In the present study ,a novel change detection approach for high resolution remote sensing images is proposed based on the optical flow dynamic texture (OFDT) ,which could achieve the land use & land cover change information automatically with a dynamic description of ground-object changes .This paper describes the ground-object gradual change process from the principle using optical flow theory ,which breaks the ground-object sudden change hypothesis in remote sensing change detection methods in the past .As the steps of this method are simple ,it could be integrated in the systems and software such as Land Resource Management and Urban Planning software that needs to find ground-object changes .This method takes into account the tempo-ral dimension feature between remote sensing images ,which provides a richer set of information for remote sensing change detec-tion ,thereby improving the status that most of the change detection methods are mainly dependent on the spatial dimension in-formation .In this article ,optical flow dynamic texture is the basic reflection of changes ,and it is used in high resolution remote sensing image support vector machine post-classification change detection ,combined with spectral information .The texture in the temporal dimension which is considered in this article has a smaller amount of data than most of the textures in the spatial di-mensions .The highly automated texture computing has only one parameter to set ,which could relax the onerous manual evalua-tion present status .The effectiveness of the proposed approach is evaluated with the 2011 and 2012 QuickBird datasets covering Duerbert Mongolian Autonomous County of Daqing City ,China .Then ,the effects of different optical flow smooth coefficient and the impact on the description of the ground-object changes in the method are deeply analyzed .The experiment result is satis-factory ,with an 87.29% overall accuracy and an 0.850 7 Kappa index ,and the method achieves better performance than the post-classification change detection methods using spectral information only .%建议了一种基于光流动态纹理（optical flow dynamic texture）的高分辨率遥感影像变化检测新方法，用一种运动的关系描述地物变化，能够在多时相高分辨率遥感影像中自动获取土地利用和土地覆盖的变化信息。利用光流理论从原理上描述了地物渐变的过程，突破了以往遥感变化检测方法中认为地物发生突变的假设。该方法的流程简单，易于在目前的土地管理、城市规划等需要发现用地变化的系统和软件中使用。该方法考虑到了多时相遥感影像间的时间维度特征，为遥感变化检测提供了更加丰富的信息，进而改善了变化检测方法主要依赖空间维度信息的现状。以光流动态纹理作为变化的基本体现，结合光谱信息共同用于高分辨率遥感影像的支持向量机分类后变化检测，方法顾及了遥感影像时间维度的纹理，相较大多数空间纹理其数据量较小；纹理计算仅需设定一个参数，自动程度较高；可缓解行业中大量人工解译的现状。通过利用中国大庆市杜尔伯特蒙古自治县2011年和2012年Q uickBird影像对该方法的有效性进行了评价。深入分析了不同的光流平滑系数α对该方法的影响，以及对地物变化描述效果的影响。实验结果显示，该方法效果理想，总体精度达到87.29％、Kappa系数达到0.8507，其精度优于单纯利用光谱信息的分类后变化检测方法。
展开▼
