| With acceleration of urbanization process and an increase numbers of public building,indoor space structure is gradually complicated,environmental monitoring,emergency rescue,indoor decoration and design and intelligent building all show a strong interest in indoor space.Accurate access and application to three-dimensional spatial data of indoor building for modeling are keys to develop indoor space.Terrestrial laser scanner technology can obtain high-density and high-precision point cloud in a relatively small space,complex structure and indoor space without GPS signal.It has become an efficient method to acquire indoor three-dimensional information and to achieve rapid modeling,which provides a new technical means for indoor building modeling.At present,the main problems faced for terrestrial laser scanning system are the large amount of data,the complex structure,the poor topological structure and the close distance of indoor wall,low recognitions and classifications of different targets.According to the present phenomenon,this paper takes the indoor scene as an example,studies the reconstruction of the indoor model and the indoor plan in the indoor scene.The specific research contents are as follows:(1)The research status of point cloud indoor and outdoor modeling at home and abroad is analyzed,the shortcomings and difficulties of indoor modeling by point cloud are pointed out,and the research idea and route in this paper are planned.(2)Development and classification of terrestrial laser scanner are summarized,according to the characteristics of indoor modeling and indoor point cloud,the collection and preprocessing of point cloud are analyzed.The filtering algorithm of point cloud is summarized based on the reasons for the noise of point cloud.(3)According to the characteristics of the building indoor structure,the indoor ceiling is detected and which is as seed to obtain point cloud of each room,and then point cloud of window and doorway are got.Point cloud of the window and doorway are projected on 2D plane,use the eight neighborhood algorithm to cluster.According to the length,width and height of the minimum bounding box of point cloud,the each doorway point cloud is extracted,and then the each window point cloud is obtained.(4)Established 3D grid based on point cloud.Point cloud is filled into the 3D grid,use the total least squares space plane fitting method to acquire normal vector of each grid point cloud and then to cluster,realize rough segmentation of facade point cloud.The improved K-means clustering algorithm is used to cluster and remove the noise points of rough segmentation facade,which is as seed to obtain the complete facade point cloud,realize fine segmentation of facade.The feature points of each room,doorway and window are obtained by the relationship among the fine segmented facade point cloud,feature points are connected orderly to form indoor model and indoor plan.(5)The effectiveness of algorithm is verified by experiment in this paper,and use proposed method detecte point cloud from indoor top,and extract point cloud of the room,doorway and window.Facade of the room,doorway and window are segmented,indoor model is built and indoor plan is drawn by experiment.The results show that the method proposed in this paper can effectively build indoor 3D model and draw 2D indoor plan. |
PDF Full Text Request