See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/371788538

Video Surveillance Fire Detection System Using CNN Algorithm

Conference Paper · June 2023

CITATIONS READS

0 76

5 authors, including:

Umakant Tupe Prasad Kulkarni

Savitribai Phule Pune University Savitribai Phule Pune University
11 PUBLICATIONS 7 CITATIONS 1 PUBLICATION 0 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Umakant Tupe on 23 June 2023.

The user has requested enhancement of the downloaded file.

 Video Surveillance Fire Detection System Using CNN Algorithm
U.L.Tupe, LakhanJadhav, ShivanandKoli, PrasadKulkarni, MayurGaikwad
[email protected], [email protected], [email protected]
Department of Information Technology, JSPM’s Rajarshi Shahu College of
Engineering, Pune
Abstract-

Fires in public places such as shopping malls, hospitals, train stations, and others can endanger both
people and resources. This has been a big concern for the past few decades. Stopping these accidents
should be a priority. So to achieve these, there are various techniques, but there are some loopholes in
those techniques. So to overcome those loopholes, we have developed a model that detects fire in
images and video frames, and as soon as it detects fire, it sends an alert message to the nearest fire
station and related authorities. The main purpose of these models is to prevent accidents due to fire
and minimise human work load. This paper uses the CNN algorithm to build these projects.

Keywords:- Fire detection, CNN, Deep learning, Image Processing etc.

I. Introduction

Rate of forest fires reports have increased yearly due to human causes and dry climate. To avoid
terrible disaster of fire, many detection techniques have been widely studied to apply in practice. Most
of traditional method are based on sensors due to its low-cost and simple installation. These systems
are not applicable for using outdoor where energy of flame affected by fire materials and the burning
process affected by environment that have potential cause of false alarms. Visual-based approach of
image or video processing was shown to be more reliable method to detect the fire since the closed
circuit television (CCTV) surveillance systems are now available at many public places, can help
capture the fire scenes. In order to detect fire from scenes of color-videos, various schemes have been
studied, mainly focus on the combination of static and dynamic characteristics of fire such as color
information, texture and motion orientation, etc. Examine the convolution to address the above issue
depth function based on a neural network (CNN) for early fire detection in a surveillance network.
Our most important original contribution. The summary is as follows.

1) Avoid the time-consuming effort of traditional ones Handmade features for fire detection and
detailed investigation .Learning the architecture for early fire detection in both CCTV surveillance
networks Indoor and outdoor environments.

2) The fire Detection framework improves the accuracy of fire detection Reduces false positives
compared to state-of-the-art Method. Therefore, this algorithm can play an important role Early fire
detection to minimize damage[2][3].
II. Literature Review

1. Image Fire Detection Based on Support Vector Machine Ke Chen, Yanying Cheng have proposed
‘Image Fire Detection Based on Support Vector Machine’. Traditional temperature and smoke fire
detectors are subject to environmental conditions such as the height of the monitoring space, air
velocity, and dust in order to identify and alert early fire quickly and effectively. By researching the
properties of fire in digital images, an image fire detection technique based on support vector machine
is developed. To begin, the inter-frame difference approach is used to extract the motion region,
which is then referred to as the suspected fire area. The uniform size is then sampled one more.
Finally, the flame colour moment feature and texture feature are extracted and used to classify and
recognise objects using a support vector machine. Data sets were created by combining Internet
resources and self-shot films, and the trained support vector machine was put to the test. The results of
the tests revealed that the algorithm is more accurate in detecting early fire [1].

2. Efficient Deep CNN-Based Fire Detection and Localization in Video Surveillance Applications
Khan Muhammad, Jamil Ahmad, et Al [6] have proposed ‘Efficient Deep CNN-Based Fire Detection
and Localization in Video Surveillance Applications’. Convolutional neural networks (CNNs) have
yielded state-of-the-art performance in image classification and other computer vision tasks. Their
application in fire detection systems will substantially improve detection accuracy, which will
eventually minimize fire disasters and reduce the ecological and social ramifications. However, the
major concern with CNN-based fire detection systems is their implementation in real-world
surveillance networks, due to their high memory and computational requirements for inference. It
proposes an original, energy-friendly, and computationally efficient CNN architecture, inspired by the
SqueezeNet architecture for fire detection, localization, and semantic understanding of the scene of
the fire. It uses smaller convolutional kernels and contains no dense, fully connected layers, which
helps keep the computational requirements to a minimum. Despite its low computational needs, the
experimental results demonstrate that our proposed solution achieves accuracies that are comparable
to other, more complex models, mainly due to its increased depth. Moreover, it shows how a trade-off
can be reached between fire detection accuracy and efficiency, by considering the specific
characteristics of the problem of interest and the variety of fire data [2]. Video Surveillance Fire
Detection System

3. An Improved Multi-Scale Fire Detection Method Based on Convolutional Neural Network Huang
Hongyu1, Kuang Ping1 ET. Al [4] have proposed ‘An Improved Multi-Scale Fire Detection Method
Based on Convolutional Neural Network’. Fire is a major disaster in the world, and the fire detection
system should accurately detect the fire in the shortest time to reduce economic loss and ecological
damage. Traditional sensors are still widely used in a large number of applications, but they do not
perform well in remote high-dome environments or the early stages of low-flame fires, and now the
method of using image and video to predict fire is becoming more and more popular. It proposes an
improved YOLOv4 fire detection method based on Convolutional Neural Networks (CNN). The
accuracy of the model through the self-built highquality fire dataset, use the changed loss function to
improve the detection ability of small-scale flames, and combine the Soft-NMS post-processing and
DIoUNMS post-processing to improve the suppression effect of the redundant Bounding box and
reduce low recall rate. The experimental results of the model show that the model has an excellent
performance in fire detection and can detect multi-scale fire in real-time [3].

4. Prototype of Fire Symptom Detection System Oxsy Giandi, Riyanarto Sarno have proposed
‘Prototype of Fire Symptom Detection System’. It is one of the smart home function it is a fire alert
detection. The symptom detection of fire in the house is important action to prevent the mass fire and
save many things. It applies the new system of fire detection using gas leak concentration to predict
the explosion and fire earlier called fire predictor and the fire appearance detector. The fire predictor
just show the gas leak concentration and make an alarm rang. The fire detector use fuzzy system to
make the fire detector classification. The output simulation system can send the data to MFC, but the
MFC reader cannot parse it in real time [4]. Video Surveillance Fire Detection System

5. Design and experimental research of video detection system for ship fire Jiang Feng, Yang Feng
have proposed ‘Design and experimental research of video detection system for ship fire’. In order to
make up for the shortcomings of traditional fire detectors and improve the reliability of fire alarm,
based on the Raspberry Pi hardware conditions and the Keras deep learning framework, it uses the
lightweight direct regression detection algorithm YOLO v3-tiny to implement a small local video
identification system for ship fire. Based on video test and fire simulation, the RpiFire system has
achieved high accuracy in high recall rate and can meet the needs of ship fire detection [5].

6. A Comprehensive Study on Fire Detection Sneha Wilson, Shyni P Varghese, etc Al [5] have
proposed ‘A Comprehensive Study on Fire Detection’. The Accidents due to undetected fire have
caused the great cost to the world. The need for efficient fire detection system is rising. Existing fire,
smoke detectors are failing because of the inefficiency of the system. A vision based system with the
video surveillance fire detection system is proposed to have high detection rate and low fault alert
rate. Real-time fire detection is achieved by analysing live camera footage. The fire flame features are
studied and using edge detection, thresholding methods fire is detected, thus establishing a fire
detection model. It uses color, motion, shape, and texture of the fire to detect hazardous fire. Color
models like HSV, YCbCr are used in the system for more effective detection. It can be used for both
indoor and outdoor scenes [6]. Video Survelliance Fire Detection System

