Fire Detection Method Research Articles

Improved Real-Time Fire Warning System Based on Advanced Technologies for Visually Impaired People.

Early fire detection and notification techniques provide fire prevention and safety information to blind and visually impaired (BVI) people within a short period of time in emergency situations when fires occur in indoor environments. Given its direct impact on human safety and the environment, fire detection is a difficult but crucial problem. To prevent injuries and property damage, advanced technology requires appropriate methods for detecting fires as quickly as possible. In this study, to reduce the loss of human lives and property damage, we introduce the development of the vision-based early flame recognition and notification approach using artificial intelligence for assisting BVI people. The proposed fire alarm control system for indoor buildings can provide accurate information on fire scenes. In our proposed method, all the processes performed manually were automated, and the performance efficiency and quality of fire classification were improved. To perform real-time monitoring and enhance the detection accuracy of indoor fire disasters, the proposed system uses the YOLOv5m model, which is an updated version of the traditional YOLOv5. The experimental results show that the proposed system successfully detected and notified the occurrence of catastrophic fires with high speed and accuracy at any time of day or night, regardless of the shape or size of the fire. Finally, we compared the competitiveness level of our method with that of other conventional fire-detection methods to confirm the seamless classification results achieved using performance evaluation matrices.

Spatio-temporal fire detection based on brightness temperature change in Himawari-8 images

ABSTRACT The geostationary satellite Himawari-8 has become the primary source for fire remote-sensing detection because of its advantages of high frequency, large width and easy access. To solve the problem of difficulty in small fire detection and with a high omission on Himawari-8 image, a spatio-temporal fire detection (STFD) method based on Himawari-8 image brightness temperature change is proposed. Calculate the mean value difference of brightness temperature between multi-time series images before the target time, and use the least square model to fit the ideal mean value difference of brightness temperature between the target time and the image at the previous time. Then, the difference value of the actual brightness temperature between the target time and the last image time is calculated. The initial brightness temperature change pixels are obtained by the comparative analysis method. Based on the spatial statistical characteristics, the potential fire point is determined by the brightness temperature value ratio of the image, temperature mean and standard temperature deviation. Then, combined with the spatio-temporal context information, the persistent fire points are detected according to the threshold conditions to supplement undetected persistent fires. The results of experiments on the Himawari-8 images in Guizhou, China and southwestern Australia indicate that STFD exhibits superior performance on small fire detection points, and the accuracy is better than the several traditional methods.

Flickering Signal Flame Detection Method Using a Matched Filter

Flame combustion is a type of decomposition in which materials are combusted using oxygen in the air and a visible flame. During combustion, the respiratory process is detected as a flickering signal. Flame detectors recognize combustion by monitoring the generated infrared (IR) radiation. However, IR signals derived from artificial lighting and sunlight are also detected. Thus, a method for distinguishing nonflame-derived IR signals, such as artificial lighting, from flame-derived IR signals is required. In this study, a method for identifying a flickering signal using a matched filter that distinguishes between nonflame- and flame-derived IR signals was investigated. Matched filters selectively irradiate a frequency of the target signal by regulating the frequency of the waveform generator. When the frequencies of the inspection and flickering signal waveforms coincide, phase information may be obtained from the target waveform, and phase deviation may be minimized. Additionally, the magnitude of the detected IR signal, including the flickering signal, can be determined from the evaluation results of the waveform in the phase-locked state. Therefore, the flickering detection method using a matched waveform filter is effective for identifying IR signals and can be used as an economical fire detection method.

Fire Detection and Localization Method Based on Deep Learning in Video Surveillance

Fire detection and localization in video surveillance had become a particularly important part of disaster rescue. Considering the fire detection and localization of slow detection speed, low detection accuracy, and low localization precision in video surveillance, we proposed a fire detection and localization method based on deep learning. The first thing we improved the SuperPoint method to extract video keyframe in video surveillance. The next thing we employed Convolutional Neural Network (CNN) model to detect the fire on the extracted video keyframes. The last thing we located the fire via superpixel and CNN on the extracted video keyframes which broke out a fire. The experimental results on open fire dataset revealed that the recall of keyframe extraction reached 0.83, the precision of fire detection reached 0.96 and the F1-score of fire localization reached 0.90. Our method realized rapid and accurate detection and precise localization of fire in video surveillance.

