Alarm system based on Internet of Things and image recognition
Through the alarm system of the Internet of Things and image recognition, combined with the target detection and color space segmentation module, the suspected flame area is extracted, and the support vector machine is used for flame recognition, which solves the problem of inaccurate fire warning in large space places, and achieves fast and accurate fire detection and alarm.
Patent Information
- Application Number
- CN202510008852.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-03
- Publication Date
- 2025-05-30
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to quickly and accurately complete fire warnings in large space places, resulting in large losses before the fire spread.
An alarm system of the Internet of Things and image recognition is adopted to collect dual-band video images through a flame detector, and a suspected flame area is extracted in combination with the target detection and color space segmentation module, a support vector machine is used to perform flame recognition, and real-time data is sent to the upper computer monitoring system through Internet of Things communication.
It realizes timely reminding managers when a fire occurs, with a low false alarm rate and can curb the fire before the fire expands, improving the accuracy of flame detection.
Smart Images
Figure CN120071540A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of fire alarm, and particularly relates to an alarm system combining Internet of Things and image recognition. Background Art
[0002] A fire is a combustion that gets out of control in terms of time or space, which endangers public safety and hinders social development. In shopping malls, factories and other places, once a fire breaks out, the loss of life and property will be extremely large. In the prior art, automatic fire prevention systems are installed in many public places or buildings. The occurrence of a fire is detected by a temperature sensor, and when high temperature is detected, the fire sprinkler is controlled to spray water for fire extinguishing.
[0003] Currently, the number of large shopping malls and super-large space buildings is increasing continuously. Such buildings are densely populated, and once a fire breaks out, it will cause great damage. Currently, in large space venues, smoke alarms are mostly used to judge the occurrence of a fire. When the alarm is successful in this way, the fire has spread greatly and caused great losses. It is very difficult to complete fire early warning quickly and accurately in large space venues by this method. Summary of the Invention
[0004] The purpose of the present invention is to provide an alarm system combining Internet of Things and image recognition to solve the technical problems raised in the background art.
[0005] To achieve the above purpose, the specific technical solution of the present invention is as follows: An alarm system combining Internet of Things and image recognition includes an Internet of Things module, which is connected to a flame detector, a fire alarm controller, an alarm device, a linkage controller, a fire extinguishing system and a host computer monitoring system; The flame detector is responsible for the acquisition and encoding of on-site dual-band video images, video image processing, flame detection and recognition, video image data storage and transmission; The fire alarm controller is used to judge according to the results of the data collected by the flame detector and control the alarm device to give an audible and visual warning; The linkage controller is used to control the fire extinguishing system on site; The Internet of Things module is used to realize the communication between the flame detector and the host computer monitoring system; The host computer monitoring system is used to connect to multiple flame detectors in the on-site monitoring environment in real time, responsible for searching and controlling the front-end flame detectors, displaying the video images of the monitoring site, saving the video and providing video playback, fire alarm response, system operation and query of alarm logs.
[0006] Preferably, the flame detector includes an ARM processor, and the ARM processor is connected to a camera module, an image storage module, a power supply module and a flame detection and recognition module; The power supply module supplies power to the camera module, the power supply module, the ARM processor, and the flame detection and recognition module respectively; The camera module is used to obtain on-site video images; The image storage module provides storage space for storing and processing images; The ARM processor is used to process image data; The flame detection and recognition module is used to judge the image data and transmit the results to the host computer monitoring system and the fire alarm controller respectively.
[0007] Preferably, the camera module includes a visible light camera and a near-infrared camera.
[0008] Preferably, the flame detection and recognition module includes a target detection module, the target detection module is respectively connected with a color space segmentation module and a threshold segmentation module, a region growing and filling module is connected between the color space segmentation module and the threshold segmentation module, the region growing and filling module is connected with a feature extraction module, the feature extraction module is connected with a feature normalization module, and the feature normalization module is connected with a support vector machine classification and recognition module.
[0009] Preferably, the target detection module is used to detect and segment moving targets in the current video image, the color space segmentation module is used to extract suspected flame regions in the moving targets of the video image according to color, and the threshold segmentation module is used to extract the suspected flame regions in the moving targets of the video image.
[0010] Preferably, the region growing and filling module is used to completely fill the segmented region.
[0011] Preferably, the feature extraction module is used to utilize the learning ability of the network to learn the shape, texture, and color features of the flame and further discriminate the flame candidate region image.
[0012] Preferably, the feature normalization module is used to normalize the extracted features, and the support vector machine classification and recognition module is used to train the normalized feature data to obtain the flame recognition model of the system.
