COMPUTER VISION-BASED EARLY FIRE DETECTION

USING OPEN CV AND MACHINE LEARNING

Submitted in partial fulfillment of the requirements for the award of
Bachelor of Engineering degree in
Computer Science and Engineering

By

PETA MANISH KUMAR REDDY(Reg. No.38110404)

POTU LAKSHMI PRANEETH REVANTH KUMAR(Reg. No.38110414)

DEPARTMENT OF COMPUTER SCIENCE AND

ENGINEERING SCHOOL OF COMPUTING

SATHYABAMA
INSTITUTE OF SCIENCE AND TECHNOLOGY
(DEEMED TO BE UNIVERSITY)Accredited with Grade “A” by
NAAC JEPPIAAR NAGAR, RAJIV GANDHI SALAI, CHENNAI –
600 119

MARCH - 2022
 SATHYABAMA
INSTITUTE OF SCIENCE AND TECHNOLOGY
(DEEMED TO BE UNIVERSITY)
Accredited with “A” grade by NAAC
Jeppiaar Nagar, Rajiv Gandhi Salai, Chennai – 600 119
www.sathyabama.ac.in

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

BONAFIDE CERTIFICATE

This is to certify that this project report is the bonafide work of PETA MANISH
KUMAR REDDY(Reg. No.38110404) and POTU LAKSHMI PRANEETH
REVANTH KUMAR(Reg. No.38110414) who carried out the project entitled
“COMPUTER VISION-BASED EARLY FIRE DETECTION USING OPEN CV AND
MACHINE LEARNING ” under my supervision from August 2021 to March 2022.

Internal Guide

Dr.S.Raja Shree M.E., P.h.D.,

Head of the Department

Dr. S. Vigneswari, M.E., Ph.D.,

Dr. L. Lakshmanan, M.E., Ph.D.,

Submitted for Viva voce Examination held on

Internal Examiner External Examiner

 DECLARATION

I PETA MANISH KUMAR REDDY and POTU LAKSHMI PRANEETH

REVANTH KUMAR hereby declare that the Project Report entitled “

COMPUTER VISION-BASED EARLY FIRE DETECTION USING OPEN

CV AND MACHINE LEARNING ” is done by me under the guidance of

Dr.S.Raja Shree M.E.,P.h.D., Department of Computer Science and

Engineering at Sathyabama Institute of Science and Technology is
submitted in partial fulfillment of the requirements for the award of
Bachelor of Engineering degree in Computer Science and Engineering.

DATE:

PLACE: CHENNAI SIGNATURE OF THE CANDIDATE

 ACKNOWLEDGEMENT

I am pleased to acknowledge my sincere thanks to Board of Management of

SATHYABAMA for their kind encouragement in doing this project and for
completing it successfully. I am grateful to them.

I convey my thanks to Dr. T. Sasikala, M.E., Ph.D., Dean, School of Computing,

Dr. S. Vigneswari, M.E., Ph.D., and Dr. L. Lakshmanan, M.E., Ph.D., Heads of
the Department of Computer Science and Engineering for providing me
necessary support and details at the right time during the progressive reviews.

I would like to express my sincere and deep sense of gratitude to my Project

Guide Dr.S.Raja Shree M.E.,P.h.D., for her valuable guidance, suggestions and
constant encouragement paved way for the successful completion of my project
work.

I wish to express my thanks to all Teaching and Non-teaching staff members of the
Department of Computer Science and Engineering who were helpful in many
ways for the completion of the project.
 ABSTRACT

The project aimed to detect fire by using the image processing

technology that will alert people by early detection of fire. As there are many
automatic fire alarm systems already existed like the sensor method, that has
some limitations and designed to sense fire with the smoke, limited areas. To
reduce limitations and to optimize with new technology, the project is proposed.
The project is implemented by using pycharm IDE and to connect the webcam as
hardware.Webcam is taken as an input source, which captures the video feed
from the surrounding and feeds into the system for analysis.The entire code is
written in pure python language using the open CV library for image processing.
The theoretical parts emphasize more in computer vision, machine learning,
image processing, color model, and the working algorithm of the project to detect
the fire. The project gives a better understanding of object detection with the
computer and the use of these technologies in different forms and uses.
 Table of contents

Chapter No. Title Page No.

ABSTRACT V
LIST OF FIGURES VII

1. INTRODUCTION 9
1.1 Overview 9
1.2 Scope and Objective 10
1.3 Domain Overview 10
1.3.1 Machine Learning Model 11
1.3.2 Data Pre-processing 12
1.3.3 Supervised Learning 14
1.3.4 Regression 16
1.3.5 Unsupervised Learning 17
1.3.6 Training and testing the 26
model of data
1.3.7 Validation Set 19

2. LITERATURE SURVEY 22
2.1 Analysis of literature 22
2.2 Literary review 23

3. SYSTEM ANALYSIS 26
3.1 Existing System 26
3.2 Problem Statement 26
3.3 Proposed System 26
3.4 System Design 28
 3.5 Haar Like Features 29
3.6 Working Principle 32
3.7 System Requirements 33

4. RESULTS 34

5. CONCLUSION AND FUTURE WORK 35

5.1 Conclusion 35
5.2 Limitation and Future Work 35

6. REFERENCE PAPERS 36

7. APPENDIX 38
A. Sample Code 38
B. Plagarism Report 43
 LIST OF FIGURES

FIGURE No. FIGURE NAME PAGE No.

