文章基本信息

标题：Integrated Registration of Range Images from Terrestrial LIDAR
本地全文：下载
作者：Wang Yanmin ; Wang Guoli
期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII Part B3b
页码：361-366
出版社：Copernicus Publications
摘要：There are generally two kinds of LIDAR, airborne and terrestrial. Terrestrial LIDAR is outstanding for its high accuracy, hyper speed and high density. A single range image of terrestrial LIDAR usually could not cover the whole object due to its multi aspects. The range images from different aspects of object can be registered into a unified point cloud model of object. The accuracy of resulted point cloud model is affected by registration accuracy directly. There are two kinds of registration. One is based on range images and another on extracted features. The former computes the distance errors of overlap part to align the range images. The other uses some frameworks or abstracted features as constraints to register the range images into a unified coordinate system. The combination of two methods can also make good results.For the methods above, neighbor range images are registered one by one. This procedure will evoke error accumulation. To solve the problem, an integrated registration method is put forward in this paper. All range images are registered and transformed to a unified coordinate system simultaneously. Parameter adjustment method is used in the new arithmetic. The spatial transformation parameters and constraint parameters are regarded as unknowns to be solved simultaneously. Integrated algorithm is actually a least square adjustment method, which assures the error distribution more rational.Point, line and plane are three feature types used as registration constraints in this paper. The point used for registration is called tic. The center point of circle or of sphere, the intersection point of two spatial lines or of three spatial planes are all points can be used for registration. Spatial transformation equation is always used as constraint of tic. Tics are the main constraints in range image registration. Line features and plane features are important registration constraints. If there are blunders in the range images, parameters of each range image will be polluted with them. Blunders are removed with comprehensive error analysis, which makes the algorithm more robust. The procedure of integrated registration can be divided into three steps: feature extract, one by one image registration and integrated registration. Feature extract procedure is to extract point, line and/or plane features from range images. The parameters of extracted feature are obtained and can be regarded as observations in constraint equation. One by one image registration is to roughly register the involved range images and get approximates of unknowns. Integrated registration is to solve all unknowns of constraints with least square adjustment method. Blunders are removed at first and second steps.The new algorithm is realized into software. The data of the Gate of Supreme Harmony in Forbidden City are registered with the new method. The precisions from old method and new method are compared. The result proves that error accumulation is efficiently prevented. The precision is distinctly improved
关键词：Registration; Range Image; Spatial Transformation; Terrestrial LIDAR