[0013] Preferably, the host computer monitoring system includes a camera search and control module, and the camera search and control module is connected with a video image browsing module, a video saving and playback module, and a log query module.
[0014] An alarm system for the Internet of Things and image recognition according to the present invention has the following advantages: 1. The present invention can obtain on-site images in real time and conduct flame detection, give an alarm according to the detection results, and send real-time data to the host computer monitoring system through the Internet of Things. When a fire occurs, it can timely remind the management personnel and provide the situation of the fire scene, can timely detect the occurrence of an early fire and give an alarm, has a low false alarm rate, and can ensure that the fire is contained before the fire spreads.
[0015] 2. The present invention extracts the suspected flame regions in visible light images by combining a target detection module and color space segmentation for the collected video images, extracts the suspected flame regions in near-infrared images by combining a target detection module and a threshold segmentation module, finally fills the extracted regions completely by using a region growing and filling module, extracts the features of the extracted suspected flame regions, extracts color moments and local binary patterns in the visible light region, and extracts circularity, contour roughness, area change rate, and centroid change rate in the near-infrared region. The extracted features are combined into a sequence and input into a support vector machine for training a flame recognition model, thus greatly improving the accuracy of flame detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 is a schematic diagram of the overall structure of an alarm system for the Internet of Things and image recognition according to the present invention; Figure 2 is a schematic diagram of the structure of a flame detector of an alarm system for the Internet of Things and image recognition according to the present invention; Figure 3 is a schematic diagram of the structure of a host computer monitoring system of an alarm system for the Internet of Things and image recognition according to the present invention; Figure 4 is a schematic diagram of the structure of a flame detection and recognition module of an alarm system for the Internet of Things and image recognition according to the present invention. DETAILED DESCRIPTION OF THE INVENTION
[0017] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.
[0018] As Figures 1-4 shown, an alarm system for the Internet of Things and image recognition according to the present invention includes an Internet of Things module, and the Internet of Things module is connected to a flame detector, a fire alarm controller, an alarm device, a linkage controller, a fire extinguishing system, and a host computer monitoring system; The flame detector is responsible for collecting and encoding dual-band video images on site, video image processing, flame detection and recognition, video image data storage and transmission; the fire alarm controller is used to judge according to the results of the data collected by the flame detector and control the alarm device to give an audible and visual warning; the linkage controller is used to control the fire extinguishing system on site; the Internet of Things module is used to realize the communication between the flame detector and the upper computer monitoring system; the upper computer monitoring system is used to connect multiple flame detectors in the on-site monitoring environment in real time, and is responsible for searching and controlling the front-end flame detectors, displaying the video images of the monitoring site, saving the video and providing video playback, fire alarm response, system operation and query of alarm logs; Further, in specific implementation, the site and objects prone to fire are closely monitored by the fire detector for a long time. Once a fire occurs in the place monitored by the fire detector, the collected fire signal will be converted into an electronic signal by it, and will be judged and processed by the ARM processor and the flame detection and recognition module. If it is in an abnormal state, a fire alarm bell will be sounded, and an audible and visual alarm device in the alarm device will give a warning, and the staff at the fire site will be reminded to stay away from the fire site as soon as possible or carry out fire extinguishing treatment. The linkage controller will also start the fire extinguishing system on site to prevent the fire from spreading and expanding.
[0019] Among them, the flame detector includes an ARM processor, and the ARM processor is connected with a camera module, an image storage module, a power supply module and a flame detection and recognition module; the power supply module supplies power to the camera module, the power supply module, the ARM processor and the flame detection and recognition module respectively; the camera module is used to obtain on-site video images; the image storage module provides storage space for storing and processing images; the ARM processor is used to process image data; the flame detection and recognition module is used to judge the image data and transmit the results to the upper computer monitoring system and the fire alarm controller respectively; the camera module includes a visible light camera and a near-infrared camera; Further, in specific implementation, after the video image is collected, it is analyzed and judged in real time whether there is a flame in the current area. The ARM processor can complete functions such as video collection and encoding / decoding, output display, denoising and sharpening processing, image stitching, and distortion correction. Based on the media processing platform, the video input module is applied to collect the original data of two cameras, and the received image data is converted into the YUV422 format by the video processing module and subjected to cropping, horizontal and vertical reduction processing. The image is enhanced, denoised, sharpened and filtered by the intelligent video engine module. Finally, the video is encoded by the video encoding module and uploaded to the upper computer monitoring system in real time for browsing and saving, and the on-site video image is stored and backed up in the flame detector.
