence, or gas evidence in the fire arbitration set still corresponds to the continuously rising segment, gas-first segment, or heat-first segment, the home fire alarm controller maintains the alarm prompt and continues to request user confirmation through the user confirmation field.

[0125] When the user confirmation in the feedback message confirms a fire, the home fire alarm controller maintains gas cutoff, partial power outage, door lock escape, lighting indication, and alarm prompts according to the terminal warning arrangement table. When the alarm confirmation in the feedback message indicates that an alarm has been received, the home fire alarm controller continuously sends information such as the room where the person is located, the number of people, their location near doors and windows, door lock status, the room where the fire occurred, and changes in the next affected room. When the execution status in the equipment handling record is inconsistent with the priority execution equipment in the terminal warning arrangement table, the home fire alarm controller continues to trigger the corresponding smart home devices according to the terminal warning arrangement table.

[0126] Through consistency verification, user confirmation, alarm confirmation, and smart home response can be consistent with the fire arbitration set, avoiding the problem of user false alarm feedback directly overriding evidence of a continuously escalating fire.

[0127] S5.5. When a sequence break occurs in the segmented transmission of the feedback message and the personnel stay verification result, the home fire alarm controller determines the missing frame sequence number and the abnormal frame sequence number according to the frame sequence number continuity verification method, and requests the retransmission of the feedback message data frame and the personnel stay verification result data frame with the corresponding sequence number through the flow control frame.

[0128] After retransmission, the home fire alarm controller re-verifies the continuity of frame sequence numbers for the feedback message and the personnel stay verification result. If the frame sequence numbers are continuous, the home fire alarm controller executes the corresponding action according to the complete feedback message and complete personnel stay verification result. If there are still breaks in the frame sequence numbers, the home fire alarm controller continues to request retransmission of the corresponding sequence number through the flow control frame. The feedback message, equipment handling record, personnel stay verification result, fire arbitration set, and terminal early warning arrangement table complete consistency verification, and after the feedback message and personnel stay verification result with sequence breaks are retransmitted through the flow control frame, the early warning action result is generated.

[0129] In summary, this invention avoids the disconnect between fire location and response targets in traditional single-point alarms by binding home early warning zones to smart home devices. It distinguishes between short-term disturbance segments, continuously rising segments, gas-preceding segments, and heat-preceding segments through a second-order exponential smoothing recurrent neural network time-series extraction method. Combined with evidence-based fire arbitration and personnel risk status correction, it can differentiate between situations such as cooking smoke, gas leaks, electrical overheating, open flame spread, and personnel entrapment. Compared to alarms triggered by a single threshold, it has stronger false alarm suppression and fire category identification capabilities. By utilizing single data frames, first data frames, continuous data frames, and flow control frames for sequence verification and selective retransmission, it maintains a complete correspondence between alarm content, confirmation content, and response records even when frames are lost in the alarm communication link. Combined with fuzzy logic fire signal pattern recognition, stacked converter encoder early warning orchestration, and infrared and thermal imaging fusion for human body verification, it enables notification terminals, smart home devices, and personnel entrapment confirmation to execute collaboratively according to the fire development sequence, improving the accuracy, communication reliability, and targeted emergency response of two-way home fire early warning.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A two-way early warning method for household fires based on the Internet of Things, characterized in that: include, Establish a binding relationship between home early warning zones and smart home devices, and use a second-order exponential smooth recurrent neural network time-series extraction method to process zone fire signals and generate zone status sheets; Fire arbitration based on evidence theory is conducted using a zoned status sheet, and fire severity levels are classified in conjunction with personnel risk status to generate a fire arbitration set; The partition status sheet and fire arbitration set are encapsulated into a two-way early warning message, and sequence verification and frame loss retransmission are performed through single data frames, first data frames, continuous data frames and flow control frames to generate a two-way early warning frame chain. Fuzzy logic fire signal pattern recognition and stacked converter encoder early warning arrangement are performed through bidirectional early warning frame chain to generate terminal early warning arrangement table; Based on the terminal warning scheduling table, user confirmation, alarm confirmation, and smart home response are triggered, and the status of personnel staying is verified by human body detection through infrared and thermal imaging fusion, and the warning response results are generated.

