A System for Automated Generation of Urban 3D Digital Map

我们开发了一系列方法来生成非常详细的城市地区3D数字地图。该方法的特点是分别建立了基于机载传感器和车载传感器的方法，并将其相结合。1.3车载激光/CCD传感器的三维数据采集和路边特征的自动提取。一种从车辆激光扫描数据中提取空间目标的自动化方法。提取的目标不仅包括道路、树木、车辆和建筑物，还包括线性特征，如护栏和管道、交通标志和电线杆。在目标物体没有被遮挡的情况下，我们可以获得相当高的提取成功率。开发3D建模的半自动方法我们开发了一种半自动方法，操作员只需手动划定目标物体的边界或位置，就可以非常容易地生成3D模型。在这个过程中，计算机确定目标物体的确切位置，并生成目标物体的精确表面，目标物体的边界大致由操作员给出。实际上，在城市地区，空间物体可能有非常复杂的形状，有时它们会被树木遮挡。这些情况使得3D建模的完全自动化变得非常困难。采用半自动化的方法，可以大大提高三维数据生成的总效率。利用机载线传感器数据生成城市物体（主要是建筑物）三维模型的三维建模方法的发展表明，通过提供现有的二维地图作为初始值或对象，可以大大提高建筑物的自动提取和三维建模的成功率。将该方法与半自动化方法相结合，可以大大提高城市物体三维建模的整个过程的效率。此外，车载激光扫描仪的数据可以在三维建模过程中集成。

We developed a series of methods to generate a very detailed 3D digital map of urban areas. The approach is characterized by the fact that approaches using air-borne sensors and vehicle borne sensors are established respectively and then integrated.1.3D data acquisition from vehicle borne laser/CCD sensors and automated extraction of roadside features.An automated method of extracting spatial objects from laser scanning data from a vehicle. Targets of extraction include not only road, trees, vehicles and buildings but also linear features such as guard rails and pipes, traffic signs and poles. In case the target objects are not occluded, we can achieve a rather high probability of extraction success.2.Development of semi-automated methods for 3D modelingWe develop a semi-automated method with which an operator can generate 3D models very easily just by delineate the boundary or the location of a target-object manually. Behind the process, a computer determines the exact location and generates accurate surfaces of the target-object the boundary of which is roughly given by the operator. Actually, in urban areas, spatial objects may have so much complicated shapes and sometimes they are occluded by trees. These situations make the full automation in 3D modeling so difficult. With the semi-automated method, the total efficiency of the 3D data generation can be so much improved.3.Development of 3D modeling methods using air-borne line sensor dataIn generating 3D models of urban objects, mainly buildings, it is demonstrated that the success ratio of automated extraction and 3D modeling of buildings can be so much improved by providing an existing 2D map as a initial value or objects. By combining this method with the semi-automated method, it is demonstrated that the entire process of 3D modeling of urban objects can be so much improved in efficiency. In addition, data from the vehicle borne laser scanners can be integrated in the process of 3D modeling.