A fire hazard intelligent identification system and method

Abstract

The present application relates to the field of fire hazard identification, in particular to a fire hazard intelligent identification system and method, which collects the surface temperature of the object in real time and associates it with the corresponding three-dimensional object. Within a predetermined time period, the scheme combines the temperature and timeline of the object to extract features and form a temperature rise behavior habit. The temperature rise behavior habit includes temperature rise duration, temperature rise period and temperature rise constant. Then, based on the real-time temperature rise behavior and the temperature rise behavior habit, the weight distribution is carried out to generate the risk coefficient R. According to the risk coefficient, the system can issue a corresponding early warning report. Real-time warning is carried out when the object shows abnormal temperature rise behavior, which realizes the advance of the prediction time and the prediction accuracy of the fire hazard identification.

Description

Technical Field

[0001] This invention relates to the field of fire hazard identification, and more particularly to an intelligent fire hazard identification system and method. Background Technology

[0002] Fire protection systems are increasingly widely used in modern society, playing a vital role in various public places, commercial buildings, residential communities, and industrial production areas. The primary purpose of a fire protection system is to issue timely alarms when a fire occurs, enabling necessary preventative measures to be taken to avoid or reduce injury to people and property damage. Furthermore, efficient fire protection systems can assist fire departments in quickly locating the source of the fire, thus extinguishing it more effectively.

[0003] Traditional fire protection systems primarily rely on devices such as smoke detectors, temperature sensors, and infrared sensors to monitor and warn of fires. These devices typically detect fires based on specific physical or chemical properties; for example, smoke detectors detect smoke particles in the air to indicate the occurrence of a fire, while temperature sensors detect sudden increases in ambient temperature to warn of fires.

[0004] However, these traditional fire suppression systems have some significant drawbacks. First, they cannot predict fire hazards in their early stages, only issuing alarms after a fire has actually occurred. Second, because they are typically based on a single sensor technology, false alarms or missed alarms are possible, which is highly detrimental to personal safety and property protection. For example, some electrical appliances may generate high temperatures during normal operation, potentially causing temperature sensors to issue false alarms. Summary of the Invention

[0005] To address the above problems, this invention provides an intelligent fire hazard identification system and method.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for intelligent identification of fire hazards, comprising:

[0008] Multiple images of the detection area from different angles are acquired in real time to determine the three-dimensional objects of each item, and the surface temperature of the item is acquired in real time and associated with the corresponding three-dimensional object.

[0009] Within a preset time period, features are extracted by combining the surface temperature of the item with the corresponding timeline to generate heating behavior habits, which include heating duration, heating constant value and heating cycle.

[0010] Real-time heating behavior is generated, and a risk coefficient R is generated based on the weight allocation of real-time heating behavior and heating behavior habits. Real-time heating behavior includes real-time heating duration, real-time temperature value, and heating period.

[0011] An early warning report is issued based on the risk coefficient R.

[0012] Furthermore, the real-time acquisition of multiple images of the detection area from different angles to determine the three-dimensional objects of each item, and the real-time acquisition of the surface temperature of the item and its association with the corresponding three-dimensional object, specifically include:

[0013] Create a 3D model of the item and compare its similarity with 3D objects in a 3D object library;

[0014] Temperature correlation is performed when the similarity is higher than the preset value;

[0015] When the similarity is lower than the preset value, a 3D object is created for the 3D model and saved in the 3D object library.

[0016] Furthermore, the establishment of the three-dimensional model includes the following steps:

[0017] Noise reduction, contrast adjustment, and color correction are performed on multiple real-time images of the detection area from different angles.

[0018] Use stereo matching and 3D reconstruction techniques to create a 3D model of the object.

[0019] Furthermore, the step of extracting features by combining the surface temperature of the item with the corresponding timeline within a preset time period to generate warming behavior habits includes the following steps:

[0020] Generate a temperature-time curve, locate the starting point when the temperature begins to rise and the ending point when the temperature begins to fall, and calculate the time difference as the duration of the temperature rise.

