Detection Of Fire Using Advance Algorithms

Abstract

Detection of forest fire should be quick and
accurate as forests are the important sources to lead a
vital life on earth. Detection of fire can be extremely
difficult using existing methods of smoke sensors
installed and they are slow and cost inefficient, so in
order to avoid large scale fires, detection from visual
scenes is required. In this work detection of fire in an
image is done by extracting features using Deep
learning algorithm and with those features as input to
machine learning algorithm, a model is build with the
help of different machine learning algorithms like
Random Forest, Support Vector Machine, XGBoost
and K-Means Clustering. Using these algorithms the
data sets are classified into fire and non fire images to
build the model and the test data of the data set is
provided as input for getting the validation accuracy of
the model. Then comparison is done among machine
learning algorithms to find which algorithm provides
more accuracy. To test the accuracy of the fire
presence classification evaluation metrics are used in
the model and find that accuracy of CNN-RF and
CNN-XGBOOST are 98.53% which is greater than
accuracy of CNN-SVM 97.06%.