Real-Time Fire Detection Using Camera Sequence Image in Tunnel Environment

Current indoor fire detection systems often have limitations, including insufficient intelligence and slow response times. Recent studies have started to explore AI vision-based methods for real-time fire and smoke detection to replace traditional fire sensors. However, indoor environments present many interference factors, and few methods can effectively demonstrate superior fire and smoke detection performance under multiple simulated interference conditions. To meet the requirements of real-time, high accuracy and anti-interference, a 3D convolutional neural network (3D-CNN)-based real-time fire and smoke detection model has been developed via red-green-blue (RGB) and near-infrared (NIR) feature fusion. RGB and NIR images with complementary information were used as dual-stream inputs to the model. The feature extraction module was enhanced by 3D attention mechanisms to simultaneously capture dynamic features between frames and static features within frames of fire and smoke. A non-linear feature gated fusion module has been developed to solve the challenges of missing multi-modal features and insufficient fusion. To validate the engineering effectiveness of the model, a large-scale multi-modal dataset was constructed in real indoor building scenarios, and a real-time indoor fire and smoke detection system was also developed and deployed at the experimental site.

Currently, the forest fires are worldwide problem which requires a complete solution. The consequences of it are related, not only, to the environmental and biodiversity destruction but also to human and economic losses. In this article, forest fire and smoke detection systems, methods and technics literature review are carried out. To accomplish this aim, a generic mobile and non-mobile monitoring and surveillance systems applied in this field analysis took place. From the review carried out, it has been found that for fire detection two types of methods have been used, one based on machine learning and another through color features extraction of images or video frames, on the other hand, the combination of both methods is used for the detection of smoke and fire or only smoke, this guarantees accurate rates of more than 90%. Finally, a detailed analysis of the use of these fire and smoke detection algorithms in the UAS/UAV’s systems was carried out paying special attention on them due to the flexibility, versatility, and maneuverability abilities among others displayed by these systems that help prevent and mitigate forest fires.KeywordsFire forestSmoke forestMachine learningColor features extraction

Fire detection is a critical task in ensuring the safety of human lives and property. In recent years, deep learning techniques, such as Convolutional Neural Networks (CNNs), have shown promising results in fire detection due to their ability to automatically learn relevant features from images. In this project, we propose a real-time fire detection system using CNNs to detect fires in images or video streams. The proposed system consists of a CNN-based architecture that includes multiple convolutional layers for feature extraction, activation functions for introducing non-linearity, pooling layers for spatial dimension reduction, and fully connected layers for global pattern learning. The architecture is trained on a labelled dataset of fire and non-fire images or videos using optimization algorithms to minimize the loss function. To achieve real-time fire detection, the trained CNN model is integrated into a fire detection system that can process images or video streams in real-time. The system captures images or video frames from a live feed, pre-processes the data, and passes it through the CNN model for fire detection. The system can generate alerts or notifications in real-time when fire is detected, allowing for prompt response and mitigation measures. The performance of the proposed fire detection system is evaluated using metrics such as accuracy, precision, recall, and F1-score. The system is tested on different datasets and under various conditions to assess its accuracy, robustness, and real-time processing capabilities. The results of this project demonstrate the effectiveness of CNNs for real-time fire detection and highlight the potential of the proposed system for enhancing fire safety in various applications, including surveillance systems, smart buildings, and industrial settings. The developed system can be used as an early warning tool for fire detection, enabling timely response and mitigation, and helping to prevent or minimize the damage caused by fires. Keywords: Fire

Healthy forest is the vital resource to regulate climate at a regional and global level. Forest fire has been regarded as one of the major reasons for the loss of forest and degradation of the environment. Global warming is increasing its intensity at an alarming rate. Real-time fire detection is a necessity to avoid large scale losses. Remote sensing is a quick and cheap technique for detecting and monitoring forest fires on a large scale. Advance Very Radiometer Resolution (AVHRR) has been used already for a long period for fire detection. The use of Moderate Resolution Imaging Radio Spectrometer (MODIS) for fire detection has recently preceded AVHRR and a large number of fire products are being developed. MODIS based forest fire detection and monitoring system can solve the problem of real-time forest fire monitoring. The system facilitates data acquisition, processing, reporting and feedback on the fire location information in an automated manner. It provides location information at 1 × 1 kilometer resolution on the active fires which are present during the satellite overpass twice a day. The users are provided with the information on SMS alert with fire location details, email notification, and online visualization of fire locations on website automatically. The whole processes are automated and provide better accuracy for fire detection.

