Integration of Machine Learning Algorithms for Land Classification Using Satellite Images

Lands might be classified using satellite images, which would support the delivery of remote goods and services and lessen the need for time-consuming manual entry of these records. Recently, the subject of land classification has undergone a revolution attributable to integrating sophisticated machine learning algorithms with satellite imagery. For semantic segmentation and classification tasks, the UNet Convolutional Neural Network (CNN) architecture has shown to be one of these algorithms’ most promising tools. The research offers a thorough approach for incorporating UNet CNN into satellite image-based land classification. This approach addresses the pressing need for accurate and automated land cover classification in various applications, including environmental monitoring, urban planning and autonomous agriculture.

The proposed methodology encompasses a series of steps to enhance the accuracy and quality of land classification results. The UNet model, with its unique u-shaped arrangement of convolutional layers, is adept at capturing detailed spatial information while maintaining global context. The method is implemented using Python software. Combining segmentation capabilities with classification layers, the model accurately delineates land types and assigns semantic labels, providing a comprehensive understanding of land cover composition within satellite imagery. The proposed Unit CNN shows a better dice coefficient value of 0.89 and an accuracy of 98.7%, which is 14.26% higher when compared with RF, SVM ANN and Faster R-CNN. Overall, the integration of UNet CNN algorithms emerges as a powerful solution for land classification, enabling precise identification and classification of various land cover types for diverse geographical applications. This research contributes to advancements in remote sensing technology, with implications for sustainable land management and environmental monitoring.