The Change Detection Of Buildings Based On Terrestrial Laser Scanning Data

In recent years, with the 3-D geographical information and virtual reality technology in depth, the traditional measurement method can not fully meet industry needs because of its shortage on efficiency and data accuracy. The emergence of 3-D laser scanning technology to make up for the lack of traditional measurement, which provides high-resolution point cloud data can record all the details of complex objects. It is an important application is that to monitor changes over time, such as updating geographic information, comparing building structural affected by natural disasters before and after changes, restoring and conserving ancient buildings. All of them have the need for effective and timely processing means. Of course, in addition to the big scene, the terrestrial laser scanner is also applied to small-scale change detection, even detecting the facial expression. Laser scanner take the advantage of high speed, good precision, ability to work around the clock, etc. It has been brought on gradually by many colleges and work units of surveying and mapping into the application of science and engineering.In this paper, data source came from the terrestrial laser scanner point clouds. Each data station's coordinate system was unified by Scale Invariant Feature Transform (SIFT) algorithm. This paper proposed a method of extracting building point cloud based on the statistical histogram of adjacent angle difference. The extraction building point cloud of the mobile scanner (car) data proposed a method based on the histograms of projection point density. Both of them finally used scanning strip method to extract buildings point cloud. The change detection of buildings used Hausdorff distance algorithm to realize auto-detection of two point cloud data. And then we improved the original algorithm by using 2-D image coordinates of the search window algorithm, and it was able to identify the pseudo-change area. The quantitative for the change of building's model used surface area as a indicator. We calculate the change in surface area and total building surface area by constructing Delaunay triangulations and excluding concaves, and finally get the damage rate of the buildings.The study results showed that the use of the scale invariant feature transform (SIFT) algorithm achieved a very good point cloud automatic registration, high precision and speed; The extraction building point cloud proposed in the paper obtained superior results. Hausdorff distance algorithm had a low detection efficiency when detecting large number of point clouds. Then we improved the method based on the 2-D image coordinates search window to raise the detection efficiency, and calculated each changed region and total building's areas and boundaries by the steps of segmentation, plane fitting, building triangulation, holes exclusion (windows,doors). The damage condition of buildings can be used for emergency response and disaster needs of insurance claims. The paper verified high accuracy of the algorithm by comparing the change detection of the actual and theoretical boundaries.