【摘要】 以北京市延庆区2004年Spot-5影像和2015年GF-1影像数据为研究对象,应用面向对象分类变化检测算法,通过选择最优尺度和结合典型地物光谱特征、纹理特征建立规则集来对两期影像进行分类,然后提取十年间延庆区公益林的变化地块,最后进行精度评价,旨在对延庆区公益林的变化及其驱动因素进行探索分析。结果表明:Spot-5影像的分类精度为87.1%,加入FC特征值规则的GF-1影像的分类精度为89.1%,高于未加入FC特征值规则的GF分类精度(84.8%),说明在规则集中加入FC特征值能提高森林分类精度;变化信息提取的结果总体精度为87.3%,漏判率、错判率都在20%以内,提取效果较佳;2004—2015年间,公益林面积呈上升趋势,且主要集中在有林地面积增加,农田、灌木地和其他土地面积减少,这与国家对林业及公益林日益增加的重视度、各项工程项目密不可分。更多还原

【Abstract】 Based on Spot-5 and GF-1 image data, object-oriented classification method was used to detect the dynamic change in Yanqing District of Beijing. To extract the changeable plots, the two images were classified by selecting the optimal multi-segmentation scale and combining typical spectral features and texture features to establish rule sets system, afterwards the accuracy evaluation and analysis of the changes were carried out. This research aims to explore and analyze the changes and driving factors of non-commercial forest. The results indicate that the classification accuracy of Spot-5 image is 87.1%, the classification accuracy of GF-1 image with FC eigenvalue rule is 89.1%, which is higher than that of GF-1 image without FC eigenvalue rule, 84.8%, which means that adding FC eigenvalue in rule sets system can improve forest classification accuracy; the overall accuracy of the results of change information extraction is 87.3%, missed rates and error rates were less than 20%; From 2004 to 2015, the area of non-commercial forest showed an upward trend, which mainly reflected on increasing forest land area and decreasing farmland, shrubs and other lands. These changes were closely related to the increasing importance of forestry and non-commercial forests and various engineering projects.更多还原