The Research On Parrel Inversion Of Forest Biomass Based On Spark

With the global warming,forest carbon storage has become a very important issue.Forest biomass is an important parameter to estimate forest carbon reserve,which is not only a sign of carbon cycle,but also an important index to evaluate forest ecosystem.Forest canopy density is also one of the main parameters in the investigation of forest ecological resources and an important factor in the carbon cycle,which is closely related to biomass.The inversion of biomass can be realized more accurately if the canopy density is obtained effectively through the inversion of remote sensing image.However,with the development of remote sensing big data,the complexity and data volume of remote sensing image processing are also increasing.Therefore,in order to achieve efficient processing of remote sensing images,this paper applies Spark parallel computing framework to the parallel processing of remote sensing images.Taking Laoshan application area of Maoershan as the data source,this paper proposes a parallel inversion algorithm based on Spark,and integrates the biomass inversion model based on canopy density with it to complete the parallel inversion of biomass.The main research contents of this paper are as follows:(1)In view of the large amount of remote sensing image data,repeated reading andstorage of data in the parallel inversion process will cause low timeliness.In this paper,a remote sensing image pyramid is built after remote sensing image preprocessing.In parallel inversion,information can be extracted directly from the pyramid model.At the same time,according to the data types in this paper,a storage method of divide-and-conquer based on different databases is proposed for remote sensing image data and image tiles.The processing process and results of the original remote sensing image and parallel inversion are stored in the distributed file system HDFS.Image pyramid tile pyramid model generation and build process using HBase column type distributed database to store data.Applications from the bottom storage to the top of the parallel computing framework are implemented on the cloud platform,which improves the efficiency of data storage and reading.(2)Aiming at the low efficiency of remote sensing image inversion,this paper proposes a parallel inversion algorithm based on Spark custom RDD.The three key steps of remote sensing image fusion,relevant factor extraction and inversion model integration in the inversion process are taken as operators in the custom RDD to design and complete the transformation of different types of RDD in the custom RDD.Parallel inversion is accomplished by calling custom RDD during program execution.(3)Forest canopy density controls the physical processes of forest such as photosynthesis,respiration and circulation,and has a strong correlation with biomass.To solve the problem of low precision of biomass inversion,a biomass inversion model based on canopy density is proposed in this paper,and the parallel inversion algorithm based on Spark custom RDD is integrated to complete the parallel inversion of biomass.Comparative experiments were conducted on the single-machine serial model,the MapReduce model and the Spark standalone model with the Spark on yarn model in the Spark parallel computing framework.The experimental results show that the proposed parallel inversion model based on canopy density has high precision,and the parallel inversion algorithm based on Spark has achieved the expected results,and the inversion efficiency is high.With the increase of computing nodes,the efficiency of parallel inversion is also improving.On the premise of consistent and accurate inversion results,the parallel inversion algorithm can accurately and efficiently complete the parallel inversion of forest biomass.