[0021] The number of times an item heats up within a time period is counted, and the timing and frequency of the item's temperature rise are determined based on the statistical results as the heating cycle.

[0022] The maximum temperature reached during the heating period is located, and the maximum value is averaged across multiple heating processes to obtain the constant temperature of the item.

[0023] Furthermore, the heating period specifically refers to the specific date and time of the real-time heating.

[0024] Furthermore, the formula for weight allocation is as follows:

[0025]

[0026] Where R is the risk coefficient, T r For real-time heating duration, T h To determine the duration of the heating, V r V represents the real-time temperature value. hThe constant temperature is α, and the real-time heating time T is α. r Duration of temperature rise T h The weight of the ratio, β is the real-time temperature value V r With constant temperature V h Weight of the ratio.

[0027] Furthermore, the issuance of the early warning report based on the risk coefficient R includes:

[0028] When the risk coefficient R is less than the preset threshold, the hazard status is set to "no abnormality found".

[0029] When the risk coefficient R is greater than or equal to the preset threshold, the hazard status is set to have abnormal fire risk.

[0030] Furthermore, when the heating period is outside the range of the heating cycle, the preset threshold is lowered.

[0031] A fire hazard intelligent identification system includes:

[0032] The data acquisition module is used to acquire multiple images of the detection area from different angles in real time to determine the three-dimensional objects of each item, and to acquire the surface temperature of the item in real time and associate it with the corresponding three-dimensional object.

[0033] The feature extraction module is used to extract features from the surface temperature of an item and the corresponding timeline within a preset time period to generate heating behavior habits.

[0034] The real-time analysis module is used to generate real-time heating behavior, assign weights based on real-time heating behavior and heating behavior habits, and generate a risk coefficient R.

[0035] The hazard reporting module is used to issue early warning reports based on the risk coefficient R.

[0036] The beneficial effects of this invention are as follows: First, it acquires multiple images of the detection area from different angles in real time to establish a three-dimensional model for each item and correlates it with its surface temperature, enabling monitoring of the temperature status and changes of various items within the detection area. Then, it generates a risk coefficient based on the weighted allocation of the items' heating behavior habits, enabling real-time early warning when items exhibit abnormal heating behavior. Compared to traditional methods, this solution can issue warnings when fire hazards first appear, rather than waiting until a fire actually occurs. This early identification allows for earlier preventative measures, significantly reducing the risk of personal injury and property damage. Furthermore, preset thresholds can be dynamically adjusted according to the item's heating period, resulting in more accurate warnings. This invention achieves both earlier prediction and higher accuracy in fire hazard identification. Attached Figure Description

[0037] Figure 1 This is a flowchart of the steps of an intelligent fire hazard identification system and method according to the present invention.

[0038] Figure 2 This is a flowchart of step S2 in this invention. Detailed Implementation

[0039] Please see Figure 1-2 As shown, the present invention relates to an intelligent identification system and method for fire hazards.

[0040] Specifically, the present invention provides an intelligent fire hazard identification system and method, comprising the following steps:

[0041] S1. Acquire multiple images of the detection area from different angles in real time to determine the 3D objects of each item, and acquire the surface temperature of the item in real time and associate it with the corresponding 3D object; Step S1 includes the following steps:

[0042] S11. Build a 3D model of the object and compare its similarity with 3D objects in the 3D object library; wherein, building the 3D model includes noise reduction, contrast adaptation and color correction of multiple real-time acquired images of the detection area from different angles;

[0043] Use stereo matching and 3D reconstruction techniques to create a 3D model of the object;

