A Spark-based parallel framework for geospatial raster data processing

With the rapid development of geospatial sensor technologies in recent years, the acquisition of higher resolution of spatial and temporal data is becoming common, resulting in an explosive increase in the volume of geospatial data. Meanwhile, the sophistication and complexity of geospatial algorithms is also increasing. The current geospatial applications often have limited capabilities in dealing with the trend. In this paper, we proposed a Spark-based parallel framework for geospatial raster data processing. The framework provides a parallel processing way for both a single high-resolution remote sensing image with large size using a RsImage model and multiple remote sensing images with small size using a RsBundleImage model. Taking the edge extraction from remote sensing images as an example, the paper demonstrates how the framework can improve the efficiency of the Canny edge detector by achieving the ratio of speedup up to 7.0 similar to 8.0.