7. Automatic Fire and Smoke Detection Method for Surveillance Systems Based on Dilated CNNs
Yakhyokhuja Valikhujaev, Akmalbek Abdusalomov, etc Al [3] have proposed ‘Automatic Fire and
Smoke Detection Method for Surveillance Systems Based on Dilated CNNs’. The technologies
underlying fire and smoke detection systems play a crucial role in ensuring and delivering optimal
performance in modern surveillance environments. In fact, fire can cause significant damage to lives
and properties. Considering that the majority of cities have already installed camera-monitoring
systems, this encouraged us to take advantage of the availability of these systems to develop cost-
effective vision detection methods. However, this is a complex vision detection task from the
perspective of deformations, unusual camera angles and viewpoints, and seasonal changes. To
overcome these limitations, It is proposed with a new method based on a deep learning approach,
which uses a convolutional neural network that employs dilated convolutions. They evaluated their
method by training and testing it on their custom-built dataset, which consists of images of fire and
smoke that they have collected from their internet and labelled manually. The performance of their
method was compared with that of methods based on well-known state-of-the-art architectures. Their
experimental results indicate that the classification performance and complexity of their method are
superior. In addition, their method is designed to be well generalized for unseen data, which offers
effective generalization and reduces the number of false alarms [7]. Video Surveillance Fire Detection
System

8. Fire Smoke Detection Based on Contextual Object Detection Xuan Zhao, Hang Ji, have proposed
‘Fire Smoke Detection Based on Contextual Object Detection’. The Smoke detection based on
automatic visual system has been applied to fire alarm in open spaces where traditional smoke
detection system is not suitable for it. However, detecting the course of smoke posed great challenges
for both systems. To address this problem, they have proposed a new method that combines context-
aware framework with automatic visual smoke detection. The strategy is evaluated on dataset and the
results demonstrate the effectiveness of the proposed method [8].

9. A New Fire Detection Method Based on the Centroid Variety of Consecutive Frames Shi Leil, Shi
Fangfei2, etc Al[5] have proposed ‘A New Fire Detection Method Based on the Centroid Variety of
Consecutive Frames’, In the field of video fire detection, traditional fire color models have the poor
adaptability and week robustness to interference. Therefore, a new fire detection method referring to
the centroid variety of fire in consecutive frames is proposed in this paper. Firstly, the areas similar to
fire are detected out through RGB-HIS color model. Then, the centroid movement of these areas are
calculated with video tracking algorithm and a new fire detection model is proposed. A series of
experiment results show that the proposed method can eliminate the influence of common
interferences and prompt the fire warning correctly, which has a certain practical significance for
indoor fire detection [9]. Video Surveillance Fire Detection System

10. Aerial Forest Fire Surveillance – Evaluation of Forest Fire Detection Model using Aerial Videos
Hanh Dang-Ngoc, Hieu Nguyen-Trung have proposed ‘Aerial Forest Fire Surveillance – Evaluation
of Forest Fire Detection Model using Aerial Videos’. The Unmanned Aerial Vehicles (UAVs), which
can provide an aerial view for fast responding in large-scale zones of disaster, are recently utilized for
forest fire monitoring. In this paper, one general model of forest fire detection using aerial videos is
investigated to prove its robustness for practical application of aerial forest fire surveillance. Fire
pixels are extracted using the color and motion characteristics of fire. The fire detection performance
is evaluated through a large database of various scene conditions to show the efficiency as well as
deficiency of the fire detection model in previous study. Their database consists of 49 aerial videos
with total of 16898 examined frames of forest fires. The accuracy rate of this forest fire detection
model is 93.97 parentage while the false alarm rate and the miss rate are 7.08 parentage and 6.86
parentage, respectively. Thick smoke which covers almost the fire is found as the main cause of miss
detection in the fire detection model. To enhance the detection performance, It has one more stage of
smoke detection. Smoke pixels are segmented using both color and motion characteristics of smoke.
The results prove that smoke detection stage give help in detecting the fire area in case of smoke [10].

