Integration Of LiDAR Point Clouds From Multi Platforms

Along with the social economy development and improvement of science and technology, demands for time-sensitive, large-scale, high-positioning-accuracy spatial data are increasing. As LiDAR has its advantages for high resolution, activity, timeliness, antijamming capability and etc., it has made great progress in its development. Relevant LiDAR technology includes satelite-borne LiDAR, airborne LiDAR, vehicle-borne LiDAR, terrestrial LiDAR and etc.. Properties of different LiDAR systems vary, leading to discrepant data characteristics. With the sustainable development of LiDAR technology, integrated applications of LiDAR point clouds from different platforms are gradually becoming a new trend. The integration of LiDAR point clouds from multi platforms has become a new research hotspot.It’s no easy task to perform the registration of point clouds from different platforms due to the data heterogeneity derived from different platforms (different point of views, different spatial resolutions, different ranges and different targets) and the discreteness of point clouds. Based on the analysis of different LiDAR systems, this research tries to break through the key points for point registration, including feature extraction and feature matching. The ultimate target is to introduce a set of methodologies for the registration of point clouds from multi platforms with high reliability, high positioning accuracy and high automation. Two key methods are introduced in this research, i.e. registration method for airborne and terrestrial LiDAR points, registration method for airborne and vehicle-borne LiDAR points. The research contents are as follows: (1) Registration method for airborne and terrestrial LiDAR points. The corners extracted from airborne LiDAR are relatively low in precision and numerous in anomalous corner points, while the corners extracted from terrestrial LiDAR are relatively high in precision with extremely few anomalous corner points. It is remaining a topic how to make full use of few conjugated corner points with greatly varying precision for precise registration of airborne and terrestrial LiDAR. In order to solve this problem, a shiftable leading point method is introduced in this research, which tries to make full use of all conjugated features. In this method, terrestrial corners are used as referenced data and leading points are generated from airborne corners. An iterative shifting procedure is conducted to shift the leading points from their original positions to their corresponding terrestrial corners. Then the shifted leading points are registered with the terrestrial corners, leading to the final registration of airborne and terrestrial LiDAR. In this method, airborne building corners are extracted from airborne LiDAR points using a reversed iterative mathematic morphological algorithm (RIMM). As for building corners from terrestrial LiDAR, based on the density of projected points method (DoPP), a quantitative estimation method for building contour density is introduced.(2) Registration of airborne and vehicle-borne LiDAR points. Airborne and vehicle-borne LiDAR are usually used for large-range spatial data. If building contours or building corners are directly used for the registration task, the computation work will be intensive and the matching of contours is easily trapped in local optimum. As a result, a new hierarchical method for the automatic registration of airborne and vehicle LiDAR data using3D road networks and building contours is proposed. The cores of this paper lie in the coarse registration method with3D road networks and the fine registration method using3D building contours. During the coarse registration with road networks, road network extraction from airborne LiDAR is emphasized, where intensity and elevation information are used for the extraction. Then the extracted road network is registered with vehicle trajectory lines for the coarse registration result. With coarse transformation matrix as constraints,3D building contours extracted are extracted from vehicle-borne LiDAR points using a high value accumulation method in local areas and then registered with airborne building contours. In the fine registration, the vector-based transformation model and collinearity equation are used.The experiments show that, the shiftable leading point method can make full use of all conjugate features. When faced with few conjugate features, this method can achieve high registration accuracy. In the hierarchical method for integration of airborne and vehicle-borne, the coarse registration method with3D road networks can contribute to a reliable coarse registration effectively, on which basis the fine registration with3D building contours bears high reliability and geometric accuracy.