Research On Combined Filtering Of Airborne Lidar Point Cloud And Building Feature Extraction

As a new geospatial information acquisition mode,the airborne LiDAR technology has the characteristics of fast and accurate acquisition of three-dimensional spatial data information,and is widely used in many fields such as urban geospatial information extraction.The post-processing technology of point cloud data has become the primary problem that scholars must solve in applied research.In this paper,we takes airborne LiDAR point cloud data as the research object and the post-processing of point cloud as the key technology,focusing on lowland object filtering and construction feature extraction.Some areas in zhengzhou were used as experimental areas to complete the implementation and comparison of relevant algorithms,which have made a qualitative progress in theory and algorithm.The main research contents and results are as follows:(1)A combined filtering algorithm for low-rise ground object filtering is proposed.The primary filtering adopts a progressive encryption triangulation algorithm to filter out high ground objects.Then,for the high difference filtering algorithms for surface fitting,the Least Square(LS)is only applicable to the global surface fitting,while for the local area fitting model,the accuracy is not high and the computational efficiency is low.An innovative method that Simplified Moving Least Squares(SMLS)is proposed for surface fitting,which can effectively solve the fitting smoothness and local fitting problems,and on this basis,the fitting height difference is obtained.Finally,based on the skewness balancing filtering algorithm,the obtained fitting height difference is used instead of the elevation to perform the skewness balancing filtering algorithm to complete the filtering.The experimental results show that the fitting accuracy and computational efficiency have significantly improved by using the SMLS method instead of the LS method.Moreover,the total error of the proposed combined filtering algorithm in the two experimental areas is reduced to 3.42% and 4.99% respectively,indicating that the algorithm has a good effect in urban areas and areas with less terrain fluctuations.(2)A progressive building feature extraction method is proposed.For the building edge points are divided into vegetation points by the traditional echo frequency method,the echo frequency characteristics combined with the standard deviation method of the first and last elevation of the neighborhood are proposed to detect the vegetation points.This method can not only remove part of the vegetation,but also effectively reduce the loss of the building edge points.In addition,on the basis of the above method,the attribute feature information of the point set in the neighborhood set is calculated,and a multi-attribute feature region growth algorithm with the attribute feature curvature as the seed and the normal angle and neighborhood height difference as the judgment criteria is proposed to separate the remaining vegetation points from the building points.Finally,the convex hull algorithm based on local ellipse constraints is used to extract contour points.Experimental results show that compared with the method proposed in literature [41],the accuracy of the proposed progressive method in the two experimental areas is improved by 4.9% and 3.6% respectively,indicating that this method has a good applicability for urban building areas.