Fire identification alarm method based on AI image and fire detector

By building a centralized management model library platform, matching and adjusting the fire recognition model, the problems of decreasing identification accuracy and difficulty in fire monitoring are solved, and efficient and accurate fire monitoring and rapid response are achieved.

CN119049215AActive Publication Date: 2024-11-29ANHUI CHUANBAI TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411514696.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-29
Publication Date
2024-11-29
Estimated Expiration
2044-10-29

AI Technical Summary

Technical Problem

In actual deployment, existing fire monitoring technology has reduced recognition accuracy due to factors such as environmental complexity and lighting changes, and false alarms or missed alarms occur frequently, and it is difficult for users to quickly adjust the model to meet emergency needs.

Method used

Build a centralized management model library platform, match highly adaptable fire identification models through user needs, and identify potential risk points through fire simulation verification and equivalent evaluation, formulate cross-complementary measures to reduce false alarms and missed alarm rates, and if necessary Make model adjustments.

Benefits of technology

The rapid deployment and dynamic adjustment of the fire identification model are realized, the false alarm and missed alarm rates are reduced, the accuracy of monitoring and response efficiency are improved, and the work volatility of technicians is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119049215A_ABST
    Figure CN119049215A_ABST
Patent Text Reader

Abstract

The invention discloses a fire identification alarm method based on an AI image and a fire detector, and belongs to the technical field of fire identification, and the method comprises the steps: constructing a model library by a platform side; identifying a target area according to a fire monitoring demand of a user; matching a fire identification model meeting the fire monitoring requirement from the model library, and marking the fire identification model as a main identification model; monitoring equipment is installed in the target area, fire verification simulation is carried out in the target area according to the monitoring equipment, and misalignment data are obtained; performing temporary fire monitoring supplement according to the misalignment data, and setting a corresponding fire comprehensive identification scheme; monitoring the target area in real time to obtain monitoring data, and processing the monitoring data according to the fire comprehensive identification scheme to obtain a fire identification result; and the platform side adjusts the main identification model according to the misalignment data, updates the adjusted main identification model at the user, and cancels temporary fire monitoring supplement.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of fire identification, and specifically relates to a fire identification alarm method and a fire detector based on AI images. Background Art

[0002] In the current field of image recognition technology, especially in the critical application of fire monitoring, traditional methods face many challenges. First of all, building an efficient and accurate image recognition model is a complex and resource-intensive task. This process not only requires a large amount of data collection and annotation work, but also requires powerful computing resources and professional technicians to conduct long-term model training and optimization. Even so, when the trained model is actually deployed to the user's site, it often suffers from uncontrollable factors such as the complexity of the field environment, changes in lighting conditions, and the presence of obstructions, resulting in a significant decrease in recognition accuracy, and then false alarms or missed alarms, which seriously affects the accuracy and reliability of fire monitoring.

[0003] In addition, when the model shows deficiencies on the user side, it becomes particularly difficult to adjust and optimize it. On the one hand, users may not have the professional technical knowledge and resources to directly modify the model; on the other hand, even if technical personnel are involved, the adjustment process is not only time-consuming but also inefficient due to the need to conduct a series of tedious steps such as field data collection and model training again, making it difficult to meet users' urgent needs for rapid response and high-precision monitoring in emergency situations.

[0004] In view of the above problems, the market urgently needs a more flexible, efficient and adaptable fire monitoring solution; based on this, the present invention provides a fire identification and alarm method and fire detector based on AI images. Summary of the invention

[0005] In order to solve the problems existing in the above scheme, the present invention provides a fire identification alarm method and a fire detector based on AI image.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A fire recognition alarm method based on AI image, the method comprising: Step 1: The platform constructs a model library, which is used to store various fire identification models and model information corresponding to each fire identification model; the model information includes the scope of application scenarios and the accuracy of each application scenario.

[0007] Step 2: Obtain the user's fire monitoring needs, and identify the target area that needs to be monitored according to the fire monitoring needs; match the fire recognition model that meets the fire monitoring needs from the model library and mark it as the main recognition model.

[0008] Furthermore, the method of matching the main recognition model that meets the fire monitoring requirements from the model library includes: Identify the user's monitoring requirements according to the fire monitoring requirements, and collect the scene information of the target area; input the scene information and monitoring requirements into the model library to match with each model information, obtain each matching fire recognition model, and mark it as a candidate recognition model; The candidate recognition models are screened to obtain a main recognition model.

