Machine Learning-Based Classification of Geographical Landscape for Precision Cartographic Representation

Abstract

The classification of aerial imagery through automation plays an essential role in areas like environmental monitoring, urban planning, and disaster response. This study introduces a method utilizing a convolutional neural network (CNN) to classify aerial landscape images, leveraging the SkyView Aerial Landscape Dataset. The suggested pipeline includes streamlined preprocessing, immediate data enhancement, and a simple but powerful CNN architecture crafted with the TensorFlow-Keras framework. The model underwent training and validation using an 80/20 split of the dataset, reaching a training accuracy of about 98% and a validation accuracy of 91%, which suggests a robust ability to generalize. Data augmentation techniques, such as random flipping, rotation, zoom, and contrast adjustments, greatly improved model robustness. The confusion matrix analysis showcases the model’s overall dependability while uncovering slight difficulties in distinguishing between visually similar classes.