Fire early warning method and system based on visual identification

Abstract

The invention provides a fire early warning method and system based on visual identification. The method comprises the following steps: constructing a visual identification model based on collected fire accident cases; capturing an on-site video to be subjected to fire early warning; fire visual features are extracted from the captured on-site video; and identifying the extracted fire visual features based on the constructed visual identification model to obtain a fire early warning result. According to the method, the fire visual features are extracted from the on-site video to be subjected to fire early warning, the constructed visual recognition model is input for fire early warning judgment, dependence on a sensor is not needed, the influence of equipment sensitivity and equipment misinformation is avoided, intelligent and accurate early warning of the fire is achieved, and powerful support is provided for environmental safety and accident prevention and control.

Description

Technical Field

[0001] The present application relates to the field of fire warning, and specifically to a fire warning method and system based on visual recognition. Background Art

[0002] Traditional fire warning methods mainly include the following:

[0003] 1. Sound and light alarm: This is one of the most common traditional fire warning devices. It uses sound and light to sound the alarm. Once a fire is detected, it will immediately activate a high-decibel alarm and a strong flash to alert people to the occurrence of a fire.

[0004] 2. Manual alarm button: This is a fire warning device that requires manual operation. When people find a fire, they can press the manual alarm button to trigger the alarm system.

[0005] 3. Telephone alarm: When a fire occurs, the alarm information is transmitted to the fire center or relevant departments through the telephone line. This requires the telephone line to remain open and relies on manual alarm.

[0006] These traditional fire warning methods mainly rely on basic sensor technology and mechanical operations. They can effectively remind people of the occurrence of fire to a certain extent, but relatively speaking, the false alarm rate is high and the ability to prevent and detect fire early is limited.

[0007] Although with the development of technology, modern fire warning systems have introduced more advanced technologies, such as wireless sensor networks, mobile smart terminals and big data analysis, which have greatly improved the accuracy and timeliness of fire warnings. However, traditional fire warning methods still have their application value in some areas, especially in some areas with poor infrastructure or limited budgets.

[0008] There are some problems in the application of traditional fire warning technology, which may affect the accuracy and timeliness of fire warning, thereby increasing fire risks and losses. At present, the existing fire warning methods still rely on personnel inspections to actively discover problems, and mainly rely on judging flames and thick smoke. It is impossible to identify harmful gases and high temperatures with the naked eye, and there are the following shortcomings:

[0009] 1. High false alarm and missed alarm rates: Traditional fire warning systems may cause false alarms or missed alarms due to insufficient equipment sensitivity, interference from environmental factors, or improper threshold settings. False alarms can cause unnecessary panic and disturbance to people, while missed alarms may prevent fires from being discovered and controlled in a timely manner.

[0010] 2. Long warning time: Traditional fire warning systems usually take a long time to detect and confirm fires, which may result in the failure to issue warnings in time at the early stages of a fire, thereby delaying the time to extinguish the fire.

[0011] 3. Low level of intelligence: Traditional fire warning systems are mostly based on simple sensors and threshold judgments, lacking intelligent analysis and processing capabilities, which makes it difficult for the system to adapt to complex and changing fire scenes and unable to accurately judge the development trend and severity of the fire.

[0012] 4. Difficulty in system maintenance: Traditional fire warning systems usually require regular equipment inspection, maintenance and calibration to ensure their normal operation and accuracy. However, due to the complexity of the system and the limitations of the skill level of maintenance personnel, this may lead to difficult or untimely maintenance, thus affecting the stability and reliability of the system. Summary of the invention

[0013] The present application provides a fire warning method and system based on visual recognition, which can solve the technical problem in the prior art that traditional fire warning may cause false alarm or missed alarm of fire due to insufficient equipment sensitivity, interference from environmental factors or improper setting of thresholds.

[0014] In a first aspect, the present application provides a fire warning method based on visual recognition, comprising the following steps:

[0015] Build a visual recognition model based on the collected fire accident cases;

[0016] Capture live video for fire warnings;

[0017] Extracting fire visual features from captured on-site videos;

[0018] The extracted fire visual features are identified based on the constructed visual recognition model to obtain fire warning results.