An Intelligent Fire Detection Algorithm and Sensor Optimization Strategy for Utility Tunnel Fires

With the rapid development of utility tunnels, fire safety after construction is increasingly important, especially in the cable compartment. An intelligent fire detection method based on the particle swarm optimization algorithm was proposed for fire state estimation of the utility tunnel, including the fire source location, the maximum temperature value, and the temperature attenuation coefficient. Additionally, a corresponding sensor optimization strategy was also established. The dispersion coefficient of the fire source location was defined as the judgment criteria of sensor optimization. The validity of the proposed algorithm and the sensor optimization strategy were demonstrated in the application of a full-scale experimental example. The maximum errors of the identified fire source location and the maximum temperature value after sensor optimization were 34.7164 m and 8.5403°C, respectively. The total number of temperature sensors was reduced by more than 50%. The proposed intelligent fire detection algorithm can provide precise guidance for fire protection and extinguishing plan. Particularly, the sensor optimization strategy can economize the cost of temperature sensors.

CCTV 기반 화염의 특성과 밀도 기반 공간 클러스터링을 이용한 화재 감지 방법

In this study, we propose fire detection method using CCTV-based flame features and density-based spatial clustering with noise (DBSCAN). To do this, first, the 1st candidate region using the color of the flame image is extracted and the 2nd candidate region using the high-frequency region and background removal is extracted. Next, the extracted 1st candidate region and 2nd candidate region are merged, and the clustering region is extracted using DBSCAN. And then, the method for judging flame and rhinitis through the number of blocks passing through the movement trajectory of the central point of the clustering region extracted using DBSCAN is proposed. Finally, the applicability of the method proposed in this paper is reviewed through experiments in indoor and outdoor environments.

Multi‐step implicit Adams predictor‐corrector network for fire detection

Fire detection methods based on the Convolutional Neural Networks (CNN) have advantages of high accuracy, wide coverage and robustness, receiving significant attention from researchers. Among CNN-based methods, ResNet has achieved better performance than other CNN frameworks in fire detection system, since it uses stacked residual blocks to enlarge the receptive field to overcome the vanishing gradient problem with residual learning. The merits of ResNet can be attributed to the similarity between ResNet and the single-step explicit solver for Ordinary Differential Equations (ODEs), for example, the Euler method. Motivated by the theory of numerical ODE that a multi-step implicit solver has higher accuracy than a single-step explicit solver, the Multi-step Implicit Adams predictor-corrector (MIAPC) network for fire detection is proposed. The MIAPC method is first mapped to a corresponding predictor-corrector Adams block which achieves higher accuracy than a single-step explicit solver. Then, Adaptive Feature Fusion (AFF) and the Spatial Attention Layer (SAL) are utilized to extract hierarchical features from stacked predictor-corrector Adams blocks, forming the corresponding Adams module. Finally, the 4 Adams modules which are made of 4, 6, 8, 10 predictor-corrector Adams blocks and followed by AFF and SAL form the crucial ODE-based approximation part in the proposed network. By adding a simple feature extraction and detection in front of and after the ODE-based approximation part, the MIAPC network is built. Experiments demonstrate that the method achieves 87% accuracy in the challenging test dataset, outperforming existing methods by at least 6%. Besides, the 5.3M model size with inference speed of 4.7 frames/second in CPU and 65.7 frames/second in GPU enables the proposed method to be used in practical applications.

A Robust Fire Detection Model via Convolution Neural Networks for Intelligent Robot Vision Sensing.

