Nomadic people optimisation based Bi-LSTM for detection and tracking of tropical cyclone

The early detection of track and intensity of cyclones reduces the impacts of destruction, and hence this paper proposes a model to determine the intensity of tropical cyclones from satellite images. The proposed system employs radial features and angular features to depict the characteristics of the tropical cyclone. Here the convolutional neural network (CNN) is employed in investigating the similarities between the actual recorded image and the image of the target cyclone, thereby extracting the optimal features of a tropical cyclone. The cyclone image considered as a target is the query image and bi-directional long short-term memory-based nomadic people optimisation (Bi-LSTM based NPO) algorithm predicts the intensity of the tropical cyclone. The prediction accuracy is enhanced by employing the proposed model. This paper utilised tropical cyclones for an image to intensity regression dataset (TCIR dataset). The experimental analysis is conducted to deliberate the estimation of tropical cyclone intensity. Finally, the comparative analysis is carried out for the proposed approach and other existing approaches to determine the error value rate of the proposed model.