Research And Implementation Of Remote Sensing Thematic Information Parallel Extraction System Based On OpenMP/OpenAcc

At present,China's space remote sensing technology is in a period of rapid development.Because of the large amount of time overhead and computer memory consumption required for massive remote sensing big data processing,the processing speed of remote sensing thematic information extraction algorithm can not meet the increasing demand for remote sensing special products.Remote sensing image parallel processing based on large-scale clustering has become the mainstream processing mode.However,how to mine the computing power inside a single node in cluster computing and how to efficiently utilize multi-GPU hardware devices has become an urgent problem in the calculation of remote sensing thematic information.At the same time,the remote sensing thematic information processing technology has also developed rapidly.From the early development of the system development technology to the remote sensing information extraction system developed by ENVI,to the remote sensing special product production service system based on cluster parallel.And the dynamic real-time remote sensing thematic application system customization platform has witnessed the rapid development of the remote sensing thematic information processing system.However,how to further improve the timeliness of remote sensing thematic information processing system and how to deal with the huge increase in the demand for remote sensing services by Internet users is still the bottleneck restricting the universalization of remote sensing applications.In view of the above mentioned problems,the main work carried out in this paper is as follows:(1)The parallelization process design of remote sensing thematic information extraction algorithm based on vegetation index is realized.The NDVI,RVI,DVI and EVI are optimized and improved.Based on the serial algorithm,the parallel specification process is proposed,and the correctness of the parallel algorithm is verified by numerical experiments.The existing parallel techniques and strategies are summarized,and a special feature of the remote sensing thematic vegetation index extraction algorithm is proposed.It is a practical and efficient remote sensing algorithm acceleration strategy,which is a hybrid multi-GPU parallel strategy of OpenMP/OpenAcc.At the same time,it analyzes the advantages and disadvantages of parallel strategy in OpenCV and GDAL remote sensing image processing.(2)The architecture of remote sensing thematic information parallel extraction system based on OpenMP/OpenAcc(hereinafter referred to as: remote sensing information parallel extraction system)is analyzed and designed.According to the requirements analysis,the system architecture is divided into four modules: remote sensing school,remote sensing information parallel extraction,visual display,and server script.The system design is composed of administrator users and ordinary users.The database entity relationship is determined according to the remote sensing information extraction service.The function designed by the remote sensing information parallel extraction system is realized,and the parallelization strategy is applied to the system.The system adopts distributed multi-service mode,based on the Spring Boot framework integration development,deploys high-performance computing services,file storage services and application development services on different server nodes in a micro-service manner,in order to more intuitively display hybrid OpenMP/ The effectiveness of the OpenAcc multi-GPU parallel strategy,the system shows the acceleration effect of the vegetation index extraction algorithm in the form of two-dimensional visualization.The experimental results show that the proposed hybrid OpenMP/OpenAcc multi-GPU parallel extraction of remote sensing thematic information has a significant acceleration effect compared with the traditional serial extraction algorithm,and the code instructions are more standard and standardized,which is more suitable for popularization.After many times of functional testing of remote sensing information parallel extraction system,the remote sensing information parallel extraction system designed in this paper has good stability and good user experience,and further validates the effectiveness of the proposed OpenMP/OpenAcc hybrid parallel strategy.