With double-temporal TM and ETM+ remote sensing data, the information of the variation of forest resources of Culai Mountain in Shandong Province, China was explored. Decision tree classification based on C5. 0 algorithm to forest change detection was applied. Three different detection methods were compared:1) to classify single-temporal data by C5. 0 respectively, and extract change information after comparing classification results;2) to create C5. 0 train rules through double-temporal raw data,then generate change detection map;3) in addition to double-temporal remote sensing data,neighborhood correlation analysis images are also added as one of the data sources of C5. 0,and generate change detection map of variation. The experimental result shows that decision tree classification based on C5. 0 algorithm could detect variation information effectively, and after adding neighborhood correlation analysis images the classification accuracy of change detection was improved.%以山东省徂徕山林场为试验区,利用两时相的TM与ETM+遥感数据对该地区的针叶林、阔叶林等森林资源的变化进行研究.将基于C5.0算法的决策树分类方法应用于森林变化检测,并对3种检测方案进行试验比较:(1)以单一时相图像作为数据源并各自分类,分类后作比较提取变化信息；(2)以两时相图像的原始波段数据作为数据源训练规则,并生成变化检测图；(3)以两时相图像加上邻近相关分析图像作为数据源训练规则,生成变化检测图.试验结果表明,基于C5.0算法的决策树分类可以有效的进行森林变化检测,并且加入邻近相关分析图像后的变化检测精度达到最高.
展开▼
