Research On Building Facade Recognition And Extraction Based On Terrestrial LiDAR Dat

The traditional measurement has the characteristics of being susceptible to the topography,weather conditions,etc.,and the labor intensity is high,and the data processing efficiency is low,which makes the traditional measurement technology products more and more unable to meet the demand.Terrestrial Laser Scanning technology emerges as the times require.It is a fast and efficient measurement technology for obtaining the target object of measurement.It has the characteristics of automatic,non-contact,high precision,etc.,and it is applied in “smart city” and “digital city”.It has become more and more mature.However,because the point cloud data acquired by the terrestrial laser scanner is massively disordered and unstructured,how to deal with massive,diverse and unstructured 3D point clouds is a challenging problem in point cloud research.Based on the research of point cloud data by domestic and foreign scholars,this paper focuses on the preprocessing and model construction of TLS data,focusing on the filtering process after data acquisition,building facade extraction,and building facade surface detail recognition.Research and made some progress.The main research contents are as follows:(1)This paper has explained the working principle of the ground laser scanner,and has already introduced the composition of the RIEGL VZ-1000 and the main flow of data acquisition.The characteristics of TLS data have been summarized and summarized.The characteristics of point cloud data of several typical features have been highlighted,including spatial distribution features and geometric features,in preparation for the subsequent experimental process.(2)This paper has discussed the existing classical point cloud filtering algorithms,including mathematical morphology filtering algorithm,slope-based filtering algorithm and filtering algorithm in PCL point cloud method library.The deficiencies and limitations of the existing filtering algorithms have been summarized and analyzed.Based on this,a filtering algorithm based on two-dimensional gamma distribution has been proposed.The number of different neighborhoods is selected,and the neighborhood mean and slope are set.Constraints separate noise points.Experimental comparison results have shown that the two-dimensional joint Gamma distribution is used to filter out noise points while retaining the subject details.(3)Building facade extraction based on RANSAC algorithm.The existing fa?ade information extraction algorithm has been deeply studied,and the advantages and disadvantages of its existing algorithms have been analyzed.On this basis,the RANSAC algorithm has been proposed for experiments,and the building facade data has been extracted.The effectiveness of this algorithm has been verified by experiments.(4)Structural feature recognition of building facade based on topographic feature lines.The detailed characteristics of the building facade data have been deeply analyzed.The semantic features of the facade details have been summarized.It has been proposed that after the elevation data is rotated and projected,the terrain feature lines of the generated TIN triangle network have been identified.Point cloud data with different elevation details and reconstruction of the geometric framework model.This article is innovative:(1)Through the in-depth study of point cloud data filtering,a filtering algorithm based on two-dimensional joint gamma distribution has been proposed,which has realized the denoising of disordered point cloud.Experimental results have shown that this algorithm can reduce the subjective setting of the threshold and the fitting effect is already good.(2)Through the research on the feature recognition of point cloud data fa?ade,an algorithm based on the semantic information of the fa?ade details has been proposed,and the point cloud data with different details has been identified in combination with the feature of the ground line.