[0009] Furthermore, the method for screening each candidate recognition model includes: Identify the application scenarios where the scene information matches each of the to-be-selected recognition models, and obtain the accuracy of each of the scene information; Performing scene change adaptation analysis on the scene information and each of the candidate recognition models to obtain a scene adaptation value corresponding to each of the candidate recognition models, wherein the scene adaptation value has a value range of [1, 10]; The accuracy is marked as μ and the scene adaptation value is marked as CA; According to the formula Calculating the screening value of each of the candidate recognition models; Where: PA is the screening value; b1 and b2 are both proportional coefficients, with a value range of 0 <b1≤1,0<b2≤1; The candidate recognition model with the largest screening value is selected as the main recognition model.

[0010] Step three: installing monitoring equipment in the target area, and performing a fire verification simulation in the target area according to the monitoring equipment to obtain various misalignment data.

[0011] Furthermore, the method of conducting fire verification simulation in the target area according to the monitoring equipment is as follows: Setting each fire simulation condition according to the scene information; performing a fire simulation in the target area according to each fire simulation condition, and obtaining a monitoring simulation image corresponding to each fire simulation condition; Analyzing each of the monitoring simulation images through the main recognition model to obtain a fire recognition result corresponding to each of the fire simulation conditions; verifying each of the fire recognition results to obtain a verification result, wherein the verification result includes a correct recognition and an inaccurate recognition; The verification result is a fire simulation condition corresponding to the recognition misalignment, a monitoring image and a fire recognition result which are integrated into misalignment data.

[0012] Furthermore, the method of setting various fire simulation conditions according to the scene information includes: Summarize the reasons for each scenario, where the reasons for the scenarios are the reasons leading to inaccurate fire recognition; conduct an equivalent evaluation of every two reasons for the scenarios to obtain the corresponding equivalent evaluation results, where the equivalent evaluation results include being regarded as equivalent and not being regarded as equivalent; screen each of the reasons for the scenarios according to each equivalent evaluation result to obtain each preliminary reason; Estimate the simulation cost and occurrence probability in the target area for each of the preliminary reasons; According to the formula Calculate the corresponding reason value; In the formula: RV is the reason value; b3 and b4 are both proportionality coefficients, and the value range is 0 ≤ b3 ≤ 1, 0 < b4 ≤ 1; YB is the simulation cost; η is the occurrence probability; Screen each of the preliminary reasons according to each of the reason values to obtain each target reason; set fire simulation conditions according to each of the target reasons.

[0013] Further, the method for conducting an equivalent evaluation of every two reasons for the scenarios includes: Establish an equivalent evaluation model, and the expression of the equivalent judgment model is: ; In the formula: s is the input data, and the input data is two reasons for the scenarios for which the equivalent evaluation is conducted; the output data is the equivalent judgment value DT(s), and the equivalent judgment value is 1 or 0; Analyze every two reasons for the scenarios through the equivalent judgment model to obtain the corresponding equivalent judgment values for every two reasons for the scenarios; When the equivalent judgment value is 1, the equivalent evaluation result is regarded as equivalent; When the equivalent judgment value is 0, the equivalent evaluation result is not regarded as equivalent.

[0014] Step Four: Conduct temporary fire monitoring supplementation according to each of the inaccurate data, and set corresponding comprehensive fire recognition schemes.

[0015] Further, the method for conducting temporary fire monitoring supplementation according to each of the inaccurate data includes: Identify the reasons for inaccuracy of each of the inaccurate data, analyze each of the reasons for inaccuracy to obtain each corresponding solution for each of the reasons for inaccuracy; define a combination standard, where the combination standard is to solve the influence of each of the reasons for inaccuracy on fire recognition; form each candidate combination method according to the combination standard and each of the solutions; Screen each of the candidate combination methods to obtain the target combination method, and conduct temporary fire monitoring supplementation according to the target combination method.

[0016] Step 5: Monitor the target area in real time to obtain monitoring data, process the monitoring data according to the comprehensive fire identification scheme to obtain fire identification results, and display the fire identification results to the user.

[0017] Step 6: The platform adjusts the main recognition model according to the inaccurate data, updates the adjusted main recognition model at the user's location, and cancels the temporary fire monitoring supplement.

[0018] Furthermore, the adjusted main recognition model is stored in a model library.

