An Enhanced Remote Sensing Image Classification for LC/LU using Deep Convolutional Neural Networks

Urban planning, agricultural decision-making, and environmental monitoring all rely heavily on remote sensing. However, analysing remote sensing images presents several challenges, such as variations in sensor technology, weather, and seasons. To address these challenges, remote sensing communities have high hopes for deep learning-based classification systems for land cover and land use (LCLU). An advanced Convolutional Neural Network (CNN) Inception V3 model is suggested in a recent research study by utilizing block vector data and remote prints to estimate the position of damage to groups of structures impost-earthquake remote monitoring imagery. The issue of challenging feature selection is resolved by the suggested method, which automatically chooses the optimal features using CNN. Clustering can be hindered by block boundaries, which can substitute image segmentation and prevent the emergence of fragmented and unsatisfactory outcomes. By incorporating distinct & and merged layers to handle vast remote sensing images, the enhanced CNN architecture is adept at its task. The UCM dataset is employed for the models training. Classification algorithms for Land Cover and Land Use by using images of remote sensing and deep learning play a crucial role in monitoring the environment, decision-making in urban and rural planning. Remote sensing imagery taken after the earthquake can be used to estimate the extent of damage to building clusters using CNN’s best proposed architecture. Researchers evaluated and compared the effectiveness of various deep learning models based on performance endpoints such as f1 score, accuracy, precision, confusion matrix and memory usage. According to the outcomes, the enhanced technique was proficient in extracting the extent of damage to building clusters in every building from before and after earthquake remote sensing images.