顾及纹理特征贡献度的变化影像对象提取算法

发布时间：2018-07-16 16:34

【摘要】：遥感影像变化检测是全球变化研究的重要内容。基于两期遥感影像的变化检测方法存在数据条件要求苛刻、难以充分利用快速发展的多源遥感影像数据等问题。目前许多变化检测的参考数据中包含了一期分类矢量数据,矢量数据中往往包含了位置、形状、大小和类别属性等先验信息,充分利用这些先验信息将可提高变化检测精度。提取变化影像对象是结合矢量数据和遥感影像进行变化检测的核心步骤。本文提出了一种顾及纹理特征贡献度的变化影像对象提取方法。该方法利用矢量数据分割遥感影像,获取影像对象,计算影像对象纹理特征值。根据信息增益原理计算纹理特征参数的特征贡献度,选择特征参数。由贡献度指数大小确定纹理特征参数权重,计算影像对象与先验要素类别的相似度系数,提取变化影像对象。试验结果表明,基于纹理特征贡献度的特征参数选择,能有效地提高变化影像对象提取结果的精度。
[Abstract]:Remote sensing image change detection is an important part of global change research. The change detection method based on two-phase remote sensing image has some problems such as demanding data conditions and making full use of the rapidly developing multi-source remote sensing image data. At present, many reference data for change detection contain one stage classified vector data, which often contain prior information such as position, shape, size and category attributes. The accuracy of change detection can be improved by making full use of these prior information. Extracting change image object is the core step of change detection combining vector data and remote sensing image. In this paper, a method of object extraction based on texture feature contribution is proposed. The method uses vector data to segment remote sensing image, obtains image object, and calculates texture feature value of image object. According to the information gain principle, the feature contribution of texture feature parameters is calculated and the feature parameters are selected. The weight of texture feature parameters is determined by the contribution index, the similarity coefficient between the image object and the prior element category is calculated, and the changing image object is extracted. The experimental results show that the selection of feature parameters based on texture feature contribution can effectively improve the accuracy of the extraction results of changing image objects.
【作者单位】：中南大学地球科学与信息物理学院;中南林业科技大学土木工程学院;有色金属成矿预测与地质环境监测教育部重点实验室(中南大学);有色资源与地质灾害探查湖南省重点实验室;
【基金】：十三五国家重点研发计划重点专项(2016YFB0501403) 国家自然科学基金(41371366)~~
【分类号】：P237

【相似文献】