The current fire-detection methods rely primarily on smoke and temperature detection, which are generally performed in the late stage of fire and thus cannot provide a timely reminder in the early stage of fire. The continuous development of artificial intelligence has enabled machine-vision fire detection. This study proposes a convolutional neural network target-detection algorithm, i.e., You Only Look Once version 4 (YOLOv4), to detect small targets. It offers outstanding characteristics and enables scenic-spot monitoring via the video extraction of real-time fire detection using a significant amount of fire data. The diverse fire scenes can provide accurate and timely detection in the early stage of fire, thus providing favorable early warning and alarm function.

The overall results of this research addressed the question of how increased time for mineworkers to evacuate mines prior to the onset of untenable fire conditions is achievable. Additionally, improved asset loss control for underground mining is feasible through the successful development of Video Based Fire Detection (VBFD) in mining, as it provides more information earlier for mine fire fighters to intervene in the growth and spread of fire, therefore maintainable coal production is achievable. VBFD may also be used to activate automatic fire suppression systems. VBFD is an emerging volume type smoke detection technology that utilises Closed Circuit Television (CCTV) cameras to capture and analyse real time video data for the purpose of smoke recognition. VBFD technology performance was compared with the performance of fire detection methods currently used in underground coalmines to determine whether valuable improvements to fire safety and asset loss control could be made. Increased available safe evacuation time for mineworkers and a longer available time to extinguish fires earlier in their growth stage were considered the primary measures of success. The scope of this research included three main studies: a. Proof Of Concept (POC) tests; b. Detailed Experiments; and c. Field Testing. The initial trials successfully demonstrated the capability of VBFD to detect small coal fires associated with fixed plant installations in a low light simulated mine environment. Encouraged by these findings, detailed experiments were developed and subsequently applied at the Safety In Mines Testing and Research Station (SIMTARS) experimental facility at Redbank, Australia. The experimental plan involved 54 individual fire test experiments that recorded the level of Carbon Monoxide (CO) concurrent at the VBFD alarm activation time. Sensitivity analysis was undertaken by varying the air velocity over the fire source, which affected the smoke plume shape. The other component of the sensitivity analysis involved variations in light levels aimed at testing the VBFD smoke detection sensitivity. The data were subsequently analysed to identify trends in the detection response of both the CO sensors and the VBFD. Some of the tests carried out exceeded the capability of the VBFD system, which was an important aspect in the research and the evolving capability of video smoke detection. The experimental fire characteristics were subsequently replicated numerically using Computational Fluid Dynamics (CFD) so it could be used in this research to estimate the evacuation response of mineworkers under more onerous fire conditions, which could not physically be undertaken in an operational mine. A following VBFD field study was proposed to review and utilise the results of installation commissioning data and maintenance testing of a VBFD system at an operational underground mine environment. The purpose of this field-testing was to inform the underground mining industry of limitations that may exist in terms of the VBFD reliability in service, such as potential maintenance issues. During the period of the subject research the Australian resources industry moved from being in a position to readily facilitate and support minor external unfunded research projects within its underground mine assets, to its current position where it was no longer willing or able to assist. As a result of the lack of a trial VBFD system in an underground mine, the subsequent field study opportunistically considered the reliability and performance of VBFD in a very similar and related environment, however this facility was not an underground mine, but a South African coal fired power station. The results of the reliability testing of 360 VBFD devices over a 12 months’ period showed that VBFD was very reliable. Future VBFD research might address two aspects. Firstly, VBFD may be used to detect flames or smoke. Whilst coal handling plant fires typically manifest during their early stages of growth as smouldering fires, other fixed plant fires have the potential to produce flames from the outset. Underground refuelling stations and transformers exhibit this phenomenon, so flame detection capability is possible and should be explored. In this research, VBFD CCTV cameras were of a very basic functionality. They were simple, relatively low cost devices operating in the visible light spectrum only. Secondly, the use of high resolution “mega pixel” cameras with the capability to operate in very low light levels should be researched. It was found that VBFD offers considerably faster response to visible smoke plumes than other forms of fire detection currently used in underground coalmines. In terms of VBFD reliability, it was more resistant to contamination from mine pollutants than typical commercial ‘point type’ photoelectric or ionisation smoke detectors. It was shown in the research that significant improvements in fire safety and asset loss control in underground mining can be made through the application of VBFD for early fire detection of fixed plant fires.