[0019] In combination with the first aspect, in one implementation, the building of a visual recognition model based on the collected fire accident cases specifically includes the following steps:

[0020] Collect fire accident cases;

[0021] Extract visual recognition related information and case fire conditions from the collected videos of fire accident cases as a dataset;

[0022] Divide the training set from the dataset;

[0023] The visual recognition model is constructed by taking the visual recognition related information in the training set as the model input and the fire warning results as the model output.

[0024] In combination with the first aspect, in one embodiment, visual recognition related information and case fire conditions are extracted from the collected videos of fire accident cases as visual recognition related information in the data set, including but not limited to smoke obstruction area, flame color, flame size, and surrounding heat wave pictures.

[0025] In combination with the first aspect, in one implementation, the visual recognition related information in the training set is used as the model input, and the fire warning result is used as the model output to construct the visual recognition model. The fire warning algorithm rules in the visual recognition model are as follows:

[0026] When the matching degree between the fluctuation video and the instantaneous change picture in the training set and the unique visual features of the fire reaches a first preset matching degree, it is determined that a fire will occur; wherein the unique visual features of the fire include flame shape, flame color, preset change rate range of the smoke occlusion area, and surrounding heat wave visual features;

[0027] When the matching degree between the fluctuation video and the instantaneous change picture in the training set and the unique visual features of the fire is between the first preset matching degree and the second preset matching degree, a feature re-matching operation is performed to obtain a re-matching result;

[0028] According to the re-comparison results, the fire warning results are obtained.

[0029] In combination with the first aspect, in one implementation, obtaining a fire warning result according to the second comparison result specifically includes the following steps:

[0030] When the result of the re-comparison is that the degree of match between the fluctuating video and the instantaneous changing image and the unique visual features of the fire is between the first preset matching degree and the second preset matching degree, the feature re-comparison operation is performed until the degree of match between the fluctuating video and the instantaneous changing image and the unique visual features of the fire is no longer between the first preset matching degree and the second preset matching degree, or the preset number of cycles is reached, and the fire warning result is obtained based on the final comparison result.

[0031] In combination with the first aspect, in one embodiment, when the re-comparison result is that the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual feature of the fire is between the first preset matching degree and the second preset matching degree, a feature re-comparison operation is performed until the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual feature of the fire is not between the first preset matching degree and the second preset matching degree, and a fire warning result is obtained according to the final comparison result, which specifically includes the following steps:

[0032] When the re-comparison result shows that the matching degree between the fluctuation video and the instantaneous change picture and the unique visual feature of the fire is greater than the first preset matching degree, the subsequent feature re-comparison operation is not performed, and it is determined that a fire will occur;

[0033] When the re-matching result is that the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual features of the fire is less than the first preset matching degree, the subsequent feature re-matching operation is not performed, and it is determined that no fire will occur.

[0034] In combination with the first aspect, in one implementation, extracting fire visual features from a captured on-site video specifically includes the following steps:

[0035] When the picture changes in the live video exceed the preset ratio, capture the fluctuating video and the instantaneous changing picture;

[0036] Extract fire visual features from fluctuating videos and instantaneous changing images.

[0037] In a second aspect, the present application provides a fire warning system based on visual recognition, comprising:

[0038] Early warning model building module, used to build a visual recognition model based on collected fire accident cases;

[0039] On-site video capture module, used to capture on-site video to be used for fire warning;

[0040] A fire visual feature capture module, which is in communication with the on-site video capture module and is used to extract fire visual features from the captured on-site video;

[0041] The fire warning module is in communication with the model building module and the fire visual feature capturing module, and is used to identify the extracted fire visual features based on the constructed visual recognition model to obtain a fire warning result.