1.1 Supervised Architecture 14

1.2 Unsupervised Architecture 14
1.3 RGB Colours 15
1.4 Prediction Of IQ 16
1.5 Original Clustered Data 17
1.6 Clustered Data 17
1.7 Classifications Of Machine Learning 18
1.8 Training Set 18
1.9 Validation Set 19
1.10 Confusion Matrix 20
3.1 Block Diagram 27
3.2 Flow Chart 28
3.3 Haar Features 30
4.1 Result ScreenShot 34
 CHAPTER 1
INTRODUCTION
1.1 OVERVIEW

Computer Vision based fire detection using image processing has the potential to
be useful in conditions in which conventional methods cannot be adopted. The fire
detection algorithm uses visual characteristics of fires like brightness, color,
spectral texture, spectral flicker, and edge trembling to discriminate them from
other visible stimuli.There are various fire detection techniques such as infrared
sensor, a thermal detector, smoke detector, flame detector, and optical smoke
detector. These methods are not always reliable as they do not always detect the
fire itself but detect one or more phenomena resulting from fire, such as smoke,
heat, infrared, ultraviolet light radiation or gas, which could be produced in other
ways and hence, produces many false alarms.By the help of computer vision and
image processing techniques, it is possible to get better results than conventional
systems because images can provide more reliable information.
In recent times, research on detection of flame and smoke using surveillance
cameras with machine vision has gained momentum. The image processing
approach involves the extraction of the smoke-plume or flame from the
background by using frame difference technologies. In the case of the
segmentation of fire features, color processing scores over gray-scale processing.
Color processing can avoid the generation of false alarms due to variations in the
lighting conditions, e.g. natural background illumination, better than gray-scale
processing. Further, a video camera is a volume sensor, and potentially monitors a
larger area. The traditional point sensor looks at a point in space. Since the point
sensor may not be affected by smoke or flame, fire would be undetected.
However, vision-based flame and smoke detection still has great technical
challenges, since flame and smoke are non-rigid objects, with none of the primitive
image features and variability in density, lighting, etc.
1.2 SCOPE AND OBJECTIVE

The proposed software product is used in various areas like in parking

areas, road traffic congestions, agriculture, farming, there are many uses of object
detection.we are using the Haar Cascade classifier, which is very popular in object
detection through the image or any other video feeds.In future, this approach can
be extended by for other disasters as well.

1.3 DOMAIN OVERVIEW

Steps of Machine Learning

● Step 1: Gathering Data

● Step 2: Preparing that Data

● Step 3: Choosing a Model

● Step 4: Training

● Step 5: Evaluation

● Step 6: Hyper parameter Tuning

● Step 7: Prediction

Introduction:

In this blog, we will discuss the workflow of a Machine learning project this
includes all the steps required to build the proper machine learning project from
scratch.We will also go over data pre-processing, data cleaning, feature
exploration and feature engineering and show the impact that it has on Machine
Learning Model Performance. We will also cover a couple of the pre-modelling
steps that can help to improve the model performance.

Python Libraries that would be need to achieve the task:

1.Numpy
2.Pandas
3.Sci-kitLearn
4. Matplotlib

Understanding the machine learning workflow

We can define the machine learning workflow in 3 stages.

1. Gathering data

2. Data pre-processing

3. Researching the model that will be best for the type of data

4. Training and testing the model

5. Evaluation

1.3.1 MACHINE LEARNING MODEL

The machine learning model is nothing but a piece of code; an engineer or data
scientist makes it smart through training with data. So, if you give garbage to the
model, you will get garbage in return, i.e. the trained model will provide false or
wrong prediction

Gathering Data:

The process of gathering data depends on the type of project we desire to make, if
we want to make an ML project that uses real-time data, then we can build an IoT
system that using different sensors data. The data set can be collected from
various sources such as a file, database, sensor and many other such sources but
the collected data cannot be used directly for performing the analysis process as
there might be a lot of missing data, extremely large values, unorganized text data
or noisy data. Therefore, to solve this problem Data Preparation is done.We can
also use some free data sets which are present on the internet. Kaggle and UCI
Machine learning Repository are the repositories that are used the most for
making Machine learning models. Kaggle is one of the most visited websites that
is used for practicing machine learning algorithms, they also host competitions in
which people can participate and get to test their knowledge of machine learning.
1.3.2 Data pre-processing:

Data pre-processing is one of the most important steps in machine learning. It is

the most important step that helps in building machine learning models more
accurately. In machine learning, there is an 80/20 rule. Every data scientist should
spend 80% time for data per-processing and 20% time to actually perform the
analysis.Data pre-processing is a process of cleaning the raw data i.e. the data is
collected in the real world and is converted to a clean data set. In other words,
whenever the data is gathered from different sources it is collected in a raw format
and this data isn’t feasible for the analysis.Therefore, certain steps are executed to
convert the data into a small clean data set, this part of the process is called as
data pre-processing.

