Building Model Reconstruction Based On3D Laser Point Cloud With Image Feature Lines Assisted

With the development of the digital city, constructing3D city buildings model has been an important content of the construction of digital cities.3D building model is one of the main constituent elements of digital city model, and how to construct building model quickly is a current research hotspot. In recent years,3D laser scanning is developing rapidly. As a kind of method of acquiring special data, it can obtain3D point cloud data of building surface rapidly and accurately, but point cloud data is huge, unstructured and lacking of texture. So it is difficult of building automatic or semi-automatic modeling using3D point cloud data directly. The ground photographic image has the characteristics of continuous features and rich texture information, but it has some shortcomings such like two-dimensional projection and low geometry precision, so using photographic images for constructing3D building model is also very difficult. Hence, if we take advantage of the complementary superiorities of both to construct3D building model, it will be beneficial to exploit the advantages of both and improve the efficiency and the degree of automation of constructing3D building model.This paper makes a research on building model reconstruction based on3D laser point cloud with image features assisted, and mainly has finished the following work.(1)The basic principle and method of registration using3D point cloud data and image data are discussed, and the registration has been realized using the point cloud data and the image data from the Riegl LMS-Z420i three-dimensional laser scanner system.(2)Feature extraction algorithm and post-processing algorithm based on the gradient direction are elaborated, and the algorithm uses gradient direction to extract the edge in the process of edge detection, then the obtained result is further processed according to the integrity rules of building elements. Experiments show that the algorithm has very good effect for building feature extraction, laying the foundation of the accurate feature extraction from point cloud data with image feature lines assisted.(3) The mapping from2D image feature lines to3D point cloud space is implemented, and through it the3D brief feature lines are obtained from building3D point cloud data, and then the3D point cloud data around3D feature lines are searched with the restraint of feature lines. Moreover the more accurate3D feature lines are obtained after a serious of operations of "area classification-plane fitting-feature line extraction" on the point cloud data around the feature lines. (4)Building model construction using the modeling software based on3D feature lines is completed, and then the images are used as texture data of texture mapping for modeling results. Finally, the theory and method mentioned in this paper have been tested and confirmed with experimental data for example.