[0044] Specifically, the images of the object captured from different angles are first preprocessed. This includes eliminating image noise, adjusting contrast for optimal results, and performing color correction to ensure image realism and consistency. For example, consider an image of an electric kettle captured under specific lighting conditions; due to lighting or camera settings, the image may appear too dark or have distorted colors. These preprocessing steps improve image quality and accuracy. Next, stereo matching techniques are used to match images from multiple perspectives to obtain the object's 3D coordinates. These coordinates are then used in 3D reconstruction techniques to ultimately generate a 3D model of the object. Once the 3D model is obtained, it is compared for similarity with existing models in a 3D object library containing 3D models of many common objects, such as furniture, appliances, and decorations. To compare two 3D models, key features need to be extracted from them. These features may include the object's shape, curvature, edges, and surface texture. For example, the features of an electric kettle might include the shape of its handle and the curvature of its spout. The extracted features need to be converted into a standardized form, often referred to as a feature descriptor. Feature descriptors provide a numerical or vector representation for each feature, enabling comparison between different models. Nearest neighbor matching is used to match the feature descriptors of the newly generated model with the feature descriptors of all models in the 3D object library. The goal of matching is to find the most similar feature pairs. A similarity score is calculated based on the matching results. This is typically done by comparing the distance or difference between the matched features. The higher the score, the greater the similarity. If the calculated similarity score is higher than a preset threshold, the newly generated model is considered highly similar to a model in the library. If the score is lower than the threshold, it is considered a new item. This step uses image processing and 3D reconstruction techniques to obtain a high-precision 3D model of the item, significantly improving the accuracy of fire hazard detection. Furthermore, dynamic interaction with the 3D object library allows the system to adaptively identify and track newly appearing items, further enhancing its real-time monitoring and prediction capabilities.

[0045] Finally, edge locking is applied to the identified item to ensure that it is not repeatedly identified unless it has moved significantly, thus saving computing resources.

[0046] S12. When the similarity is higher than the preset value, perform temperature correlation;

[0047] Specifically, once a 3D model of an object successfully matches a model in a 3D object library, the surface temperature data of that object is acquired in real time. This is achieved through a temperature sensor or infrared camera. The acquired real-time temperature data is then associated with the corresponding 3D model. Therefore, each 3D model has an associated temperature dataset that changes over time. For subsequent analysis and processing, the system stores the associated temperature data in a database. This allows for the retrospective analysis of past temperature changes, even in the future.

[0048] By establishing temperature correlations, a complete temperature history can be built for each identified item. This not only helps in real-time monitoring of fire hazards but can also be used for subsequent data analysis.

[0049] S13. When the similarity is lower than the preset value, create a 3D object for the 3D model and save it in the 3D object library.

[0050] Specifically, when the similarity between a detected item and any model in the 3D object library is lower than a preset value, it means that the item is a new, previously unrecognized object. At this point, a new 3D model of the item is constructed using image data captured from multiple angles. Key features of the newly constructed 3D model (such as shape, curvature, edges, and surface texture) are extracted and stored. These features will play a crucial role in future similarity comparisons. The newly created 3D model and its related data are added to the system's 3D object library, thus expanding the library's content. This ensures that the library is always up-to-date, capable of recognizing and processing more types of items, and also provides new objects for subsequent learning of warming behavior habits. By creating 3D models for new items not in the library and storing them in the 3D object library, the system can continuously learn and adapt to new environments and items. This dynamic update mechanism ensures the system's adaptability and long-term effectiveness. Over time, the 3D object library becomes richer and more diverse, thereby improving the system's recognition capabilities and accuracy. This also helps to more accurately monitor and identify fire hazards, further improving safety.

[0051] S2. Within a preset time period, feature extraction is performed by combining the surface temperature of the item with the corresponding timeline to generate heating behavior habits;

[0052] Among these warming behaviors are the duration of warming (T). h Temperature constant V h and heating cycle;

[0053] Step S2 includes the following steps:

[0054] S21. Generate a temperature-time curve, locate the starting time point when the temperature begins to rise and the ending time point when the temperature begins to fall, and calculate the time difference as the duration of the temperature rise.

