| In recent years,aerial photogrammetry technology has become one of the important means of data acquisition in digital city construction in the new era with low cost and wide range of high-precision scene information.UAV platform is equipped with GPS IMU and non-measuring camera to collect high-precision positionThe ground image data and attitude data can be used to generate digital orthophoto image digital surface model and dense point cloud data by using the Motion recovery target Structure algorithm(SFM)and multi-view Stereo vision reconstruction algorithm(MVS)of computer vision,In the aspect of photogrammetry,most of the applications are based on the digital surface model of digital orthophoto image and the real three-dimensional model.Point cloud,as the transitional product of image and basic geographic information products,is often ignored,so that the application and research of image dense point cloud are relatively few.However,related research and application of LIDAR point cloud are relatively mature,although there are differences in data sources and compatibility of filtering algorithms between image matching point cloud and LIDAR point cloud.However,in general,the point clouds obtained by the two methods have something in common,so this study mainly uses the processing method of LIDAR point cloud for image matching point cloud processing,and makes use of the rich texture and spectral information of the image matching point cloud to conduct automatic classification and visual classification editing.The process of dividing ground points and non-ground points in point cloud data is called point cloud filtering.At present,all kinds of classical point cloud filtering algorithms need to set more complex parameters in the algorithm,and professionals need to have a deep understanding of the operation area,so as to achieve a certain filtering effectDue to the existence of uncontrollable factors such as natural and human factors,the error of image matching point cloud is inevitable in the process of data collection and processing.Therefore,an appropriate gross error elimination algorithm of point cloud should be adopted to remove the noise points in point cloud before point cloud filtering.The research contents are as follows:(1)From the camera imaging model,2 d photos to the three-dimensional space coordinate transformation,camera calibration,feature point extraction and matching,restore structure movement,more visual image dense matching six aspects about Unmanned Aerial Vehicle(UAV)tilt image matching for principle and the key technology of 3 d point cloud,and organize the uav image data acquisition and image processing involves cloud of specific methods.(2)Filtering and classification are the basic links of point cloud application.In order to carry out high-precision filtering and classification of image matching point cloud,the characteristics of point cloud are fully analyzed before the beginning of filtering,and the error of point cloud is eliminated by using elevation statistics and dension-based algorithm;For filtering,several current classical point cloud filtering algorithms are studied in depth.Qualitative and quantitative evaluation methods are used to analyze the filtering results of each filtering algorithm in different ground features and geomorphic regions,so as to find the best algorithm suitable for this study.Classification is to conduct feature statistics on the point cloud first,select representative features and use machine learning random forest algorithm for classification.(3)Combine the acquired digital orthophoto image with the filtered ground point data to make the digital line plot.Firstly,the point cloud data with high density is diluted to generate the digital elevation model and contour data.Then the generated contour data is combined with the vectorized part of the ground objects in the digital orthophoto map,and the digital line map is made after partial modification.Finally,check points are selected for accuracy analysis. |
PDF Full Text Request