2. The two-way early warning method for household fires based on the Internet of Things as described in claim 1, characterized in that: The establishment of a binding relationship between home early warning zones and smart home devices specifically involves... The family warning zones are divided according to the functional areas of the family, and each family warning zone is assigned a zone code to generate a family zone relationship table. Fire detection equipment, personnel detection equipment, door lock equipment, gas equipment, circuit breaker equipment, lighting equipment, and alarm notification equipment are bound to the corresponding zone codes to limit the source of fire signals and the objects to be dealt with, and an early warning zone binding table is generated.

3. The two-way early warning method for household fires based on the Internet of Things as described in claim 2, characterized in that: The generation of the partition status sheet specifically refers to... Smoke signals, temperature signals, carbon monoxide signals, and combustible gas signals are arranged in chronological order according to the warning zone of the same household to generate a zone fire signal sequence; Trend extraction is performed on the fire signal sequence of the region, and it is divided into short-term disturbance segment, continuous rise segment, gas-preceding segment and heat-preceding segment to generate a time series segment set of the region; According to the household early warning zone, the time sequence fragment set of the zone, abnormal room, adjacent room, personnel detection status, door lock status, gas status, circuit breaker status, lighting status and executable device status are divided into zones and matched to form a zone status sheet.

4. The IoT-based two-way early warning method for home fires as described in claim 3, characterized in that: The aforementioned method of using a zoned status sheet for evidence-based fire arbitration specifically involves... Extract smoke evidence, heat evidence, gas evidence, and diffusion evidence from the zoning status sheet to generate a fire evidence group; The fire evidence group is defined according to the time sequence fragment set of the partition, so that short-term disturbance fragments correspond to false alarm evidence, continuous rise fragments correspond to local fire evidence, gas-preceding fragments correspond to gas leak evidence, and heat-preceding fragments correspond to electrical overheating evidence, thus generating a time sequence evidence group. Evidence theory synthesis rules are used to synthesize chronological evidence groups, and the fire category is determined among false alarms, fires awaiting confirmation, local fires, spreading fires, and emergency fires, generating a basic set for fire arbitration.

5. The IoT-based two-way early warning method for home fires as described in claim 4, characterized in that: The generation of the fire arbitration set specifically refers to... Extract the personnel detection status and door lock status from the partition status sheet, determine the room where the personnel are located, the personnel's tendency to stay, and the door lock escape status, and generate the personnel risk status; The risk status of personnel is used to correct the level of the fire arbitration base set. When the personnel are in the same room as the fire room and the door is locked, it corresponds to an emergency fire. When the personnel are in the same room as the adjacent affected room, it corresponds to a spreading fire, thus generating a fire level sequence. The abnormal rooms corresponding to the fire type are identified as fire rooms, and the adjacent rooms corresponding to the spreading fire are identified as adjacent affected rooms, thus generating the fire room relationship; The consistency of the fire severity sequence, the relationship between fire-affected rooms, and the corresponding time-series fragments of the fire evidence group is verified to generate a fire arbitration set.

6. The two-way early warning method for household fires based on the Internet of Things as described in claim 5, characterized in that: The process of encapsulating the partition status sheet and fire arbitration set into a two-way early warning message is as follows: Extract the fields of fire room, adjacent affected room, room where people are located, personnel detection status, door lock status, smart home device status and time segment from the partition status sheet and the relationship between fire rooms to generate a partition early warning field group; Extract the fire severity sequence and fire evidence set from the fire arbitration set to generate an arbitration field set; Based on the fire severity level sequence, determine the smart home handling field, user confirmation field, and alarm confirmation field, and determine the two-way early warning message number according to the fire room, the room where the person is located, the time segment field, and the fire severity level sequence to generate the confirmation handling field group; The partitioned early warning field group, arbitration field group, and confirmation and handling field group are encapsulated into the same message payload to generate a two-way early warning message with a two-way early warning message number.