III. METHODOLOGY

Figure 4.1 shows, theproposed system input as video then Pre-processing refers to all the
transformations on the raw data before it is fed to the machine learning or deep learning algorithm.
For instance, training a convolutional neural network on raw images will probably lead to bad
classification performance CNN is a neural network that extracts input image features and another
neural network classifies the image features. The input image is used by the feature extraction
network. The extracted feature signals are utilized by the neural network for classification.

• CNN (Regions with CNN feature) is one representative work for the region based methods. It
performs the semantic segmentation based on the object detection results.

• To be specific, CNN first utilizes selective search to extract a large quantity of object proposals and
then computes CNN features for each of them. The convolutional neural network (CNN) is a class of
deep learning neural networks. CNNs represent a huge breakthrough in image recognition. They’re
most commonly used to analyse visual imagery and are frequently working behind the scenes in video
classification. Video Surveillance Fire Detection System

• Preprocessing- Preprocessing refers to the transformations applied to our data before feeding it to
the algorithm. Data Preprocessing is a technique that is used to convert the raw data into a clean data
set.

• Feature Extraction- Aims to reduce the number of features in a dataset by creating new features from
the existing ones (and then discarding the original features). These new reduced set of features should
then be able to summarize most of the information contained in the original set of features.
• Data Segmentation- is the process of taking the data you hold and dividing it up and grouping
similar data together based on the chosen parameters so that you can use it more efficiently within
marketing and operations. Examples of Data Segmentation could be: Gender. Customers vs.
Prospects.

• Classification using CNN-Convolutional Neural Network is a Deep Learning algorithm specially

designed for working with Images and videos. It takes images as inputs, extracts and learns the
features of the image, and classifies them based on the learned features.

• Detect the fire or Not.

Figure 3.1: System Architecture

IV. PROPOSED SYSTEM

In Data Flow Diagram, It shows the flow of data in the proposed system in DFD0 it also shows that
base DFD in which rectangle present input as well as output and circle show our proposed system, In
DFD1 It shows actual input and actual output of system input of proposed system is text or image and
output is rumor detected likewise in DFD 2. The Proposed system present operation of user as well as
admin.

DFD - 1.0: Figure 4.2 shows DFD Level 0 which is also called a Context Diagram. It’s a basic
overview of the Fire Detection system or process being analyzed or modeled. It’s designed to be an
at-a-glance view, showing the system as a single high-level process, with its relationship to external
entities. Here we have Input in Image/Video format which is the

Processed and gives the output i.e. Fire detect or not.

 Figure 4.2: Data Flow-1.0 diagram

DFD - 1.1: Figure 4.3 shows level 1 DFD notates each of the main sub-processes that together form
the complete Fire Detection system. We can think of a level 1 DFD as an “expanded view” of the
context diagram. The proposed System have Image Processing, Feature Extraction,

Classification.

Figure 4.3: Data Flow-1.2 diagram

DFD - 2.1: Figure 4.4 shows the highest abstraction level DFD is known as Level 2 DFD depicts
basic modules in the system and flow of data among various modules. In DFD 2 level show flow of
fire detection system.
 Figure 4.4: Data Flow-2.1 diagram

V.MATHEMATICAL MODEL

Let S be the Whole system S= I, P,O

I - input

P - Procedure

O - Output

Input( I)

I= Input as Fire image/video Dataset

Where,

Step 1. Dataset contain Fire images/Video dataset.

Procedure (P),

P= 1. Train the dataset (Fire dataset)

Step 2. Pre-processing

Step 3. Feature Extraction

Step 4(a). Segmentation

Step 4(b). Classification (used CNN algorithm)

Step 5. Create Model.

Output (O)-

O= System detect Fire or Not.

