Fire automatic detection system based on image recognition

Abstract

The invention relates to the technical field of fire early warning, and discloses an automatic fire detection system based on image recognition. The fire automatic detection system comprises an image acquisition module used for acquiring image information of a monitoring area in real time; the image preprocessing module processes and analyzes the acquired image data; the fire feature extraction module extracts colors, shapes and motion trails of flames and smog by using a computer vision technology; the fire identification and classification module uses a machine learning algorithm to classify and identify the extracted features; the early warning and alarm module is used for immediately sending out an early warning or alarm signal once a fire disaster is detected; and the recording and playback module is used for recording a fire detection process and supporting subsequent query and analysis. Through the image recognition technology, the system can recognize flames and smoke more accurately, and false alarm and missing alarm are avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of fire early warning, and in particular to an automatic fire detection system based on image recognition. Background Art

[0002] With the acceleration of urbanization and increasing population density, large-scale commercial complexes, high-rise buildings, and underground spaces are constantly emerging, increasing the risk of fire. The consequences of a fire in these places are disastrous, placing increasing demands on fire detection systems. People are increasingly concerned about the safety of life and property, raising expectations for early warning and rapid response to fires.

[0003] Traditional fire detection systems are susceptible to environmental interference, such as dust and water vapor, which can cause false alarms. In some dusty industrial environments, smoke detectors often issue false alarm signals due to the high dust content. Furthermore, installation locations are limited, typically requiring installation on ceilings or walls, which can make them ineffective for detecting fires in certain locations. For example, smoke detectors are significantly less effective in high-ceiling spaces or areas obscured by shelves.

[0004] Therefore, there is an urgent need for a technology to replace the existing fire warning method to solve the problem of how to improve the reliability of fire detection, improve the accuracy and timeliness of fire warning, and reduce fire losses. Summary of the Invention

[0005] The embodiment of the present application provides an automatic fire detection system based on image recognition, thereby solving the problems in related technologies of improving the reliability of fire detection, increasing the accuracy and timeliness of fire warnings, and reducing fire losses.

[0006] Among them, an automatic fire detection system based on image recognition provided according to an embodiment of the present application includes: an image acquisition module, an image preprocessing module, a fire feature extraction module, a fire identification and classification module, an early warning and alarm module, and a recording and playback module;

[0007] The image acquisition module is used to acquire video images of the area to be monitored in real time;

[0008] The image preprocessing module is used to perform noise reduction and enhancement processing on the collected images;

[0009] The fire feature extraction module uses computer vision technology to extract the features of flames and smoke after image processing;

[0010] The fire feature extraction module adopts a multi-threshold segmentation unit, sets multiple thresholds, compares the color value of each pixel in the image with the set multiple thresholds, and divides the pixels in the image into different categories;

[0011] When the hue value of a pixel is in the range of [0,30] or [330,360], the saturation value is in the range of [0.4,1], and the lightness value is in the range of [0.4,1], the pixel is marked as a flame area;

[0012] When the hue value is in the range of [0,180], the saturation value is in the range of [0,0.2], and the lightness value is in the range of [0.6,1], the pixel is marked as a smoke area;

[0013] If it is within other ranges, it is marked as a suspected area;

[0014] The fire identification and classification module uses a machine learning algorithm to classify and identify the features extracted;

[0015] The early warning and alarm module is used to send fire warning information to users;

[0016] The recording and playback module is used to record the fire detection process and support subsequent query and analysis.

[0017] Furthermore, the image acquisition module includes multiple camera modules, and the areas to be detected between the multiple camera modules are connected to each other;

[0018] The image acquisition module calibrates the white balance and color gain parameters of the camera module according to a standard color card every three months.

[0019] Furthermore, the image preprocessing module includes a noise reduction processing unit and an image enhancement unit. The noise reduction processing unit adopts a Gaussian filtering algorithm to determine the filter size and standard deviation, and performs a convolution operation on each pixel in the image to perform noise reduction.

[0020] Furthermore, the image enhancement unit obtains a grayscale component image in the original image, and through a CPLD processor, performs grayscale value correction on the pixel points of the grayscale component image according to the relative brightness and darkness relationship between the pixel points of the grayscale component image and the pixel points in the grayscale component image after Gaussian smoothing, and synthesizes the grayscale component image that has undergone grayscale value correction to obtain an enhanced image.