[0019] Carry out real-time monitoring of the target area and obtain corresponding monitoring data.

[0020] A fire detector comprises a monitoring module, an analysis module and a transmission module; The monitoring module monitors the target area and obtains monitoring data; The analysis module is used to analyze the monitoring data to obtain fire identification results; The transmission module is used for data transmission.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention realizes the centralized storage, rapid deployment and dynamic adjustment of fire identification models by constructing a centralized model library platform. Specifically, the platform is responsible for collecting, establishing and maintaining a fire identification model library that contains multiple scene adaptability. These models have been strictly tested and optimized and have high recognition accuracy and generalization ability. When a user proposes a fire monitoring demand, the platform can quickly analyze, match and recommend the most suitable fire identification model according to the user's needs. Then, according to the actual scene simulation, potential false alarm and missed alarm risk points are identified and predicted. For these risk points, the platform can immediately formulate and implement cross-complementary measures, such as adding cross-monitoring points, introducing multiple side analysis algorithms, etc., to minimize the false alarm rate and missed alarm rate. At the same time, once the temporary supplementary measures achieve the expected effect, the temporary supplementary measures can be cancelled to achieve efficient layout and high-precision recognition of fire identification; and the platform will have sufficient time to adjust the fire identification model applied by the user, improve the recognition accuracy and the work continuity of the platform's technical personnel, and reduce the volatility of the technical personnel's work. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 is a flow chart of the method of the present invention; Figure 2 This is an example diagram of a mobile phone fire simulation of the present invention; Figure 3 This is an example diagram of a lamp fire simulation of the present invention; Figure 4 This is an example diagram of a vehicle headlight fire simulation according to the present invention. DETAILED DESCRIPTION

[0024] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] like Figures 1 to 4 As shown, a fire recognition alarm method based on AI image, the method includes: Step 1: The platform builds a model library, which is used to store various fire recognition models established by the platform or obtained from other channels, such as fire recognition models adjusted and set according to the needs of previous users; and sets corresponding model information for each fire recognition model. The model information includes relevant model information such as the scope of application scenarios and the accuracy of each application scenario. The simulation information is set according to its actual usage records and verification conditions. The fire recognition model is used to identify fires in collected high-definition images; it is generally established based on deep neural networks.

[0026] Step 2: Obtain the user's fire monitoring requirements. The fire monitoring requirements are used by the user to describe which areas to monitor for fire and the corresponding monitoring requirements; determine the target area that needs to be monitored for fire and the scene information corresponding to the target area based on the fire monitoring requirements; match the corresponding fire recognition model from the model library based on the scene information and monitoring requirements, and mark it as the main recognition model; that is, first eliminate those that do not meet the monitoring requirements, and then match them through the application scenario range to obtain various fire recognition models, screen them, and obtain the main recognition model.

[0027] In one embodiment, the fire identification models may be screened based on existing screening methods; for example, based on the accuracy of each application scenario.

[0028] In one embodiment, each fire recognition model that needs to be screened is marked as a candidate recognition model, and a method for screening each candidate recognition model includes: Identify the accuracy of application scenarios that each candidate recognition model matches; The scene information of the target area is analyzed for changes, that is, in the subsequent scene environment changes, the more complex the scene is, the greater the impact of the corresponding changes is, and some fire identification models have low adaptability to changes, resulting in a large adjustment workload after being applied to the actual scene; statistical analysis is performed on a large amount of historical record data, and a scene adaptation value is set, with a value range of [1, 10]. The more adaptable it is to its changes, the higher its scene adaptation value is. A value equal to 1 indicates that it is within the scope of the application scene, and a value smaller than 1 indicates that it is not within the scope of the application scene; the adaptation adjustment workload after various changes that may be possible is determined based on the historical record data, and they are sorted. The adaptation value of the performance that does not need to be adjusted (no adjustment workload) is 10, and the corresponding scene adaptation value is inserted according to the difference in the adjustment workload; subsequently, the corresponding scene adaptation value can be estimated by estimating the adjustment workload of each fire identification model in different application scenarios, or statistical simulation is performed based on the historical record data to determine the scene adaptation value corresponding to each application scenario of each fire identification model within the scope of the application scenario; in one embodiment, because the subsequent adjustment is mainly performed by the staff of the platform, the scene adaptation value of each fire identification model in different application scenarios can be directly set manually, and the setting is performed from the perspective of the staff.