As we know that data pre-processing is a process of cleaning the raw data into
clean data, so that can be used to train the model. So, we definitely need data
pre-processing to achieve good results from the applied model in machine
learning and deep learning projects.Most of the real-world data is messy, some of
these types of data are:

● Missing data: Missing data can be found when it is not continuously created
or due to technical issues in the application (IOT system).
● Noisy data: This type of data is also called outliners, this can occur due to
human errors (human manually gathering the data) or some technical
problem of the device at the time of collection of data.
● Inconsistent data: This type of data might be collected due to human errors
(mistakes with the name or values) or duplication of data.
How can data pre-processing be performed?

These are some of the basic pre — processing techniques that can be used to
convert raw data.

● Conversion of data: As we know that Machine Learning models can only

handle numeric features, hence categorical and ordinal data must be
somehow converted into numeric features.
● Ignoring the missing values: Whenever we encounter missing data in the
data set then we can remove the row or column of data depending on our
need. This method is known to be efficient but it shouldn’t be performed if
there are a lot of missing values in the dataset.
● Filling the missing values: Whenever we encounter missing data in the data
set then we can fill the missing data manually, most commonly the mean,
median or highest frequency value is used.
● Machine learning: If we have some missing data then we can predict what
data shall be present at the empty position by using the existing data.
● Outliers detection: There are some error data that might be present in our
data set that deviates drastically from other observations in a data set.
[Example: human weight = 800 Kg; due to mistyping of extra 0]
Searching the suitable model :

Our main goal is to train the best performing model possible, using the
pre-processed data.The Machine learning is divided into two parts namely
Supevised Learning and Unsupervised Learning.So these methods will help us
with the best model.

Fig 1.1 Supervised Learning Fig 1.2 Unsupervised Learning

1.3.3 Supervised Learning:

In Supervised learning, an AI system is presented with data which is labelled,

which means that each data tagged with the correct label.

The supervised learning is categorized into 2 other categories which are

“Classification” and “Regression”.

Classification:

Classification problem is when the target variable is categorical (i.e. the output
could be classified into classes — it belongs to either Class A or B or something
else).
A classification problem is when the output variable is a category, such as “red” or
“blue” , “disease” or “no disease” or “spam” or “not spam”.

Fig 1.3 RGB Colours

As shown in the above representation, we have 2 classes which are plotted on the
graph i.e. red and blue which can be represented as ‘setosa flower’ and ‘versicolor
flower’, we can image the X-axis as there ‘Sepal Width’ and the Y-axis as the
‘Sepal Length’, so we try to create the best fit line that separates both classes of
flowers.

These some most used classification algorithms.

● K-Nearest Neighbor

● Naive Bayes

● Decision Trees/Random Forest

● Support Vector Machine

● Logistic Regression
1.3.4 Regression:

Regression is a technique for investigating the relationship between independent

variables or features and a dependent variable or outcome. It's used as a method
for predictive modelling in machine learning, in which an algorithm is used to
predict continuous outcomes.

1.4 Prediction of IQ

As shown in the above representation, we can imagine that the graph’s X-axis is
the ‘Test scores’ and the Y-axis represents ‘IQ’. So we try to create the best fit line
in the given graph so that we can use that line to predict any approximate IQ that
isn’t present in the given data.

These are some most used regression algorithms.

● Linear Regression

● Support Vector Regression

● Decision Tree/Random Forest

● Gaussian Processes Regression

● Ensemble Methods

1.3.5 Unsupervised Learning:

The unsupervised learning is categorized into 2 other categories which are “Clustering”
and “Association”.

Clustering:

A set of inputs is to be divided into groups. Unlike in classification, the groups are
not known beforehand, making this typically an unsupervised task.

Fig 1.5 Original Clustered Data Fig 1.6 Clustered Data

Methods used for clustering are:

● Gaussian mixtures

● K-Means Clustering

● Boosting
● Hierarchical Clustering

● K-Means Clustering

● Spectral Clustering

Overview of models under categories:

Fig 1.7 Classifications of Machine Learning

1.3.6 Training and testing the model on data:

For training a model we initially split the model into 3 three sections which are
‘Training data’ ,‘Validation data’ and ‘Testing data’.You train the classifier using
‘training data set’, tune the parameters using ‘validation set’ and then test the
performance of your classifier on unseen ‘test data set’. An important point to note
is that during training the classifier only the training and/or validation set is
available. The test data set must not be used during training the classifier. The test
set will only be available during testing the classifier.
 Fig 1.8 Training set

Training set:

The training set is the material through which the computer learns how to process
information. Machine learning uses algorithms to perform the training part. A set of
data used for learning, that is to fit the parameters of the classifier.

1.3.7 Validation set:

Cross-validation is primarily used in applied machine learning to estimate the skill

of a machine learning model on unseen data. A set of unseen data is used from
the training data to tune the parameters of a classifier.

Fig 1.9 Validation Set

Once the data is divided into the 3 given segments we can start the training
process.

In a data set, a training set is implemented to build up a model, while a test (or
validation) set is to validate the model built. Data points in the training set are
excluded from the test (validation) set. Usually, a data set is divided into a training
set, a validation set (some people use ‘test set’ instead) in each iteration, or
divided into a training set, a validation set and a test set in each iteration.

The model uses any one of the models that we had chosen in step 3/ point 3.
Once the model is trained we can use the same trained model to predict using the
testing data i.e. the unseen data. Once this is done we can develop a confusion
matrix, this tells us how well our model is trained. A confusion matrix has 4
parameters, which are ‘True positives’, ‘True Negatives’, ‘False Positives’ and
‘False Negative’. We prefer that we get more values in the True negatives and true
positives to get a more accurate model. The size of the Confusion matrix
completely depends upon the number of classes.

