A Segment-based Approach For Filtering In Dtm Derivation Of Airborne Lidar Point Cloud

Airborne Light Detecting and Ranging technique (Airborne LiDAR) can directlycarry out high precision and low cost of three-dimensional topographic survey. As anew type of measurement technology, it has been widely used in earth surfacedetection、terrain simulation and reconstruction. However, airborne LiDAR can obtain3D coordinate points which distribute discretely on the surface, these coordinatepoints includes not only the bare ground、ground vegetation and building, but alsoincluding the features of various kinds of unknown nature. How to make use ofLiDAR data to obtain the high accuracy digital terrain model(DTM) is one of the hotissures of current research. Point cloud filtering processing is the key to obtain highaccuracy DTM. Many applications use the data interpolated to filter to extract DTM.Although this method can obtain a relatively smooth surface model, but at the sametime it can make terrain feature information lost and miscalculation.The paper make a research about airborne LiDAR point cloud filtering andclassification, and precision of filtering algorithm. It establishes the reasonableassumption model, split the data into several blocks, makes the data into a series ofpoint set with feature attribute characteristics in order to achieve the filtering effect.The main results are as follows:(1) The paper introduces six kinds of filtering algorithms of scholars at home andabroad, making the comprehensive comparing the advantages and disadvantages. Thefiltering algorithm based on TIN was selected as the reference object.(2) The paper chooses two pieces of test data provided from ISPRS. The twosamples with terrain complexity were chosen respectively. TypeⅠand Ⅱ error issmaller in the segment-based filtering algorithm, building and vegetation extractioneffect is better.(3) Compared segment-based filtering with eight kinds of classic filtering testresults in the four samples, the error which is segment-based algorithm is relativelysmaller. The type Ⅰ error in sample1is lower than other algorithms, the stability oferrors of other samples is relatively weak. It speculates complexity of terraincharacteristics and terrain gradient are the influencing factors of type Ⅱ error.(4) The paper Compares DTM that is interpolated with DTM of the test data, thevolatility of difference curve confirms segment-based filtering algorithm is with high precision in terms of generating DTM.