Spatio-temporal initialization for IMU to camera registration

Inertia-visual sensor systems become more and more popular in mobile robotics. They allow for global and drift-free localization at high dynamics. Cameras and inertial measurement units (IMUs) are complementary sensors which mutually enhance if correctly fused. However, this complementary nature brings major problems for the IMU to camera registration. Several solutions to compute the spatial alignment are described in literature. In this work, we want to stress the importance of temporal alignment and compare two methods for determining the temporal displacement of sensor measurements. The presented temporal registration can be used as independent preprocessing step without any knowledge about the spatial relation. Further, we present closed-form methods to initialize the angular alignment of the IMU and the camera which can also be applied to setups with gyroscopes only. If high accuracies are required this result can be used to initialize any filter or batch-optimization method to improve convergence and reduce processing time. Simulations and experiments illustrate the presented methods and underline the importance of temporal alignment.