An automatic method for precise 3D registration of high resolution satellite images and Airborne LiDAR Data. Issue 24 (17th December 2019)

• Record Type: Journal Article
• Authors: Safdarinezhad, Alireza; Mokhtarzade, Mehdi; Valadan Zoej, Mohammad Javad
• Is Part Of: International journal of remote sensing. Volume 40:Issue 24(2019)
• Page Start: 9460
• Page End: 9483
• Publication Date: 2019-12-17
• Subjects: Remote sensing -- Periodicals; Télédétection -- Périodiques; 621.3678
• DOI: 10.1080/01431161.2019.1633698
• Languages: English
• ISSNs: 0143-1161

Abstract

ABSTRACT: Precise 3D registration of Light Detection and Ranging (LiDAR) data and High-Resolution Satellite Image (HRSI) is the prerequisite in the many remote sensing applications. An automatic registration process involves two main steps including the detection of corresponding entities and the estimation of a relating mathematical model. Typical matching techniques, which are generally developed to match consistent data sources, are prone to fail in case of matching between heterogeneous data sources (e.g. LiDAR and optical images). This paper proposes a three-step method to give in hand an accurate and automatic 3D registration technique between LiDAR data and the HRSI. The first step introduces a new product called Optical Consistent LiDAR Product (OCLP) which is meant to be consistent with HRSI from the radiometric point of view. The OCLP is generated using raster maps of the LiDAR heights and intensities along with information of the sun position at the acquisition time of HRSI. This new product proved to be very promising as a matching entity with HRSI. In the second step, a 3D model is estimated robustly through the matched points identified by the well-known Scale Invariant Feature Transform (SIFT) technique between the OCLP and the HRSI. The last step of the proposed method aims to strengthen this 3D model which is accomplished via iterative closest edge points matching technique. To do so, the coarse 3D model is iteratively improved based on the matching results obtained on the image edges. The proposed method was implemented on 20 different datasets containing various urban textures. The results indicate the effectiveness of the proposed method since in some cases it could achieve to sub-pixel accuracies which are the utmost expectation for a registration technique.