[0055] Specifically, whenever the system detects a change in the surface temperature of an object, it records the temperature value and the corresponding timestamp. These data points are arranged in chronological order, forming a continuous temperature-time curve. By analyzing the temperature-time curve, the system can identify trends in temperature rise and fall. When the temperature begins to rise, the point is marked as the starting time point; when the temperature begins to fall and remains stable, the point is marked as the ending time point.

[0056] S22. Count the number of times the item heats up within a time period, and determine the temperature rise habit time point and frequency of the item as the heating period based on the statistical results;

[0057] Specifically, whenever the system detects an item's heating behavior (i.e., the temperature starts to rise, persists for a period of time, and then falls), it records it as a separate heating event. These events are stored chronologically and associated with the corresponding items. Based on a preset time period (e.g., one day, one week, or one month), the system counts the number of heating events for each item within that time period. For each item, the system analyzes the number of heating events in different time periods (e.g., morning, afternoon, and evening) to determine the item's main heating habit time points. Simultaneously, based on the number of heating events throughout the entire time period (one month), the system calculates the item's heating frequency; this set of frequencies and habit time points is encapsulated as a heating cycle.

[0058] By statistically analyzing the heating events of items, the system can understand the heating habits of each item, including the main heating times and frequencies. This provides important reference information for subsequent risk assessment. For example, if an item (such as a hair dryer) suddenly heats up at an unusual time (such as 3 a.m.), it may indicate an anomaly (perhaps caused by the user forgetting to unplug it). Similarly, a sudden increase in the frequency of an item's heating could also be a potential risk signal.

[0059] S23. During the heating duration, locate the maximum temperature reached, average the maximum values ​​during multiple heating processes, and obtain the constant temperature of the item.

[0060] Specifically, for each heating event, the system detects and records the item's temperature value during the heating duration. Among these temperature values, the system identifies the highest value reached. This highest value represents the item's peak temperature during this heating event. All peak temperatures from heating events are stored and archived, associated with the corresponding item and its heating event. The system then averages all recorded peak temperatures to obtain an average peak temperature, which is the item's heating constant. This heating constant represents the item's average peak temperature across multiple heating events.

[0061] By statistically analyzing and averaging the peak temperatures of each temperature rise event, the system can obtain the constant temperature rise value of the item. This constant temperature rise value provides crucial benchmark data for subsequent risk assessment and early warning. For temperature rise events that exceed the constant temperature rise value, the system can consider them to pose a high risk and thus take corresponding early warning measures.

[0062] S3. Generate real-time heating behavior, assign weights based on real-time heating behavior and heating behavior habits, and generate a risk coefficient R.

[0063] Among them, the real-time heating behavior includes the real-time heating duration T. r Real-time temperature value V r During the warming period; the specific formula for weight allocation is as follows:

[0064]

[0065] Where R is the risk coefficient, and T is the risk factor. r Real-time heating duration, which is the heating duration T. h , is the real-time temperature value V r , which is the constant temperature V h α is the real-time heating time T r Duration of temperature rise T h The weight of the ratio, β is the real-time temperature value V r With constant temperature V h Weight of the ratio.

[0066] Specifically, by using the start and end times of the real-time heating, the system calculates the duration of this heating event, i.e., the real-time heating duration T. r During a temperature rise event, the system continuously records the temperature of the item; this value is the real-time temperature value V. r .

[0067] In the weighting process, the initial values ​​of α and β are determined manually. The system then adjusts these values ​​based on the item's heating frequency. If an item heats up frequently (its heating frequency exceeds a certain preset threshold), α is decreased. This means that frequently heating items will be more tolerant of real-time heating behavior and will not be overly treated as risk events. If an item rarely heats up (its heating frequency is below a certain preset threshold), β is increased.

