Abstract
Abstract. An automatic global registration approach for point clouds and remote sensing image based on straight line features is proposed which is insensitive to rotational and scale transformation. First, the building ridge lines and contour lines in point clouds are automatically detected as registration primitives by integrating region growth and topology identification. Second, the collinear condition equation is selected as registration transformation function which is based on rotation matrix described by unit quaternion. The similarity measure is established according to the distance between the corresponding straight line features from point clouds and the image in the same reference coordinate system. Finally, an iterative Hough transform is adopted to simultaneously estimate the parameters and obtain correspondence between registration primitives. Experimental results prove the proposed method is valid and the spectral information is useful for the following classification processing.
Highlights
Since LiDAR data and remote sensing image are obtained by various sensors under different environmental conditions, these two data have distinct imaging mechanisms, data properties and geographical references
In view of the advantages and disadvantages of these abovementioned algorithms, an automatic global registration between airborne LiDAR data and remote sensing image based on straight line features is proposed
Through the transformation of collinear condition equation, the endpoint of straight line feature of LiDAR data should be transformed into the image plane coordinate system and the distance between this endpoint and corresponding line feature in the image is zero, which is defined as the similarity measure
Summary
Since LiDAR data and remote sensing image are obtained by various sensors under different environmental conditions, these two data have distinct imaging mechanisms, data properties and geographical references. Scholars at home and abroad have done a certain amount of research on them, which can be roughly classified into three types of algorithms: The corresponding relationship is established betweeen LiDAR data and the three-dimensional point cloud in the object space, which are genenrated by matching multiple aerial images in the same viewpoint via image mathing method Such methods are essentially a 3D-3D registration transformation. The disadvantage is the eror in the process of interpolation and inaccurate point selection, which can reduce the final registration accuracy These registration methods are implemented by directly using LiDAR data and remote sensing image. In view of the advantages and disadvantages of these abovementioned algorithms, an automatic global registration between airborne LiDAR data and remote sensing image based on straight line features is proposed
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have