[0042] In conjunction with the second aspect, in one implementation, the early warning model building module includes:

[0043] Fire case collection unit, used to collect fire accident cases;

[0044] A data set acquisition unit, which is in communication with the fire case collection unit and is used to extract visual recognition related information and case fire conditions from the collected videos of fire accident cases as a data set;

[0045] A training set division unit, connected in communication with the data set acquisition unit, and configured to divide a training set from the data set;

[0046] The model building unit is connected to the training set division unit for building a visual recognition model by taking the visual recognition related information in the training set as the model input and taking the fire warning result as the model output.

[0047] In a third aspect, the present application provides a computer-readable storage medium, on which a fire warning program based on visual recognition is stored, wherein when the fire warning program based on visual recognition is executed by a processor, the steps of the fire warning method based on visual recognition as described above are implemented.

[0048] The beneficial effects brought by the technical solution provided by the embodiment of the present application include at least:

[0049] By extracting fire visual features from the on-site video to be used for fire warning, and inputting them into the constructed visual recognition model for fire warning judgment, it does not need to rely on simple sensors. It only needs to deploy cameras to capture on-site videos and then make recognition judgments. It is suitable for low-cost fire warning deployment and can monitor the occurrence of fires in real time and intuitively. It is not limited by physical sensors, which improves the accuracy and reliability of fire detection. The fire warning algorithm based on visual recognition can identify the unique characteristics of flames and other unique visual features of fire, such as color, shape, dynamic changes, etc. These features enable the algorithm to distinguish between flames and other objects, reduce the false alarm rate, and realize intelligent and accurate warning of fire conditions based on visual judgment;

[0050] The fire warning algorithm based on visual recognition can identify the unique visual characteristics of the fire, reducing the false alarm rate;

[0051] Fire warning algorithms based on visual recognition can achieve all-round monitoring. By installing multiple cameras, it can cover a wider area and detect fire hazards in a timely manner. Compared with traditional fire warning systems that can only monitor a limited range, visual recognition algorithms have greatly improved the coverage and response speed of fire warnings.

[0052] Fire warning algorithms based on visual recognition can also provide rich fire information. By analyzing video images, the algorithm can determine the starting location of the fire, the size of the fire, and the direction of spread, etc., providing valuable decision-making support for the fire department, helping to formulate more accurate rescue plans and improve the efficiency of fire response. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A schematic diagram of the method flow of the fire warning method based on visual recognition in this application;

[0054] Figure 2 This is another method flow chart of the fire warning method based on visual recognition in this application;

[0055] Figure 3 This is a functional module block diagram of the fire warning system based on visual recognition in this application. DETAILED DESCRIPTION

[0056] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0057] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.

[0058] In the description of the embodiments of the present application, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.

[0059] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or. For example, A / B can mean A or B. The “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0060] In some processes described in the embodiments of the present application, multiple operations or steps that appear in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or in parallel, and the sequence number of the operation is only used to distinguish the different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0061] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0062] First, as Figure 1 As shown, the present application provides a fire warning method based on visual recognition, which is suitable for areas with poor infrastructure or limited budget, including the following steps:

[0063] Step S1: constructing a visual recognition model based on the collected fire accident cases;

[0064] Step S2: Capturing on-site video of the fire warning to be carried out; wherein the visual information includes video and / or pictures;

[0065] Step S3: extracting fire visual features from the captured on-site video;

[0066] Step S4: Identify the extracted fire visual features based on the constructed visual recognition model to obtain fire warning results.

[0067] This application extracts fire visual features from the on-site video to be used for fire warning, and inputs them into the constructed visual recognition model for fire warning judgment. It does not need to rely on sensors, but only needs to deploy cameras to capture on-site videos for recognition and judgment. It is suitable for low-cost fire warning deployment and can monitor the occurrence of fire in real time and intuitively. It is not limited by physical sensors, and improves the accuracy and reliability of fire detection. The fire warning algorithm based on visual recognition can recognize the unique characteristics of flames and other unique visual features of fire, such as color, shape, dynamic changes, etc. These features enable the algorithm to distinguish between flames and other objects, reduce the false alarm rate, and realize intelligent and accurate warning of fire based on visual judgment.

