Research Of Fast Extraction Algorithm Of Enhanced Vegetation Index Based On Nested Block And Pipeline

With the gradual development and maturity of our country's remote sensing technology,We get more and more rich in the types of remote sensing images,data scale is growing.Faced with such a huge amount of data,the traditional image processing model has been unable to meet the requirement of efficient processing of image processing algorithms in remote sensing applications.Therefore,it is very meaningful to study the fast processing model of remote sensing image.Enhanced Vegetation Index is one of the most widely used vegetation indices in the field of ecological quantitative remote sensing monitoring.It improves the EVI formula by introducing the relevant parameters,Weaken the impact of the atmosphere and soil on it to a certain extent.So it can stably reflect the growth of vegetation in the observed area.When EVI extract object is the large-scale remote sensing image,due to the limitation of computer hardware resources,such image data can not be loaded into memory for processing,Such remote sensing data need to be reasonably divided.As the traditional block algorithm can not quickly make a reasonable block of remote sensing data,and the block efficiency is low.The traditional EVI multithreading extracted method can not accelerate the whole processing chain of EVI extracted,so the extraction efficiency is not high.Combined with the above questions,this thesis studies the process of remote sensing image EVI extraction,The parallelism analysis was performed for each step in the extraction process: for the data block part,implement the parallel blocking based on OpenMP;for the data calculation part,implement parallel computing based on CUDA parallel library;for the entire extraction chain,design the EVI fast extraction algorithm based on Pipeline thought.And through the experimental comparison analysis,The main achievements of this thesis are as follows:(1)A nested parallel block mechanism is designed based on OpenMP,aiming at the problem of low efficiency of existing image block algorithm,improving the efficiency of the block by using the OpenMP nested parallel technology and give full play to the advantages of CPU multi-core.Aiming at the problem of unable to quickly scientifically determine the block number,a calculating formula of optimal block number is designed for EVI extraction algorithm,the scale parameters of the calculation task and the performance parameters of the calculation resource are substituted into the formula,and the optimalnumber of blocks can be obtained.The parallel block mechanism not only be used in the EVI extraction algorithm,but also be applied to other areas related to image block.(2)A fast EVI extraction algorithm of remote sensing image is designed based on pipeline,for the five successive processing steps of the EVI extraction process,rely on the advantages of pipeline in multi-level continuous processing of remote sensing images,The EVI extraction process is designed in pipeline parallel.In order to reduce the time delay caused by the different speed between the pipeline nodes,a double buffer queue is designed between the algorithm processing steps to store the intermediate result,and further improve the parallelism of the steps of EVI extraction process.