[0021] Furthermore, the fire feature extraction module extracts the flame and smoke features after image processing, which also include shape features, texture features and motion features.

[0022] Furthermore, the fire identification and classification module performs weighted analysis on the fire using the features extracted by the fire feature extraction module, with the flame feature importance weight being 0.4, the smoke feature importance weight being 0.4, and the other fire extracted features importance weight being 0.2.

[0023] Multiply the extracted characteristic values of each fire by the corresponding weight and add them together to obtain the comprehensive fire value and set the fire threshold range;

[0024] When the comprehensive fire value is less than the set fire threshold range, it is determined that no fire has occurred;

[0025] When the comprehensive fire value is within the set fire threshold range, the fire is determined to be a small fire;

[0026] When the comprehensive fire value is greater than the set fire threshold range, the fire is determined to be a large fire.

[0027] Furthermore, the fire recognition and classification module uses a machine learning algorithm to learn from a large number of labeled fire images and non-fire images, and automatically extracts features that can effectively distinguish between fire and non-fire. During the training process, the algorithm automatically searches for various fire-related features in the image and converts them into feature vectors that can be used for classification. The trained model is used to classify new images to determine whether they are fire images.

[0028] Furthermore, the machine learning algorithm is subjected to deep learning based on a convolutional neural network, and a deep convolutional neural network model is constructed to allow the model to automatically learn the characteristic representation of fire from images.

[0029] Furthermore, the alarm modes of the early warning and alarm module include sound and light alarm, SMS and email notification and linkage with the fire protection system.

[0030] Furthermore, the recording and playback module is equipped with a dedicated storage device, and stores various types of data in a data storage format. When storing, the data is stored in an orderly manner according to the time sequence and data category, and an index structure is established;

[0031] The recorded data can be played back in chronological order. The video can be played back at normal speed, fast speed or slow speed.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] Using image recognition technology, the system can more accurately identify flames and smoke, avoiding false alarms and missed alerts. Compared to traditional smoke sensors, image recognition technology captures more fire characteristics, improving fire detection accuracy. It also processes video images in real time, enabling timely fire detection and early warning. This is crucial for fire prevention and control, buying valuable time for firefighting and rescue efforts. This ensures the system's high stability and reliability, enabling it to operate in a variety of scenarios. Even in complex environments, such as dense smoke or low light, the system can accurately identify fires through image preprocessing and feature extraction.

[0034] The technical solution of the present application is further described in detail below through the accompanying drawings and examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0036] The present application can be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

[0037] Figure 1 This is a structural block diagram of an automatic fire detection system based on image recognition proposed in this application. DETAILED DESCRIPTION

[0038] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangements of components and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present application.

[0039] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0040] The following description of at least one exemplary embodiment is merely illustrative in nature and is not intended to limit the present disclosure, its application, or uses.

[0041] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0042] It should be noted that like reference numerals and letters refer to like items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0043] In addition, the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the fact that ordinary technicians in this field can implement it. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0044] It should be noted that all directional indications in the embodiments of the present application (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0045] The following combination Figure 1 The following describes an automatic fire detection system based on image recognition according to an exemplary embodiment of the present application. It should be noted that the following application scenarios are provided solely to facilitate understanding of the spirit and principles of the present application, and the embodiments of the present application are not limited in this respect. Rather, the embodiments of the present application can be applied to any applicable scenario.

[0046] This application proposes an automatic fire detection system based on image recognition.

[0047] Among them, an automatic fire detection system based on image recognition provided according to an embodiment of the present application includes: an image acquisition module, an image preprocessing module, a fire feature extraction module, a fire identification and classification module, an early warning and alarm module, and a recording and playback module;

[0048] The image acquisition module is used to collect video images of the area to be monitored in real time;

[0049] The image preprocessing module is used to perform noise reduction and enhancement on the collected images;

[0050] The fire feature extraction module uses computer vision technology to extract the features of flames and smoke after image processing;

[0051] The fire feature extraction module uses a multi-threshold segmentation unit to set multiple thresholds, compare the color value of each pixel in the image with the set multiple thresholds, and divide the pixels in the image into different categories;

[0052] When the hue value of a pixel is in the range of [0,30] or [330,360], the saturation value is in the range of [0.4,1], and the lightness value is in the range of [0.4,1], the pixel is marked as a flame area;

[0053] When the hue value is in the range of [0,180], the saturation value is in the range of [0,0.2], and the lightness value is in the range of [0.6,1], the pixel is marked as a smoke area;

[0054] If it is within other ranges, it is marked as a suspected area;

[0055] The fire identification and classification module uses machine learning algorithms to classify and identify the extracted features;

[0056] The early warning and alarm module is used to send fire warning information to users;

[0057] The record and playback module is used to record the fire detection process and support subsequent query and analysis.

