Fire Detection Research Articles

Single-Valued Neutrosophic MCDM Approaches Integrated with MEREC and RAM for the Selection of UAVs in Forest Fire Detection and Management

In recent times, the world has experienced a rise in the frequency of forest fires. These fires cause severe economic damage and pose a significant threat to human lives. Therefore, it is essential to search for solutions that can help combat fires and detect them early. Once a fire reaches a certain level, it becomes challenging to control it. Various systems have been proposed to collect data and detect forest fires, such as satellites and other traditional methods. However, these solutions have been ineffective in terms of cost, coverage of large areas, accuracy, and the safety of human lives. To address these limitations, Unmanned Aerial Vehicles (UAVs) or drones have been used for detecting, combatting, and early warning of forest fires. UAVs are one of the modern technologies that have achieved great progress in monitoring natural disasters and have been widely used in monitoring, detecting, and predicting fires. They can fly without a human pilot on board, which makes them ideal for preserving human life. In addition, they are equipped with firefighting tools and various tools for remote sensing. This is to take high-quality photos or videos of the area to be detected. Different types of UAVs are used to fight fires, and here decision-makers face a problem in choosing between these types. Therefore, this research proposes a new MCDM model integrated with neutrosophic sets for selecting the optimal UAV to combat forest fires; therefore it helps in effectively detecting and fighting the fire. The proposed model integrates a Method based on Removal Effects of Criteria (MEREC) and Root Assessment Method (RAM) with the context of neutrosophic sets that effectively deal with ambiguity for selecting the optimal UAV which use in the detection and combat forest fires.

To ensure the safety of storage: Enhancing accuracy of fire detection in warehouses with deep learning models

Warehouses are important places with storage functions in process units, containing many flammable and explosive materials. Nevertheless, fire protection systems often fail to provide adequate information when fires break out, hindering firefighting and leading to catastrophic outcomes. This research aims to develop a deep learning (DL) model capable of precisely locating the fire source and estimating its intensity, providing graphical representations of fire information. We propose a novel Multi-head Attention-based Long Short Term Memory (MHA-LSTM) model and apply Computational Fluid Dynamics (CFD) to generate datasets for model training on warehouse environments. The results demonstrate losses on the validation set and test set are less than 0.003, and all errors are below 0.15, indicating that MHA-LSTM has good generalization ability for fires in various locations. Compared to 3 common DL models, such as LSTM, the loss of MHA-LSTM on the test set is reduced by at least 51.6 %, greatly improves the accuracy of fire detection in complex scenarios. Finally, based on firefighters' situational awareness (SA) requirements, graphical examples of estimated fire information are generated to provide accurate and visualized fire data. This study utilizes artificial intelligence (AI) to overcome the limitations of complex internal structures in warehouses, offering a feasible approach to improve process safety and the accuracy of fire protection systems.

Generative Algorithms for Fusion of Physics-Based Wildfire Spread Models with Satellite Data for Initializing Wildfire Forecasts

Abstract Increases in wildfire activity and the resulting impacts have prompted the development of high-resolution wildfire behavior models for forecasting fire spread. Recent progress in using satellites to detect fire locations further provides the opportunity to use measurements toward improving fire spread forecasts from numerical models through data assimilation. This work develops a physics-informed approach for inferring the history of a wildfire from satellite measurements, providing the necessary information to initialize coupled atmosphere–wildfire models from a measured wildfire state. The fire arrival time, which is the time the fire reaches a given spatial location, acts as a succinct representation of the history of a wildfire. In this work, a conditional Wasserstein generative adversarial network (cWGAN), trained with WRF–SFIRE simulations, is used to infer the fire arrival time from satellite active fire data. The cWGAN is used to produce samples of likely fire arrival times from the conditional distribution of arrival times given satellite active fire detections. Samples produced by the cWGAN are further used to assess the uncertainty of predictions. The cWGAN is tested on four California wildfires occurring between 2020 and 2022, and predictions for fire extent are compared against high-resolution airborne infrared measurements. Further, the predicted ignition times are compared with reported ignition times. An average Sørensen’s coefficient of 0.81 for the fire perimeters and an average ignition time difference of 32 min suggest that the method is highly accurate. Significance Statement To initialize coupled atmosphere–wildfire simulations in a physically consistent way based on satellite measurements of active fire locations, it is critical to ensure the state of the fire and atmosphere aligns at the start of the forecast. If known, the history of a wildfire may be used to develop an atmospheric state matching the wildfire state determined from satellite data in a process known as spinup. In this paper, we present a novel method for inferring the early stage history of a wildfire based on satellite active fire measurements. Here, inference of the fire history is performed in a probabilistic sense and physics is further incorporated through the use of training data derived from a coupled atmosphere–wildfire model.

