面向对象分类方法在森林采伐遥感监测中的应用

Application of classification object-oriented to forest cutting monitoring based on remote sensing

利用多尺度分割技术和面向对象分类方法对SPOT5遥感数据进行土地的分类及森林采伐信息提取。在面向对象的图像分析中,采用图像分割——基于规则的分类——基于分类的分割的多尺度分割方法,在综合最优分割尺度下,用最邻近分类器对SPOT5影像进行分类;采用两期图像特征比较,提取森林采伐区信息,并结合二类调查成果和伐区设计资料,使用交互式补充判读和修正。结果显示:研究区各地类的分类精度都在85%以上,对森林采伐图斑判读的加权综合正判率达到90.8%,其中皆伐图斑个数正判率92.8%,非皆伐图斑个数正判率83.3%。利用多期高分辨率遥感图像可以进行森林采伐监测,研究结果为提高森林采伐限额监测效率、采伐区识别准确度和面积估算精度提供了有效途径。

The multi-size segmentation and classification object-oriented techniques of remote sensing have been used to extract information of land classification and forest harvesting from the data of SPOT5.In an object-oriented image analysis,a segmentation,classification based on the rules,and segmentation based on the classification method has been adopted.In the optimal comprehensive scale,nearest neighborhood filter has been used to make the classification of image of SPOT5;The forest harvesting information has been extracted with comparing two term image features,and combined the forest management inventory results and forest cutting designing and interactive interpretation.The results show that： the classification of land use type in research area was of precision in 85% or more,the interpretation of forest cutting sub-compartment the weighting accuracy was of 90.8%,amongst,the accuracy of clear cutting sub-compartment was 92.8 % and the accuracy of non-clear cutting sub-compartment was 83.3%.The study indicates that： multi-term high resolution remote sensing images can be used for monitoring the forest harvesting.It provides an effective way to raise the efficiency of the forest cutting inspection and monitoring,and improve the interpretation accuracy of the forest cutting area and the estimation precision of the areas.