[0058] Specifically, the image preprocessing module performs noise reduction and enhancement on captured images, effectively removing noise interference caused by environmental and equipment factors, while also enhancing key image elements such as contrast and clarity. This results in higher-quality images processed by subsequent modules and reduces feature extraction errors or recognition biases caused by poor image quality.

[0059] Specifically, the fire feature extraction module leverages computer vision technology and employs a multi-threshold segmentation unit. By setting multiple thresholds, it meticulously divides pixel categories and accurately extracts flame and smoke features. This multi-threshold approach is more sophisticated than a single threshold and can more accurately distinguish between flames, smoke, and other areas that appear suspicious but not fire-related, effectively avoiding misjudgments and improving the accuracy of fire assessments.

[0060] Specifically, the fire identification and classification module uses machine learning algorithms to classify and identify the extracted features. The machine learning algorithm can learn the subtle differences and patterns between fire features and non-fire features from a large amount of data, further improving the accuracy of fire situation judgments, reducing the probability of false alarms and missed alarms, and ensuring that the system can accurately identify real fire incidents and classify them reasonably.

[0061] As you can understand, image recognition technology enables the system to more accurately identify flames and smoke, avoiding false alarms and missed alerts. Compared to traditional smoke sensors, image recognition technology captures more fire characteristics, improving fire detection accuracy. It also processes video images in real time, enabling timely fire detection and early warning. This is crucial for fire prevention and control, buying valuable time for firefighting and rescue efforts. This ensures the system's high stability and reliability, enabling it to operate in a variety of scenarios. Even in complex environments, such as dense smoke or low light, the system can accurately identify fires through image preprocessing and feature extraction.

[0062] Specifically, the image acquisition module includes multiple camera modules, and the areas to be detected between the multiple camera modules are interconnected;

[0063] The image acquisition module calibrates the white balance and color gain parameters of the camera module according to the standard color card every three months.

[0064] Specifically, the image acquisition module calibrates the camera module's white balance and color gain parameters every three months using a standard color chart. White balance calibration ensures the camera accurately reproduces the true color of objects under varying lighting conditions, preventing color casts caused by lighting factors and ensuring the precise capture of the color characteristics of fire-related elements such as flames and smoke.

[0065] It's understandable that regular calibration allows the camera module to maintain stable image acquisition performance over extended periods of use. Over time and as the environment changes, camera imaging may be affected by factors such as device aging and changes in lighting conditions. Regular calibration is a proactive response to these potential influencing factors. It ensures that the image acquisition module consistently outputs high-quality, color-accurate images regardless of environmental conditions. This allows subsequent modules to operate based on reliable image data, enhancing the entire automatic fire detection system's ability to cope with complex and changing environments and its long-term stable operation.

[0066] Specifically, the image preprocessing module includes a noise reduction processing unit and an image enhancement unit. The noise reduction processing unit adopts a Gaussian filtering algorithm to determine the filter size and standard deviation, and performs a convolution operation on each pixel in the image to perform noise reduction.

[0067] It is understandable that the noise reduction processing unit uses the Gaussian filtering algorithm to perform convolution operations on each pixel in the image by reasonably determining the filter size and standard deviation to achieve noise reduction. Gaussian filtering is very effective in removing common Gaussian noise in images. It can give different weights based on the distance and grayscale value difference between the pixel and other pixels in the neighborhood. While smoothing the image and removing noise, it can also better preserve the details and edge information of the image. This effectively eliminates the noise that may exist in the original image collected due to environmental interference, the equipment itself, etc., making the image clearer and cleaner, providing a high-quality image foundation for subsequent image enhancement and more in-depth fire feature extraction, and reducing feature extraction errors or misjudgments caused by noise interference.