[0068] The fire warning algorithm based on visual recognition can identify the unique visual characteristics of the fire, reducing the false alarm rate;

[0069] Fire warning algorithms based on visual recognition can achieve all-round monitoring. By installing multiple cameras, it can cover a wider area and detect fire hazards in a timely manner. Compared with traditional fire warning systems that can only monitor a limited range, visual recognition algorithms have greatly improved the coverage and response speed of fire warnings.

[0070] Fire warning algorithms based on visual recognition can also provide rich fire information. By analyzing video images, the algorithm can determine the starting location of the fire, the size of the fire, and the direction of spread, etc., providing valuable decision-making support for the fire department, helping to formulate more accurate rescue plans and improve the efficiency of fire response.

[0071] In one embodiment, the step S1: constructing a visual recognition model based on the collected fire accident cases specifically includes the following steps:

[0072] Step S11: Collect fire accident cases; further, if it is necessary to implement fire warning for a specific industry, collect industry-related fire accident cases to carry out more accurate visual recognition warning of fire conditions in the industry and reduce visual recognition interference factors;

[0073] Step S12: extracting visual recognition related information and case fire conditions from the collected videos of fire accident cases as a data set; wherein the visual recognition related information includes but is not limited to smoke obstruction area, flame color, flame size, and surrounding heat wave pictures;

[0074] Step S13: dividing a training set from the data set;

[0075] Step S14: Use the visual recognition related information in the training set as the model input and the fire warning result as the model output to build a visual recognition model.

[0076] In one embodiment, step S13 also includes further dividing a test set from the data set. After step S14, the visual recognition related information in the test set is used as the model input, and the fire warning result is used as the model output. The visual recognition model trained based on the training set is tested and verified to further improve the robustness of the model.

[0077] In one embodiment, the step S14: using the visual recognition related information in the training set as the model input and the fire warning result as the model output to construct the visual recognition model specifically includes the following steps:

[0078] Step S141: When the matching degree between the fluctuation video and the instantaneous change picture in the training set and the unique visual features of the fire reaches a first preset matching degree, it is determined that a fire will occur; wherein the unique visual features of the fire include flame shape, flame color, preset change rate range of the smoke shielding area, and surrounding heat wave visual features;

[0079] Step S1421: When the matching degree between the fluctuation video and the instantaneous change picture in the training set and the unique visual features of the fire is between the first preset matching degree and the second preset matching degree, a feature re-matching operation is performed to obtain a re-matching result; in this step, the feature re-matching is to compare the fluctuation video and the instantaneous change picture in the training set with the unique visual features of the fire;

[0080] Step S1422: Obtaining fire warning results according to the re-comparison results;

[0081] Step S143: Compare the fire warning results output by the model with the case fire conditions in the training set, and adjust the fire warning algorithm rules in the visual recognition model based on the comparison results, specifically, the preset values ​​of the first preset matching degree and / or the second preset matching degree, until the fire warning results output by the visual recognition model are consistent with the case fire conditions, so as to obtain a visual recognition model with higher prediction accuracy.

[0082] In one embodiment, obtaining the fire warning result according to the re-comparison result specifically includes the following steps:

[0083] When the result of the re-comparison is that the degree of match between the fluctuating video and the instantaneous changing image and the unique visual features of the fire is between the first preset matching degree and the second preset matching degree, the feature re-comparison operation is performed until the degree of match between the fluctuating video and the instantaneous changing image and the unique visual features of the fire is no longer between the first preset matching degree and the second preset matching degree, or the preset number of cycles is reached, and the fire warning result is obtained based on the final comparison result.

[0084] In a more specific embodiment, in the feature comparison or feature re-comparison, the comparison of the fluctuating video and the instantaneous changing picture with the unique visual features of the fire is specifically implemented as follows:

[0085] When one or more objects in the fluctuating video and the instantaneous changing picture match the flame shape, or, the picture appearing in the fluctuating video and the instantaneous changing picture matches the flame color, or, smoke occlusion occurs, or, heat waves appear around, it is determined that a fire has occurred on the scene, and then the next step of fire prediction and determination is performed. Specifically, when the number of objects in the shape of flames in the fluctuating video and the instantaneous changing picture increases at a rate that exceeds a number increase threshold, or, the rate of increase in the smoke occlusion area exceeds an area increase threshold, or, the area of ​​the surrounding heat waves exceeds a heat wave increase threshold, it is determined that a fire will occur or is occurring on the scene.

