New registration techniques to improve mobile lidar mapping accuracy

移动的测绘系统（MMS）是三维现实捕获系统，其利用集成的导航和成像传感器从移动车辆（陆地或空中）收集地理参考空间数据。MMS数据是日常生活的一部分。谷歌街景和谷歌地球视图是MMS技术如何被大众使用的最著名的例子。MMS的应用跨越许多学科领域，包括电力线路测绘、基础设施监测和工业现场建模。根据MMS数据构建的模型允许用户虚拟地检查现场并获得测量结果，以支持明智的工程决策。该项目的重点是解决两个与行业相关的科学问题，这将提高MMS数据模型的准确性和实用性。第一个是自动MMS校准。所提取的测量值的准确性主要取决于MMS上的各种传感器的相对位置和方向的精确知识，这些传感器在数据采集之前被校准。已知方位随着时间的推移是不稳定的，这损害了系统精度。将开发一种使用MMS固有物体（例如杆）进行自动系统校准的新方法，以便在驾驶车辆进行数据收集时解决新的方位参数。第二是图像自动配准。通常在MMS数据捕获之后很久才收集站点的图像，以便执行变化检测。在这种情况下，图像可以从智能电话收集，并且其取向是未知的。将开发新的方法，自动将这一图像定位到已经收集的MMS数据，以便可以更新网站模型。

Mobile mapping systems (MMSs) are three-dimensional reality capture systems that collect georeferenced spatial data with integrated navigation and imaging sensors from a moving vehicle (terrestrial or aerial). MMS data are a part of everyday life. Google Street View and Google Earth View are the best-known examples of how MMS technology is used by the general population. The applications of MMS span many discipline areas including power line mapping, infrastructure monitoring and industrial site modelling. Models constructed from MMS data allow users to virtually inspect sites and derive measurements in support of informed engineering decisions. This project focuses on solving two industry-relevant, scientific problems that will enhance the accuracy and the utility of MMS data models. The first is automated MMS calibration. The accuracy of extracted measurements is critically dependent on precise knowledge of the relative location and orientation of the various sensors on board the MMS, which is calibrated prior to data acquisition. The orientation is known to be unstable over time, which compromises system accuracy. A new method for automated system calibration using objects inherent to an MMS (e.g. poles) will be developed to allow new orientation parameters to be solved as the vehicle is driven for data collection. The second is automated image registration. Imagery of a site is often collected long after MMS data capture in order to perform change detection. In this situation the imagery may be collected from a smart phone and its orientation is unknown. New methods will be developed to automatically orient this imagery to already-collected MMS data so that site models can be updated.