Accurate fire identification can help to control fires. Traditional fire detection methods are mainly based on temperature or smoke detectors. These detectors are susceptible to damage or interference from the outside environment. Meanwhile, most of the current deep learning methods are less discriminative with respect to dynamic fire and have lower detection precision when a fire changes. Therefore, we propose a dynamic convolution YOLOv5 fire detection method using a video sequence. Our method first uses the K-mean++ algorithm to optimize anchor box clustering; this significantly reduces the rate of classification error. Then, the dynamic convolution is introduced into the convolution layer of YOLOv5. Finally, pruning of the network heads of YOLOv5’s neck and head is carried out to improve the detection speed. Experimental results verify that the proposed dynamic convolution YOLOv5 fire detection method demonstrates better performance than the YOLOv5 method in recall, precision and F1-score. In particular, compared with three other deep learning methods, the precision of the proposed algorithm is improved by 13.7%, 10.8% and 6.1%, respectively, while the F1-score is improved by 15.8%, 12% and 3.8%, respectively. The method described in this paper is applicable not only to short-range indoor fire identification but also to long-range outdoor fire detection.

UFS-Net: A unified flame and smoke detection method for early detection of fire in video surveillance applications using CNNs

Fire is a recurring event that usually causes a lot of social, environmental, ecological, and economic damage in different environments. Therefore, machine vision-based fire detection can be one of the most important tasks in modern surveillance systems. Most of the existing computer vision-based fire detection methods are only able to detect a single case of flame or smoke. In this research, a unified flame and smoke detection approach, termed “UFS-Net,” based on deep learning is proposed. An efficient and tailored convolutional neural network architecture is designed to detect both fire flames and smoke in video frames. UFS-Net is capable of identifying fire hazards by classifying video frames into eight classes: 1) flame, 2) white smoke, 3) black smoke, 4) flame and white smoke, 5) flame and black smoke, 6) black smoke and white smoke, 7) flame, white smoke and black smoke, and 8) normal status. To further increase the reliability of UFS-Net, a decision module based on a voting scheme is applied. In addition, a rich annotated dataset named “UFS-Data” that includes 849,640 images and 26 videos, captured/collected from various data sources and artificial images made in this research, is prepared for training and evaluation of UFS-Net. Extensive experiments conducted on “UFS-Data” and other benchmark datasets (i.e., “Mivia,” “BoWFire,” and “FireNet”), and the comparisons with state-of-the-art methods, confirm the high performance of UFS-Net. All the implementation source codes and the “UFS-Data” are made publicly available at https://github.com/alihosseinice/UFS-Net.

A visualized fire detection method based on convolutional neural network beyond anchor

A visualized fire detection method based on convolutional neural network beyond anchor

Multi-parameter fire detection method based on feature depth extraction and stacking ensemble learning model

The losses caused by fire are becoming more and more serious, with the continuous development of social economy. If fire is detected as early as possible, the losses can be reduced more than 80%. The detection accuracy of traditional single-parameter fire detection is low, which often leads to false and missing fire alarm. This paper proposes a multi-parameter fire detection method based on feature depth extraction and stacking ensemble learning model. Fire feature parameters, such as temperature, smoke concentration and CO concentration, are collected as the original data. The algorithms, based on XGboost, gradient boosted decision tree (GBDT), random forest and decision tree, are used to analyze and sort the importance of nine types of data, such as the slope of smoke concentration, the amplitude of CO concentration etc. Three types of data with high importance are selected to extract depth features using Hilbert Huang transform (HHT), one-dimensional convolution neural network (1D-CNN) and long short term memory neural network (LSTMNN). The extracted features are input into stacking ensemble learning model. Stacking ensemble learning model is a hierarchical combination learning classifier and improves the performance by learning the classification bias of the upper model. Experimental results show that the proposed method has higher accuracy, stability and efficiency against several single models (e.g., XGBoost, GBDT, and LSTMNN).

Real-Time Fire Detection Method for Electric Vehicle Charging Stations Based on Machine Vision

