Object-based 3D information incorporated classification and change detection

高达80%用于决策的信息都有一些地理空间背景。自1999年以来发射了许多现代高分辨率卫星，以提高获取和更新全球详细地理空间信息的能力。然而，由于从图像中提取信息的有效技术非常有限，许多图像在被有效使用甚至使用之前就已经过时了。 土地覆盖分类和变化检测是遥感影像信息提取中应用最广泛的两种技术。基于对象的分类是一项新兴的技术。它显示了巨大的潜力，但也遇到了许多具有技术挑战性的问题。基于我目前NSERC项目的发现和结果，这项拟议的研究旨在开发新的解决方案，以缓解基于对象的分类和变化检测这三个基本领域存在的问题，即对象分割、对象分类和图像配准。将开发用于监督分割、基于对象和3D信息融合的监督分类以及3D信息融合的变化检测的新技术。将利用立体图像的3D信息，因为所有现代高分辨率卫星和机载传感器都具有立体能力。将开发基于对象和结合3D信息的图像配准，以克服现有的图像配准问题。 这项研究的成功成果将有助于缩小大量收集的遥感图像与非常有限的图像信息提取能力之间的差距。它将帮助用户有效地将遥感图像用于广泛的应用，如环境监测、自然资源调查、规划和军事任务。在拟议的研究领域培养高素质的研究人员或工程师，对于保持加拿大在地理信息学方面的领先地位具有重要的战略意义。

As much as 80 percent of the information used for decision making has some geospatial context. Numerous modern high resolution satellites have been launched since 1999 to improve the capacity of acquiring and updating detailed geospatial information globally. However, due to very limited availability of effective technologies to extract information from the images, many images have become outdated before being effectively used or even being used. Land cover classification and change detection are two most widely used techniques in information extraction from the images. Object-based classification is a newly emerging technology. It has demonstrated tremendous potential, but also encountered many technically challenging problems. Based on the findings and results of my current NSERC projects, this proposed research intends to develop new solutions to mitigate problems existing in the three essential areas of object-based classification and change detection, i.e. object segmentation, object classification, and image registration. New techniques for supervised segmentation, object-based and 3D information incorporated supervised classification, and 3D information incorporated change detection will be developed. 3D information from stereo images will be utilized, as all the modern high resolution satellites and airborne sensors have stereo capacity. Object-based and 3D information incorporated image registration will be developed to overcome existing image registration problems. The successful outcome of this research will contribute to the gap reduction between vastly collected remote sensing images and very limited capacity of information extraction from the images. It will help users to effectively utilize remote sensing images for a broad range of applications, such as environment monitoring, natural resource investigation, planning, and military missions. The education of high quality researchers or engineers in the proposed research area is strategically important for keeping Canada's leading position in geoinformatics.