基于多时相遥感数据的东亚飞蝗生境分类研究

东亚飞蝗的发生及成灾与其存在的生境有十分密切的关系,因此对东亚飞蝗生境进行分类是有效防治蝗灾的基础。以河北省渤海新区为研究区,利用环境减灾小卫星多时相CCD数据,采用4种组合方案,分别使用最大似然法和RuleGen决策树进行了东亚飞蝗生境的遥感分类。结果表明:最大似然法和决策树分类方法总体精度相差不大,但在类别数据较多时,决策树分类方法的执行效率变低。利用5月20日单时相的光谱数据进行分类的总体精度仅有76.43%,Kappa系数〖JP2〗0.7396;加入NDVI时间序列信息后,总体分类精度可以达到93.93%,Kappa系数0.9323。因此,使用多时相信息可以较好地解决异物同谱问题,降低混合像元带来的影响,提高生境分类的精度。〖JP〗

Oriental Migratory Locust Habitat Classification based on Multi-temporal Remote Sensing Data

There are close relationships between locust occurrence and their habitat.Therefore,the key of efficiently control of locust plague carrys out locust habitat monitoring.In this study,Bohai development zone in Hebei province was selected as the study area and multi-temporal data from the HJ-1 satellites were chosen to monitor the locust habitat.The locust habitat were monitored by the use of classification based on four different schemes of images band combination and two kinds of classifiers,including the maximum likelihood classifier and RuleGen Decision Tree classifier.The results show that there is no obvious difference between the two kinds of classifiers in terms of the overall accuracy of locust habitat classification.Precisely,the execution efficiency of RuleGen decision tree becomes low when the different category data exist.On the other hand,the overall accuracy of locust habitat classification with the single temporal combination schemes dated on May 20 is only 76.43%,Kappa coefficient of 0.7396,but after adding NDVI time-series information,the overall accuracy can reach 93.93%,Kappa coefficient of 0.9323.Therefore,taking advantage of multi-temporal data,which have similar spectral characteristics easy to classify,the impact of mixed pixel can be reduced largely and the classification accuracy of locust habitat can further be improved.