The paper starts with the introduction of the basic requirements of classical intrusion detection and alarm systems and fire detection systems. The drawbacks of such conventional systems are highlighted. Techniques of computer vision are employed to remove the drawbacks and at the same time increase the reliability and response rate of the systems. For security and low level fire detection, a fuzzy logic based image comparison algorithm is deemed adequate. In order to confirm the existence of fire or smoke, techniques related to optical flow are employed as high level fire or smoke detection, which generate a velocity field for the image so that the decision can be judged by using fizzy logic. Details of implementation and some experimental results have been included in the paper for illustration.

The fire detection system combines the simultaneous measurements of smoke, carbon monoxide, and carbon dioxide. The security of campus against intruders moving in laboratories, class rooms, staffrooms or washrooms. The fire alarm system consists of Fire detectors (with can be smoke detector, heat, or Infra-Red detectors), control unit and alarm system. A fire detection system is developed based on the simultaneous measurements of temperature and smoke. The fire detection system with the alarm algorithm detected fires that were not alarmed by smoke sensors, and alarmed in shorter times than smoke sensors operating alone. Previous fire detection algorithms used data from sensors for temperature, smoke, and combustion products. The smoke sensor alarms when the analog output signal exceeds or equal the threshold value. The node includes analog sensors to measure smoke, carbon monoxide (CO) and temperature. A fire alarm system should reliably and in a timely way notify building occupants about the presence of fire indicators, such as smoke or high temperatures.

Smoke control is one of the important aspects in atrium fire. For an efficient smoke control strategy, it is very important to identify the smoke and fire source in a very short period of time. However, traditional methods such as point type detectors are not effective for smoke and fire detection in large space such as atrium. Therefore, video smoke and fire detection systems are proposed. For the development of the system, automatic extraction and tracking of flame are two important problems needed to be solved. Based on entropy theory, region growing and Otsu method, a new automatic integrated algorithm, which is used to track flame from video images, is proposed in this paper. It can successfully identify flames from different environment, different background and in different form. The experimental results show that this integrated algorithm has stronger robustness and wider adaptability. In addition, because of the low computational demand of this algorithm, it is also possible to be used as part of a robust, real-time smoke and fire detection system.

I. ABSTRACT Fire hazards pose a significant threat to life and property, making early detection crucial for minimizing damage. This project focuses on AI-based fire detection and alarming using computer vision and deep learning techniques. The system employs CCTV cameras or laptop cameras to capture real-time video feeds, which are analysed using Python, OpenCV, and deep learning models to detect fire accurately. A Convolutional Neural Network (CNN) or YOLO (You Only Look Once) object detection model is trained on fire datasets to distinguish fire from non-fire scenarios, reducing false alarms. Once fire is detected, the system triggers an alarm sound and can also send notifications via SMS, email, or IoT-based alerts to concerned authorities. This AI-powered solution is designed for high accuracy and efficiency, making it suitable for smart surveillance, industrial safety, and residential fire monitoring. The integration of edge computing devices like Raspberry Pi or Jetson Nano can further enhance real-time processing. The proposed system aims to improve fire safety measures by providing automated, real-time fire detection, reducing response time, and preventing large-scale disasters. II. KEYWORDS Fire Detection, Real-time Fire Monitoring, Computer Vision (CV).

<strong class="journal-contentHeaderColor">Abstract.</strong> With the advancement of computer vision, artificial intelligence, and remote sensing technologies, deep learning algorithms are increasingly used in terrestrial, airborne, and spaceborne-based fire detection systems. The performance and generalization of these data-driven fire detection algorithms, however, are restricted by the limited number and source of fire detection datasets. A large-scale fire detection benchmark dataset covering complex and varied fire scenarios is urgently needed. This work constructs a 100,000-level Flame and Smoke Detection Dataset (FASDD) based on multi-source heterogeneous flame and smoke images. It holds rich variations in image size, resolution, illumination (day and night), scenario (indoor and outdoor), image range (far and near), viewing angle (top view and side view), platform (surveillance cameras, drones, and satellites), and data source (Internet, social media, and open-access fire datasets). To the best of our knowledge, FASDD is currently the most versatile and comprehensive dataset for fire detection. It provides a challenging benchmark to drive the continuous evolution of fire detection models. Additionally, we formulate a unified workflow for preprocessing, annotation, and quality control of fire samples. Out-of-the-box annotations are published in four different formats for training deep learning models. Extensive performance evaluations based on classical methods show that most of the object detection models trained on FASDD can achieve satisfactory fire detection results, and especially YOLOv5x achieves nearly 80 % mAP@0.5 accuracy on heterogeneous images spanning two domains of computer vision and remote sensing. And the application in wildfire location demonstrates that deep learning models trained on our dataset can be used in recognizing and monitoring forest fires. Deep learning models trained with FASDD can be simultaneously deployed on satellites, drones, and ground sensors, thus realizing collaborative fire observation and detection in a space-air-ground integrated network environment. The dataset is available from the Science Data Bank website at <a href="https://doi.org/10.57760/sciencedb.j00104.00103" target="_blank" rel="noopener">https://doi.org/10.57760/sciencedb.j00104.00103</a> (Wang et al., 2022a).