[0029] The accuracy is marked as μ and the scene adaptation value is marked as CA; According to the formula Calculate the screening value of each candidate recognition model; Where: PA is the screening value; b1 and b2 are both proportional coefficients, with a value range of 0 <b1≤1,0<b2≤1; The candidate recognition model with the largest screening value is selected as the main recognition model.

[0030] Step three: Install monitoring equipment in the target area, including corresponding acquisition equipment, transmission equipment and other hardware and software equipment, which are used to capture images of the target area and perform corresponding fire identification analysis, such as color images 1080P, near-infrared images 1080P, etc.; adapt to all-day image acquisition; after the monitoring equipment is installed, conduct fire verification simulation in the target area to determine the inaccurate data, which are various situation data corresponding to false alarms and missed alarms.

[0031] In one embodiment, the method for performing fire verification simulation in a target area is as follows: Determine various possible fire situations in the target area and the corresponding scenario situations for various fire situations according to the scenario information, and integrate them into various fire simulation conditions; as Figures 2 to 4 shown, scenario simulations under various interference conditions; Perform fire simulation in the target area according to various fire simulation conditions, collect corresponding monitoring images, mark them as monitoring simulation images, analyze the monitoring simulation images through the main recognition model to obtain corresponding fire recognition results; check each fire recognition result to obtain a check result, and the check result includes correct recognition and misrecognition; Identify the fire simulation conditions, monitoring images, and fire recognition results corresponding to misrecognition in the check result, and integrate the corresponding fire simulation conditions, monitoring images, and fire recognition results into misalignment data.

[0032] In one embodiment, the method for setting various fire simulation conditions according to the scenario information is to set them in an existing manner, such as setting according to the work experience of the staff.

[0033] In one embodiment, the method for setting various fire simulation conditions according to the scenario information includes: Obtain various scenario reasons for false alarms, missed alarms, etc. in image-based fire recognition based on existing big data technology, such as various reasons such as light interference, smoke interference, occlusion, etc.; perform equivalent evaluation on pairwise scenario reasons, including AB and BA, that is, reverse evaluation; obtain corresponding equivalent evaluation results, and the equivalent evaluation results include regarded as equivalent and not regarded as equivalent; screen each scenario reason according to each equivalent evaluation result, and mark the remaining scenario reasons as primary selection reasons; regarded as equivalent means that if it is assumed that a fire can be recognized under this scenario reason, then it can also be recognized under another scenario reason, then the above two scenario reasons are regarded as equivalent, and one is retained. If it is not equivalent when analyzed from another angle, then this scenario reason is removed and the one with a larger scope is retained; Estimate the occurrence probability of each primary selection reason in the target area, and estimate the simulation cost of simulating this primary selection reason, mainly the costs of corresponding simulation equipment, materials, etc.; all are predicted using existing prediction technologies. According to the formula Calculate the corresponding reason value; In the formula: RV is the reason value; b3 and b4 are both proportionality coefficients, and the value ranges are 0 ≤ b3 ≤ 1, 0 < b4 ≤ 1; YB is the simulation cost; η is the occurrence probability; Mark the primary selection reasons with reason values greater than the threshold X1 as target reasons; set fire simulation conditions according to each target reason, and the fire simulation conditions are the corresponding conditions for simulating this target reason.

[0034] In one embodiment, an equivalent evaluation may be performed on the causes of each pair of scenarios based on the prior art.

[0035] In one embodiment, a method for performing an equivalence evaluation on the causes of two scenarios includes: An equivalent evaluation model is established. The equivalent evaluation model is used to simulate and judge the causes of the two corresponding scenarios to determine whether they meet the equivalent judgment criteria. Specifically, a large amount of existing historical fire identification data is used to make judgments to understand whether fire identification can be performed under the cause of scenario A, and whether fire identification can also be performed under the cause of scenario B. Then, it is determined whether the equivalent judgment criteria are met. The judgment can also be made based on other intelligent technologies. The expression of the equivalent judgment model is: ; Where: s is the input data, the input data is the two scene reasons for the equivalent evaluation; the output data is the equivalent judgment value DT(s); Analyze the causes of each pair of scenarios through the equivalent judgment model to obtain the equivalent judgment values ​​corresponding to the causes of each pair of scenarios; When the equivalence judgment value is 1, the equivalence assessment result is deemed to be equivalent; When the equivalence judgment value is 0, the equivalence evaluation result is not considered equivalent.

