INTEGRATION OF LOGICAL FEATURES WITH NEURAL NETWORKS FOR CONTROLLED VS UNCONTROLLED FIRE CLASSIFICATION: A COMPARATIVE STUDY

Abstract

This paper presents an analysis of the performance of a convolutional neural network (CNN) for the classification of controlled and uncontrolled fires. The study focuses on the incorporation of custom features such as standard deviation, spikes, fall, vertical intensity arrays (VIA), and arc length to improve the accuracy of the model. These features were individually concatenated with the features selected by the neural network to test the cumulative performance. The paper also puts forth the comparison between a logical (decision tree) classifier and a black box (neural net) classifier and the corresponding performance analysis.