Research On Parallel Mosaicking For Massive Remote Sensing Images Based On Spark

With the continuous development of earth observation technology,the analysis of massive remote sensing images has become the focus of military,meteorology,and transportation in recent years.As an important part of remote sensing image processing,image mosaicking has played an important role in the analysis of cross-regional remote sensing images.In order to solve the problems of low utilization rates of the nodes and frequent data I/O in the traditional parallel algorithms of remote sensing images,we propose a parallel mosaicking algorithm based on self-defined RDD(Resilient Distributed Datasets),in which the Spark distributed memory computing framework has been used.In this paper,we take full advantage of the Spark,which is conducive to the processing of iterative data,and build remote sensing images parallel mosaicking processing model through the operation of the Spark RDD.The entire mosaicking process is completed by the self-defined RDD,which is aimed at the remote sensing image mosaicking processing,so as to improve the efficiency of image parallel mosaicking.The main works of our research lie in:(1)The implementation of the self-defined RDD in Spark cluster.We override the compute and getPartitions methods in RDD and self-define the RDD for remote sensing image processing.In the self-defined RDD,we implement three operators of image overlapping region estimation.image registration and image fusion.In the implementation of overlapping region estimation operator,we face the problem of low processing efficiency caused by the serially performed image overlapping region estimation operation in the traditional remote sensing image parallel mosaicking algorithm.According to the logical separability and data independence of computational steps such as the Fourier transform and inverse Fourier transform in the phase correlation method for overlapping region estimation,we improve the traditional phase correlation method through the parallel execution of multiple nodes in a single instruction to achieve parallel processing on the multiple nodes of the cluster.In the implementation of image registration,we refine the processing steps of the SIFT algorithm,decouple the relationship between images in the processing of the algorithm,and achieve the parallel processing through the task parallelism.In the implementation of image fusion.according to the characteristics of each processing step in Poisson fusion.we perform fusion calculation of multiple images at the same time in the cluster,so as to achieve parallel processing of image fusion operations.(2)Research on parallel mosaicking for remote sensing images based on the self-defined RDD.We use the three key steps of the image mosaicking,including overlapping region estimation,image registration and image fusion.which are the Transformation-type operators of the self-defined RDD.And we divide the RDD generated by the calling operator according to the number of remote sensing images in each partition.Finally,the parallel processing of image mosaicking is realized by calling the operators of self-defined RDD with the method of implicit conversion.The experimental results show that the parallel mosaicking algorithm of massive remote sensing image based on self-defined RDD has good stability and scalability,and compared with the parallel mosaicking algorithm based on MPI,the algorithm can effectively improve the image mosaicking efficiency of massive remote sensing images on the basis of guaranteeing the image mosaicking effects.