Deep neural network for crop classification using multitemporal data: A case study of Sialkot, Pakistan

Abstract

Agriculture is a fundamental sector in Pakistan’s economy, providing employment to a large portion of the population and significantly contributing to national income. However, traditional methods of crop yield estimation, such as crop-cutting surveys, are outdated, labor-intensive, and often result in inaccuracies. These methods also tend to be time-consuming, which can hinder timely decision-making in agricultural planning and resource management. The advent of freely available spatial data, particularly from remote sensing technologies, coupled with artificial intelligence, presents an opportunity to transform the way crop yields are monitored and predicted. This study investigates the integration of Geographic Information Systems (GIS) and remote sensing imagery with deep learning techniques specifically Convolutional Neural Networks (CNN) and Satellite UNET, to enhance the accuracy and efficiency of crop type classification and yield estimation. By leveraging artificial intelligence, the proposed method not only automates the process but also improves the precision of yield predictions compared to traditional approaches. Initial findings suggest that the application of deep learning models to remotely sensed data allows for real-time monitoring, enabling quicker and more informed decisions regarding food production and resource management. The primary objective of this study is to develop a reliable, scalable tool for crop yield estimation in Pakistan, facilitating timely responses to potential food shortages and contributing to the effective management of food security. This innovative approach could revolutionize agricultural practices in Pakistan, offering a modern solution to address both current and future challenges in the sector.