Research On Filtering Terrestrial Laser Scanning Data

The application of laser scanning in the field of surveying and mapping becomes more andmore widely. Three-dimensional laser scanning has the advantage of speediness, high accuracyand good concealment when used in topographic survey. This article aims at the key step ofeliminating the off-terrain points form terrain point clouds, and the main research is focused onthe following aspects:1. Mainly introduces the constitution and operating principle, scanning approaches and dataprocessing steps.2. After a detailed account of trend surface, accounting for point clouds usually is a largeamount and processing is often inefficiency, filtering is processed in girding. Consideringoff-terrain points such as points on vegetation as gross error, robust estimate is adopted calledmoving robust trend surface. Experiments show that this arithmetic can get a better result.3. Mathematic morphology and its application are recommended, processing in regulargrids instead of in single points considering its efficiency, considering mathematic morphologywindow size which is difficult to set makes a large difference to filter result, a method that awindow size parameter is added to each gird. A method called “shadow search” is proposed tosearch out surface features position and size after an in-depth analysis terrain points’characteristic, realization adaptive filter, and experiments show that its filter precision is higherthan fixed window size method.4. After go into the space distributing difference between terrestrial laser scanning andairborne laser scanning point cloud, sector division of terrain point cloud is proposed in thispaper according to its space distributing presents a characteristic that dense in the center andsparse in the out part of data. Sector parameters are derived and neighborhood point search ispresented. The result of test shows that the lowest method of sector simplification can reducepoint number and filter out the off-terrain point at the same time.The applicability of the tree filter algorithms are presented at last, their sensitivity to data isanalyzed, and their combining use means are given.