[0086] In one embodiment, when the re-comparison result is that the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual feature of the fire is between the first preset matching degree and the second preset matching degree, the feature re-comparison operation is performed until the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual feature of the fire is not between the first preset matching degree and the second preset matching degree, and the fire warning result is obtained according to the final comparison result, which specifically includes the following steps:

[0087] When the re-comparison result shows that the matching degree between the fluctuation video and the instantaneous change picture and the unique visual feature of the fire is greater than the first preset matching degree, the subsequent feature re-comparison operation is not performed, and it is determined that a fire will occur;

[0088] When the re-matching result is that the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual features of the fire is less than the first preset matching degree, the subsequent feature re-matching operation is not performed, and it is determined that no fire will occur.

[0089] In a specific embodiment, the visual recognition algorithm rules of the visual recognition model constructed based on the collected fire accident cases in step S1 are as follows:

[0090] When the matching degree between the fluctuating video and instantaneous changing images in the training set and the unique visual features of the fire reaches 80% (the parameters can be customized), a fire alarm is generated and fire safety operations are performed (the operation content can be customized);

[0091] When the matching degree between the fluctuating video and instantaneous changing images in the training set and the unique visual features of the fire is less than 50%, the matching degree is still less than 50% after multiple comparisons in a loop. In this case, the abnormal changing images will be pushed to the remote user, allowing the remote user to decide whether to initiate a fire warning and perform corresponding fire emergency operations.

[0092] In one embodiment, the fire scene is artificially simulated and the actual parameters on site are measured, which are compared with the parameters of the visual recognition model, and the parameters in the visual recognition algorithm rules of the existing visual recognition model are corrected. At the same time, this parameter is used as a baseline parameter, and the corrected parameters are used to analyze the sorted fire videos and pictures, and specific judgment parameters are optimized in turn, thereby improving the accuracy of fire warning and judgment.

[0093] In one embodiment, the step S2: capturing the on-site video for fire warning, specifically includes the following steps:

[0094] Deploy high-definition cameras at the scene where fire warning is to be carried out. Preferably, high-definition cameras of multiple angles are deployed in multiple areas of the scene where fire warning is to be carried out; specifically, the scene area where fire warning is to be carried out is the photovoltaic station area, and high-definition cameras are deployed in key areas of the photovoltaic station (such as photovoltaic module installation area, cable access point, box inverter, etc.) to ensure that the camera's field of view can cover the entire station, and the camera's resolution is high enough to capture subtle changes when a fire occurs. High-definition cameras play a vital role in the monitoring and early warning of harmful gases and smoke with their excellent image capture capabilities. It can capture and transmit clear and accurate on-site images in real time, and even subtle harmful gas diffusion or thick smoke can be accurately presented in the picture. Through the intelligent analysis system, high-definition cameras can quickly identify abnormal conditions and issue warnings in time, providing strong support for environmental safety and accident prevention and control; ensure that all cameras can be stably connected to the central server or monitoring center for real-time transmission of video data; provide a stable power supply for the camera, and consider setting up a backup power supply to cope with possible power outages.

[0095] In one embodiment, step S3: extracting fire visual features from the captured on-site video, specifically includes the following steps:

[0096] When the picture changes in the live video exceed the preset ratio, capture the fluctuating video and the instantaneous changing picture;

[0097] Extract fire visual features from fluctuating videos and instantaneous changing images.

