UAV Image Matching Point Cloud Filtering And 3D Reconstruction

With the rapid development of "smart city" in China, the demand of city’s 3D model which is an important infrastructure for smart city grow in the city planning, land use, traffic management, social services and other areas. Oblique Photography Technique based on UAV remote sensing platform has the advantage of wide range, high efficiency in 3D modeling and has been gradually popularized. Image matching point cloud are 3D feature point from UAV image, the core idea of the 3D model construction based on Oblique Photography Technique is to build triangular grid model which has high precision and high sense of reality, but the modeling principle of oblique modeling results in the following problems:data of models is too large, large amount of data result in difficult in spatial analysis and failing to achieve the model to be single. In this background, this paper will take the image point cloud data as the object of study, try to solve the problem of how to improve the image matching point cloud accuracy, point cloud filtering and 3D reconstruction based on point cloud information. The research contents can be concluded as follows:(1) Extract the UAV image matching point cloud. Clear the important influence of camera calibration and POS-supported aerial triangulation in improving the accuracy of matching point cloud based on the summary and analysis of image matching point cloud extraction technology. This paper achieves camera calibration and UAV image distortion correction based OpenCV programming, extracts image matching point cloud using Pix4Dmapper software platform and carries out the accuracy evaluation experiment of point cloud data to ensure the accuracy of image matching point cloud.(2) Image matching point cloud filtering. Analyze the problem of morphological filtering algorithm based on regular grid and optimize the algorithm using the virtual grid to make the algorithm solve the problem of the original point cloud accuracy loss in regular grid. Finally analyze the filtering accuracy of the improved algorithm by filtering experiment.(3) The reconstruction of 3D building based on parametric 3D modeling technology. In order to obtain the basic parameters of building model, this paper first extracts the contour line of the building and building point by Canny operator and convert the contour line to vector data, then combines the vector data with the nonground points data to measure the building point including building height, roof type and other information about building size. At last, this paper realizes parametric modeling based on parameter.