1.10 Confusion Matrix

True positives : These are cases in which we predicted TRUE and our predicted
output is correct.

True negatives : We predicted FALSE and our predicted output is correct.

False positives : We predicted TRUE, but the actual predicted output is

FALSE.
False negatives : We predicted FALSE, but the actual predicted output is TRUE.

We can also find out the accuracy of the model using the confusion matrix.

Accuracy = (True Positives +True Negatives) / (Total number of classes)

i.e. for the above example:

Accuracy = (100 + 50) / 165 = 0.9090 (90.9% accuracy)

Evaluation:

Model Evaluation is an integral part of the model development process. It helps to

find the best model that represents our data and how well the chosen model will
work in the future.

To improve the model we might tune the hyper-parameters of the model and try to
improve the accuracy and also looking at the confusion matrix to try to increase
the number of true positives and true negatives.

Conclusion:

In this blog, we have discussed the workflow a Machine learning project and gives
us a basic idea of how a should the problem be tackled.
 CHAPTER 2
LITERATURE SURVEY

2.1 ANALYSIS OF THE LITERATURE

Literature survey is the main advance in programming improvement measure.

Prior to building up the instrument it is important to decide the time factor,
economy and friends strength. When these things are fulfilled, at that point the
subsequent stage is to figure out which working framework and language can be
utilized for building up the device. When the developers begin assembling the
apparatus the software engineers need parcel of outer help. This help can be
gotten from senior developers, from book or from sites. The major part of the
project development sector considers and fully survey all the required needs for
developing the project. Before developing the tools and the associated designing it
is necessary to determine and survey the time factor, resource requirement, man
power, economy, and company strength. Prior to building the framework the above
thought are considered for building up the proposed framework. The significant
piece of the undertaking advancement area considers and completely survey all
the necessary requirements for building up the venture. For each undertaking
Literature survey is the main area in programming improvement measure.
Prior to building up the instruments and the related planning it is important to
decide and survey the time factor, asset prerequisite, labor, economy, and friends
strength. When these things are fulfilled and completely surveyed, at that point the
following stage is to decide about the product details in the separate framework,
for example, what kind of working framework the venture would require and what
are largely the important programming are expected to continue with the
subsequent stage like building up the apparatuses, and the related activities. Here
we have taken the general surveys of different creators and noted down the
fundamental central issues with respect to their work. In this venture literature
survey assumes a prevailing part in get assets from different areas and all the
connected points that are exceptionally valuable under this segment. The most
awesome aspect if this is the manner in which things get all together and
encourages us to suite our work according to the current information.
2.2 LITERARY REVIEWS

Towards Distributed Event Detection in Wireless Sensor

Networks(IEEE2020)
Norman Dziengel et al. distributed event detection in wireless sensor networks
(WSNs) is the process of observing and evaluating an event using multiple sensor
nodes without the help of a base station or other means of central coordination
and processing. Current approaches to event detection in WSNs transmit raw data
to an external entity for evaluation or rely on simplistic pattern recognition
schemes. This implies either high communication overhead or low event detection
accuracy, especially for complex events. In this paper, we present our currently
on-going work on a system for distributed event detection that particularly suits the
specific characteristics of WSNs. Adapting traditional pattern recognition
algorithms to highly embedded devices, it uses the distributed sampling of sensor
nodes to optimize the accuracy of the event detection process. Four different
algorithms for distributing, classifying and fusing “fingerprints” of the raw data
sampled on each sensor are proposed and quantitatively evaluated in a
small-scale experiment.

Detecting and monitoring time-related abnormal events using a wireless

sensor network and mobile robot( IEEE, 2019)

YuanYuan Li and Lynne E. Parker et al. presented an anomaly detection system

that is able to detect time-related anomalies by using a wireless sensor network
and a mobile robot. The sensor network uses an unsupervised fuzzy Adaptive
Resonance Theory (ART) neural network to learn and detect intruders in a
previously unknown environment. Upon the detection of an intruder, a mobile
robot travels to the position where the intruder is detected to investigate by using
its camera. The wireless sensor network uses a hierarchical
communication/learning structure, where the mobile robot is the root node of the
tree. Our fuzzy ART network is based on Kulakov and Davcev’s implementation
[8]. However, we enhance their work by extending the fuzzy ART neural network
with a Markov model to learn a time series and detect time-related anomalies.
Finally, a mobile robot is employed to verify whether the detected anomalies
were caused by intruders.

A Novel Mechanism for Fire Detection in Subway Transportation Systems

Based on Wireless Sensor Networks(vol. 2019)

Z.-J. Zhang, J.-S. Fu, H.-P. Chiang, and Y.-M. Huang et al. said fire is a common
and disastrous phenomenon in subway transportation systems because of closed
environment and large passenger row. Traditional methods detect and forecast re
incidents by fusing the data collected by wireless sensor networks and compare
the fusion result with a threshold. However, this approach has a signicant
shortcoming. Fusion of data decreases the amount of useful information since the
distribution of data is not considered. In this paper, a novel mechanism based on
wireless sensor networks is proposed and we can use the mechanism to detect re
in subway transportation systems. e core algorithm in this mechanism is an
alternative of data fusion in wireless sensor networks. e mechanism uses the
quantity information acquired from the data and also considers the distribution
information of the data in order to predict the occurrence of re incidents. e
simulation results show that re incidents can be detected fast and eciently.

