Automated Remote Sensing Image Classification using Enhanced Fireworks Algorithm with Deep Learning Model

Remote Sensing Image (RSI) classification had a crucial role in different applications, like environmental monitoring, land cover mapping, and disaster management. Conventional classification methods frequently depend on manual parameter tuning and handcrafted features that can be less effective and time-consuming in capturing complicated patterns in RS data. Currently, DL methods have shown incredible performance in image classification tasks. Still, the challenges lie in optimizing the deep learning (DL) method and choosing the optimal hyper parameters for attaining the optimal classification accuracy. In this article, we devise an Automated Remote Sensing Image Classification using Enhanced Fireworks Algorithm with Deep Learning (ARSIC-EFWADL) model. In the presented ARSIC-EFWADL method, a four-stage process is followed. Initially, the median filtering (MF) approach is used to preprocess the images. Secondly, the ARSIC-EFWADL technique uses a densely connected networks (Densenet) model for feature extraction. Next, the EFWA is used for the parameter tuning process. At last, Adaptive Neuro Fuzzy Inference System (ANFIS) methodology is implemented for classification purposes. To exhibit the improvised classification results of the ARSIC-EFWADL method, wide-ranging simulations were involved. The outcomes of the experiments inferred the enhanced performance of the ARSIC-EFWADL method over compared approaches under several evaluation metrics.