[0036] Step 4: Perform temporary fire monitoring supplements based on the inaccurate data and set up corresponding comprehensive fire identification plans.

[0037] When temporary fire monitoring is supplemented, the coordination relationship with the main identification model can be determined, thereby realizing fire identification of the monitoring data and solving various inaccuracy causes.

[0038] In one embodiment, temporary fire monitoring supplement is performed based on each inaccurate data, and the supplement is performed manually, that is, based on the inaccurate data, it is determined under what circumstances the main recognition model cannot meet the fire recognition requirements, and the staff selects a supplementary monitoring method to overcome the cause; illustratively, such as the cause of light interference, it is possible to supplement the existing model for light recognition, and comprehensively identify the fire with the main recognition model, and eliminate the light in the recognition of the main recognition model; it is also possible to supplement monitoring images from other angles and remove the influence through multi-angle monitoring.

[0039] In one embodiment, the method for performing temporary fire monitoring supplement according to each misaligned data includes: Identify each inaccuracy data, determine each inaccuracy cause based on each inaccuracy data, and determine the inaccuracy cause in combination with the target cause corresponding to the fire simulation condition; analyze various solutions that can solve the corresponding inaccuracy cause, combine them according to each solution, and obtain each candidate combination method. Each combination can solve each inaccuracy cause, that is, combine them based on solving all inaccuracy causes to form each candidate combination method; screen each candidate combination method to obtain the target combination method, and perform temporary fire monitoring supplement according to the target combination method.

[0040] In one embodiment, the selection of the combinations can be performed based on various existing screening methods, such as screening based on cost, implementation difficulty, and the like.

[0041] Step 5: Conduct real-time monitoring of the target area, obtain corresponding monitoring data, process the monitoring data according to the comprehensive fire identification plan, obtain corresponding fire identification results, and display the obtained fire identification results to the user.

[0042] Step 6: The platform adjusts the main recognition model according to the inaccurate data. When the main recognition model can overcome the inaccurate causes corresponding to the inaccurate data, the adjusted main recognition model will be updated at the user's location; the temporary fire monitoring supplement will be canceled, and the adjusted main recognition model will be stored in the model library.

[0043] A fire detector comprises: a monitoring module, an analysis module and a transmission module; The monitoring module monitors the target area and obtains monitoring data; The analysis module is used to analyze the monitoring data to obtain fire identification results; The transmission module is used for data transmission, such as wireless or wired data transmission.

[0044] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0045] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A fire recognition and alarm method based on AI image, characterized in that: Methods include: The platform constructs a model library, which is used to store various fire identification models and model information corresponding to each fire identification model; the model information includes the scope of application scenarios and the accuracy of each application scenario; Obtaining the user's fire monitoring requirements, and identifying the target area that needs to be monitored according to the fire monitoring requirements; matching the fire recognition model that meets the fire monitoring requirements from the model library and marking it as the main recognition model; Installing monitoring equipment in the target area, performing fire verification simulation in the target area according to the monitoring equipment, and obtaining various misalignment data; Perform temporary fire monitoring supplements based on the inaccurate data and set up corresponding fire comprehensive identification plans; Performing real-time monitoring on the target area to obtain monitoring data, processing the monitoring data according to the comprehensive fire identification scheme to obtain a fire identification result, and displaying the fire identification result to a user; The platform adjusts the main recognition model according to the inaccurate data, updates the adjusted main recognition model at the user's location, and cancels the temporary fire monitoring supplement.

2. The fire identification and alarm method based on AI image according to claim 1 is characterized in that: Methods for matching the main recognition model that meets the fire monitoring requirements from the model library include: Identify the user's monitoring requirements according to the fire monitoring requirements, and collect the scene information of the target area; input the scene information and monitoring requirements into the model library to match with each model information, obtain each matching fire recognition model, and mark it as a candidate recognition model; The candidate recognition models are screened to obtain a main recognition model.

