面向多期点云所构建真实表面模型的精细三维变化检测方法

There are many reasons for us to propose a new solution of high quality three-dimensional change detection towards real geometric surface models constructed by multi-temporal point clouds. The first reason is that the theory of the current three-dimensional change detection systems is lack of prefectly abstraction. The second reason is that the automation level of practice is relatively low. And last reason is that we can easily acquire high-precision, high-density point clouds with a lot of ways. Furthermore, many key technologies of our proposed mehtod are discussed in detail,which can be listed as follows:1)co-registration of multi-temporal point clouds;2)real surface reconstruction of point clouds;3)region segmentation for three-dimensional triangulation network and automatic extraction for key elements;4)rapidly change detection by means of point to point intersection and error change regions removal. In view of the design intent of this project,three important issues will be resolved:1)the main technical methods of three-dimensional change detection will be successfully established.2)the main drawbacks of the traditional methods can be overcomed, such as low efficiency, low accuracy.More importantly, our proposed method can be easily applied into many applications, such as urban change detection of national geographical conditions monitoring project, high quality dynamic monitoring in the local area,three-dimensional assessment of disater and so on.

针对目前三维变化检测理论方法体系不够完善、生产作业自动化水平较低等现实问题，结合现有高精度、高密集点云获取手段飞速发展的技术现状，提出面向同一覆盖区域多期点云所构建的真实几何表面模型进行精细三维变化检测的理论方法与技术路线。对项目所涉及的多期点云整体配准优化、三维点云真实表面重构、空间三角网区域分割与关键要素自动提取、模型间逐点法向快速求交及伪变化剔除等关键技术展开研究。项目研究工作的顺利开展与实施，不仅能初步建立一套适合局部范围内多期点云的三维变化检测技术方法体系，提高三维变换检测自动化程度，从而逐步克服现有三维变化检测应用所存在的检测精度不足、效率低等显著缺陷，而且相关理论成果还能直接为开展地理国情监测中城市关键要素监测提供理论参考与技术支撑、高质量动态监测工程提供新途径，并有望在准实时三维灾情评估等应急响应领域发挥重要作用。

项目在执行过程中，基本能按照原定计划执行，较好地完成了既定的研究目标。在研究内容方面，项目解决了具有海量观测数量的多期点云整体配准问题，总结不同配准策略、不同分辨率以及不同表面几何结构对配准精度影响分析，得出一整套可行的多期点云配准方案，并开发出实用的算法模块，并在点云预处理软件中形成工具；项目开发了海量点云的全自动三角网构建算法，能在无人工交互条件下解决海量点云的自动化构网，具有很强的实用性；项目还重点研究了实现基于不同期精细三维表面模型的变化检测方法，其中对基于三角形的表面区域分割算法、区域拓扑构建以及目标区域引导下的变化要素提取与实体化表达等核心难点进行算法设计，并针对实际数据展开测试。另外，从产业化角度来看，针对大范围城市地表进行3D变化监测是行业应用的重要趋势之一。传统以带地理编码的2D遥感影像为核心数据源的变化监测技术手段，势必会引入3D数据源作为信息辅助，并将逐步向3D变化监测过渡和升级。因此，本项目在该领域中只是初步尝试了一些核心的、基础的算法理论探索，更希望在未来的研究中能进一步得以完善，并尽最大可能地朝着产业化方向迈进。最后，项目组通过对相关成果的提炼与整理，发表多篇学术论文，申报多项专利和编著专著，出版专著1部，参加学术会议或行业会议交流多次，基于此项目还培养了至少5名硕士研究生。