[0068] The above weighting formula measures the degree of danger when the actual object's heating time is too long or exceeds the heating constant value too much, achieving comprehensive monitoring. By comprehensively considering various parameters of real-time heating behavior and their weights, a risk coefficient R is calculated. This coefficient represents the risk level of the current heating event relative to the object's usual heating behavior.

[0069] S4. Issue an early warning report based on the risk coefficient R;

[0070] Specifically, when the risk coefficient R is less than the preset threshold, the hazard status is set to "no abnormality found".

[0071] When the risk coefficient R is greater than or equal to the preset threshold, the hazard status is set to have abnormal fire risk.

[0072] First, the system needs a preset risk threshold. This threshold is manually determined based on past data and experience, aiming to distinguish between normal conditions and potential fire hazards. If the calculated risk coefficient R is lower than the preset threshold, it means that no obvious fire hazard has been detected. The system will set the hazard status to "No abnormality detected" and continue monitoring. If the risk coefficient R is greater than or equal to the preset threshold, it means that there is a certain fire risk. The system will set the hazard status to "Abnormal fire risk" and immediately issue an early warning report. At this time, relevant personnel should immediately review the report, determine the source of the risk, and take necessary countermeasures.

[0073] Furthermore, when the heating period is outside the range of the heating cycle, the preset threshold is lowered;

[0074] This setting is designed to pay closer attention to and be more vigilant about temperature rises that occur outside the normal temperature rise timeframe. These unusual temperature rises may indicate anomalies or potential risks. Consider an industrial piece of equipment that typically operates between 9:00 AM and 5:00 PM daily, experiencing several expected temperature rises during this period – this is its temperature rise cycle. However, if this equipment suddenly starts heating up at 3:00 AM, this is outside its normal temperature rise cycle. Such an abnormal temperature rise could be due to equipment malfunction, power failure, or other reasons. In such cases, to ensure sufficient attention is paid to this unusual temperature rise, the system will lower the preset risk threshold. For example, the system might normally set a risk threshold of 5, but when an unusual temperature rise is detected, the threshold might be adjusted to 2. This means that even if the equipment temperature does not exceed the normal temperature rise value V, the risk threshold will be lowered. h It did not run for the duration T of the heating period. h The system will issue an early warning. This ensures that action is taken in time before potential risks arise, thereby improving safety. By dynamically adjusting thresholds, the intelligent fire hazard identification system can respond more sensitively to abnormal heating behaviors outside the expected heating cycle, thus providing more accurate and timely warnings.

[0075] The present invention also provides a fire hazard intelligent identification system, comprising:

[0076] The data acquisition module is used to acquire multiple images of the detection area from different angles in real time to determine the three-dimensional objects of each item, and to acquire the surface temperature of the item in real time and associate it with the corresponding three-dimensional object.

[0077] The feature extraction module is used to extract features from the surface temperature of an item and the corresponding timeline within a preset time period to generate heating behavior habits.

[0078] The real-time analysis module is used to generate real-time heating behavior, assign weights based on real-time heating behavior and heating behavior habits, and generate a risk coefficient R.

[0079] The hazard reporting module is used to issue early warning reports based on the risk coefficient R.

[0080] This system allows multiple components to work together to achieve real-time and accurate identification and early warning of fire hazards, greatly improving the proactiveness of fire prevention.

[0081] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for intelligent identification of fire hazards, characterized in that, Include: Multiple images of the detection area from different angles are acquired in real time to determine the three-dimensional objects of each item, and the surface temperature of the item is acquired in real time and associated with the corresponding three-dimensional object. Within a preset time period, features are extracted by combining the surface temperature of the item with the corresponding timeline to generate heating behavior habits, which include heating duration Th, heating constant Vh, and heating cycle. Real-time heating behavior is generated, and a risk coefficient R is generated based on the weight allocation of real-time heating behavior and heating behavior habits. Real-time heating behavior includes real-time heating duration Tr, real-time temperature value Vr, and heating period. The specific formula for weight allocation is: R = α(Tr / Th) + β(Vr / Vh), where R is the risk coefficient, Tr is the real-time heating duration, Th is the heating duration, Vr is the real-time temperature value, Vh is the heating constant value, α is the weight of the ratio of real-time heating duration Tr to heating duration Th, and β is the weight of the ratio of real-time temperature value Vr to heating constant value Vh. The weights α and β are adjusted according to the heating frequency of the item. If the item is frequently heated and its heating frequency exceeds a certain preset threshold, α is decreased; if the item is basically not heated and its heating frequency is below a certain preset threshold, β is increased. An early warning report is issued based on the risk coefficient R.