Forest Fire Smoke Detection Based on Visual Smoke Root and Diffusion
(IEEE2019)
The damage caused by forest fire to forestry resources and economy is quite
serious. As one of the most important characters of early forest fire, smoke is
widely used as a signal of forest fire. In this paper, we propose a novel forest fire
smoke detection method based on computer vision and diffusion model. Unlike
the video-based methods that usually rely on image characters extraction, we try
to find the shape of smoke that is at the generation stage. To combine vision and
diffusion model together, the basic concept of smoke root is proposed. In the
frame processing stage, none characters of fire smoke are extracted (like texture,
color, frequency information etc.), and continuous frames are only used to extract
stable points in dynamic areas as the smoke root candidate points. In the diffusion
model simulation stage, all smoke root candidate points information is adopted by
the model to generate the simulation smoke. Finally, the match algorithm based
on
color, dynamic areas and simulation smoke is implemented to get the final results.
In order to reduce the complexity of computation, we ignored the simulation
process of the smoke details, such as texture and turbulence, and only retained
the contour features in two-dimensional form.

Flutter Analysis Based Video Smoke Detection(IEEE2011)

Based on the flutter characteristic of smoke, a novel video smoke detection

algorithm is proposed. Firstly, the motion region is detected from video frames.
Then four flutter features of the motion region are extracted over a sliding time
window, including the flutter direction and three types of flutter intensities. Finally,
the neuro-fuzzy inference system is used to detect smoke, in which fuzzy rules
and membership functions are trained according to the valid sample set.
Experimental results show that the new algorithm has excellent performance on
Receiver Operating Characteristic (ROC) curve which is very important for real
application.

AN EARLY FIRE-DETECTION METHOD BASED ON IMAGE PROCESSING.

(IEEE2004 )

The paper presents an early fire-alarm raising method based on video processing.
The basic idea of the proposed of fire-detection is to adopt a RGB (red, green,
blue) model based chromatic and disorder measurement for extracting fire-pixels
and smoke-pixels. The decision function of fire-pixels is mainly deduced by the
intensity and saturation of R component. The extracted fire-pixels will be verified if
it is a real fire by both dynamics of growth and disorder, and further smoke. Based
on iterative checking on the growing ratio of flames, a fire-alarm is given when the
alarm-raising condition is met. Experimental results show that the developed
technique can achieve fully automatic surveillance of fire accident with a lower
false alarm rate and thus is very attractive for the important military, social security,
commercial applications, and so on, at a general cost.
 CHAPTER 3
SYSTEM ANALYSIS

3.1 EXISTING SYSTEM

The methods and classifiers used in the existing system are:
● Correlation based analysis and transformation methods
● Data Mining Approach to Predict Forest Fires using Meteorological Data
● SVM classifier
● Circle-based Approximation to Forest Fires with Distributed Wireless
Sensor Networks using clustering
● Event Detection in Wireless Sensor Networks for Disaster Management

3.2 PROBLEM STATEMENT:

The main motivation for fire prediction is to provide proper resource allocation and
to help in best possible way to firefighters of Fire Management team. The main
factors of fire are Meteorological conditions. the climatic information is gotten from
nearby sensors which are fused in the closest meteorological stations. Land with a
possible high fire risk has many indicators that can be used to measure the
forecast by closely evaluating the indications. Every year, fire destroys millions of
hectares of land. These fires have burned vast areas and generate more carbon
monoxide than total vehicle traffic.Monitoring potential danger areas and early
warning of fire can greatly reduce response time, as well as the potential for
damage and firefighting costs.

3.3 PROPSOED SYTEM:

Here In Fire Detection, We Are Using The Haar Cascade Classifier, Which Is Very
Popular In Object Detection Through The Image Or Any Other Video Feeds.For
Detection We Use Open Cv Which Is A Library Of Python Bindings Designed To
Solve Computer Vision Problems.We Use Adaboost Algorithm Which Is A
Learning Algorithm Used To Train The Classifier And Selects The Best Subset Of
Features.Cascade Classifier Is Used For The Accuracy Of Identification. It Is
Composed Of Several Stages Consisting Of A Strong Classifier.

ADVANTAGES:

The system can be used in various areas like in parking areas, road traffic congestions,
agriculture, farming, there are many uses of object detection.
In future, this approach can be extended by for other disasters as well.

BLOCK DIAGRAM:

Fig 3.1 Block Diagram

FLOW DIAGRAM:

Fig 3.2 Flow diagram

3.4 SYSTEM

DESIGN Modules:

● Detection
● Haar-like Features
● Integral Image
● Training and Learning
● Adaptive Boosting (Ada Boost)
● Cascading
Detection:

The procedure for detecting fire classifies images based on the value of simple
features from an image. While detecting the fire, first of all, image is converted into
grayscale since it is easy to work with, and it has less data rather than RGB color
images
The algorithm outlines a box and searches for the fire in the image,, the box is
searching for Haar-like features.
Along with small steps, the box detects the features like edges, brightness level
from the image of fire, and then data collected by boxes are put together, which
helps to determine where the fire is located

3.5 Haar-like Features:

Haar features are similar to convolutional kernels, which are used to detect the
presence of those features in the given image.
The grayscale has some features some of the regions are black, some regions are
white, and some region is slightly darker or vice-versa from each other which
helps the machine to understand an image.
He Haar features like edge features, line features, center-surround features
rectangle is moved around the image.
The color white and black is decided according to the region. Which compares the
brightness and extracts features by assigning some values as it goes and inspects
through that region in an image.
⦁ The construction of the haar is like that of a confirmed egg and is utilized to

decide if these properties are displayed in a given picture.