A Study on Comparative Analysis of Object Detection-based Real-Time Wildfire and Class B Fire Detection Performance

Wildfires cause human casualties, property damage, and significant damage to ecosystems, so prompt fire detection and response are essential. In this study, training and evaluation experiments were performed on the YOLOv3, Faster R-CNN, and Cascade R-CNN models to analyze their real-time wildfire detection performance. The performance of each model was evaluated using precision, recall, mAP 50, mAP 50-95, the model parameter, and frames per second (FPS). The experiment results showed that YOLOv3 had slightly lower overall performance than other models, but the difference was not significant in terms of mAP 50. YOLOv3 had the highest inference speed (48.7 FPS) and an overall mAP 50 of 0.832. On the contrary, Faster R-CNN's overall mAP 50 was 0.835, while Cascade R-CNN showed higher performance with 0.840. Compared to YOLOv3, both models showed relatively low inference speeds (36.2 and 26.3 FPS, respectively). Therefore, the YOLOv3 model is deemed appropriate for real-time fire detection owing to its high inference speed and insignificant wildfire detection performance difference in terms of mAP 50. Future wildfire detection models are expected to improve by optimizing the network structures (e.g., FPN and Head) of 1-Stage models, such as YOLOv3.

Multi-Height and Heterogeneous Sensor Fusion Discriminant with LSTM for Weak Fire Signal Detection in Large Spaces with High Ceilings

Fire is a significant cause of fatalities and property loss. In tall spaces, early smoke dispersion is hindered by thermal barriers, and initial flames with limited smoke production may be obscured by ground-level structures. Consequently, smoke, temperature, and other fire sensor signals are weakened, leading to delays in fire detection by sensor networks. This paper proposes a multi-height and heterogeneous fusion discriminant model with a multilayered LSTM structure for the robust detection of weak fire signals in such challenging situations. The model employs three LSTM structures with cross inputs in the first layer and an input-weighted LSTM structure in the second layer to capture the temporal and cross-correlation features of smoke concentration, temperature, and plume velocity sensor data. The third LSTM layer further aggregates these features to extract the spatial correlation patterns among different heights. The experimental results demonstrate that the proposed algorithm can effectively expedite alarm response during sparse smoke conditions and mitigate false alarms caused by weak signals.

Unveiling Anomalies in Surveillance Videos

Abstract: The implementation of Convolutional Neural Network (CNN) algorithm for the detection of anomalies like fire and potholes in images or videos. Leveraging deep learning technique, the model aims to accurately identify fire, potholes, etc. within various contexts. The CNN architecture is trained on a dataset comprising diverse images to enhance its ability to generalize. This proposed project not only addresses safety concerns by promoting helmet usage and proper maintenance of roads but also contributes to the broader field of computer vision and object detection. This implementation involves training the CNN on annotated datasets, fine-tuning the model for optimal performance, and integrating it into a practical application for efficient anomaly identification. With a focus on real-time detection, the system utilizes image processing techniques and a trained CNN model to analyze visual data from cameras or video feeds. The system's versatility allows integration into surveillance systems, industrial sites, or any environment where fire detection is crucial for safety. By automating the detection process, the project contributes to minimizing human error and ensuring consistent monitoring. The proposed solution holds the potential to significantly impact safety protocols, particularly in industries where protective headgear is paramount. Potholes pose a significant threat to both drivers and pedestrians, leading to accidents and infrastructure damage. The proposed solution leverages CNN's ability to effectively process visual data, making it well-suited for image recognition tasks. Our approach involves training the CNN on a diverse dataset of road images to enable it to accurately identify potholes. The model will learn distinctive features and patterns associated with potholes, allowing for robust detection under various environmental conditions. Real-time implementation on embedded systems or cameras along roadways will enable instantaneous identification and alerting.

