摘要:
为了研究利用高分一号宽幅影像(GF-1 WFV)监测森林覆盖变化的方法,选取四川省甘孜州雅江县为研究区,利用2014年和2016年2期GF-1WFV数据,采用迭代加权多元变化检测(iteration re-weight multivariate alter-ation detection,IR-MAD)法对数据进行辐射归一化;分别对2期影像进行核主成分分析(kernel principal component analysis,KPCA)方法变换,采用最大类间方差法(OTSU)确定自动识别阈值,对2期GF-1WFV影像中的森林覆盖变化区域进行检测和精度验证;并与变化矢量分析(change vector analysis,CVA)法检测结果进行对比分析.研究结果表明:所用2种变化检测算法的总体检测精度都超过了80%,其中,KPCA法的总体精度为89.27%,未变化区用户精度达93.88%,变化区用户精度为80.28%;基于KPCA法的精度均较优于传统CVA检测算法,说明 KPCA算法通过数据变换后,可减少变量间的相关性、增强影像信噪比,从而提高了对变化区域的识别精度.
Abstract:
In order to study the methods for forest cover change monitoring by using GF-1 images, the Yajiang County in Sichuan Province was selected as the research area to extract the information of forest coverage based on the two GF-1 WFV data. Firstly,the data were normalized by using the iteration re-weight multivariate alteration detection(IR-MAD)method. The two images were transformed by kernel principal component analysis(KPCA) method,and formed differencing image. Then, the changed area was extracted using the method of maximum between class variances(OTSU)for automatic threshold selection. Finally, the change detection results were validated using OTSU with the field sample data, and the extracted results were verified by way of precision test, and comparatively analyzed with the change vector analysis(CVA). The research results show that the overall accuracy of the two change detection methods is higher than 80%,and the overall accuracy of the KPCA method is 89.27%. The user precision of unchanged area is 93.88%,and the user's accuracy of changed area is 80.28%. The accuracy of the KPCA method is better than that of the algorithm based on the traditional CVA method. It is shown that, after the data transformation, KPCA algorithm can reduce the correlation between the variables and enhance the signal to noise ratio of the image,thus improving the recognition accuracy for the changed area.
下载: