基于DBSCAN算法的虚假卫星林火热点挖掘研究

Research on the false forest fire hotspot mining of satellite based on DBSCAN algorithm

利用卫星遥感数据自动识别林火热点的过程中,需要尽可能排除卫星林火虚假热点以提高林火热点识别的精度。利用DBSCAN算法对湖南省范围内的卫星林火热点信息数据库(2015年10月—2017年12月)进行了数据挖掘,并建立了虚假林火热点历史经验库,对挖掘结果进行了分析和判别验证。结果表明:运用得到的虚假林火热点历史经验数据库进行比对,能够对固定热源类、农用火源类和水面反射类的卫星林火虚假热点进行有效过滤,为提高森林火灾卫星动态监测的准确度提供了一种有效的解决办法。

During the process of automatic identification of forest fire hotspots by satellite remote sensing data, it is necessary to eliminate the false hotspot of forest fires as far as possible so as to improve the accuracy of forest fire hotspot identification. In this paper, the DBSCAN algorithm is used to data mining of the satellite forest fire hot spot information database （2015.10-2017.12） in Hunan Province, and the historical experience base of the false forest fire hot spot is set up, and the mining results are analyzed and verified. The results show that the false hot spot of fixed heat source, agricultural fire source and water surface reflection can be effectively filtered by the comparison of the false forest fire historical experience database obtained. It provides an effective solution to improve the accuracy of the dynamic monitoring of forest fire satellite.