During the charging process of electric vehicles (EV), the circuit inside the charger plug is connected in series, the charger input voltage does not match the rated input voltage, the temperature caused by the severe heating of the charging time is too high for too long, and other factors are very likely to trigger a fire in the vehicle charging pile. In this paper, an improved You Only Look Once v4 (YOLOv4) real-time target detection algorithm based on machine vision is proposed to monitor the site based on existing monitoring equipment, transmit live video information in real-time, expand the monitoring range, and significantly reduce the cost of use. During the experiment, the improved neural network model was trained by a homemade fire video image dataset, and a K-means clustering algorithm iwasintroduced to recalculate the anchor frame size for the specific object of flame; the existing dataset was used to perform multiple divisions by using a tenfold cross-validation algorithm, thus avoiding the selection of chance hyperparameters and models that do not have generalization ability because of special divisions. The experimental results show that the improved algorithm is fast and accurate in detecting large-size flames in real-time and small-size flames at the beginning of a fire, with a detection speed of 43 fps/s, mAP value of 91.53%, and F1 value of 0.91. Compared with YOLOv3 and YOLOv4 models, the improved model is sensitive to detecting different sizes of flames. It can suppress false alarms well in a variety of complex lighting environments. The prediction frame size fits the area where the target is located, the detection accuracy remains stable, and the comprehensive performance of the network model is significantly improved to meet the demand of real-time monitoring. It is significant for developing the EV industry and enhancing emergency response capability.

CNN-based fire detection method on autonomous ships using composite channels composed of RGB and IR data

Since fires grow exponentially after an outbreak, it is crucial to extinguish them quickly. In the case of fires occurring on commercial ships and naval vessels, smoke and heat in the narrow enclosed compartments of the ships make it difficult to extinguish fires, threatening human safety, and causing massive property damage. So-called autonomous ships are operated without crew members in charge of fire suppression. Therefore, it is crucial to improve the performance and reliability of the automatic fire detection and suppression system. However, the standard fire detection system installed on ships is not adequate for the level of fire risk. Existing fire detection methods still have false alarm problems, and it takes a long time for heat and smoke from the fire to reach the detector. In this study, the use of combined channel data (RGB-IR) is proposed as an image-based fire detection method applicable to ships. In nature, many animals obtain multiple wavelength information, which is advantageous for hunting or risk preparedness in a wild environment. In this way, it is assumed that both the characteristics of RGB data in the visible area and IR data in the infrared area may be utilized through the combined channel data. A fire detection model with composite channel data input was built using deep learning with a convolution neural network (CNN), a fire detection model using composite channel data input was constructed, and hyper-parameters were tuned during the training process to determine the optimal model and compare it with the existing RGB and IR models, respectively. Compared with the model using only conventional RGB or IR data, the fire detection accuracy of the model using RGB-IR combined channel data increased and the false detection rate decreased.

Fire and smoke detection with deep learning: A review

This paper outlines a state-of-the-art method for smoke and fire detection utilizing Convolutional Neural Networks (CNNs). The current smoke detectors installed in buildings pose a challenge for effective fire detection. The inefficiency of traditional methods in terms of speed and cost led to the exploration of using Artificial Intelligence (AI) to identify and alert from Closed Circuit Television (CCTV) footage. In this paper, an analytical overview of AI is conducted by using a selfcreated dataset of video frames containing flames and smoke. The data undergoes pre-processing before being used to train a CNN-based machine learning model. The goal of this review study is to understand the available literature in the field and propose a highly accurate, cost-effective, and simple system for fire detection in various scenarios.

A fire detection and localisation method based on keyframes and superpixels for large-space buildings

A fire detection and localisation method based on keyframes and superpixels for large-space buildings

Optimized Deep Learning Model for Fire Semantic Segmentation

Recent convolutional neural networks (CNNs) based deep learning has significantly promoted fire detection. Existing fire detection methods can efficiently recognize and locate the fire. However, the accurate flame boundary and shape information is hard to obtain by them, which makes it difficult to conduct automated fire region analysis, prediction, and early warning. To this end, we propose a fire semantic segmentation method based on Global Position Guidance (GPG) and Multi-path explicit Edge information Interaction (MEI). Specifically, to solve the problem of local segmentation errors in low-level feature space, a top-down global position guidance module is used to restrain the offset of low-level features. Besides, an MEI module is proposed to explicitly extract and utilize the edge information to refine the coarse fire segmentation results. We compare the proposed method with existing advanced semantic segmentation and salient object detection methods. Experimental results demonstrate that the proposed method achieves 94.1%, 93.6%, 94.6%, 95.3%, and 95.9% Intersection over Union (IoU) on five test sets respectively which outperforms the suboptimal method by a large margin. In addition, in terms of accuracy, our approach also achieves the best score.