3. The fire identification and alarm method based on AI image according to claim 2 is characterized in that: Methods for screening the candidate recognition models include: Identify the application scenarios where the scene information matches each of the to-be-selected recognition models, and obtain the accuracy of each of the scene information; Performing scene change adaptation analysis on the scene information and each of the candidate recognition models to obtain a scene adaptation value corresponding to each of the candidate recognition models, wherein the scene adaptation value has a value range of [1, 10]; The accuracy is marked as μ and the scene adaptation value is marked as CA; According to the formula Calculating the screening value of each of the candidate recognition models; Where: PA is the screening value; b1 and b2 are both proportional coefficients, with a value range of 0 <b1≤1,0<b2≤1; The candidate recognition model with the largest screening value is selected as the main recognition model.

4. The fire identification and alarm method based on AI image according to claim 2 is characterized in that: The method for conducting fire verification simulation in the target area according to the monitoring equipment is: Setting each fire simulation condition according to the scene information; performing a fire simulation in the target area according to each fire simulation condition, and obtaining a monitoring simulation image corresponding to each fire simulation condition; Analyzing each of the monitoring simulation images through the main recognition model to obtain a fire recognition result corresponding to each of the fire simulation conditions; verifying each of the fire recognition results to obtain a verification result, wherein the verification result includes a correct recognition and an inaccurate recognition; The verification result is a fire simulation condition corresponding to the recognition misalignment, a monitoring image and a fire recognition result which are integrated into misalignment data.

5. The fire recognition and alarm method based on AI image according to claim 4 is characterized in that: Methods for setting various fire simulation conditions according to scene information include: Summarize the reasons for each scenario, where the reasons for the scenario are the reasons for the misalignment of fire recognition; conduct an equivalent evaluation of every two scenario reasons to obtain the corresponding equivalent evaluation results for every two scenario reasons. The equivalent evaluation results include being regarded as equivalent and not being regarded as equivalent; screen each of the scenario reasons according to each equivalent evaluation result to obtain each primary reason; Estimate the simulation cost and occurrence probability in the target area for each of the primary reasons; According to the formula Calculate the corresponding cause value; In the formula: RV is the reason value; b3 and b4 are both proportionality coefficients, and the value range is 0 ≤ b3 ≤ 1, 0 < b4 ≤ 1; YB is the simulation cost; η is the occurrence probability; Screen each of the primary reasons according to each of the reason values to obtain each target reason; set fire simulation conditions according to each of the target reasons.

6. The fire identification and alarm method based on AI image according to claim 5 is characterized in that: The method for conducting an equivalent evaluation of every two scenario reasons includes: Establish an equivalent evaluation model, and the expression of the equivalent judgment model is: ; In the formula: s is the input data, and the input data is two scenario reasons for which an equivalent evaluation is conducted; the output data is the equivalent judgment value DT(s), and the equivalent judgment value is 1 or 0; Analyze every two scenario reasons through the equivalent judgment model to obtain the corresponding equivalent judgment values for every two scenario reasons; When the equivalent judgment value is 1, the equivalent evaluation result is regarded as equivalent; When the equivalent judgment value is 0, the equivalent evaluation result is not regarded as equivalent.

7. The fire recognition and alarm method based on AI image according to claim 1 is characterized in that: The method for supplementing temporary fire monitoring according to each misalignment data includes: Identify the reasons for the misalignment of each of the misalignment data, analyze each of the reasons for the misalignment to obtain each solution corresponding to each of the reasons for the misalignment; define a combination standard, where the combination standard is to solve the influence of each reason for the misalignment on fire recognition; form each candidate combination method according to the combination standard and each of the solutions; Screen each of the candidate combination methods to obtain a target combination method, and conduct temporary fire monitoring supplement according to the target combination method.

8. The fire identification and alarm method based on AI image according to claim 1 is characterized in that: Store the adjusted main recognition model in the model library.

9. A fire detector, characterized in that: Execute a fire recognition and alarm method based on AI images according to any one of claims 1 to 8, including a monitoring module, an analysis module, and a transmission module; The monitoring module monitors the target area to obtain monitoring data; The analysis module is used to analyze the monitoring data to obtain a fire recognition result; The transmission module is used for data transmission.

Citation Information

Patent Citations

  • Landslide susceptibility evaluation method based on fractal-machine learning hybrid model

    CN111898315A

  • Multi-scene adaptive model fusion method and face recognition system

    CN113361488A

  • Fire hazard level assessment method and system based on artificial intelligence

    CN117437470A

  • Mountain torrent disaster early warning misalignment risk analysis method based on Monte Carlo simulation method

    CN118411034A

  • Space-utilizing multipurpose mattress

    KR1020240016753A