[0098] In a more specific embodiment, step S3: extracting fire visual features from the captured on-site video, specifically includes the following steps:

[0099] Develop a software system that can receive and process high-definition video streams in real time, including decoding, frame extraction, image processing and other operations of video data, and capture the fluctuating video and instantaneous changing images when the image changes in the live video exceed the preset ratio

[0100] Based on the data generated by historical fires (such as videos, high-definition pictures) and publicly available online fire video data, computer vision and image processing technology are used to extract fire visual features, such as flame color, shape, dynamic changes, smoke-obstructed area, and heat wave phenomena generated in the surrounding area when the fire occurs, and record them.

[0101] In one embodiment, step S4: identifying the extracted fire visual features based on the constructed visual recognition model to obtain a fire warning result, specifically includes the following steps:

[0102] Real-time detection and early warning: Specifically, the trained visual recognition model is integrated into the video stream processing software to achieve real-time fire detection of photovoltaic stations. When the model detects fire characteristics, the early warning mechanism is triggered to generate a fire warning signal;

[0103] Early warning signal transmission and display: Specifically, early warning signals are transmitted through the network to the monitoring center or the mobile devices of relevant personnel so that fires can be discovered and handled in a timely manner. Early warning signals can include video clips, pictures, location information, etc., to help personnel quickly locate the fire and take appropriate countermeasures.

[0104] Alarm and linkage; specifically, when the fire warning result is judged to be a fire, the alarm device (such as alarm, sound and light alarm, etc.) can be automatically triggered according to the level of the warning signal and the preset alarm rules, or it can be linked with the fire protection system to automatically start the fire extinguishing device.

[0105] In one embodiment, step S4: after the extracted fire visual features are identified based on the constructed visual recognition model and the fire warning result is obtained, the following steps are also included:

[0106] Further optimize the constructed visual recognition model, including:

[0107] Test the entire visual recognition-based fire warning method or system in a real environment and obtain test results, including recognition performance, warning response time, false alarm rate and other indicators in different scenarios;

[0108] According to the test results, the fire warning method or system based on visual recognition is optimized and adjusted, including adjusting the camera angle and position, optimizing the algorithm parameters of the visual recognition model, and improving the alarm rules.

[0109] The fire warning method based on visual recognition in this application is flexible and scalable. With the development of technology, the algorithm can be continuously optimized and updated to adapt to different environments and fire scenarios. At the same time, the algorithm can also be combined with other intelligent systems to achieve more intelligent fire warning and rescue.

[0110] In one embodiment, step S4: after the extracted fire visual features are identified based on the constructed visual recognition model and the fire warning result is obtained, the following steps are also included:

[0111] Maintain and upgrade the visual recognition-based fire warning system, including:

[0112] Regular maintenance; specifically, regularly clean and maintain the camera to ensure the quality of images captured from the scene is not affected, and check whether the network connection and power supply are stable and reliable.

[0113] System upgrades; specifically, as technology continues to evolve and new algorithms emerge, the system can be regularly upgraded and improved to improve the accuracy and reliability of fire warnings.

[0114] In specific applications, this application can implement a specific solution for fire warning in photovoltaic station areas using visual recognition technology using high-definition cameras.

[0115] It should be noted that in practical applications, many factors need to be considered, such as changes in ambient light, the impact of obstructions, the identification of different types of fires, etc. These need to be fully considered and processed during the algorithm training and system optimization of the visual recognition model.

[0116] like Figure 2 As shown, the fire warning method based on visual recognition provided by this application is implemented as follows:

[0117] Build a visual recognition model that embeds fire warning algorithm rules;

[0118] Collect and organize fire accident cases in the industry (including videos and pictures of fires) to develop visual recognition models to extract key data before or at the early stage of fires, such as smoke-blocked area, flame color and size, and surrounding heat wave pictures;

[0119] Build a visual recognition environment for production, use high-definition equipment to cover the area to be detected, and transmit data to the central processor in a timely manner. The central processor obtains the fire scene video, processes the video stream, and extracts the fire visual features from it;

[0120] The visual recognition model in the central processing unit processes the output of visual features of fires, performs fire warnings and judgments, conducts self-learning on the detected information, adds the detected fire information to the existing database, and improves the accuracy of fire judgments; for fire information that has not been detected, manual judgment assistance is added to improve the basic information for fire warnings, and human assistance is provided to extract key fire data and add it to the database.