Development and field deployment of a mid-infrared CO and CO2 dual-gas sensor system for early fire detection and location

In order to realize early fire detection and location, a mid-infrared carbon monoxide (CO) and carbon dioxide (CO2) dual-gas sensor system was developed, which mainly includes a gas pretreatment module, a CO2 sensor module, a CO sensor module, and a laptop monitoring platform. CO2 and CO absorption lines located at 4.26 μm and 4.66 μm, respectively, were selected to ensure good selectivity of the sensor system. A series of experiments were carried out to evaluate the sensor performance. The 10–90% response time of the CO and CO2 sensor modules was measured to be ∼ 30 s at a flow rate of 1 L/min, and the limits of detection (LoD) of CO2 and CO were assessed to be 5.66 parts per million by volume (ppmv) and 0.94 ppmv, respectively, when the averaging time was 0.25 s. According to the correlation between CO2 and CO concentration in the early fire stage, a method of early fire detection was studied and proposed using the normalized concentration ratio between CO and CO2 (C(CO)/C(CO2)) as the key alarm parameter. Based on gas turbulent diffusion (GTD) model combined with particle swarm optimization (PSO) algorithm, a mobile early fire location method was presented. Correlative experiment results verified that the reported sensor system has a good performance for early fire detection and location.

An adaptive Particle Swarm Optimization algorithm for fire source identification of the utility tunnel fire

In order to overcome serious utility tunnel fires and minimize economic loss, it is essential to establish an effective and practical method for finding the fire source. An adaptive Particle Swarm Optimization (PSO) algorithm was proposed for fire source identification of utility tunnel fires, whose inertia weight and learning factors were adjusted dynamically. In comparison with traditional fire detection methods, the proposed algorithm is immune to the specific environment and complicated fire propagation mechanisms, the fire source location can be identified only based on several temperature values. The proposed algorithm was applied in both numerical and experimental scenarios. The results showed that the adaptive algorithm showed greater global searchability in the early exploration, and stronger local searchability in the later exploration compared with the basic one, the improvement ratio of convergence performance could achieve 50%. Additionally, all of the cases demonstrated that the fire source location can be identified by the proposed algorithm with good accuracy. Critically, the spatial temperature distribution of the whole utility tunnel was obtained based on the fire source location and the attenuation coefficient. The reasonable results indicate that the proposed algorithm can provide a reference for fire protection and intervention in the utility tunnel.

Computer Vision Method for Forest Fires Detection Based on RGB Images Obtained by Unmanned Motor Glider

The article proposes a method (algorithm) of forest fire detection by means of RGB images obtained by using an unmanned aerial vehicle (motor glider). It includes several stages associated with background detection and subtraction and recognition of fire areas by means of RGB colour space. The proposed method was tested using images of forest fires. It is proposed to use unmanned aerial vehicles capable to monitor large areas continuously for several hours. The results of calculations are shown, which demonstrate that the proposed method allows us to detect areas of images occupied by forest fires and may be used in automatic forest fire monitoring systems.

An Improvement of the Fire Detection and Classification Method Using YOLOv3 for Surveillance Systems.

Currently, sensor-based systems for fire detection are widely used worldwide. Further research has shown that camera-based fire detection systems achieve much better results than sensor-based methods. In this study, we present a method for real-time high-speed fire detection using deep learning. A new special convolutional neural network was developed to detect fire regions using the existing YOLOv3 algorithm. Due to the fact that our real-time fire detector cameras were built on a Banana Pi M3 board, we adapted the YOLOv3 network to the board level. Firstly, we tested the latest versions of YOLO algorithms to select the appropriate algorithm and used it in our study for fire detection. The default versions of the YOLO approach have very low accuracy after training and testing in fire detection cases. We selected the YOLOv3 network to improve and use it for the successful detection and warning of fire disasters. By modifying the algorithm, we recorded the results of a rapid and high-precision detection of fire, during both day and night, irrespective of the shape and size. Another advantage is that the algorithm is capable of detecting fires that are 1 m long and 0.3 m wide at a distance of 50 m. Experimental results showed that the proposed method successfully detected fire candidate areas and achieved a seamless classification performance compared to other conventional fire detection frameworks.