[0020] Among them, the flame detection and recognition module includes a target detection module, which is respectively connected to a color space segmentation module and a threshold segmentation module. There is a region growing and filling module connected between the color space segmentation module and the threshold segmentation module. The region growing and filling module is connected to a feature extraction module, the feature extraction module is connected to a feature normalization module, and the feature normalization module is connected to a support vector machine classification and recognition module. The target detection module is used to detect and segment the moving targets in the current video image. The color space segmentation module is used to extract the suspected flame regions in the moving targets of the video image according to the color. The threshold segmentation module is used to extract the suspected flame regions in the moving targets of the video image; Further, in specific implementation, the flame detector is usually fixed at a certain position for video acquisition. The videos collected within a period of time are continuous, and there is similarity between each frame. When there is no object movement in the environment, the change between frames is extremely weak. Once there is a moving object in the environment, there will be a drastic change between several frames. When a fire occurs, the flame generated by the combustion of combustibles usually flickers, and the flames will jump and shake. Therefore, the flame can be regarded as a moving object. Through moving target detection, the moving targets in the current video image can be well detected and segmented, excluding the interference of the static area. After the moving targets in the region are extracted through the work of the color space segmentation module, there will still be other moving interference parts that are not flames. At this time, it is necessary to further segment the target region to extract the suspected flame regions. When the flame is burning, its color has a special distribution law, which is generally very different from the surrounding environment. The flame color has a sense of hierarchy, mainly white inside and mainly red-yellow outside. And as the temperature rises, the flame color will change from red to yellow to white. Therefore, the distribution law of the flame color can be found in different color spaces of the visible light video image and the region can be segmented accordingly. When the threshold segmentation module works, since most of the visible light bands are filtered out in the near-infrared video image and only the infrared band spectrum is allowed to pass through, it is impossible to normally segment and extract the target region from the color space. At this time, the threshold segmentation module is used to extract the suspected flame regions.
[0021] Among them, the region growing and filling module is used to completely fill the segmented region; Further, in specific implementation, the pixel points with similar properties in the connected region are divided into the same region. The specific principle is to start from the initial region, select a central point as the seed pixel point, and merge the pixel points with similar properties in the surrounding connected region into this region. In this way, the region continues to grow and expand until there are no similar pixel points in the connected region, and the region growth stops. The gray mean value, texture, and color of the region can be used as the judgment of the similarity between pixels.
[0022] Among them, the feature extraction module is used to utilize the learning ability of the network to learn the shape, texture, and color features of the flame and further discriminate the flame candidate region image; Furthermore, in specific implementation, the basic features of the flame include static features, dynamic features, and texture features. The static features include the brightness, color, and shape contour (circularity, number of sharp corners, eccentricity) of the flame. The dynamic features include shape similarity, flickering characteristics, area change rate, and edge jitter characteristics. The texture features include local binary pattern and gray-level co-occurrence matrix.
[0023] Among them, the feature normalization module is used to normalize the extracted features, and the support vector machine classification and recognition module is used to train the normalized feature data to obtain the flame recognition model of the system; Furthermore, in specific implementation, this model is used for the classification and recognition of flames and interferences. Before training the model, flame and interference video images are selected for feature data extraction. The video images come from the network and are obtained by shooting with the system flame detector. There are a total of 30 videos, including 15 flame videos and 15 interference videos, and the video duration is greater than 1 minute. Among them, the flame video images include fuel fire, alcohol fire, and lighter fire, and the interference video images include flashlight, incandescent lamp, red light, moving car light, and crowd. 12-dimensional color moment features, 1-dimensional circularity feature, 1-dimensional contour roughness feature, 1-dimensional area change rate feature, 2-dimensional centroid change rate feature, and 10-dimensional rotation-invariant unified local binary pattern histogram features are extracted, for a total of 27-dimensional feature data. Finally, 72,212 groups of sample data are extracted, including 36,540 groups of flame data and 35,671 groups of interference data. 70% of the extracted sample data is used for support vector machine training, and 30% is used to verify the trained model to obtain the accuracy rate of the flame recognition model. The training and verification steps are repeated 10 times, and the average value of the accuracy rates of the 10 models is used as the final score of this model; In this process, the process of the flame detection and recognition module is as follows: (1) Extraction of suspected flame region: Motion target detection is performed on the two-channel video images collected, the motion region is segmented, then region extraction is performed on the visible light image in the color space, region extraction is performed on the near-infrared image using the threshold segmentation module, and finally the region growing filling module is used for internal filling of the region to obtain a complete suspected flame region; (2) Feature extraction: After the suspected flame region is extracted, feature data of this region is extracted, including 27-dimensional feature data such as color moment and local binary pattern of the visible light image region, and circularity, contour roughness, area change rate, and centroid change rate of the near-infrared image region; (3)Flame recognition: Normalize the extracted features, input the normalized features into a vector machine for training to obtain the flame recognition model of the system, and use this model to classify and identify flames and interferences.