7. The IoT-based two-way early warning method for home fires as described in claim 6, characterized in that: The generation of the bidirectional early warning frame chain specifically involves... The payload length of the two-way early warning message is detected. When the payload length meets the single-frame carrying condition, the two-way early warning message is encapsulated into a single data frame. When the payload length exceeds the single-frame carrying condition, the initial payload of the two-way early warning message is encapsulated into the first data frame, and the remaining payload is divided into continuous data frames according to the sending order to generate the initial early warning frame group. Write the frame sequence number to the consecutive data frames in the initial warning frame group, and store the frame sequence number, frame payload and bidirectional warning message number into the frame buffer to generate a warning frame group to be verified. Verify the continuity of the frame sequence number of the early warning frame group to be verified. When the frame sequence number is continuous, splice the frame payload to generate a reconstructed early warning message. When there is a break in the frame sequence number, write the missing frame sequence number and the abnormal frame sequence number into the flow control frame to generate a retransmission request frame. Based on the retransmission request frame, extract the corresponding consecutive data frames from the frame buffer, and limit the retransmission batch by the data block size in the flow control frame to generate a selective retransmission frame group. The selective retransmission frame group is filled back into the early warning frame group to be verified, and the continuity of the reconstructed early warning message is verified again according to the frame sequence number to obtain the frame payload sequence that has passed the verification. The bidirectional early warning message is reconstructed by combining the single data frame payload, the first data frame payload, and the consecutive data frame payloads in the frame payload sequence, and a bidirectional early warning frame chain is generated.

8. The IoT-based two-way early warning method for home fires as described in claim 7, characterized in that: The aforementioned fuzzy logic fire signal pattern recognition via a two-way early warning frame chain specifically involves... Fire signal identification data is generated by extracting the fire room, adjacent affected rooms, fire level, zonal time sequence fragment set and fire evidence group from the two-way early warning frame chain. The fire signal identification data is classified into categories based on smoke persistence, temperature persistence, signal fluctuation, time segment type, and changes in adjacent rooms. The fire signal patterns are identified in categories such as cooking smoke, electrical overheating, gas leak, open flame spread, and people entrapment, and pattern recognition results are generated.

9. The two-way early warning method for household fires based on the Internet of Things as described in claim 8, characterized in that: The generated terminal early warning arrangement table is specifically as follows: The pattern recognition results, the room where the fire occurred, the adjacent affected rooms, the rooms where people were located, the door lock status, the smart home device status, and the set of time-series fragments of the partition were constructed into an early warning arrangement sequence according to the time order. The stacked converter encoder is used to extract the room impact order, personnel risk order and equipment handling order in the early warning arrangement sequence, and to determine the next affected room, the priority notification terminal and the priority execution device; A terminal warning scheduling table is generated based on the order of execution of notification terminals and smart home devices according to the next affected room, priority notification terminals, and priority execution devices.

10. The IoT-based two-way early warning method for home fires as described in claim 9, characterized in that: The generation of early warning and handling results specifically includes, According to the terminal early warning scheduling table, user confirmation, alarm confirmation, alarm prompt, gas cut-off, partial power outage, door lock escape and lighting indication are triggered, and feedback messages and equipment handling records are generated. When the pattern recognition result is a person staying, the user terminal does not return confirmation content, or the community alarm terminal requests verification of the person's location, the infrared human body contour information and thermal imaging heat source distribution information in the person detection state are extracted, and spatial registration is performed according to the family early warning zone to generate a person staying verification map. In the personnel retention verification map, the overlapping areas of human body contour information and heat source distribution information are selected, and the room where the personnel are located, the number of personnel, and the location near the doors and windows are determined by combining the fire room, adjacent affected rooms, and door lock status, and the personnel retention verification results are generated. The consistency of the personnel stay verification results, feedback messages, equipment handling records, fire arbitration set and terminal early warning arrangement table is checked, and flow control frames are retransmitted for feedback messages and personnel stay verification results that have sequence breakpoints, and early warning handling results are generated.

Images