[0068] Specifically, the image enhancement unit obtains the grayscale component image in the original image, and through the CPLD processor, performs grayscale value correction on the pixel points of the grayscale component image according to the relative brightness and darkness relationship between the pixel points of the grayscale component image and the pixel points in the grayscale component image after Gaussian smoothing, and synthesizes the grayscale component images that have undergone grayscale value correction to obtain an enhanced image.

[0069] It's understandable that correcting grayscale values by comparing the relative brightness of pixels in the original grayscale component image with the Gaussian-smoothed grayscale component image can accurately capture the true brightness differences between different parts of the image. For fire detection, flames typically have relatively high brightness, and the brightness of smoke areas also differs from the surrounding background. This grayscale value correction, based on detailed pixel comparison, can specifically enhance the characteristics of these fire-related areas, making flames brighter and more prominent, and smoke more clearly visible, making them more distinct from the background. This helps the subsequent fire feature extraction module extract key features more efficiently and accurately.

[0070] Specifically, the fire feature extraction module extracts the features of flames and smoke after image processing, including shape features, texture features and motion features.

[0071] Specifically, the fire identification and classification module performs weighted analysis on fires using the features extracted by the fire feature extraction module, with the flame feature importance weight set to 0.4, the smoke feature importance weight set to 0.4, and the other fire extracted features importance weight set to 0.2.

[0072] Multiply the extracted characteristic values of each fire by the corresponding weight and add them together to obtain the comprehensive fire value and set the fire threshold range;

[0073] When the comprehensive fire value is less than the set fire threshold range, it is determined that no fire has occurred;

[0074] When the comprehensive fire value is within the set fire threshold range, the fire is determined to be a small fire;

[0075] When the comprehensive fire value is greater than the set fire threshold range, the fire is determined to be a large fire.

[0076] Specifically, the fire feature extraction module uses not only color features but also shape, texture, and motion features to characterize flames and smoke from multiple dimensions. Shape features can capture unique patterns such as the irregular, upward spread of flames and the cloud-like diffusion of smoke. Texture features help distinguish the flickering dynamics of flames and the hazy, gradual changes in smoke. Motion features determine whether fire characteristics are met based on the movement patterns of flames and smoke during combustion and diffusion. This multi-dimensional feature extraction approach provides a more comprehensive and detailed characterization of fires, avoiding misjudgments that can occur when relying solely on a single feature.

[0077] As you can understand, by comparing the comprehensive fire value with the set fire threshold range, fires are classified into different categories: no fire, small fire, and large fire, thus achieving a graded assessment of fire scale. This graded assessment has important practical significance, allowing for targeted response measures for fires of different sizes. For example, if a fire is determined to be small, on-site personnel can use simple fire-fighting equipment to promptly put out the fire and prevent it from spreading. However, if a fire is determined to be large, a more comprehensive fire emergency plan must be quickly activated, such as notifying the fire department and organizing a large-scale evacuation. This grading mechanism helps improve the efficiency of fire response and minimize the casualties and property losses caused by fires.

[0078] Specifically, the fire identification and classification module uses machine learning algorithms to learn from a large number of labeled fire images and non-fire images, and automatically extracts features that can effectively distinguish between fire and non-fire. During the training process, the algorithm automatically searches for various fire-related features in the image and converts them into feature vectors that can be used for classification. The trained model is used to classify new images to determine whether they are fire images.

[0079] Specifically, deep learning is performed on the machine learning algorithm based on the convolutional neural network. By constructing a deep convolutional neural network model, the model automatically learns the characteristic representation of fire from images.

[0080] Specifically, the warning and alarm module's alarm modes include sound and light alarms, SMS and email notifications, and linkage with the fire protection system.

[0081] Specifically, the recording and playback module is equipped with a dedicated storage device, which uses a data storage format to store various types of data. When storing, the data is stored in an orderly manner according to the time sequence and data category, and an index structure is established;

[0082] The recorded data can be played back in chronological order. The video can be played back at normal speed, fast speed or slow speed.

[0083] It's understandable that specialized storage devices and standardized data storage formats ensure that all types of data are fully preserved, preventing data loss or corruption due to improper storage. From the start of the fire detection system's operation, all relevant data is recorded in an orderly manner, forming a complete "data chain." This is crucial for long-term fire monitoring and understanding the system's operating status at different times. This makes the entire fire detection process traceable, enabling detailed information to be retrieved at any time.

