Optimizing Wildfire Detection in India: A Convolutional Neural Network Approach Utilizing VIIRS Data

Wildfire is a major natural hazard and an essential factor in an environmental change and unpredictable in many countries of the earth. Wildfire is inextricably linked with environmental impacts. Wildfires detected from coarse spatial resolution sensors like Visible Infrared Imaging Radiometer Suite (VIIRS) and Moderate Resolution Imaging Spectroradiometer (MODIS) provides an accurate and cost effective solution in detecting and monitoring the wildfire areas. VIIRS sensor data makes wildfire observations that are about three times greater than MODIS. The traditional contextual fire detection method using VIIRS data provides less accuracy since it mainly depends on threshold value for classifying the pixels as fire pixel or not. In this paper, wildfire detection method using convolutional neural network method (Wildfire CNN) is proposed for effective wildfire detection of VIIRS data. The proposed Wildfire CNN method is tested with the dataset which contains fire and non-fire spots and the performance metrics such as recall rate, precision rate, omission error, commission error, F-measure and accuracy rate are considered for the analysis of the model. The proposed Wildfire CNN method experiment shows higher accuracy of 99.3% and lower omission error of 0.512% when compared to traditional contextual wildfire detection method. The proposed Wildfire CNN method also proves to detect small fires in Alaska forest dataset when compared to other machine learning models.