Boresight Calibration of the Aerial Multi-Head Camera System Young-Jin Lee * , Alper Yilmaz Photogrammetric Computer Vision Lab., The Ohio State University, Columbus, OH, USA 43210 ABSTRACT This paper introduces a novel geometric constraint to boresight calibration for aerial multi-head camera systems. Using the precise EOPs (exterior orientation parameters) estimated for each physical camera and the surface information of the area of interest, multi head camera provides a synthetic image at each time epoch. The camera EOPs can be computed directly from the navigation solution provided by an onboard GPS/INS system and camera platform geometric calibration parameters, which represent the geometric relationship between the camera heads. For direct acquisition of EOPs from the navigation system, the camera frame and the INS frame should be precisely aligned. Boresight can be defined as mounting angles between the INS frame and camera frame. Since small but unknown misalignment angles could cause large errors on the ground, which suggests that they should be precisely estimated. In this paper, unknown boresight angles are estimated by using camera platform geometric calibration parameters as constraints. Since each physical camera of the multi-head camera system is tightly affixed to the platform, the geometry between camera frames remains constant. Simulation results show that the constrained method provides better estimation in terms of both accuracy and precision compared to traditional approach which does not use the constraint. Keywords: boresight, multi-head camera, direct georeferencing, constrained bundle adjustment 1. INTRODUCTION An aerial multi-head digital camera system uses multiple digital cameras with different viewing angles in order to maximize ground coverage. Due to technological and economical reasons, a single digital camera cannot substitute a film-based aerial mapping camera [12]. Several commercial aerial mapping digital camera systems have been introduced, which use multiple digital cameras to overcome these problems. Aerial multi-head camera systems provide a single synthetic image from a set of images acquired from each physical camera. The synthetic image can be generated with precisely estimated exterior orientation parameters (EOP) of images acquired simultaneously from each physical camera, and surface information of the target area. The EOP of each image can be directly calculated from a navigation solution and camera platform geometric calibration parameters that represent the geometric relationship between the camera heads. To acquire precise EOPs of the images directly, the coordinate frame of the navigation system and that of the master camera (or camera system) should be precisely aligned. There is positional displacement between the INS (inertial navigation system) frame and camera frame. These offset components are measured before a flight mission and considered during navigation data processing [2], [11]. However, measuring small rotation angles between axes of the two frames is more difficult because axes of the two frames cannot be observed physically [11]. The boresight can be defined as the mounting angles between the two frames [10]. Small but unknown misalignment angles could cause large errors on the ground, which suggests that they should be precisely estimated. In [5], the authors introduced an automatic on-the-fly boresight estimation method without ground control points (GCP) for direct georeferencing of an aerial three-head camera system. They estimated all nine boresight angles (three angles per camera) which are a combination of three boresight angles of the master camera and parts of two sets of camera platform geometric platform calibration parameters of two slave cameras. Reference [3] introduced an in-flight calibration method, which includes boresight misalignment angles for direct georeferencing of a RC20 camera system, and investigated correlations between the parameters to improve the quality of the direct georeferencing results. In [4], the authors mentioned the importance of boresight misalignment calibration and suggested flight planning for calibration. Reference [1] introduced a new boresight misalignment angle calibration method using combinations of transformation matrices between coordinate frames. * lee.3043@osu.edu; phone 1 614 247-6699