In recent years the complexity and cost of Fire and Gas detection systems have increased rapidly with no measurable reduction in fatality rates or other losses. A principal reason for this disappointing outcome despite substantial investment is that systems designs have lost the focus of what they are trying to achieve. Another reason worth exploring is that investment in other areas may result in greater benefits. For instance, diverting funds to preventing lifting accidents, helicopter failures and materials handling incidents, could result in greater OVERALL benefit since these account for a larger proportion of risks. Nevertheless, Fire and Gas detections systems have a major role to play in preventing escalation, and this paper shows how they can be made more efficient in both cost and performance by combining four methods: targeting perceived hazards,integration from concept stage with other safety designs,control of design with effective performance specification, andcontrol of modifications throughout the system's life.

DEVELOPMENT OF COMPUTER VISION-BASED AUTOMATIC FLAME DETECTION ALGORITHM USING MATLAB SOFTWARE ENVIRONMENT

Authorities and policymakers in Korea have recently prioritized improving fire prevention and emergency response. Governments seek to enhance community safety for residents by constructing automated fire detection and identification systems. This study examined the efficacy of YOLOv6, a system for object identification running on an NVIDIA GPU platform, to identify fire-related items. Using metrics such as object identification speed, accuracy research, and time-sensitive real-world applications, we analyzed the influence of YOLOv6 on fire detection and identification efforts in Korea. We conducted trials using a fire dataset comprising 4000 photos collected through Google, YouTube, and other resources to evaluate the viability of YOLOv6 in fire recognition and detection tasks. According to the findings, YOLOv6’s object identification performance was 0.98, with a typical recall of 0.96 and a precision of 0.83. The system achieved an MAE of 0.302%. These findings suggest that YOLOv6 is an effective technique for detecting and identifying fire-related items in photos in Korea. Multi-class object recognition using random forests, k-nearest neighbors, support vector, logistic regression, naive Bayes, and XGBoost was performed on the SFSC data to evaluate the system’s capacity to identify fire-related objects. The results demonstrate that for fire-related objects, XGBoost achieved the highest object identification accuracy, with values of 0.717 and 0.767. This was followed by random forest, with values of 0.468 and 0.510. Finally, we tested YOLOv6 in a simulated fire evacuation scenario to gauge its practicality in emergencies. The results show that YOLOv6 can accurately identify fire-related items in real time within a response time of 0.66 s. Therefore, YOLOv6 is a viable option for fire detection and recognition in Korea. The XGBoost classifier provides the highest accuracy when attempting to identify objects, achieving remarkable results. Furthermore, the system accurately identifies fire-related objects while they are being detected in real-time. This makes YOLOv6 an effective tool to use in fire detection and identification initiatives.

Flame detection is a specialized task in fire detection and forest fire monitoring systems. In this paper, a static and dynamic texture analysis of flame in forest fire detection is proposed. The flames are initially segmented, based on the color in YCbCr (luminance, chrominance blue and chrominance red components) color space called candidate flame region. From the candidate flame region, the static and dynamic texture features are extracted. Static texture features are obtained by hybrid texture descriptors. Dynamic texture features are derived using 2D wavelet decomposition in temporal domain and 3D volumetric wavelet decomposition. Finally, extreme learning machine classifier is used to classify the candidate flame region as real flame or non-flame, based on the extracted texture features. The proposed flame detection system is applied to various fire detection scenes, in the real environments and it effectively distinguishes fire from fire-colored moving objects. The results show that the proposed fire detection technique achieves the average detection rate of 95.65% which is better compared to other state-of-art methods.