In their work "Global Optimization through Rotation Space Search", Richard Hartley and Fredrik Kahl introduce a global optimization strategy for problems in geometric computer vision, based on rotation space search using a branch-and-bound algorithm. In its core, Lemma 2 of their publication is the important foundation for a class of global optimization algorithms, which is adopted over a wide range of problems in subsequent publications. This lemma relates a metric on rotations represented by rotation matrices with a metric on rotations in axis-angle representation. This work focuses on a proof for this relationship, which is based on Rodrigues' Rotation Theorem for the composition of rotations in axis-angle representation.

Vision-based driver assistance systems have great potential for preventing fatalities. This work addresses the problem of 3D monocular vehicle tracking and turn rate estimation in situations where vehicles need to be tracked along intersections and curves. To estimate the tracked vehicle's turn rate, an approach based on image feature correspondences and a simplified geometric vehicle model is used. The model is robustly and efficiently fitted to the matched image features using an improved RANSAC scheme that automatically enforces physically plausible vehicle motions and speeds up the overall system at the same time. Temporal integration of the computed turn rates is performed by an Extended Kalman Filter with the bicycle motion model. Experiments with real world data show the applicability and robustness of the proposed concepts.

This paper introduces globally optimal hand-eye self-calibration of camera orientation under a wide range of cost-functions, which includes L2 and Blake-Zisserman. The main contributions are new feasibility tests to integrate this hand-eye calibration problem into a branch-and-bound parameter space search. The presented method constitutes the first guaranteed globally optimal estimator, which is not limited to L∞-type cost-functions. The algorithm directly minimizes the desired cost-function based on reprojection errors of interest point correspondences and does not depend on any structure and motion preprocessing to estimate camera poses. The system is evaluated in both synthetic and real world scenarios. Typical cost-functions are exemplarily integrated to validate the algorithm's flexibility.

This paper introduces simultaneous globally optimal hand-eye self-calibration in both its rotational and translational components. The main contributions are new feasibility tests to integrate the hand-eye calibration problem into a branch-and-bound parameter space search. The presented method constitutes the first guaranteed globally optimal estimator for simultaneous optimization of both components with respect to a cost function based on reprojection errors. The system is evaluated in both synthetic and real world scenarios. The employed benchmark dataset is published online1 to create a common point of reference for evaluation of hand-eye self-calibration algorithms.

This paper introduces simultaneous global optimization of both camera orientation and vehicle wheel circumference without requiring any information about the translations in the system. The main contribution are new objective function bounds to integrate this problem into a branch-and-bound parameter space search. The presented method constitutes the first guaranteed globally optimal estimator for both components of the problem with respect to a cost function based on reprojection errors. The algorithm operates directly on image measurements and does not depend on any structure and motion preprocessing to estimate camera poses. The complete system is implemented and validated on both synthetic and real automotive datasets.

This paper introduces robust globally optimal hand-eye self-calibration of camera orientation for the automotive domain. The main contribution are new feasibility problems to integrate this problem into a branch-and-bound parameter space search. The algorithm constitutes the first guaranteed globally optimal maximizer for the support of all three orientation parameters with respect to an a priori defined threshold of reprojection errors. The algorithm operates directly on interest point correspondences and does not depend on any structure and motion preprocessing to estimate camera poses. The complete system is implemented and validated on both synthetic and real automotive datasets.

This paper introduces an approach to extrinsic autocalibration of vehicle mounted cameras for Bird's Eye View applications. We extend existing self-calibration methods by estimation of two position parameters while assuming the camera height to be known. The presented algorithm is derived from classic hand-eye calibration and uses independent temporal feature correspondences to introduce constraints on the camera position. An error measure of the constraints' degree of satisfaction is used to formulate the parameter search as a quadratic optimization problem. Temporal accumulation of correspondences over time improves robustness to potentially sparse interest points and pitch or roll movements of the vehicle.

When humans return to the moon, Astronauts will be accompanied by robotic helpers. Enabling robots to safely operate near astronauts on the lunar surface has the potential to significantly improve the efficiency of crew surface operations. Safely operating robots in close proximity to astronauts on the lunar surface requires reactive obstacle avoidance capabilities not available on existing planetary robots. In this paper we present work on safe, reactive navigation using a stereo based high-speed terrain analysis and obstacle avoidance system. Advances in the design of the algorithms allow it to run terrain analysis and obstacle avoidance algorithms at full frame rate (30Hz) on off the shelf hardware. The results of this analysis are fed into a fast, reactive path selection module, enforcing the safety of the chosen actions. The key components of the system are discussed and test results are presented.

The Ulm Sparrows RoboCup team has successfully competed now for eight years in the RoboCup tournaments. It is mainly a student-driven team and serves both as a research project as well as a teaching tool at the University of Ulm. We try to develop solutions not only applicable in the RoboCup scenario but in other robotic projects too. Three main contributions are presented within this paper: our middleware for mobile robots Miro including its framework based building blocks for common robotic tasks, our visual object detection system and our current and ongoing work in the field of communication and interaction in teams of heterogeneous robots.