⦁ Dim has a few dark regions and some have white regions, some are hazier or
some are a method for aiding the vehicle feel the picture.

⦁ He Haar has side highlights, line elements, and elements around the focal
point of the square shape.
⦁ White and dark are territorially supported. It represents light and disposes of
the trademark by allocating a particular worth as it goes through the photo and
controlling the region.

Fig 3.3 Haar Features

Formula:

Δ = dull - white = 1 𝑛 ∑ 𝐼 (𝑥) 𝑛 𝑑𝑎𝑟𝑘 - 1 𝑛 𝐼 𝐼 (𝑥)

For positive pictures, Δ: 1 - 0 = 1

For the genuine picture, Δ: 0.67 - 0.15 = 0.52

Integral Image:

The integral image plays an important role. It gives a boost to the calculation,
which makes the approach as fast as possible because there are thousands of
pixels that need to be calculated.
While applying rectangular haar features the sum of pixels in unshaded rectangles
sides are subtracted from the sum of the pixel in the shaded side of rectangles.
Even for a small size images there are lots of features (over 160,000 for a 24x24
image).
 Since due to large number of features the algorithm requires iterating overall
features, the features must be computed efficiently. So to solve this issue integral
image is introduced.
The sum of the pixel in rectangle D can be calculated with reference to four arrays.
The value of the image at location 1 is the sum of pixels in a rectangle.
Respectively value at position 2 is A+B, value at position 20 3 is A+C, and value at
position 4 is A+B+C+D.

Training and Learning:

Extracting features from an image to detection. The machine is now trained to

identify the features from the input. So it is fed with lots of information so that it can
predict objects from the information.
To train the machine, It needs a lot of pictures of fire, different fireplaces,
conditions of fire, also known as a positive image. Concerning that, many negative
images were uploaded.
The negative image refers to pictures in which fire is not included in an image.
Negative images are uploaded so it can differentiate between the two classes,
which helps to an algorithm which features are more likely to be a fire and which
are not expected to be fire.

Adaptive Boosting (Ada Boost):

Adaboost algorithm is a learning algorithm used to train the classifier and selects
the best subset of features. The algorithm learns from the data that is given and
determine the false positive and true negatives in data.
F(x)= a1f1(x) + a2f2(x) + ....
Here F(x) is the strong classifier, and a1f1(x), a2f2(x) are weak classifiers where
a1, a2 are the weights, and f1, f2 are the features.
So the strong classifier is made up of many weak classifiers since one weak
classifier is not good as adding more classifiers makes the algorithm stronger and
accurate, which is called an ensemble.
Cascading:

Cascade classifier is used for the accuracy of identification. It is composed of

several stages consisting of a strong classifier. Those strong classifiers are
passed by So all the features are grouped in several stages where each stage has
a certain number of features.
The use of these several stages is used to determine whether the given input sub
window has features of fire or not if there are no features of fire, then the given
sub window is discarded and fails to go for other stages.
there are some stages stage 1 and stage 2. Usually, the first few stages will
contain very less numbers of features.
If the window fails, it is discarded if not apply the second stage of features and
continue the process. The window or stage which passes all the features of fire,
then it is detected as fire.

3.6 WORKING PRINCIPLE:

The first step is the training of the classifier, as mentioned early. Training for the
highly accurate classifier needs lots of time and processing power, so here we
only used a little number of pictures.
After training the fire cascade classifier, the captured frame from the webcam is
converted into grayscale. The reason for converting frame into grayscale is
because the frame captured by webcam is in RGB color.
Since RGB images have three channels of colors, so if the image is converted into
grayscale, there will be only one channel, either black or white, which is easy to
process.
After the conversion, the fire classifier is used, which will help to find the features
and location of images. Parameters like scale factor and min neighbor are passed.
These factors are an essential factor in detecting fire. A scale factor is used for
creating of scale pyramid because while training the classifier fixed size of the
image is trained, so the scale factor will allow rescaling the size of an input frame
to detect the fire.
Another parameter min 24 neighbor will determine the quality of an image here
for thesis min neighbor factor

3.7 SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

System : Pentium i3 Processor.

Hard Disk : 500 GB.
Monitor : 15’’ LED
Input Devices : Keyboard, Mouse
Ram : 2 GB

SOFTWARE REQUIREMENTS:

Operating system: Windows 7.

Coding Language: python
Tool: Anaconda Navigator
Libraries: OpenCV
 CHAPTER 4
Results and Discussion

Fig 4.1 Result Screen Shot

Since the test was performed during room lighting, the estimations must be
changed by the room light to be effective.

⦁ In case of a fire, a boisterous commotion. The outcomes are likewise

displayed as FIRE.