[0084] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0085] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. An automatic fire detection system based on image recognition, characterized in that: include: Image acquisition module, image preprocessing module, fire feature extraction module, fire identification and classification module, early warning and alarm module, and recording and playback module; The image acquisition module is used to acquire video images of the area to be monitored in real time; The image preprocessing module is used to perform noise reduction and enhancement processing on the collected images; The fire feature extraction module uses computer vision technology to extract the features of flames and smoke after image processing; The fire feature extraction module adopts a multi-threshold segmentation unit, sets multiple thresholds, compares the color value of each pixel in the image with the set multiple thresholds, and divides the pixels in the image into different categories; When the hue value of a pixel is in the range of [0,30] or [330,360], the saturation value is in the range of [0.4,1], and the lightness value is in the range of [0.4,1], the pixel is marked as a flame area; When the hue value is in the range of [0,180], the saturation value is in the range of [0,0.2], and the lightness value is in the range of [0.6,1], the pixel is marked as a smoke area; If it is within other ranges, it is marked as a suspected area; The fire identification and classification module uses a machine learning algorithm to classify and identify the features extracted; The early warning and alarm module is used to send fire warning information to users; The recording and playback module is used to record the fire detection process and support subsequent query and analysis.

2. The image recognition automatic fire detection system according to claim 1, characterized in that: The image acquisition module includes multiple camera modules, and the areas to be detected between the multiple camera modules are connected to each other; The image acquisition module calibrates the white balance and color gain parameters of the camera module according to a standard color card every three months.

3. The image recognition automatic fire detection system according to claim 1, characterized in that: The image preprocessing module includes a noise reduction processing unit and an image enhancement unit. The noise reduction processing unit adopts a Gaussian filtering algorithm to determine the filter size and standard deviation, and performs a convolution operation on each pixel in the image to perform noise reduction.

4. The automatic fire detection system based on image recognition according to claim 3, characterized in that: The image enhancement unit obtains a grayscale component image in the original image, and performs grayscale value correction on the pixel points of the grayscale component image according to the relative brightness and darkness relationship between the pixel points of the grayscale component image and the pixel points in the grayscale component image after Gaussian smoothing through a CPLD processor, and synthesizes the grayscale component images that have undergone grayscale value correction to obtain an enhanced image.

5. The automatic fire detection system based on image recognition according to claim 1, characterized in that: The fire feature extraction module extracts the flame and smoke features after image processing, including shape features, texture features and motion features.

6. The automatic fire detection system based on image recognition according to claim 1, characterized in that: The fire identification and classification module performs weighted analysis on fires using the features extracted by the fire feature extraction module, with the flame feature importance weight being 0.4, the smoke feature importance weight being 0.4, and the other fire extracted features importance weight being 0.

2. Multiply the extracted characteristic values of each fire by the corresponding weight and add them together to obtain the comprehensive fire value and set the fire threshold range; When the comprehensive fire value is less than the set fire threshold range, it is determined that no fire has occurred; When the comprehensive fire value is within the set fire threshold range, the fire is determined to be a small fire; When the comprehensive fire value is greater than the set fire threshold range, the fire is determined to be a large fire.

7. The automatic fire detection system based on image recognition according to claim 6, characterized in that: The fire recognition and classification module uses a machine learning algorithm to learn from a large number of labeled fire images and non-fire images, and automatically extracts features that can effectively distinguish between fire and non-fire. During the training process, the algorithm automatically searches for various fire-related features in the image and converts them into feature vectors that can be used for classification. The trained model is used to classify new images to determine whether they are fire images.

8. The automatic fire detection system based on image recognition according to claim 7, characterized in that: The machine learning algorithm is deeply learned based on a convolutional neural network. By constructing a deep convolutional neural network model, the model is allowed to automatically learn the characteristic representation of fire from images.

9. The automatic fire detection system based on image recognition according to claim 1, characterized in that: The alarm modes of the early warning and alarm module include sound and light alarm, SMS and email notification and linkage with the fire protection system.

10. The automatic fire detection system based on image recognition according to claim 1, characterized in that: The recording and playback module is equipped with a dedicated storage device and uses a data storage format to store various types of data. When storing, the data is stored in an orderly manner according to the time sequence and data category, and an index structure is established; The recorded data can be played back in chronological order. The video can be played back at normal speed, fast speed or slow speed.

Images