Abstract
Rivers are managed by preparing materials including various drawings and registers. In particular, cross sections are very important materials because they are used for not only grasping the current conditions of levees but also river improvement plans and so on. However, since cross sections are surveyed and made into drawings by human operation, they are created only on the distant posts at 200m intervals due to limitations of survey costs. Currently, therefore, a technology for generating cross sections from LP data acquired by airborne laser surveying is needed. However, there are some problems for LP data ; that they contain noises of vegetation and so on which disturb ground surface survey, and that it is impossible to measure changing points of cross sections due to the problem of point cloud density. Existing researches on dealing with the problems for LP data include a method of extracting point cloud data of the ground surface by obtaining the lowest point at every specific range from LP data, and a method of estimating break lines from LP data. However, the former has "a problem that the accuracy of point cloud data is dependent on the grid width" and the latter "a problem that it makes an error in extraction of a break line at bends". To solve these two problems, this research proposes a method for noise reduction which is independent of grid width, and a method for estimating break lines properly at bends as well. In order to prove usefulness of the proposed methods, evaluation experiments were conducted to compare the cross sections of proposed methods with those of existing methods, and evaluated them.