⦁ Assuming that there is no fire, the outcome won't be NO FIRE

 Chapter 5
Conclusion and Future Work

5.1 Conclusion:

The project aimed to detect fire with a different approach rather than using an
existing system. As technology is getting better and better as to keep it up with the
technology and to minimize the limitations also, the new system has created. By
using image processing technology for detecting the fire, these limitations can be
reduced because in this system camera acts like a human eye, as it detects a fire,
the video is captured, and the image is processed using the software alert user. It
can be used everywhere eg. Hospital, railway station, forest etc. The designed
prototype successfully detects fire. gives the review analysis, designing system,
and algorithm, test, and result.
Currently, we not used the systems like a smoke detector and sprinkler water
discharge systems, but in future it can be included.

5.2 Limitation & Future scope:

Due to safety reasons, the effectiveness of the system cannot be tested in the full
phase, so here for the prototype demo, the lighter is used as the source of fire as it
has the same features and characteristics of the fire. There is some error in
detection when the fire is far away from the camera because of the size of the
flame. The system can be made more accurate and precise by using other
machine learning algorithms like YOLO(You Only Look Once), Tensorflow, Keras
algorithm, neural network like CNN, RCNN, which are more advanced than the
current algorithm.
REFERENCE PAPERS:

[1] N. Dziengel, G. Wittenburg, and J. Schiller, "Towards distributed event

detection in Wireless Sensor Networks," in Adjunct Proc. of 4th
IEEEfACM IntI. Conf. on Distributed Computing in Sensor Systems
(DCOSSa:AZ'08), Santorini Island, Greece, 2020.
[2] S. Jarupadung, "Distributed event detection and semantic event processing,"
in The 6th ACM International Conference on Distributed Event- Based
Systems (DEBS 2020)(Doctoral Symposium), 2020.
[3] Y. Li and L. E. Parker, "Detecting and monitoring time-related abnormal events
using a Wireless Sensor Networks and mobile robot," in Intelligent Robots and
Systems, 2019. IROS 2019. IEEEfRSJ International Conference on. IEEE,
2019.
[4] P. Radivojac, U. Korad, K. M. Sivalingam, and Z. Obradovic, "Learning from
class-imbalanced data in Wireless Sensor Networks," in Vehicular
Technology Conference, 2020. VTC 2020-Fall. 2020 IEEE 58th, vol. 5,IEEE,
2020.
[5] Z.-J. Zhang, J.-S. Fu, H.-P. Chiang, and Y.-M. Huang, "A novel mechanism for
fire detection in subway transportation systems based on Wireless Sensor
Networks," International Journal of Distributed Sensor Networks, vol. 2019,
2019.
[6] Boroujeni, N. S. (2019). Monocular vision system for unmanned aerial
vehicles. doi:10.22215/etd/2013-07242.
[7] Akyurek, S., Yılmaz, M.A., and Taskıran, M. (2012)., “İnsansız Hava
Araçları(Muharebe Alanında ve Terörle
[8] Mücadelede Devrimsel Dönüşüm”, technical report No: 53, Ankara, TR.
[9] Chen, T., Wu, P., & Chiou, Y. (2014). An early fire-detection method based on
image processing. 2004
[10]International Conference on Image Processing (ICIP04), Singapore.
[11] Gao, Y., & Cheng, P. (2019). Forest Fire Smoke Detection Based on Visual
Smoke Root and Diffusion
[12] Model.Fire Technology. doi:10.1007/s10694-019-00831-x.
[13] Gao, Y., & Cheng, P. (2019). Forest Fire Smoke Detection Based on Visual
 Smoke Root and Diffusion
[14] Model.Fire Technology. doi:10.1007/s10694-019-00831-x.
[15]Wang, T., Liu, Y., & Xie, Z. (2011). Flutter Analysis Based Video
Smoke Detection. Journal of Electronics
[16]& Information Technology,33(5),
1024-1029.
doi:10.3724/sp.j.1146.2010.00912.
[17]Lin et. al, Smoke detection in video sequences based on dynamic texture
using volume local binary
[18]patterns. (2017). KSII Transactions on Internet and Information Systems, (11).
doi:10.3837/tiis.2017.11.019.
[19]Lin et. al, Wang, Z., Wang, Z., Zhang, H., & Guo, X. (2017). A Novel Fire
Detection Approach Based on
[20]CNN-SVM Using Tensorflow. Intelligent Computing Methodologies Lecture
Notes in Computer Science,682-693.
[21]Shaqura, M., & Shamma, J. S. (2017). An Automated Quadcopter CAD based
Design and Modeling
[22]Platform using Solidworks API and Smart Dynamic Assembly. Proceedings of
the 14th International
[23]Conference on Informatics in Control, Automation and Robotics.
doi:10.5220/0006438601220131.
[24] O. T. Cetinkaya, S. Sandal, E. Bostancı, M. S. Güzel, M. Osmanoğlu and N.
Kanwal, "A Fuzzy Rule Based Visual Human Tracking System for Drones,"
2019 4th International Conference on Computer Science and Engineering
(UBMK),
[25] Samsun, Turkey, 2019, pp. 1-6, doi: 10.1109/UBMK.2019.8907104.
 APPENDIX

A.SAMPLE CODING

import cv2
import numpy as np
import math
import pygame
pygame.init()

pygame.mixer.music.load("alarm.wav")
windowName = "OpenCV Video Player"
cv2.namedWindow(windowName)

cap = cv2.VideoCapture(0)

ret, lframe = cap.read()