2. The intelligent identification method for fire hazards according to claim 1, characterized in that, The process of acquiring multiple images of the detection area from different angles in real time to determine the three-dimensional objects of each item, and acquiring the surface temperature of the item in real time and associating it with the corresponding three-dimensional object, specifically includes: Create a 3D model of the item and compare its similarity with 3D objects in a 3D object library; Temperature correlation is performed when the similarity is higher than the preset value; When the similarity is lower than the preset value, a 3D object is created for the 3D model and saved in the 3D object library.

3. The intelligent identification method for fire hazards according to claim 2, characterized in that, The process of creating the 3D model includes the following steps: Noise reduction, contrast adjustment, and color correction are performed on multiple real-time images of the detection area from different angles. Use stereo matching and 3D reconstruction techniques to create a 3D model of the object.

4. The intelligent identification method for fire hazards according to claim 1, characterized in that, The step of extracting features from the surface temperature of an item and the corresponding timeline within a preset time period to generate a temperature rise behavior habit includes the following steps: Generate a temperature-time curve, locate the starting point when the temperature begins to rise and the ending point when the temperature begins to fall, and calculate the time difference as the duration of the temperature rise. The number of times an item heats up within a time period is counted, and the timing and frequency of the item's temperature rise are determined based on the statistical results as the heating cycle. The maximum temperature reached during the heating period is located, and the maximum value is averaged across multiple heating processes to obtain the constant temperature of the item.

5. The intelligent identification method for fire hazards according to claim 1, characterized in that, The heating period refers specifically to the date and time of the real-time heating.

6. The intelligent identification method for fire hazards according to claim 1, characterized in that, The issuance of early warning reports based on the risk coefficient R includes: When the risk coefficient R is less than the preset threshold, the hazard status is set to "no abnormality found". When the risk coefficient R is greater than or equal to the preset threshold, the hazard status is set to have abnormal fire risk.

7. The intelligent identification method for fire hazards according to claim 6, characterized in that, When the heating period is outside the range of the heating cycle, the preset threshold is lowered.

8. A fire hazard intelligent identification system, characterized in that, include: The data acquisition module is used to acquire multiple images of the detection area from different angles in real time to determine the three-dimensional objects of each item, and to acquire the surface temperature of the item in real time and associate it with the corresponding three-dimensional object. The feature extraction module is used to extract features from the surface temperature of an item and the corresponding timeline within a preset time period to generate heating behavior habits. The real-time analysis module generates real-time heating behavior and assigns weights based on the real-time heating behavior and heating behavior habits to generate a risk coefficient R. The specific formula for weight allocation is: R = α(Tr / Th) + β(Vr / Vh), where R is the risk coefficient, Tr is the real-time heating duration, Th is the heating duration, Vr is the real-time temperature, Vh is the constant heating value, α is the weight of the ratio of the real-time heating duration Tr to the heating duration Th, and β is the weight of the ratio of the real-time temperature Vr to the constant heating value Vh. The weights α and β are adjusted based on the heating frequency of the item. If the item is frequently heated and its heating frequency exceeds a certain preset threshold, α is decreased; if the item is rarely heated and its heating frequency is below a certain preset threshold, β is increased. The hazard reporting module is used to issue early warning reports based on the risk coefficient R.

Images