Deep Learning-Based Land Use and Land Cover Classification for Change Detection Studies

Abstract Accurate land use land cover (LULC) classification is crucial for monitoring environmental changes, urban planning, and natural resource management. Traditional methods and conventional machine learning algorithms can be time-consuming and less effective in handling large datasets. Hence, this paper presents the deep learning-based classification method, to classify Linear Imaging Self-Scanning Sensor-III ((LISS-III) satellite imagery into various land cover categories with improved accuracy and efficiency. We developed a Convolutional Neural Networks (CNN)-based deep learning model and trained on a diverse dataset of LISS-III images, covering the geographical region of Nanjangud taluk, Myusuru district, India. The developed model is applied to LISS-III images of the two years 2010 and 2020 respectively for LULC classification. Further, classified maps are used for change detection studies to identify and quantify land cover changes over time. The results indicate that the CNN approach achieved an overall accuracy of 94.08% for the 2010 data and 95.30% for the 2020 data, demonstrating the model's robustness and precision in LULC classification. The results highlight significant improvements in detecting small and large-scale changes, thereby providing valuable insights for environmental monitoring and policy-making.