Automated Forest Fire Detection using Atom Search Optimizer with Deep Transfer Learning Model

Abstract

Automated Forest Fire Detection (AFFD) contains the technology used to recognize and alert authorities on latent wildfires in a forested region. AFFD methods are latent to enhance response times and decrease the damage led by wildfires. But, these systems are utilized in conjunction with typical fire management practices like fire prevention and suppression measures, to provide the best achievable outcomes. There are several algorithms to AFFD, comprising computer vision (CV), remote sensing, and machine learning (ML). This article develops an Automated Forest Fire Detection using Atom Search Optimizer with Deep Transfer Learning (AFFD-ASODTL) model. The goal of the AFFD-ASODTL technique lies in the effectual recognition of forest fires accurately and promptly. In the presented AFFD-ASODTL technique, residual network (ResNet50) model is applied for feature vector generation. Besides, the ASO technique is exploited for the optimal hyperparameter tuning of the ResNet model. Meanwhile, Quasi-Recurrent Neural Network (QRNN) model is used for forest fire classification. To exhibit the optimum resultant of the AFFD-AS ODTL system, a comprehensive set of simulations is carried out. The comparative study highlighted the improvised results of the AFFD-ASODTL method over other models.