面向小班对象的森林资源变化遥感监测方法——以福建省厦门市为例

快速、客观、有效的森林资源变化监测技术是林业资源管理部门迫切需要解决的技术难题.本研究基于面向对象影像分析的思想,提出面向小班对象的森林资源变化遥感监测方法.首先,通过结合森林小班图层的大尺度影像分割,得到小班影像对象;其次,在森林小班专题对象内部,进行小尺度分割,自动提取变化图斑边界,得到变化与未变化小班影像对象;最后,通过对每一林地专题内部变化小班的分类解译,直接获得森林资源的地类变化信息.该方法可以根据影像分辨率的高低和森林小班图的尺度,满足不同比例尺森林覆盖变化监测的精度要求.以福建省厦门市为例,选用2011年RapidEye卫星影像和2007年森林小班图层进行森林覆盖变化信息提取.结果表明,所提出的方法在确保精度的同时,时间效率提高1～2倍,满足林业部门对森林覆盖变化信息快速准确获取的要求.

A Method to Extract Forest Cover Change by Object -Oriented Classification

Object-oriented classification is currently a trend in the detection of forest cover change. Here, we propose a method for extracting forest cover change based on the latest remote sensing images and old forest maps using object-oriented classification. First, basic forest objects can be obtained by chessboard segmentation. Second, changed objects in the forest basic object are segmented from unchanged objects based on the first segment result with a smaller scale parameter by multi-resolution segmentation (Fractal Net Evolution Approach). Third, forest cover change class can be extracted by the rule set and expert knowledge or by classification arithmetic. Using Xiamen, Fujian province as a case study, forest cover change object and information from 2007 to 2011 was detected and classified. Classification accuracy reached 94.5%. The results show that change detection using this kind of forest basic object-orientation is an effective semi-automatic change detection method. The work efficiency is about doubled compared with traditional methods. The total rate of forest cover in Xiamen increased by 0.25% from 2007 to 2011, and reached 43.05%. Most types for forest cover change, such as conversion from forest to cutting scar or fire scar or non-forest land, can be easily identified by RapidEye satellite images. Sub-meter resolution remote sensing images, such as QuickBird or WorldView 2, are still necessary for more detailed forest cover change inventory.