Research On Parallelization Of Dynamic K-means Algorithm In Remote Sensing Image Mining

With the development of science and technology,especially Computer Technology Aerospace and Sensing Technology advances.Remote Sensing Technology emerged at that time,and has become one of the most important way to manage and use territory resource.Remote sensing technology has accurate,rapid,and short characters,which has become the most important source of land and resources observation data.The dynamic monitoring of land resources has become the most important field of remote sensing data application,and it is also a hot spot in scientific research field.As the modern remote sensing technology can access and gather large number of remote sensing image information more quickly and conveniently,the traditional manual monitoring cannot meet the current technical requirements,With the development of computer technology,a large number of remote sensing image classification technology has emerged,which becomes the main processing method of remote sensing image data.This article focus on researching on parallelization of dynamic k-means algorithm in remote sensing image mining,which use the BP Neural Network Algorithm to clarify the satellite remote sensing image by dynamic splitting and merging in every iteration process to determine the clustering result.and improved algorithms proposed are processed in parallel on the Hadoop cloud compute platform.then,use a method called band integration mask to do the change detection.The content and innovation in this paper are as follows:(1)To resolve the constraints in initialization process of the traditional K-means algorithm,the final clustering center is determined dynamically by splitting and merging steps in each iteration.The results were verified by the test section.(2)Combined with BP neural network algorithm to optimize the weight of splitting and merging algorithm to determine division rules for each iteration process.(3)This paper introduces the feasibility and idea of parallel algorithm,and improved algorithms proposed are processed in parallel on the Hadoop cloud computing platform,which makes the algorithm get the ability to dealing with the huge data.Finally,the reliability and efficiency of the algorithm are verified by research.The parallelization experiment on Hadoop platform proves that the parallel algorithm can definitely improve the efficiency of remote sensing data processing.