Study On Semi-automatic Building Extraction Method Based On Airborne LiDAR Points Cloud Data

Cities are areas on the surface of earth where population, economy, technology，infrastructure and information are the most densely distributed. Digital city is graduallybecoming the focus of GIS and informationization, and is of great application value in daily life,municipal administration, social security and many other fields. Buildings are the most primaryelement in cities and the construction of building models in urban areas is of some practicalsignificance. The development of airborne LiDAR measuring technique provides a brand newmeans of obtaining geo-spatial information of high temporal and spatial resolution. AirborneLiDAR data contain large amounts of3-D information about buildings, especially about roofs，which provide many clues for reasoning and great convenience for the3-D reconstruction ofbuildings.In this paper, the development history of airborne LiDAR technique, as well as the status ofresearch and development trend of the hardware and data processing of the later period is mainlyintroduced. Expatiation is given about the status of research on the extracting methods ofairborne LiDAR buildings. Introduction to the composition and work modes, observationprinciples and data features of airborne LiDAR is given. Study is made on the semi-automaticextracting methods of buildings of simple regular by the roof type of buildings. The details are asfollows:(1) Summary of the research on the extraction of airborne LiDAR buildings home andabroad, as well as the introduction to the system composition, work modes, earth observationprinciple and data features.(2) Introduction to the relevant concepts of filtering and classifications of airborne LiDAR,which are also carried out using specialized software; study on the methods of points cloud patchsegmentation with KD tree to the points on the buildings after filtering and classifications;classifications of points on building roofs using methods of connectivity in spatial grid.(3) Detection of edge points on buildings using α-shape; outline abstraction of buildingroofs using edge-tracking, Douglas-Peucker method and least-squares linear fitting method;regularization of the outlines of buildings.(4) Detection of seed points belonging to the same plane on multi-roof buildings such asthe flat, the slope, the gable and other type of roof using3D Hough Transformation; buildingroof extraction using region growing and least-squares fitting plane; extraction and reconstruction of ridges and key points using the method of roof surface intersecting combiningthe information of building outlines.(5) Experiment with TerraSolid training data and other data, which are points cloud patchsegmentation of buildings, extraction of building outline, extraction of simple regular buildingroof surface and reconstruction of some simple regular buildings.