Live Data Monitoring in Industry

Abstract: In contemporary industrial environments, real-time monitoring of critical parameters is vital for ensuring safety, efficiency, and compliance with regulations. This thesis details the development and implementation of an advanced live data monitoring system designed to track temperature, humidity, fire presence, and oxygen levels. Key components of the system include a DHT sensor for temperature and humidity measurement, a fire sensor for fire detection, an oxygen sensor for monitoring atmospheric oxygen, and a buzzer for immediate auditory fire alerts. Central to the system is the ESP8266 WiFi module, which facilitates the wireless transmission of sensor data to ThingSpeak, a cloud-based data collection and visualization platform. This setup allows for continuous, remote monitoring of environmental conditions, providing critical insights and alerts to industrial operators. The project delves into the hardware configuration, sensor calibration, data acquisition processes, and wireless communication protocols essential for effective real-time monitoring. It also addresses the challenges of integrating multiple sensors and ensuring reliable data transmission in an industrial setting. The results demonstrate the system's capability to provide accurate, timely data, enhancing safety protocols and operational efficiency. By utilizing IoT technology, the proposed system offers a scalable solution for industrial environments aiming to implement advanced monitoring and alert systems. This work contributes to the field of industrial automation by illustrating a practical application of IoT in improving monitoring and safety measures, highlighting both the technical achievements and practical implications of the implemented system.

Robust Forest Fire Detection using Deep Convolutional Neural Networks

Forest fires pose significant threats to ecosystems, wildlife, and human lives, necessitating proactive measures for early detection and rapid response. This paper presents FireGuard, an efficient model for forest fire detection using deep convolutional neural networks (CNNs). Leveraging the power of deep learning and image processing techniques, FireGuard analyzes aerial imagery and satellite data to detect signs of smoke and fire outbreaks in forested areas. The model utilizes a lightweight CNN architecture optimized for real-time performance and resource-constrained environments, making it suitable for deployment on unmanned aerial vehicles (UAVs), surveillance cameras, and satellite platforms. Experimental results demonstrate the effectiveness of FireGuard in accurately identifying forest fires with high precision and recall, outperforming traditional methods and existing deep learning models. By providing early warnings of potential fire incidents, FireGuard enables timely intervention by fire fighting agencies, thereby mitigating the impact of forest fires and preserving natural habitats.

Forest fire detection utilizing ghost Swin transformer with attention and auxiliary geometric loss

Forest fires are a devastating natural disaster. Existing fire detection models face limitations in dataset availability, multi-scale feature extraction, and locating obscured or small flames and smoke. To address these issues, we develop a dataset containing real and synthetic forest fire images, sourced from a UAV (Unmanned Aerial Vehicle) perspective. Additionally, we propose the Ghost Convolution Swin Transformer (GCST) module to extract multi-scale flame and smoke features from different receptive fields by integrating parallel Ghost convolution and Swin Transformer. Subsequently, we design a lightweight reparameterized rotation attention module, which captures interactions across channel and spatial dimensions to suppress background noise and focus on obscured flames and smoke. Finally, we introduce a loss function, called Efficient Auxiliary Geometric Intersection over Union (EAGIoU), which employs an auxiliary bounding box to accelerate the model's convergence while integrating the geometrical principles of the predicted and real bounding boxes to accurately locate small flames and smoke. Extensive experimental results demonstrate that our method achieves 75.2 % mAP@0.5 and 42 % mAP@0.5:0.95 with a frame rate of 239 frames per second, indicating a significant improvement in accuracy and real-time performance compared to state-of-the-art techniques. The code and datasets are available at https://github.com/luckylil/forest-fire-detection.

Advancing Early Warning Systems for Fire Detection: A Comprehensive Approach in Machine Learning

This research conducts a comprehensive investigation of the efficacy of various machine learning algorithms for fire detection. The algorithms that were examined include logistic regression, decision tree, random forest, support vector classifier, gradient boosting, K-nearest neighbors, Gaussian naive Bayes, multilayer perceptron classifier, and XGBoost classifier. Through in-depth experiments, this study rigorously assesses the performance of these algorithms in identifying and predicting fires based on pertinent input features. Among the algorithms that were investigated, logistic regression is the best performer, with a high accuracy rate of 99%. The findings from this research offer valuable insights for optimizing fire detection systems, providing a nuanced understanding of the practical applicability of machine learning techniques in real-time fire monitoring scenarios. The primary objectives of this study are to elucidate specific challenges in fire detection, evaluate the performance of various machine learning algorithms, and contribute to the foundational knowledge that is essential for enhancing fire management strategies. The research addresses the limited precision of existing fire detection systems and aims to rectify this issue through a systematic exploration of advanced machine learning approaches. The overarching goal is to bolster the foundations of fire management, facilitating the development of proactive measures and prompt responses to mitigate the profound impact of wildfires. By presenting a detailed examination of the strengths and weaknesses of various machine learning algorithms, this research strives to foster a robust and effective approach to fire detection, thereby advancing the field and ensuring the safety of communities at risk.