1. CNN layer is calculated as : “inputsize-(filtersize-1)”

2. To calculate padding: inputsize + 2 * paddingsize - (filtersize - 1).

3. To calculate the number of parameters the network learned. (n*m*l+1)*kWhere, n = Width of

filter. m = Height of filter. l = Number of input feature maps. k = Number of output feature maps
[2].

Figure 4.5: Mathematical Equation

It is a process where we take a small matrix of numbers (called kernel or filter), we pass it over our
image and transform it based on the values from filter. Subsequent feature map values are calculated
according to the following formula, Where, f = input image, h = kernel, m = indexes of rows, n =
indexes of columns.

VI. Results and discussion:

Figure 4.11: Main Window

As shown in Figure 4.10. This is the Main Window of the proposed Fire Detection System. Here, First
the user has to create his/her account. After creating an account user has to login with his/her Userid
i.e Username and Password. These data is saved into the database. The proposed system uses SQlite
database for storing the data and the records.
Figure 4.12: Registration Window

As shown in figure 4.11, The Proposed System has Registration Window through which user has to
register his/her account. The System has predefined format which user has to follow

While registration.

Figure 4.13: Login Window

As shown in figure 4.12, Here The Proposed System consist of Login window through which a user
can login to his account. The Username is the one which he/she has provided while registration. Only
valid account will be get logged in and he/she can access the Fire Detection

System.

Figure 4.14: Fire Detection Window

As shown in figure 4.13, This is the Fire detection window. Here we have a Fire detection button
through which we will take the input data which will be in terms of Video. The System will perform
various modules to determine whether fire is there or not.

Figure 4.15: Fire Detection Frame 1

As shown in figure 4.14, here we have taken input data set as a video and with the help of the video
frames we are detecting the fire at initial stage it is observing the environment and according to this it
will give alert message.

Figure 4.16: Fire Detection Frame 2

As shown in figure 4.15, In this Frame as we can see that the fire has been detected an according to
that the abnormal activity detected it will send alert message to the user.

VII. Conclusion and future work

7.1 CONCLUSION

In summary, an aerial based forest fire detection method has been examined through a large database
of videos of forest fires of various scene conditions. To enhance the detection rate, at first the
chromatic and motion features of forest fire are extracted and then corrected using rule to point out the
fire area. Secondly, to overcome the challenge of heavy smoke that covers almost the fire, smoke is
also extracted using our proposed algorithm. Proposed System framework proves its robustness with
high accuracy rate of detection and low false alarm rate in practical application of aerial forest fire
surveillance. Future studies may focus on making challenging and specific scene understanding data-
sets for fire detection methods and detailed experiments. Furthermore, reasoning theories and
information hiding algorithms can be combined with fire detection systems to intelligently observe
and authenticate the video stream and initiate appropriate action, in an autonomous way.
7.2 FUTURE WORK

Future studies may focus on making challenging and specific scene understanding datasets for fire
detection methods and detailed experiments. Furthermore, reasoning theories and information hiding
algorithms can be combined with fire detection systems to intelligently observe and authenticate the
video stream and initiate appropriate action, in an autonomous

VIII. References

[1] C. Kao and S. Chang, “An Intelligent Real-Time Fire-Detection Method Based on
VideoProcessing,” IEEE (2003 Int.)

[2] C. Ha, U. Hwang, G. Jeon, J. Cho, and J. Jeong, “Vision-based fire detection
algorithm using optical flow”, 2012.

[3] C. E. Premal and S. S. Vinsley, “Image Processing Based Forest Fire Detection using
YCbCr Colour Model”, 2014.

[4] N. I. Binti Zaidi, N. A. A. Binti Lokman, M. R. Bin Daud, H. Achmad, and K. A.

Chia,“Fire recognition using RGB and YCbCr color space”, 2015.

[5] K. Poobalan and S. Liew, “Fire Detection Algorithm Using Image Processing
Techniques,”, December, 2015.

