顾及多特征的配准与建筑物变化检测一体化处理研究

With the continuous and accelerated development of urbanization in recent years, a timely and accurate monitoring of land use and land cover around urban,especially the building change of most active, has become a problem to be solved urgently. With the improvement of image resolution, spatial features changes caused by surface relief because of receiving status of image with multi-temporal are more obvious. The difficulties of image registration are more outstanding, which seriously influence the change detection accuracy. Traditional change information extraction technique cannot be well applied to building change detection in high resolution images, whose automation is low. Registration error leads to large amount of false detection and the change information influences the precision of registration in return. Key technologies of building automatic change detection are researched, which include building automatic feature extraction method and building change detection well applied to high resolution images, and unifying approach to registration and change detection. Through this research, a overall solution about automatic registration and building change detection well applied to high resolution images is expected to be proposed. The research results also would be expected to provide technique support for building monitoring by means of remote sensing.

随着城市化进程的持续推进以及城镇化道路的加速发展，及时准确地监测城市周边土地利用/覆盖的变化，尤其是最活跃的建筑物变化尤为重要。随着影像分辨率的提高，不同时相影像成像角度的不同使地表起伏引起的空间特征变化更加明显，影像的配准更加困难，严重影响变化检测精度；现有建筑物变化检测算法存在精度低、鲁棒性差和自动化程度低的问题，不能很好地适用于高分辨率影像的建筑物变化检测；配准误差导致变化检测结果存在大量误检，变化信息反过来影响配准精度。本项目针对上述难点问题开展研究，重点解决适应高分辨率影像特点的建筑物特征自动提取、建筑物变化自动检测和配准与变化检测一体化处理等难点问题。通过本课题的研究可望提出一套适应于高分辨率影像的配准与建筑物变化检测的自动化处理方案和流程，为利用遥感手段进行建筑物快速监测工作提供技术支持。

建筑物作为城镇环境的重要组成部分，及时准确地监测其变化对于土地利用规划、城镇化进程监测等都具有重要的应用价值。高分辨率影像中地物空间特征更加详细，但与此同时不同成像角度、地表起伏等引起的多时相影像空间特征变化更加明显，给影像配准和变化检测带来了更大困难。为了提高影像配准与变化检测的精度，项目从影像配准与变化检测本质出发，研究了影像配准、变化检测的新模型与新方法，其主要内容包括以下几个方面：（1）适应高分辨率卫星数据特点的面向建筑物变化检测的特征自动提取方法；（2）适应高分辨率卫星数据特点的建筑物变化自动检测方法；（3）配准与变化检测一体化处理方法；（4）顾及图像异质性的变化检测方法。项目通过资源三号、WorldView-2、TerraSAR等卫星影像数据对提出的新模型和新方法进行了大量实验，并与同类方法进行比较，获得了较好的实验效果。本项目提出了一套适应于高分辨率影像的配准与建筑物变化检测的自动化处理方案和流程，这些成果有望为测绘国土和城市规划等部门快速准确地监测建筑物信息提供技术支持。