漳平市森林可燃物类型变化遥感动态监测

森林可燃物类型的空间分布是林火蔓延、灭火可视化建模与仿真中需要考虑的重要因素之一。在对森林可燃物类型划分的研究进行回顾的基础上,提出了考虑树种信息的分类方法。根据漳平市2003年小班图层中的优势树种信息,获得该市2003年四种森林可燃物类型即竹林、阔叶树、杉木林以及马尾松的空间分布专题图。在此基础上,采用面向对象分类技术,对ASTER影像进行分类,探测每种可燃物类型的内部变化和外部变化。该技术利用影像分割技术构建分类对象,使每个对象具有光谱信息的同时,具备大小、形状、拓扑关系、类别层次等诸多信息。对分类结果进行评价的结果表明,利用面向对象分类技术,充分利用了光谱特征以及类别相关特征,提高了分类精度,分类的面积精度达到89.3%。由于影像分割过程应用了专题图层,对象的边界不会超越专题图层的边界,使得对现有图层的更新尤为容易。最后,利用矢量格式的遥感分类结果对原森林小班图层进行更新,获得新的可燃物类型图层,作为林火可视化模型的输入图层。该研究不仅提供了现势性强的森林可燃物类型图层,而且在不破坏原小班边界的基础上,对发生变化的区域进行刻画。对于森林资源管理者,提供了实地调查过程的目标区域的信息。

Forest Fuel Types Classification and Change Detection by Remote Sensing in Zhangping City

Forest fuel type's spatial distribution is one of the key factors to be considered during forest fire spread and fire-fighting visualization modeling and simulation.Based on review of the status of the study on forest fuel type classification,a classification method considering tree types was put forward.According to the attribute of predominant tree types of 2003 forest stand layers,four forest fuel types,i.e.,bamboo forest,broad-leaved forest,Chinese fir forest,and Masson pine forest and their spatial distribution in Zhangping was acquired firstly.The object-oriented classification technology,which using the techniques of image segmentation to construct objects containing information of size,shape,topography and class hierarchy besides spectral information,and involving procedures of segmentation,construction of class hierarchy,and classification,was then applied to the ASTER images to detect interior and exterior change of each type's.Accuracy assessment of the classification result indicated that by the technology of object-oriented classification,both the spectral features and class-related features were used,and the accuracy was improved in terms of area,reaching 89.3%.Because thematic layers were involved during the image segmentation,the boundaries of objects did not get across the boundaries of each thematic layer,this makes the updating of the existing layers become very convenient.Finally,the vector format data of remote sensing classification result was used to update the original forest stand layer in order to get the available inputs of fire visualization modeling.This study not only provided up-to-date input layers of forest fuels types for fire visualization modeling,but also detected the areas which have been changed without changing the boundaries of the original forest stands.For forest resource managers,it provided target areas for validation by field survey.