CSC-RS: Leveraging cloud-native serverless computing for large-scale remote sensing data processing

The rapid growth in the volume of remote sensing (RS) image data necessitates efficient computational methods to handle large-scale datasets. Traditional RS image processing methods, although effective for small datasets, face significant challenges when scaled up. In this research, we present a novel cloud-native serverless computing framework for RS data processing (CSC-RS), which integrates traditional parallel techniques with modern serverless computing technology. We demonstrate our 5 design principles and the architecture of CSC-RS, then introduce the three-level parallel structure for RS processing acceleration. As for implementation, CSC-RS has a high-performance RS processing core optimized with vectorization and shared memory parallelization techniques. By encapsulating these core functions within a cloud-based serverless wrapper, the framework flies onto the cloud, achieving higher scalability, efficiency, and resource utilization. We validated CSC-RS through a case study of extracting potential sandstorm sources in the middle reaches of the Yarlung Zangbo River over the past 30 years. Results demonstrate significant improvements in processing speed and scalability. CSC-RS was 8.04 times faster than the original unoptimized processing program in this case study. These improvements not only enhance the technical capabilities of remote sensing applications but also provide valuable insights for local sandstorm management and ecological protection. • Leverage a cloud-native serverless computing approach. • Introduce a three-level hybrid parallel structure for acceleration. • Applicable and efficient for complex and large-scale remote sensing processing. • Validated in a real-world case study and support ecological protection.