[0024] Among them, the host computer monitoring system includes a camera search and control module, and the camera search and control module is connected to a video image browsing module, a video saving and playback module, and a log query module; Furthermore, in specific implementation, the camera search and control module is responsible for searching for the front-end flame detectors and adding them to the device information table of the system, managing the users of the system and their corresponding permissions. The video image browsing module is responsible for real-time displaying and saving the video images of the monitoring site, receiving the fire alarm signals sent by the flame detectors and giving alarms, and controlling the connected flame detectors. The video saving and playback module is responsible for playing back the locally saved monitoring videos and the videos and pictures of the fire scene during the alarm. The log query module is responsible for querying the operation logs and alarm logs of the system and printing and saving the logs.
[0025] It can be understood that the present invention is described through some embodiments. Those skilled in the art know that, without departing from the spirit and scope of the present invention, various changes or equivalent replacements can be made to these features and embodiments. Additionally, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the scope protected by the present invention.
Claims
1. An Internet of Things and image recognition alarm system, including an Internet of Things module, characterized in that: The Internet of Things module is connected to a flame detector, a fire alarm controller, an alarm device, a linkage controller, a fire extinguishing system and a host computer monitoring system; The flame detector is used for collecting and encoding dual-band video images on site, processing video images, detecting and identifying flames, and storing and transmitting video image data; The fire alarm controller is used to judge and control the alarm device to provide sound and light warning according to the result of the data collected by the flame detector; The linkage controller is used to control the fire extinguishing system on site; The Internet of Things module is used to realize the communication between the flame detector and the host computer monitoring system; The host computer monitoring system is used to connect to multiple flame detectors in the on-site monitoring environment in real time, and is responsible for searching and controlling the front-end flame detectors, displaying video images of the monitoring site, saving videos and providing video playback, fire alarm response, system operation and alarm log query.
2. According to claim 1, the alarm system of Internet of Things and image recognition is characterized in that: The flame detector includes an ARM processor, and the ARM processor is connected to a camera module, an image storage module, a power module and a flame detection and identification module; The power module supplies power to the camera module, the power module, the ARM processor and the flame detection and recognition module respectively; The camera module is used to obtain live video images; The image storage module provides storage space for storing and processing images; The ARM processor is used to process image data; The flame detection and recognition module is used to judge the image data and transmit the results to the host computer monitoring system and the fire alarm controller respectively.
3. The Internet of Things and image recognition alarm system according to claim 2, characterized in that: The camera module includes a visible light camera and a near infrared camera.
4. The Internet of Things and image recognition alarm system according to claim 1, characterized in that: The flame detection and recognition module includes a target detection module, which is respectively connected to a color space segmentation module and a threshold segmentation module, a region growth and filling module is connected between the color space segmentation module and the threshold segmentation module, the region growth and filling module is connected to a feature extraction module, the feature extraction module is connected to a feature normalization module, and the feature normalization module is connected to a vector machine classification and recognition module.
5. The Internet of Things and image recognition alarm system according to claim 4, characterized in that: The target detection module is used to detect and segment the moving target in the current video image, the color space segmentation module is used to extract the suspected flame area in the moving target in the video image according to the color, and the threshold segmentation module is used to extract the suspected flame area in the moving target in the video image.
6. The Internet of Things and image recognition alarm system according to claim 4, characterized in that: The region growing and filling module is used to completely fill the segmented region.
7. The Internet of Things and image recognition alarm system according to claim 4, characterized in that: The feature extraction module is used to utilize the learning ability of the network to learn the shape, texture, and color features of the flame and further distinguish the flame candidate area image.
8. The Internet of Things and image recognition alarm system according to claim 4, characterized in that: The feature normalization module is used to perform normalization processing on the extracted features, and the vector machine classification and recognition module is used to train the normalized feature data to obtain a flame recognition model of the system.
9. The Internet of Things and image recognition alarm system according to claim 1, characterized in that: The host computer monitoring system comprises a camera search and control module, and the camera search and control module is connected to a video image browsing module, a video storage and playback module and a log query module.