© 2022 IJRAR June 2022, Volume 9, Issue 2 [Link].

org (E-ISSN 2348-1269, P- ISSN 2349-5138)

FIRE DETECTION USING IMAGE

PROCESSING
1
Manasa MG, 2Ashutosh Tiwari, 3Ashutosh Yadav, 4Kishan R, 4Nandan N
1
Assistant Professor, Dept. of Electronics and Communication Engineering, Sapthagiri College of
Engineering, Bengaluru, India
2345
Student, Dept. of Electronics and Communication Engineering, Sapthagiri College of
Engineering, Bengaluru, India

Abstract: Fire accidents are commonly occurring all around the world, and a large number of people suffer the loss of life and
property due to these accidents. These accidents occur due to a lack of fire safety measures in the buildings. A large number of the
buildings does not have proper fire safety and prevention systems, the reason for which is the hassle and money the person has to
go through to spend on additional fire detection devices such as smoke detectors and heat detectors. The alternative is introduced
in this project, using the preinstalled surveillance system that most buildings already have and modifying it to detect fire and alert
the prevention system to take measures to extinguish the occurring fire.

Keywords: Haar cascade classifier, Haar features, Open CV.

I. INTRODUCTION
Fires are a persistent threat to ecosystems, infrastructure, and human life. There have been a number of fires in the
past. More and more high-rise structures are springing up all around us as the urbanization process accelerates. This
may increase the frequency of fires, resulting in significant loss of life and property. Efforts should be taken to
decrease fire hazards in locations where fire would constitute an undue threat to property, human life, or vital
biological communities. Quick fire detection is becoming increasingly crucial due to the enormous damage caused by
flames. Some fire detectors have recently been utilized in a variety of locations; they detect flames using smoke,
temperature, and photosensitive character traits. However, they are insufficient to meet the demands of a vast space, a
severe climate, or an outdoor location, for example. The project's goal is to address the current issue of undependable
fire detection systems in industrial warehouses. The goal of the initiative is to use security cameras to help detect
fires. Because the cameras are already in place, this system aims to reduce the negative effects of false alarms, making
it a cost-effective and quick method of detecting fire. Open Source Computer Vision, often known as Open-CV, is
used in the system. Due to security and insurance concerns, warehouses have a surveillance network comprised of
cameras. A costly hazard detection system, comprising of fire detection and smoke detection systems, is generally
installed in warehouses. In the event of a fire, these devices may not always be very reliable. Furthermore, these
devices generate numerous false alarms as a result of cigarette or incense smoke, among other things. As a result, we
developed an algorithm that could use an existing surveillance system to scan a live video feed and identify fire. This
also saves money on the procurement and upkeep of costly and undependable fire detection systems.

II. LITERATURE SURVEY

Gajanand S. Birajdar [Link], [1]. Propose an early fire detection system to help minimize further negative repercussions and
save people's lives. This article provides a comprehensive overview of intelligent fire detection and evacuation systems, which
include fire detection, evacuation assistance, crowd monitoring, and prediction. Multiple sensors must be deployed in various
locations throughout the structure, and a Lab-view-based system is proposed for detecting and alerting fire incidents.
Panagiotis Barmpoutis [Link], [2]. This paper includes a review of the optical remote sensing technologies utilised in early fire
warning systems, as well as an in-depth look at each technology's flame and smoke detection algorithms. Three types of systems
are identified: terrestrial, airborne, and space-borne-based systems, and various models aimed at detecting fire occurrences with
high accuracy in difficult conditions are investigated. Finally, the advantages and disadvantages of optical remote sensing-based
fire detection systems.

