Constructing 3D city models by merging aerial and ground views - Computer Graphics and Applications, IEEE

geometry and texture of urban surfaces could aid applications such as urban planning, training, disaster simulation, and virtual-heritage conservation. A standard technique used to create large-scale city models automatically or semi-automatically is to apply stereo vision on aerial or satellite imagery. In recent years, advances in resolution and accuracy have also made airborne laser scanners suitable for generating digital surface models (DSM) and 3D models. Although you can perform edge detection more accurately with aerial photos, airborne laser scans require no cameraparameter estimation and feature detection to obtain 3D geometry. Previous work has attempted to reconstruct polygonal models by using a library of predefined building shapes or by combining the DSM with digital ground plans or aerial images. While you can achieve submeter resolution with this technique, you can only capture the building roofs and not their facades. Several research projects have attempted to create models from ground-based views at a high level of detail to enable virtual exploration of city environments. While most of these approaches result in visually pleasing models, they involve an enormous amount of manual work, such as importing the geometry obtained from construction plans or selecting primitive shapes and correspondence points for image-based modeling or complex data acquisition. Other attempts to acquire close-range data in an automated fashion have used image-based or 3D laser scanner approaches. However, these approaches don’t scale to more than a few buildings because they acquire data in a slow, stop-and-go fashion. In previous work, we proposed an automated method that would rapidly acquire 3D geometry and texture data for an entire city. This method uses a vehicle equipped with 2D laser scanners and a digital camera to acquire data while driving at normal speeds on public roads. Other researchers proposed a similar system using 2D laser scanners and line cameras. In both systems, the researchers acquire data continuously and quickly. In another article, we presented automated methods to process this type of data efficiently to obtain a detailed model of the building facades in downtown Berkeley. However, these facade models don’t provide information about roofs or terrain shape because they consist only of surfaces visible from the ground level. In this article, we describe an approach to register and merge our detailed facade models with a complementary airborne model. Figure 1 shows the data-flow diagram of our method. The airborne modeling process on the left in Figure 1 provides a half-meter resolution model with a bird’s-eye view of the entire area, containing terrain profile and building tops. The groundbased modeling process on the right results in a detailed model of the building facades. Using the DSM obtained from airborne laser scans, we localize the acquisition vehicle and register the ground-based facades to the airborne model by means of Monte Carlo localization (MCL). We merge the two models with different resolutions to obtain a 3D model.