Automatic Registration of Optical Images with Airborne LiDAR Point Cloud in Urban Scenes Based on Line-Point Similarity Invariant and Extended Collinearity Equations

Abstract

This paper proposes a novel method to achieve the automatic registration of optical images and Light Detection and Ranging (LiDAR) points in urban areas. The whole procedure, which adopts a coarse-to-precise registration strategy, can be summarized as follows: Coarse registration is performed through a conventional point-feature-based method. The points needed can be extracted from both datasets through a matured point extractor, such as the Forster operator, followed by the extraction of straight lines. Considering that lines are mainly from building roof edges in urban scenes, and being aware of their inaccuracy when extracted from an irregularly spaced point cloud, an "infinitesimal feature analysis method" fully utilizing LiDAR scanning characteristics is proposed to refine edge lines. Points which are matched between the image and LiDAR data are then applied as guidance to search for matched lines via the line-point similarity invariant. Finally, a transformation function based on extended collinearity equations is applied to achieve precise registration. The experimental results show that the proposed method outperforms the conventional ones in terms of the registration accuracy and automation level.

Coarse-to-Fine Registration of Airborne LiDAR Data and Optical Imagery on Urban Scenes

“…The same contest in 2015 [12], [13] involved extremely high resolution LiDAR data and RGB imagery collected from the same aircraft with the sensors being rigidly fixed to the same platform. In other words, the solutions submitted to these contests, as well as many others [14]- [16], have not intended to cope with the inherent obstacles of the context where datasets are collected from different platforms with different acquisition configuration (i.e., different flying This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/…”

“…According to our literature review, a coarse registration, which is necessary to reposition the two datasets, has not been rigorously studied by existing works. This step is often inadvertently bypassed using the dataset geospatial coordinates provided by a GPS/IMU system [7], [16], [19].…”

Robust Building-Based Registration of Airborne Lidar Data and Optical Imagery on Urban Scenes

“…In other words, the solutions submitted to these contests, as well as many others, e.g. [16]- [18], have not intended to cope with the inherent obstacles of the context where datasets are collected from different platforms with different acquisition configuration (e.g. different flying track, height, orientation, and so on) at different moments and even in different seasons, with different spatial resolutions and levels of detail.…”

“…The versatility of our proposed method is reflected through its capability of registering the datasets that are not acquired simultaneously, nor from the same platform and same acquisition configuration, nor having same spatial resolution. These assumptions are crucial to the existing works [16]- [18]. In this regard, we propose a coarse-to-fine registration approach.…”

“…A smaller distance indicates a more precise registration. The distance proposed in [18] between two line segments p (with A and B are its end-points) and q is defined as follows,…”