Registration Of Airborne LiDAR Point Cloud And Aerial Images Using Multi-Features

Aerial photogrammetry has always been the mostly used method for the acquiring of large-area and high-accurate digital terrain data in traditional survey and mapping. However, the emergence of airborne Light Detecting And Ranging (LiDAR) changes this situation. With the ability of fast acquisition of large-area and dense point cloud data, the LiDAR system simplifies the workflow of terrain survey to a very great extent. As two different survey methods, the aerial photogrammetry and airborne LiDAR actually are complementary to each other. The integration of aerial photogrammetry and airborne LiDAR plays an important role in othroimage generation and real texture three dimensional city modeling. Due to the existence of some systematic errors and random errors, the acquired LiDAR points and the orientation parameters of aerial images are error-contaminated. The directly acquired LiDAR points and aerial images can not be registrated perfectly, which is the key problem that blocks the integration of these two data. Up to now, there are various registration methods for airborne LiDAR points and aerial images. However, the need for a good solution to the registration of large-area urban airborne LiDAR points and aerial images is still urgent. Therefore, the research of the solution to the registration of large-area urban airborne LiDAR points and aerial images is meaningful both in theory and practice.With the purpose of researching the registration of large-area urban airborne LiDAR points and aerial images, this paper gives a feasible solution. The main works conducted in this paper are as follows:1) An innovative registration workflow is proposed, i.e. using the aero-triangulated aerial images as the reference data during the registration of airborne LiDAR points and aerial images. The existed registration method all use LiDAR points as reference data, which will be problematic when the LiDAR points are error-contaminated. The workflow proposed in this paper is as follows. Firstly, the aerial images are aero-triangulated with the assistance of photogrammetric ground control points, POS (Positioning and Orientation System) data and LiDAR points. Secondly, conjugate features such as conjugate point features and conjugate building corner features are matched between the aero-triangulated aerial images and LiDAR points. Thirdly, LiDAR strip adjustment is conducted to correct the coordinates of the LiDAR points. During the LiDAR strip adjustment, the three dimensional coordinates of these conjugate features are intersected using the aero-triangulated images and will serve as control points. Finally, the LiDAR points are aligned with aerial images and the ground control data.2) A method for automatic matching of tie points on aerial images assisted by POS data and LiDAR points is proposed. The main works are as follows.①The framework of object space constrained matching of aerial image tie points is established. By automaticly determine the searching area and correct the distortion of matching window, the matching efficiency, accuracy and reliability can be increased.②With the purpose of speeding up the searching procedure in normalized grey corss correlation matching, a fast searching method is proposed. About a25%reduction of the matching time as to the traditional method is reported in the experiment.③During the object space constrained image tie points matching, the accuracy of the coarse exterior orientation parameters of aerial images is the critical factor to the efficiency and reliability of the matching results. When the accuracy of the POS data is high, the mounting angular errors come to be the main factors that determine the accuracy of coarse exterior orientation parameters. Therefore, this paper proposed a mounting angles calibration method using virtual ground control points, which can improve the accuracy of coarse exterior orientation parameters without true ground control points.④In building serried areas, the matching of tie points on buildings is very important to the reliability of photogrammetry bundle block adjustment. So a building corner points matching method is proposed in this paper, which utilizes both edge information and texture information of building.3) A method for the matching of conjugate points and conjugate lines between airborne LiDAR points and aerial images is developed. The main works are as follows.①Matching conjugate points between LiDAR intensity images and down-sampled aerial images according to the object space constrained matching method.②A TIN (Triangulated Irregular Network) based building contours extraction method is proposed for extracting building contours in irregularly spaced LiDAR points. After the noise and wall points are excluded, the TIN of the LiDAR points are built. Building contours are tracked and extracted in the TIN. The extracted building contours are with good horizontal and vertical accuracy, because it avoid the uncertainty caused by the resampling of LiDAR points. The extracted building contours are then regularized and building corner features are obtaind and will be used as registration primitives. Conjugate building corner features are matched between airborne LiDAR points and aerial images, which provide conjugate building corner points between LiDAR points and aerial images.4) A LiDAR strip adjustment based registration of aerial images and LiDAR points is proposed. The LiDAR strips, aerial images and ground control points are aligned after the LiDAR strip adjustment using conjugate features matched between aero-triangulated images and LiDAR strips. A sensor system driven LiDAR strip adjustment model is proposed. This method takes the POS system errors into consideration, which is very practicle. The airborne LiDAR strips usually have big coordinate errors due to the POS systematic errors and the sensor mounting errors. If the magnitudes of errors are very large, the LiDAR strip distortion can not be modeled and corrected using linear models. In this paper a rigorious LiDAR strip adjustment method is developed based on the geo-location equation of LiDAR system. The sensor mounting errors and POS systematic errors are modeled by second order polynomials. The LiDAR strip adjustment is conducted using the features matched between airborne LiDAR points and aerial images as weighted control points.In order to verify the effectiveness of the proposed methods, real datasets are used to test the key methods in the registration workflow. At the end of this paper, the whole workflow is tesed using large-area urban airborne LiDAR points and aerial images and the results show the effectiveness of the proposed method.