Ignition detection method based on real-time measurements of a hazardous parameter in the environment

The object of this study is a selective average of the dangerous parameters of the gas environment during the ignition of materials. Theoretical substantiation of the fire detection method based on testing the null hypothesis regarding the current difference of the specified selective averages of an arbitrary dangerous parameter of the gas environment has been performed. In this case, the significance of the current difference with respect to selective averages allows detection of ignition in real-time observation of an arbitrary dangerous parameter of the gas environment. The method makes it possible to set the level of significance for the current difference and at the same time provide for the maximum power of fire detection. Laboratory experiments were conducted to verify the proposed method for detecting ignition based on the current difference of the selective averages of the measured dangerous parameters of the gas environment corresponding to the training and control general population. The results of verification showed that at a given level of significance, the method allows detecting current fires of materials based on significant differences in sample means. It was established that the current difference in the concentration of carbon monoxide during ignition and after ignition of alcohol, paper, wood shavings, and textiles is –0.459 and 8.296, –0.152 and 4.299, –0.027 and 6.9, –0.262 and 2.3, respectively. Current smoke density differences are 0.043 and 0.391, 0.012, and 0.923, –0.139, and –0.235, 0.034, and 0.129, and temperatures are –0.01 and 10.635, 0.53 135 and 2.726, respectively. This means that the current difference is significant and is due not to a random nature but to the appearance of a persistent effect from the ignition of the material. In practice, research results can be used to detect fires in real time in order to prevent them from growing into an uncontrolled fire

Comparative analysis of single-board computers for the development of a microarchitectural computing system for fire detection

Objectives. The purpose of the work is to select the basic computing microplatform of the onboard microarchitectural computing complex for the detection of anomalous situations in the territory of the Republic of Belarus from space on the basis of artificial intelligence methods.Methods. The method of comparative analysis is used to select a computing platform. A series of performance tests and comparative analysis (benchmarking) are performed on the selected equipment. The methods of comparative and benchmarking analysis are performed in accordance with the terms of reference to the current project.Results. A comparative analysis and performance testing of Raspberry Pi 4 Model B and Cool Pi 4 Model B single-board computers, as well as AI-accelerator Google Coral USB Accelerator with Google Edge TPU have been performed. The comparative analysis showed that Raspberry Pi 4 Model B and Cool Pi 4 Model B fully meet the terms of reference to the current project. At the same time Cool Pi 4 Model B handles neural network calculations well, but four times slower than similar calculations on Google Coral USB Accelerator. Neural network computations on the Raspberry Pi 4 Model B are 22 times slower than similar computations on the Google Coral USB Accelerator. Cool Pi 4 Model B outperforms Raspberry Pi 4 Model B by the factor of two to three for data copying and compression and almost six times faster for neural network computations.Conclusion. Despite the fact that Raspberry Pi 4 Model B meets the terms of reference to the project as a computational basis, when developing an on-board microarchitectural computing system for detecting anomalous situations, it is worth using more powerful alternatives with built-in AI-accelerators (e.g., Radxa Rock 5 Model A) or with an additional external AI-accelerator (e.g., a combination of Cool Pi 4 Model B and Google Coral USB Accelerator). Using a Raspberry Pi 4 Model B with an additional AI-accelerator is also acceptable and will speed up computations by several dozen times. AI-accelerators provide the fastest neural network computations, but there are features related to the novelty of the technology that will be explored in further development.

Experimental study of the temperature profile induced by the tunnel wall fire plume at various source-ceiling heights

The smoke temperature profile beneath the ceiling of tunnel is one of the most important parameters to evaluate the tunnel fire risk and guide the early fire detection. This paper investigates experimentally the temperature profile induced by the tunnel wall fire plume at various source-ceiling heights. A series of experiments were carried out in a reduced-scale tunnel and a burner combustion inside. The burner elevation above the tunnel floor was changed at the center of the tunnel, and it attaches the side-wall of the tunnel to simulate the wall fire plume evolution with fire growth inside the tunnel. The temperatures beneath the ceiling were measured by the thermocouple array at the corner between the ceiling and side-wall of the tunnel. The maximum temperature beneath the ceiling could be correlated by the heat release rate and the source-ceiling height. A new characteristic parameter model considering the virtual point source at the side of the tunnel was proposed firstly considering the tunnel fire scenario to describe the tunnel plume temperature profile. The obtained experimental data and proposed temperature profile model in this work provide a basic understanding of the heat/thermal impact on the tunnel wall fire plume.

