Along with the rising of on-board and borne satellite technology, hyperspectral remotesensing technology gets enormous development these years, which is widely appliedwithin various fields like mining, forestry, etc. The emerging of hyperspectral remotesensing technology is a revolution within remote sensing field, and it makes theundetectable material within wide band could be detected and analyzed. Classificationmeans to assort each pixel into one class within a bunch. The result of classification isseveral sub-areas come out, while each of them represents an actual land feature. Thispaper originally takes traditional K-Means classification algorithm as basic functionand universal gravitation to optimize it. This way, the advantages of K-Means arereserved while its shortcomings have been overcome.Taking hyperspectral remote sensing data from Xinjiang Dongtianshan as studyobjects, this paper figures out an unsupervised classification function based onK-Means algorithm. According to relative comparison and verification, its effectshave been proved. It is believed that this function shows some inspiring results infuture and could be applied within some extent. |