[6] Khan Muhammad, Jamil Ahmad, Zhihan Lv, Paolo Bellavista, Po Yang,
SungWookBaik, “Efficient Deep CNN-Based Fire Detection and Localization in Video
Surveillance Applications”, 2018,IEEE.

[7] Oxsy Giandi, Riyanarto Sarno “Prototype of Fire Symptom Detection System”,
©2018IEEE.

[8] Sneha Wilson, Shyni P Varghese, Nikhil G A, Manolekshmi I, Raji P G, “A

Comprehensive Study on Fire Detection”, (2018).

[9] Jiang Feng, Yang Feng, “Design and experimental research of video detection system
for ship fire”, ©2019 IEEE.

[10] Ke Chen,Yanying Cheng, “Research on Image Fire Detection Based on Support

VectorMachine”,(2020),IEEE

[11] HUANG HONGYU1, KUANG PING1, LI FAN1, SHI HUAXIN1, “An Improved
Multi-

[12] Scale Fire Detection Method Based On Convolutional Neural Network”, © 2020
IEEE

[13] M. Mueller, P. Karasev, I. Kolesov, and A. Tannenbaum, "Optical flow estimation for
flame detection in videos," IEEE Transactions on Image Processing, vol. 22, pp. 2786-2797,
2013.

[14] [15] B. U. Töreyin, Y. Dedeoğlu, U. Güdükbay, and A. E. Cetin, "Computer vision

based method for real-time fire and flame detection," Pattern recognition letters, vol. 27, pp.
49-58, 2006.

[15] R. C. Luo and K. L. Su, "Autonomous fire-detection system using adaptive sensory
fusion for intelligent security robot," Ieee/Asme Transactions on Mechatronics, vol. 12, pp.
274-281, 2007.

[16] P. V. K. Borges and E. Izquierdo, "A probabilistic approach for vision-based fire
detection in videos," IEEE transactions on circuits and systems for video technology, vol. 20,
pp. 721-731, 2010.

[17] I. Mehmood, M. Sajjad, and S. W. Baik, "Mobile-Cloud Assisted Video

Summarization Framework for Efficient Management of Remote Sensing Data Generated by
Wireless Capsule Sensors," Sensors, vol. 14, pp. 17112-17145, 2014.

[18] A. Sorbara, E. Zereik, M. Bibuli, G. Bruzzone, and M. Caccia, "Low cost optronic
obstacle detection sensor for unmanned surface vehicles," in Sensors Applications Symposium
(SAS), 2015 IEEE, 2015, pp. 1-6.

[19] B. C. Ko, K.-H. Cheong, and J.-Y. Nam, "Fire detectionbased onvision sensorand
support vector machines," Fire Safety Journal, vol. 44, pp. 322-329, 2009.

[20] K. Dimitropoulos, P. Barmpoutis, and N. Grammalidis, "Spatio-temporal flame

modeling and dynamic texture analysis for automatic video-based fire detection," IEEE
transactions on circuits and systems for video technology, vol. 25, pp. 339-351, 2015.

[21] Z. Zhang, J. Zhao, D. Zhang, C. Qu, Y. Ke, and B. Cai, "Contour based forest fire
detection using FFT and wavelet," in Computer Science and Software Engineering, 2008
International Conference on, 2008, pp. 760-763. [23] T. Celik and H. Demirel, "Fire detection
in video sequences using a generic color model," Fire Safety Journal, vol. 44, pp. 147-158,
2009.

[22] G. Marbach, M. Loepfe, and T. Brupbacher, "An image processing technique for fire
detection in video images," Fire safety journal, vol. 41, pp. 285-289, and 2006.
 [23] W. Phillips Iii, M. Shah, and N. da Vitoria Lobo, "Flame recognition in video,"
Pattern recognition letters, vol. 23, pp. 319-327, 2002.

[24] D. Han and B. Lee, "Development of early tunnel fire detection algorithm using the
image processing," in International Symposium on Visual Computing, 2006, pp. 39-48.

View publication stats