Study On Data Filtering And Buildings Abstraction Of Airborne LiDAR Points Cloud Data

With the fast development of GPS, photogrammtry and remote sensing technology,geospatial information science applications field has increased the demand on obtainingaccurate, real-time and reliable data. The emerging Light Detection and Ranging(LiDAR)technology has broken the basic framework of photogrammtry. It is a new earth observationtechnology which is an integration of laser ranging, GPS spacetrack, INS, computer and othertechnologies. Airborne LiDAR technology is not limited by sunshine and weather condition, ithas the characteristics of all-weather, rapidity and accuracy. It has greatly facilitated inacquiring geo-spatial information with a high spatial-temporal resolution. As a new surveyingmeans, airborne LiDAR has its unique technical advantages, it makes traditional single-pointpositioning data acquisition become constant automatical data acquisition, it increases theaccuracy and speed of observation, and it can directly acquire high-accurate3D coordinate ofground objects in flying area, it also can accurately describe the fluctuation of terrain, edge ofroads, canopy of plants and complex structure of buildings. It has a wide application prospectin digital earth, smarter earth, construction of digital city and other fields.Points cloud filtering and buildings extraction are the most critical steps in AirborneLiDAR data processing. Original points cloud data is a set of rambling points, they need to beprocessed in a proper way. Extract surface and objects information accurately from masspoints cloud can provide more effective information for achieving roads management, urbanplanning and other themes. At present, the study of theory and method on points cloudfiltering and buildings extraction technology has been top topic focused by many researchersat home and abroad. This paper takes LiDAR points cloud as a basis. Without other auxiliarydata, we focus on the discussion of transformation between different formats and3D displayof the LiDAR points cloud, interpolating data to produce DSM depth images and distanceimages through setting different interpolation, filitering the depth images, and automaticallyextracting buildings on distance images and filtered depth image. Meanwhile, make specificverity and comparison through experiments, then draw a conclusion.Main missions and research emphasis are as follows: 1. The development and research condition of Airborne LiDAR system and dataprocessing technology are reviewed. Composition of LiDAR system and its operatingprinciple are introduced. Comparing the principle of LiDAR technology with the principle oftraditional aerial photogrammetry and the InSAR technology. Meanwhile, the characteristic ofLiDAR points cloud data, errors and processing flow are analyzed in detail. This paper alsomakes a contrast on principle of existing several data filtering algorithms, sketches theself-adaptive filters, then carry through a summary.2. Take ENVI and ArcGIS as main operating platforms, load a plug-in what can be usedto process LiDAR data in ENVI, implement direct transformation between different formatsof points cloud data, then display in3D mode with three kinds of expressions. Importing thetransformed points cloud data to ArcGIS, taking intensity and elevation respectively asinterpolation to produce DSM depth images and distance images, and implementing filteringprocessing on DSM depth images. Make experiments through four kinds of self-adaptivefilters, then compare and analyze the results.3. Design a processing flow based on the matlab, make a binarization on filtered DSMdepth images, and make morphological opening operation what does erosion first, dilationcomes second. It can implement automatical extraction of some buildings. Meanwhile, takethe DSM distance images to make a binarization and opening operation, it can also implementautomatical buildings extraction what is based on DSM depth images. Finally, make a contraston compare and analysis on two different experiment results and planform of original pointscloud, discuss existing drawbacks, then draw a conclusion.