Lightweight GWO-LSTM-Based Land Cover Classification using Remote Sensing Images

The classification of land use and land cover (LULC) using remote sensing data is essential for many environmental models and land-use inventories. A lightweight deep learning classifier is implemented in this research to improve the performance of LULC classification, assisting in the prediction of declining environmental quality, haphazard elements, wildlife habitat, and so on. LULC classification is evaluated using Eurosat dataset and it uses algorithms like the Haralick texture features, histogram of the oriented gradient, and local Gabor binary pattern histogram sequence for feature extraction. The Grey Wolf Optimization (GWO) method is applied to select the best features, offering fast convergence tares and ease of implementation. A Long Short-Term Memory (LSTM) network is then used to categorize the LULC. The research outcomes show that the GWO method with an LSTM classifier efficiently differentiates the classification of LULC in terms of 99.80% accuracy, 98.99% precision, and 99.53% recall when compared to the Deep SHAP, HFEL-CCGSA, Human group-based PSO with LSTM and IMO-mLSTM.