基于图像理解的土地利用分类信息提取

Land Use Classification Information Extraction Based on Image Understanding

摘要:

基于图像理解理论原理，采用面向对象分类软件作为基础软件平台，模拟人类大脑对图像的认知过程，概括出不同地类的知识表达，并构建函数作为类别的特征描述，更大限度地挖掘中低分辨率卫星数据的信息。研究结果表明，基于图像理解的土地利用分类信息提取方法有效地提高了地类的检出效率和分类精度，为中巴卫星应用于国土资源宏观监测的可行性提供了依据。

Abstract:

Based on the principle of image understanding theory, using objectoriented classification software as basic software platform, simulate cognitive process of images of the human brain, knowledge expression of different land types have been summarized. Regarding construction function as the feature of category, more information of middlelow resolution satellite remoe sensing data can be digged up in maxmium. As showed by the result, the land use classification information extraction method based on image understanding can effectively improve the classification efficiency and accuracy. It will provide the basis for the feasibility of CBERS applying in land resources monitoring.