Сидорова В.С.  

Choice Dimension and Detailed of Remote Sensing Data for Histogram Hierarchical Clustering

Earth remote sensing data (RSD) are characterized by high volume, large dimension, complexity and the absence of a priori information often. Therefore, clustering is actual, which allows to allocate the data congestions. Applicable data RSD algorithms are divided into K-medium and histogram. Histogram algorithm divides data on unimodal clusters. It do not require the assignment  the number of clusters before clusterig. Popular fast Narendra algorithm , however it demand decrease the level of detail of data space. The automate choice of detail is suggested by Sidorova V. S. The parameter of detailedness is defined by the value of the cluster separation.  The criterion of the clustering quality is the degree of the clusters separation.  Hierarchical algorithm  allows different parts of the data to automatically find their utmost detail given the complexity of the objects. In this paper also proposed to reduce the dimension of data through rotating multispectral space into eigen space.The algorithm was used to map the pollution of waste productions of the Omsk region by eight spectral channels of satellite "Landsat-8" (resolution 15m, 02.08.2014).

This work was partially supported by the Russian Foundation for Basic Research (grant № 13- 07- 00068 ), and the program number 43 the Presidium of the Russian Academy of Sciences (project number 32 ) .

 


To reports list