Eike Rehder

Goal-Directed Pedestrian Prediction

Recent advances in road safety have lead to a constant decline of injured traffic participants in Europe per year. Still, the number of injured pedestrians remains nearly constant. As a countermeasure, active pedestrian safety is the focus of current research, for which accurate pedestrian prediction is a prerequisite. In this scope, we propose a method for dynamics-and environment-based pedestrian prediction. We introduce the pedestrians destination as a latent variable and thus convert the prediction problem into a planning problem. The planning is executed based on the current dynamics of the pedestrian. The distribution over the destinations is modeled using a Particle Filter. Experimental results show a significant improvement over existing approaches such as Kalman Filters.

Head Detection and Orientation Estimation for Pedestrian Safety

Head detection and orientation estimation are a vital component in the intention recognition of pedestrians. In this paper we propose a novel framework to detect highly occluded pedestrians and estimate their head orientation. Detection is performed for pedestrians heads only. For this we employ a part-based classifier with HOG/SVM combinations. Head orientations are estimated using discrete orientation classifiers and LBP features. Results are improved by leveraging orientation estimation for head and torso as well as motion information. The orientation estimation is integrated over time using a Hidden Markov Model. From the discrete model we obtain a contiunous head orientation. We evaluate our approach on image sequences with ground truth orientation measurements. To our best knowledge we outperform state of the art results.

Detection of ascending stairs using stereo vision

Environment perception is an important task in computer vision for many applications in robotics. Especially for robots navigating through different levels of a building, stair detection constitutes an important perception task. In this paper, we propose a stair detection algorithm using range data. Firstly, we introduce a parameter, which describes local surface orientations w.r.t. a global reference. Secondly, a matched filter is used to detect relevant edges in the orientation data. Afterwards, line segments are determined using these edge data which are further used to estimate stairs. The proposed method is invariant against rotations of the sensor. We show that the system can handle typical outdoor stair types and outperforms the accuracy of state-of-the-art stair detection methods. Moreover, the method is used in real time to assist visually impaired people who wear the camera system on a helmet.

Submap-based SLAM for road markings

Coherent road maps are a prerequisite for autonomous navigation. In case of an unknown environment, grid map and SLAM techniques are widely used. This paper takes a novel approach to vision based mapping of road markings by registration of local occupancy gridmaps for map stitching. We show that with reasonably accurate ego motion measurements, seamless global maps can be constructed from local grid maps. The approach is evaluated on real world data obtained from an autonomous model racing car.

Analysis of Regionlets for Pedestrian Detection.

Human detection in camera images is an important task for many autonomous robots as well as automated driving systems. The Regionlets detector was one of the best-performing approaches for pedestrian detection on the KITTI dataset when we started this work in 2015. We analysed the Regionlets detector and its performance. This paper discusses the improvements in accuracy that were achieved by the different ideas of the Regionlets detector. It also analyses what the boosting algorithm learns and how this relates to the expectations. We found that the random generation of regionlet configurations can be replaced by a regular grid of regionlets. Doing so reduces the dimensionality of the feature space drastically but does not decrease detection performance. This translates into a decrease in memory consumption and computing time during training.

Guided depth upsampling for precise mapping of urban environments

We present an improved model for MRF-based depth upsampling, guided by image-as well as 3D surface normal features. By exploiting the underlying camera model we define a novel regularization term that implicitly evaluates the planarity of arbitrary oriented surfaces. Our method improves upsampling quality in scenes composed of predominantly planar surfaces, such as urban areas. We use a synthetic dataset to demonstrate that our approach outperforms recent methods that implement distance-based regularization terms. Finally, we validate our approach for mapping applications on our experimental vehicle.

Online stereo camera calibration from scratch

Stereo cameras are among the most promising sensors for automated driving. For their deployment, however, calibration should be automated and possible in-situ. We propose a restructuring of bundle adjustment into an incremental online calibration system. It allows us to estimate all observable camera parameters on the fly. Both simulations and experiments with real world cameras show its capability to calibrate stereo rigs in real time while driving. With this method, cameras can be employed with almost no calibration overhead. Only the non-observable parameter of scale has to be defined in advance.