Global calibration of non-overlapping cameras: State of the art

Abstract

Multi-camera vision systems (MVS) have a larger field of view than a single camera and are often applied to acquire geometric information of large-scale objects or scenes. Data integration from multiple cameras is a very important part for vision systems which is usually solved by using a global calibration method. After installing the MVS, in some cases, cameras do not have a common field of view (CFV) or only a narrow CFV. Accurate global calibration of cameras with non-overlapping field of view (FOV) is a very challenging task. A variety of global calibration of non-overlapping multi-camera methods (GCNM) have been proposed to estimate the relative positions and orientations of cameras based on different types of media or techniques such as large-range measuring devices, large-scale calibration targets, optical mirrors, motion model, laser projection, visual measuring instruments, etc. However, the GCNM is not yet a completely solved problem. Choosing which type of GCNM method to use is highly dependent on the specific vision system. Thus, in this paper, we present a comparative review of different GCNM methods and analyze accuracy, range, defects, and applications. Researchers and developers can take it as background information for their future works.