The Building Extraction And Parametric 3D Reconstruction Based On Oblique Image

In the processing of rapid development of "digital city" construction, the importance of three-dimensional city models have become more and more prominent, as the important basic data of the "digital city" construction, the demand for urban planning, social services,environmental protection and other industries is increasing. Therefore, research on building extraction and parametric three-dimensional reconstruction based on the oblique image have a special value of application.The oblique photogrammetry technology based on the UAV platform can high effictively real-time obtain three-dimensional city models for large area, the model has high accuracy and realistic, and can well reflect the real situation of features, and overcome the distortion and missing of a large number of urban information in the traditional three-dimensional modeling method, this technology has been gradually applied to various fields. But this modeling principle of the oblique photogrammetry technology carry out triangular mesh model construction based on ultra high density point cloud data, resulting in some problems of the model. One of the keys problems is that the features aren't separated(monomer) in the model,which cannot meet the further management and application of geographic information system.Therefore, in order to solve the building separation problem of the model, this paper uses the image matching point cloud data as the research object and aimed at how to solve the filter processing of the image matching point cloud data, and the building edge extraction, the parametric three-dimensional reconstruction and other problems made a thorough research.The concrete work includes:(1) The filtering processing of the image matching point cloud data. This part focuses on the research of the mathematical morphology filtering algorithm, and aimed at the detects of the traditional morphological filtering algorithm, which include the use of regular gridorganization to point cloud data and window size by manual setting, improve the algorithm by the virtual grid technology and automatically determine the iteration number based on the ratio of the ground point cloud; the improved algorithm based on the traditional mathematical morphology filtering algorithm, which overcomes the problems of the original point cloud information loss result of regular grid resample and the bad self adaptability of the filtering window; and carry out the quantitative analysis of the traditional mathematical morphology filtering algorithm and the improved algorithm, the experimental results show that the improved morphological filtering algorithm can effectively improve the accuracy of filtering results.(2) The building edge extraction. This part focuses on the analysis of the Canny edge detection algorithm, and aimed at the defects of the traditional Canny algorithm, which includ the use of the Gauss filter parameters and threshold value by manual setting, only considering the gradient x and y direction of the gradient calculation, improve the algorithm by using adaptive median filtering, adding the gradient of the tilt direction(45 degree and 135 degree),and Ostu algorithm; the improved algorithm based on the traditional Canny algorithm, which solve the problems of the tilt direction edge information loss, and the bad self adaptability of Gauss filter parameters and high and low threshold; and then combine with the Hough transform, and carry out the quantitative analysis of the traditional and improved Canny algorithm, the experimental results show that the improved Canny algorithm can effectively improve the detection effect.(3) The three dimensional reconstruction based on parametric technique. This part focuses on the analysis of the acquisition method of the basic data of the parametric three-dimensional reconstruction, mainly based on the point cloud data filtering processing and the building edge extraction results, and combine the non ground point cloud data with building contour to complete the extraction and measurement of building point cloud, obtain the modeling parameter information, which include the type and height of buildings, and the type and height of roof; finally complete the building rule file preparation by combining with the UAV image texture information, so as to realize the 3D reconstruction of the building.