[0121] Second, as Figure 3As shown, the present application provides a fire warning system based on visual recognition, including: a warning model construction module 100, an on-site video capture module 200, a fire visual feature capture module 300 and a fire warning module 400; the warning model construction module 100 is used to construct a visual recognition model based on collected fire accident cases; the on-site video capture module 200 is used to capture on-site videos to be used for fire warning; the fire visual feature capture module 300 is communicated with the on-site video capture module 200 and is used to extract fire visual features from the captured on-site videos; the fire warning module 400 is communicated with the model construction module 100 and the fire visual feature capture module 300 and is used to identify the extracted fire visual features based on the constructed visual recognition model to obtain a fire warning result.

[0122] In one embodiment, the early warning model building module includes:

[0123] Fire case collection unit, used to collect fire accident cases;

[0124] A data set acquisition unit, which is in communication with the fire case collection unit and is used to extract visual recognition related information and case fire conditions from the collected videos of fire accident cases as a data set;

[0125] A training set division unit, connected in communication with the data set acquisition unit, and configured to divide a training set from the data set;

[0126] The model building unit is connected to the training set division unit for building a visual recognition model by taking the visual recognition related information in the training set as the model input and taking the fire warning result as the model output.

[0127] Among them, the functional implementation of each module in the above-mentioned fire warning system based on visual recognition corresponds to the various steps in the above-mentioned fire warning method embodiment based on visual recognition, and its functions and implementation processes will not be repeated here one by one.

[0128] On the third aspect, an embodiment of the present application provides a fire warning device based on visual recognition. The fire warning device based on visual recognition can be a personal computer (PC), a laptop computer, a server, or other device with data processing capabilities.

[0129] The communication interface includes input / output (I / O) interface, physical interface and logical interface, etc., which are used to realize the interconnection of devices inside the visual recognition-based fire warning device, and the interface used to realize the interconnection of the visual recognition-based fire warning device with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, an optical fiber interface, an ATM interface, etc.; the user device can be a display, a keyboard, etc.

[0130] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0131] The processor may be a general-purpose processor, which may call the fire warning program based on visual recognition stored in the memory and execute the fire warning method based on visual recognition provided in the embodiment of the present application. For example, the general-purpose processor may be a central processing unit (CPU). The method executed when the fire warning program based on visual recognition is called may refer to the various embodiments of the fire warning method based on visual recognition of the present application, which will not be described in detail here.

[0132] In a fourth aspect, an embodiment of the present application also provides a readable storage medium.

[0133] The readable storage medium of the present application stores a fire warning program based on visual recognition, wherein when the fire warning program based on visual recognition is executed by a processor, the steps of the fire warning method based on visual recognition as described above are implemented.

[0134] Among them, the method implemented when the fire warning program based on visual recognition is executed can refer to the various embodiments of the fire warning method based on visual recognition in this application, and will not be repeated here.

[0135] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present application.

[0137] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A fire warning method based on visual recognition, characterized in that: The following steps are involved: Build a visual recognition model based on the collected fire accident cases; Capture live video for fire warnings; Extracting fire visual features from captured on-site videos; The extracted fire visual features are identified based on the constructed visual recognition model to obtain fire warning results.

2. The fire warning method based on visual recognition according to claim 1, characterized in that: The construction of a visual recognition model based on the collected fire accident cases specifically includes the following steps: Collect fire accident cases; Extract visual recognition related information and case fire conditions from the collected videos of fire accident cases as a dataset; Divide the training set from the dataset; The visual recognition model is constructed by taking the visual recognition related information in the training set as the model input and the fire warning results as the model output.

3. The fire warning method based on visual recognition according to claim 2, characterized in that: Visual recognition related information and case fire conditions are extracted from the collected videos of fire accident cases as visual recognition related information in the dataset, including but not limited to smoke obstruction area, flame color, flame size, and surrounding heat wave pictures.

