In recent days, the major threat in the world is forest fire that affects the biodiversity, climate change, and so forth. So detection process is more essential to monitor the forest region. To detect the forest fire, the paper proposes a novel detection technique of support vector machine (SVM)-Krill herd that can effectively detect the fire region using different kinds of features. Land surface temperature, fire intensity, water vapor, and top of atmosphere temperature are being extracted as some of the features that can be exposed. These features are preferred to classify the LANDSAT image into two classes using SVM optimized by Krill herd. The Euclidean distance is chosen to find the similarity between the test and trained image and then predict the giving query image containing a fire or not based on its training samples. With the help of the feature extractor parameters, the performances have to be analyzed. When compared with existing fire detection algorithms like active fire detection, SFIDE, convolutional neural network (CNN), hybrid intelligent, and PSO-SVM algorithm, the proposed SVM-Krill herd-based detection method increases the accuracy by 1.562%, 0.675%, 1.290%, 0.876%, and 1.038%. The proposed detection method of SVM-Krill herd achieves 99.21% accuracy and high precision as 98.41%.

No abstract