Analysis of the possibility of increasing the level of reliability of fire alarms systems through integration with video-based flame detection system

Classic Fire Alarm Systems (FAS) are characterized by a high level of false alarms in relation to the number of confirmed reports. To increase the level of reliability of operation of this type of system, it was proposed to integrate it with a Video-Based Flame Detection System (VBFDS). For this purpose, a video-based fire detection algorithm was designed. In addition, methods popular in the literature for increasing the reliability of such systems, such as color filtering, the Tracking Growth Object (TGO) factor, and the use of the Naive-Bayes (NB) classifier, were tested. The purpose of the article is to analyze these methods in comparison with the basic version of the algorithm, as well as the possibility of integrating Video-Based Flame Detection System (VBFDS) with classic FAS.

Comparing machine learning algorithms to predict vegetation fire detections in Pakistan

Vegetation fires have major impacts on the ecosystem and present a significant threat to human life. Vegetation fires consists of forest fires, cropland fires, and other vegetation fires in this study. Currently, there is a limited amount of research on the long-term prediction of vegetation fires in Pakistan. The exact effect of every factor on the frequency of vegetation fires remains unclear when using standard analysis. This research utilized the high proficiency of machine learning algorithms to combine data from several sources, including the MODIS Global Fire Atlas dataset, topographic, climatic conditions, and different vegetation types acquired between 2001 and 2022. We tested many algorithms and ultimately chose four models for formal data processing. Their selection was based on their performance metrics, such as accuracy, computational efficiency, and preliminary test results. The model’s logistic regression, a random forest, a support vector machine, and an eXtreme Gradient Boosting were used to identify and select the nine key factors of forest and cropland fires and, in the case of other vegetation, seven key factors that cause a fire in Pakistan. The findings indicated that the vegetation fire prediction models achieved prediction accuracies ranging from 78.7 to 87.5% for forest fires, 70.4 to 84.0% for cropland fires, and 66.6 to 83.1% for other vegetation. Additionally, the area under the curve (AUC) values ranged from 83.6 to 93.4% in forest fires, 72.6 to 90.6% in cropland fires, and 74.2 to 90.7% in other vegetation. The random forest model had the highest accuracy rate of 87.5% in forest fires, 84.0% in cropland fires, and 83.1% in other vegetation and also the highest AUC value of 93.4% in forest fires, 90.6% in cropland fires, and 90.7% in other vegetation, proving to be the most optimal performance model. The models provided predictive insights into specific conditions and regional susceptibilities to fire occurrences, adding significant value beyond the initial MODIS detection data. The maps generated to analyze Pakistan’s vegetation fire risk showed the geographical distribution of areas with high, moderate, and low vegetation fire risks, highlighting predictive risk assessments rather than historical fire detections.

FSH-DETR: An Efficient End-to-End Fire Smoke and Human Detection Based on a Deformable DEtection TRansformer (DETR).

Fire is a significant security threat that can lead to casualties, property damage, and environmental damage. Despite the availability of object-detection algorithms, challenges persist in detecting fires, smoke, and humans. These challenges include poor performance in detecting small fires and smoke, as well as a high computational cost, which limits deployments. In this paper, we propose an end-to-end object detector for fire, smoke, and human detection based on Deformable DETR (DEtection TRansformer) called FSH-DETR. To effectively process multi-scale fire and smoke features, we propose a novel Mixed Encoder, which integrates SSFI (Separate Single-scale Feature Interaction Module) and CCFM (CNN-based Cross-scale Feature Fusion Module) for multi-scale fire, smoke, and human feature fusion. Furthermore, we enhance the convergence speed of FSH-DETR by incorporating a bounding box loss function called PIoUv2 (Powerful Intersection of Union), which improves the precision of fire, smoke, and human detection. Extensive experiments on the public dataset demonstrate that the proposed method surpasses state-of-the-art methods in terms of the mAP (mean Average Precision), with mAP and mAP50 reaching 66.7% and 84.2%, respectively.