IJRAR22B3504 International Journal of Research and Analytical Reviews (IJRAR) [Link] 416
© 2022 IJRAR June 2022, Volume 9, Issue 2 [Link] (E-ISSN 2348-1269, P- ISSN 2349-5138)
[Link] [Link], [3]. Propose the use of satellite-based systems, which have used Earth-orbiting satellites and even
air-floating devices to observe and identify forest fires. Optical Sensor and Digital Camera, in which the processing unit tracks
motion in images and checks for the presence of numerous pixels containing smoke or fire light, then sends the data to another
algorithm to determine whether or not to generate an alarm for the operator. For localization purposes, most optical systems require
integration with geographical maps; if a fire is detected, the system notifies the fire service.
Fonollosa [Link], [4]. Proposes the use of gas chemical sensors to detect interior fires. Certain types of fire, as well as chemical
volatiles, are visible before smoke particles. As a result, systems based on chemical sensing can respond to fire alarms more quickly
than traditional smoke-based fire detectors. The study looked at the harmful pollutants caused by fires and established fire detection
system guidelines.
Yashaswini R [Link], [5]. Propose a computer-programmable firefighting robot capable of handling out a complex set of actions
automatically or manually via an external control device. It's also a re-programmable, multi-functional device capable of doing a
variety of jobs that humans find difficult. The firefighting robot is a robotic vehicle that can be used to put out fires instead of
humans. The proposed robot can be moved in all directions using an android phone as an external controller, and it also includes
Wi-Fi live video streaming for robotic vehicle monitoring. When the sensor detects a fire, the robot begins spraying water to
extinguish it.
Rosniza Ramli [Link], [6]. This paper explores the past, present, and future detection and monitoring technologies for gas leaks.
Over the last several years, there has been a surge in interest in technology, methods, and techniques for detecting and monitoring
gas leaks in order to reduce the amount of fires and injuries associated with these subjects. The use of a mix of GSM and IoT-based
systems to detect gas leaks has grown in importance as this medium is more efficient, practical, and cost-effective in monitoring
capabilities.
Seraphim S. Moumgiakmas [Link], [7]. Propose that artificial intelligence, in the form of computer vision, combined with
unmanned aerial vehicles (UAVs), has greatly aided in identifying and avoiding this risk. Computer Vision is used by the UAVs to
detect fire. The RGB camera is combined with a thermal camera in the UAV. UAVs are utilised because of their mobility, which
allows them to avoid and fly above more difficult barriers. They are useful for a variety of purposes, unlike ground-based robots.
Saurabh Manwani [Link], [8]. Proposes a system that uses a basic algorithm to detect fire. The image frame is first obtained
from the live video stream. After that, the RGB colour model is applied to the frame. After that, the RGB frame is transformed into
an HSV frame. Thresholding, median blurring (to remove noise), Background Subtraction, Sobel edge detection, and motion
detection windows are then applied to this frame. The Bitwise AND operation is then used to combine the results of all of these. To
obtain the final result, segmentation algorithms are performed to this outcome, resulting in the detection of the lack or presence of
fire in the frame. The window monitor displays an acceptable answer, and the alarm buzzer sounds.
III. OBJECTIVES
1. To detect fire using HAAR cascade classifier.
2. To send an alarm to intimate the fire accident and extinguish the fire using a sprinkler system.

IV. METHODOLOGIES

1.1. Methodology 1
The block diagram of the fire detection device is shown in Figure 1.0.

 The Haar Cascade classifier is utilized here for fire detection, which is widely common in object detection using image or
other video sources.

 The fire detection technique is based on a machine learning concept that use Haar-like characteristics in conjunction with
a cascade classifier.

 To train a cascade classifier, a large number of photos, referred to as positive and negative images, were uploaded to the
computer.

 Positive images refer to the pictures in which fire is included, and negative images are just the background images that
mean those images, which don't contain a fire in the images.

 Because the cascade classifier is based on appearance-based approach, nearly twice as many negative photos must be
provided to train it.

IJRAR22B3504 International Journal of Research and Analytical Reviews (IJRAR) [Link] 417
© 2022 IJRAR June 2022, Volume 9, Issue 2 [Link] (E-ISSN 2348-1269, P- ISSN 2349-5138)

Figure 1.0. block diagram

1.2. Methodology 2

 For the prevention system, an automated system controlled by Arduino is used.

 An alarm and a water pump connected to a network of water pipelines installed with sprinklers are all connected to the
Arduino.
 When a fire is detected, the Arduino activates the alarm and activates the water pumping system.

V. CONCLUSION
The project's goal was to find a new way to detect fire rather than relying on an existing method. Currently,
smoke detectors and sprinkler water discharge systems are used, both of which are extremely useful and effective.
These systems, however, have several drawbacks. This new system was developed in order to stay up with technology
and minimize constraints. These restrictions can be overcome by employing image processing technology to detect
fires since, in this system, the camera operates like a human eye, capturing video, processing the image, and then
sounding an alarm when it detects a fire.

REFERENCES
[1] Gajanand S. Birajdar, “Development in building fire detection and evacuation system”, IJECE vol 10, December 2020.

[2] Panagiotis Barmpoutis, “Early forest fire detection using optical remote sensing”, IJRTE vol. 2, 2020.

[3] Ahmad [Link], “Forest fire detection”, IJODSN vol. 4, 2014.

[4] Jordi Fonollosa, “Chemical sensor systems and associated algorithm for fire detection”, IJITEE vol. 8, 2018.

[5] Yashaswini R, “Firefighting robot using various Wi-Fi module”, IJIRSET vol. 2, no. 7, July,2017.

[6] Rosniza ramli, “LPG gas leakage detection and monitoring system”, JSSTE vol. 2, 2021.

[7] Seraphim S. Moumgiakmas, “Computer vision for fire detection on UAV”, IJPAM vol. 11, no.11, 2021.

[8] Saurabh Manwani, “Fire detection on surveillance system using image processing”, IJERT vol.6, issue 5, May 2017.

IJRAR22B3504 International Journal of Research and Analytical Reviews (IJRAR) [Link] 418