Abstract
Abstract. Registration of point clouds is a necessary step to obtain a complete overview of scanned objects of interest. The majority of the current registration approaches target the general case where a full range of the registration parameters search space is assumed and searched. It is very common in urban objects scanning to have leveled point clouds with small roll and pitch angles and with also a small height differences. For such scenarios the registration search problem can be handled faster to obtain a coarse registration of two point clouds. In this paper, a fully automatic approach is proposed for registration of approximately leveled point clouds. The proposed approach estimates a coarse registration based on three registration parameters and then conducts a fine registration step using iterative closest point approach. The approach has been tested on three data sets of different areas and the achieved registration results validate the significance of the proposed approach.
Highlights
Terrestrial LiDAR scanning of urban objects typically require many scans from different positions and with different orientations
These targets should appear in both point clouds where they are detected and matched to offer corresponding points for the registration such as in (Akca, 2003) and (Wang et al, 2014). (Kang et al, 2009) created panoramic reflectance images form the point clouds and used these images to find the corresponding points needed for registration
The Exterior Orientation Parameters (EOPs) of the images are used to initialize the point clouds registration in (Al‐Manasir and Fraser, 2006). (Pandey et al, 2012) proposed mutual information (MI) based approach to fuse the information acquired by the LiDAR and the images towards an automatic registration
Summary
Terrestrial LiDAR scanning of urban objects typically require many scans from different positions and with different orientations. These multiple scans help to overcome the occlusions encountered by individual scans and offer a complementary view of the object of interest from different directions These multiple scans needs to be referenced to a single coordinate frame to form a complete point cloud of the object of interest. Towards the automation of the registration process, many researches proposed registration algorithms based on special targets to be placed in the scene These targets should appear in both point clouds where they are detected and matched to offer corresponding points for the registration such as in (Akca, 2003) and (Wang et al, 2014). (Al-Durgham et al, 2013) proposed a registration approach that uses the extracted 3D linear features from the point clouds and assess hypothesized corresponding linear features to find the matching needed for registration. To reduce the computation load of matching the points between the point clouds, (Barnea and Filin, 2008) exploited the 3D
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