The research on infrared radiation affected by smoke or fog in different environmental temperatures

Infrared thermal imaging camera as a non-contact monitoring of the object to be measured is widely used in fire detection, driving assistance and so on. Although there are many related studies, there is a lack of research on the influence of fog or smoke on infrared imaging under different environmental temperatures. To address this shortcoming, The temperature of both the environment and the target in this experiment is controlled by PID technology. The smoke or fog environment is generated using a smoke cake or an ultrasonic fog machine. The temperature of the target was measured by infrared thermal imaging camera. It was observed that as the temperature of the environment increases, the measured temperature of the target also increases. However, the change in temperature is more pronounced in the fog environment compared to either the smoke environment or the normal environment. It has been found through research that environmental radiation causes temperature changes in fog droplets. Therefore, Infrared radiation is less affected in the smoke environment and more affected in the fog environment. Additionally, when the environmental temperature is close to the target's temperature, the infrared image becomes blurred.

Early Smoke Recognition Algorithm for Forest Fires

Forest fires require rapid and precise early smoke detection to minimize damage. This study focuses on employing smoke recognition methods for early warning systems in forest fire detection, identifying smoke as the primary indicator. A significant hurdle lies in the absence of a large-scale dataset for real-world early forest fire smoke detection. Early smoke videos present characteristics such as smoke plumes being small, slow-moving, and/or semi-transparent in color, and include images where there is background interference, posing critical challenges for practical recognition algorithms. To address these issues, this paper introduces a real-world early smoke monitoring video dataset as a foundational resource. The proposed 4D attention-based motion target enhancement network includes an important frame sorting module which adaptively selects essential frame sequences to improve the detection of slow-moving smoke targets. Additionally, a 4D attention-based motion target enhancement module is introduced to mitigate interference from smoke-like objects and enhance recognition of light smoke during the initial stages. Moreover, a high-resolution multi-scale fusion module is presented, incorporating a small target recognition layer to enhance the network’s ability to detect small smoke targets. This research represents a significant advancement in early smoke detection for forest fire surveillance, with practical implications for enhancing fire management.

The Dynamic Change in the Reliability Function Level in a Selected Fire Alarm System during a Fire.

This article discusses fundamental issues associated with the functional reliability of selected fire alarm systems (FASs) in operation during building fires. FASs operate under diverse external or internal natural environmental conditions, and the operational process of FAS should take into account the impacts of physical phenomena that occur during fires. Their operation is associated with the constant provision of reliability. FAS designers should also consider the system's reliability when developing fire control matrices, tables, algorithms, or scenarios. All functions arising from an FAS control matrix should be implemented with a permissible reliability level, RDPN(t), prior to, as well as during, a fire. This should be assigned to the controls saved in the fire alarm control unit (FCP). This article presents the process by which high temperatures generated during a fire impact the reliability of FAS functioning. It was developed considering selected critical paths for a specific scenario and the control matrix for an FAS. Such assumptions make it possible to determine the impact of various temperatures generated during a fire on the reliability of an FAS. To this end, the authors reviewed that the waveform of the R(t) function changes for a given FAS over time, Δt, and then determined the fitness paths. The critical paths are located within the fire detection and suppression activation process, using FAS or fixed extinguishing devices (FEDs), and the paths were modeled with acceptable and unacceptable technical states. The last section of this article defines a model and graph for the operational process of a selected FAS, the analysis of which enables conclusions to be drawn that can be employed in the design and implementation stages.

Smart city fire surveillance: A deep state‐space model with intelligent agents

AbstractIn the realm of smart city development, the integration of intelligent agents has emerged as a pivotal strategy to enhance the efficacy of search methodologies. This study introduces a novel state‐space navigational model employing intelligent agents tailored specifically for fire surveillance in urban environments. Central to this model is the fusion of a convolutional neural network and multilayer perceptron, enabling accurate fire detection and localisation. Leveraging this capability, the intelligent agent proactively navigates through the search space, guided by the shortest path to the identified fire location. The utilisation of the A* algorithm as the search mechanism underscores the efficiency and efficacy of our proposed approach. Implemented in Python and Gephi, our method surpasses traditional search algorithms, both informed and uninformed, demonstrating its effectiveness in navigating urban landscapes for fire surveillance. This research study contributes significantly to the field by offering a robust solution for proactive fire detection and surveillance in smart city environments, thereby enhancing public safety and urban resilience.