论文部分内容阅读
传统的实地调查和人工解译方法已经不能满足区域尺度的林隙获取,高空间分辨率遥感影像的出现为区域尺度的林隙获取提供了可能。本研究采用QuickBird高空间分辨率光学遥感影像,结合面向对象分类技术对福建省三明市将乐县将乐国有林场进行林隙分割与分类。在面向对象分类过程中,采用10种尺度(10~100,步长为10)对QuickBird遥感影像进行分割,应用参考对象相交面积(RA_(or))和分割对象相交面积(RA_(os))进行分割结果评价。对每个尺度分割结果应用16个光谱特征,采用向量机分类器(SVM)进行林隙、非林隙和其他类型分类。结果表明:通过RA_(or)和RA_(os)等值法获得最优分割尺度参数为40。不同尺度参数之间的分类总精度最高相差22%。在最优尺度下,应用SVM分类器对林隙、非林隙和其他类型分类的总精度高达88%(Kappa=0.82)。采用高空间分辨率遥感数据并结合面向对象的方法,可以代替传统的实地调查和人工解译对区域尺度的林隙进行识别分类。
The traditional methods of field survey and manual interpretation can not meet the requirements of the gap at the regional scale. The emergence of high spatial resolution remote sensing images provides the possibility of obtaining gaps at the regional scale. In this study, QuickBird high spatial resolution optical remote sensing image was used, and object-oriented classification technology was used to segment and classify the gaps of the gaps in the Qile National Forest Farm, Jiangle County, Sanming City, Fujian Province. In the process of object-oriented classification, QuickBird remote sensing images are segmented using 10 scales (10-100 in 10 steps) and the reference object intersection area (RA_ (or)) and the segmentation target intersection area (RA_ (os) Evaluate the segmentation results. Sixteen spectral features are applied to each scale segmentation result, and the gap, non-gap and other types are classified by SVM. The results show that the optimal segmentation scale parameter is 40 through the RA and os equivalent methods. The highest accuracy of the classification accuracy among different scale parameters is 22%. At the optimal scale, the total accuracy of the SVM classifier for gap, non-gap and other types of classification is as high as 88% (Kappa = 0.82). Remote sensing data with high spatial resolution and object-oriented method can be used instead of traditional field survey and manual interpretation to identify and classify the gap at the regional scale.