The classification of land use and land cover (LULC) is an elementary remote sensing activity that is important for sustainable planning for urbanization, resource management, and also for environmental monitoring. In this study we have presented the architecture based on the Multi-Kernel Atrous Spatial Pyramid Pooling (MKASPP) technique, which will help us to address problems of complicated boundary delineation and spatial variability from high-resolution satellite imagery to enhance segmentation accuracy and feature extraction as well. By integrating spatial pyramid pooling and multi-scale Atrous convolutions, this MKASSP module has achieved both large-scale and detailed-level information. The suggested framework was evaluated by using high-resolution datasets for segmenting a few significant classes of water bodies and vegetation. The results from comprehensive testing have shown that MKASSP has superior performance compared to other models like FastFCN and FCN-8, by achieving an F1 score of 93% and 96% along with a 96% Intersection over Union (IoU) & 94%, respectively. Even visual analysis has shown that the proposed method has performed well at demarcating complicated boundaries, mostly in diverse urban landscapes. This study has demonstrated MKASPP’s reliability in LULC classification due to its computational efficiency and ability to work across different geographic regions. This implies that the modern segmentation process can show significant potential for enhancement of remote sensing applications in dynamic environments.