Ll, A, B =
cv2.split(lframe)
SI = np.zeros(np.shape(Ll))
print(SI)
NOl = 0
NOc = 0
CGO = 0
x=1
dt = 0
nop = 0
count = 0
mc = 0
while (ret):
ret, frame = cap.read()
if ret:
Ll, A, B = cv2.split(lframe)
ret, frame = cap.read()
cframe = frame
cv2.imshow("FIRE", frame)
b, g, r = cv2.split(frame)
rt = 230
gb = cv2.compare(g, b, cv2.CMP_GT)
rg = cv2.compare(r, g, cv2.CMP_GT)
rrt = cv2.compare(r, rt, cv2.CMP_GT)
rgb = cv2.bitwise_and(rg, gb)
im = cv2.bitwise_and(rgb, rrt)
# cv2.imshow("RGB",im)

t=5
p=1
k = cv2.getStructuringElement(cv2.MORPH_CROSS, ((t, t,)))
# print(k)
dil = cv2.dilate(im, k, iterations=p)
er = cv2.erode(im, k, iterations=p)
fin = cv2.bitwise_and(er, dil)
# cv2.imshow("FIRE",frame)
# cv2.imshow("segmented fire",cv2.bitwise_and(frame,frame,mask=fin))

img_ycrcb = cv2.cvtColor(frame, cv2.COLOR_BGR2YCR_CB)

Y, Cr, Cb = cv2.split(img_ycrcb)
Ym = np.mean(Y)
Crm = np.mean(Cr)
Cbm = np.mean(Cb)

I1 = cv2.compare(Y, Ym, cv2.CMP_GT)

I2 = cv2.compare(Cb, Cbm, cv2.CMP_LT)
I3 = cv2.compare(Cr, Crm, cv2.CMP_GT)
I12 = cv2.bitwise_and(I1, I2)
I23 = cv2.bitwise_and(I2, I3)
I123 = cv2.bitwise_and(I12, I23)
# cv2.imshow("I123",I123)
cbcrdiff = cv2.absdiff(Cb, Cr)
asd = cv2.compare(cbcrdiff, 40, cv2.CMP_GT)
# cv2.imshow("asd", asd)

img_cie = cv2.cvtColor(frame, cv2.COLOR_BGR2Lab)

L, a, b = cv2.split(img_cie)
Lm = np.mean(L)
am = np.mean(a)
bm = np.mean(b)
R1 = cv2.compare(L, Lm, cv2.CMP_GT)
R2 = cv2.compare(a, am, cv2.CMP_GT)
R3 = cv2.compare(b, bm, cv2.CMP_GT)
R4 = cv2.compare(b, a, cv2.CMP_GT)
R12 = cv2.bitwise_and(R1, R2)
R34 = cv2.bitwise_and(R3, R4)
R14 = cv2.bitwise_and(R1, R4)
# cv2.imshow("R14",R14)

kl = cv2.getStructuringElement(cv2.MORPH_CROSS, ((t, t,)))

e_R14 = cv2.erode(R14, kl, iterations=p)
d_R14 = cv2.dilate(R14, kl, iterations=p)
bin_cie = cv2.bitwise_and(e_R14, d_R14)
# cv2.imshow("cie_res",cv2.bitwise_and(frame,frame,mask=bin_cie))

im = cv2.bitwise_and(bin_cie, im)
# cv2.imshow("RGBCIE",im)

cframe = cv2.cvtColor(frame, cv2.COLOR_BGR2Lab)

Lc, A, B = cv2.split(cframe)

DIfd = cv2.absdiff(Lc, Ll)

# cv2.imshow("difd",DIfd)
u_DIfd = np.mean(DIfd)
sd_DIfd = np.std(DIfd)

if u_DIfd + sd_DIfd >= 10:

Tfd = u_DIfd + sd_DIfd
else:
Tfd = 10

_, FD = cv2.threshold(DIfd, Tfd, 255, cv2.THRESH_BINARY)

# cv2.imshow("difd2",FD)
# print(FD)

cntr = cv2.compare(FD, 0, cv2.CMP_EQ) / 255

SI = np.add(cntr, SI)
# print(SI)
MPM = cv2.bitwise_and(FD, im)
MPM = cv2.bitwise_and(MPM, asd)
MPM = cv2.erode(MPM, kl, iterations=p)
# cv2.imshow("MPM", MPM)
CF = cv2.bitwise_and(im, MPM)
# cv2.imshow("candi", CF)

retval, labels, stats, centroids = cv2.connectedComponentsWithStats(CF)

for i in range(retval):
if i > 0:
if stats[i][4] > 2:
NOc = NOc + stats[i][4]
# print(NOc)
if NOc > NOl:
CGO = CGO + 1
NOl = NOc
NOc = 0
fps = 24
if x % fps == 0:
 dt = CGO * 1.0 / fps
print(dt)
CGO = 0
x=1
x=x+1
# print(dt)
lframe = cframe
if dt > 0.3:
count = count + 1
if count > 10:
mc = mc + 1
if mc % 15 == 1:
print("fire")
try:
pygame.mixer.music.play()
except:
pass
count = 0
if dt < 0.2:
print("not fire")
k = cv2.waitKey(1)
if k == ord('q'):
break

cv2.destroyAllWindows()
cap.release()
B.PLAGARISM REPORT