基于高光谱特征选择和RBFNN的城市植被胁迫程度监测

以Hyperion星载高光谱数据为例,基于指数提取-特征选择-分类识别-模式分析的思路,分析广州市的城市植被胁迫状况。提取与胁迫相关的高光谱植被指数,对其进行相关分析,滤除相关性很高的植被指数,利用选取的特征应用RBF(径向基函数)神经网络对城市的植被胁迫程度进行分类,对广州市受胁迫植被的空间分布及其原因进行分析。研究表明:运用特征选取和RBF神经网络可以较好的区分城市植被受胁迫的程度;城市植被受胁迫的程度与城市交通污染、人为干扰相关性比较大;受胁迫植被的强度分布呈现从城市中心向外的梯度变化,在大块绿地外围呈环状分布。

Urban Vegetation Stress Level Monitoring Based on Hyperspectral Feature Selection and RBF Neural Network

Urban vegetation system has great ecological value to socialeconomic-natural ecosystem,but it was often submitted to different stresses caused by air pollution,water pollution,"heat island" problem,etc.,which debases its ecological service functions,so it is important to develop methods to monitor urban vegetation stress level.Using the Hyperion hyper-spectral data,which has advantage in monitoring vegetation physiological characters on large scale,an urban vegetation stress level monitoring method was developed based on vegetation stress feature selection and RBFNN(Radial Basis Function Neural Network).Firstly,14 hyperspectral vegetation indices were extracted from reflectance image of Hyperion and a feature selection based on correlation analysis was conducted to reduce the redundancies.Then an urban vegetation stress level classifier based on RBFNN was trained on the selected features and vegetation stress level was classified and mapped.Finally,the spatial distribution characteristics of vegetation stress in urban and some reasons were analyzed.The result shows that the RBFNN vegetation stress level classifier is able to identify vegetation stress level quickly and accurately.Vegetation stress level is correlated largely with urban traffic pollution and human disturbance,and vegetation in commercial and residential areas of urban center are apparently experiencing higher stress than vegetation in suburban regions;the stress level shows a ringy distribution around large greenbelts.