4. The fire warning method based on visual recognition according to claim 2, characterized in that: The visual recognition related information in the training set is used as the model input, and the fire warning result is used as the model output to construct the visual recognition model. The fire warning algorithm rules in the visual recognition model are as follows: When the matching degree between the fluctuation video and the instantaneous change picture in the training set and the unique visual features of the fire reaches a first preset matching degree, it is determined that a fire will occur; wherein the unique visual features of the fire include flame shape, flame color, preset change rate range of the smoke occlusion area, and surrounding heat wave visual features; When the matching degree between the fluctuation video and the instantaneous change picture in the training set and the unique visual features of the fire is between the first preset matching degree and the second preset matching degree, a feature re-matching operation is performed to obtain a re-matching result; According to the re-comparison result, the fire warning result is obtained; Compare the fire warning results output by the model with the case fire conditions in the training set. According to the comparison results, adjust the fire warning algorithm rules in the visual recognition model until the fire warning results output by the visual recognition model are consistent with the case fire conditions, and obtain the visual recognition model trained based on the training set.

5. The fire warning method based on visual recognition according to claim 4, characterized in that: The method of obtaining the fire warning result according to the second comparison result specifically includes the following steps: When the result of the re-comparison is that the degree of match between the fluctuating video and the instantaneous changing image and the unique visual features of the fire is between the first preset matching degree and the second preset matching degree, the feature re-comparison operation is performed until the degree of match between the fluctuating video and the instantaneous changing image and the unique visual features of the fire is no longer between the first preset matching degree and the second preset matching degree or the preset number of cycles is reached, and the fire warning result is obtained based on the final comparison result.

6. The fire warning method based on visual recognition according to claim 5, characterized in that: When the re-comparison result is that the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual feature of the fire is between the first preset matching degree and the second preset matching degree, the feature re-comparison operation is performed until the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual feature of the fire is not between the first preset matching degree and the second preset matching degree, and the fire warning result is obtained according to the final comparison result, which specifically includes the following steps: When the re-comparison result shows that the matching degree between the fluctuation video and the instantaneous change picture and the unique visual feature of the fire is greater than the first preset matching degree, the subsequent feature re-comparison operation is not performed, and it is determined that a fire will occur; When the re-matching result is that the matching degree between the fluctuating video and the instantaneous changing picture and the unique visual features of the fire is less than the first preset matching degree, the subsequent feature re-matching operation is not performed, and it is determined that no fire will occur.

7. The fire warning method based on visual recognition according to claim 1, characterized in that: Extracting fire visual features from the captured on-site video includes the following steps: When the picture changes in the live video exceed the preset ratio, capture the fluctuating video and the instantaneous changing picture; Extract fire visual features from fluctuating videos and instantaneous changing images.

8. A fire warning system based on visual recognition, characterized in that: include: Early warning model building module, used to build a visual recognition model based on collected fire accident cases; On-site video capture module, used to capture on-site video to be used for fire warning; A fire visual feature capture module, which is in communication with the on-site video capture module and is used to extract fire visual features from the captured on-site video; The fire warning module is in communication with the model building module and the fire visual feature capturing module, and is used to identify the extracted fire visual features based on the constructed visual recognition model to obtain a fire warning result.

9. The fire warning system based on visual recognition according to claim 8, characterized in that: The early warning model building module includes: Fire case collection unit, used to collect fire accident cases; A data set acquisition unit, which is in communication with the fire case collection unit and is used to extract visual recognition related information and case fire conditions from the collected videos of fire accident cases as a data set; A training set division unit, connected in communication with the data set acquisition unit, and configured to divide a training set from the data set; The model building unit is connected to the training set division unit for building a visual recognition model by taking the visual recognition related information in the training set as the model input and taking the fire warning result as the model output.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a fire warning program based on visual recognition, wherein when the fire warning program based on visual recognition is executed by a processor, the steps of the fire warning method based on visual recognition as described in any one of claims 1 to 7 are implemented.

Images