Martin Pellkofer

A system for traffic sign detection, tracking, and recognition using color, shape, and motion information

This paper describes a computer vision based system for real-time robust traffic sign detection, tracking, and recognition. Such a framework is of major interest for driver assistance in an intelligent automotive cockpit environment. The proposed approach consists of two components. First, signs are detected using a set of Haar wavelet features obtained from AdaBoost training. Compared to previously published approaches, our solution offers a generic, joint modeling of color and shape information without the need of tuning free parameters. Once detected, objects are efficiently tracked within a temporal information propagation framework. Second, classification is performed using Bayesian generative modeling. Making use of the tracking information, hypotheses are fused over multiple frames. Experiments show high detection and recognition accuracy and a frame rate of approximately 10 frames per second on a standard PC.

Reliable Detection of Overtaking Vehicles Using Robust Information Fusion

Early detection of overtaking vehicles is an important task for vision-based driver assistance systems. Techniques utilizing image motion are likely to suffer from spurious image structures caused by shadows and illumination changes, let alone the aperture problem. To achieve reliable detection of overtaking vehicles, the authors have developed a robust detection method, which integrates dynamic scene modeling, hypothesis testing, and robust information fusion. A robust fusion algorithm, based on variable bandwidth density fusion and multiscale mean shift, is introduced to obtain reliable motion estimation against various image noise. To further reduce detection error, the authors model the dynamics of road scenes and exploit useful constraints induced by the temporal coherence in vehicle overtaking. The proposed solution is integrated into a monocular vision system onboard for obstacle detection. Test results have shown superior performance achieved by the new method

EMS-Vision: a perceptual system for autonomous vehicles

A survey is given of an expectation-based multifocal saccadic vision (EMS-Vision) system. EMS-Vision is the 3rd generation dynamic vision system for road vehicle guidance following the 4D approach. It combines a wide field of view (f.o.v.) nearby (>100/spl deg/, L/sub 0.05/=36 m, peripheral part) with central areas of high resolution: a 3-chip-color-camera with a f.o.v. of 23/spl deg/ (L/sub 0.05/=100 m) and a high sensitivity b/w-camera with a f.o.v. of 5.5/spl deg/ (L/sub 0.05/=300 m, foveal part). At L/sub 0.05/ a single pixel in the image corresponds to 5 cm in the real world. By active gaze control, this foveal cone can be inertially stabilized, be redirected to a point of interest in the wide f.o.v. (saccade), and locked onto a moving object for reducing motion blur (fixation). This vertebrate type of vision system allows new performance levels in machine vision. The system has been implemented on commercial off-the-shelf (COTS) components in test vehicles.

An integrated framework of vision-based vehicle detection with knowledge fusion

This paper describes an integrated framework of on-road vehicle detection through knowledge fusion. In contrast to appearance-based detectors that make instant decisions, the proposed detection framework fuses appearance, geometry and motion information over multiple image frames. The knowledge of vehicle/non-vehicle appearance, scene geometry and vehicle motion is utilized through prior models obtained by learning, modeling and estimation algorithms. It is shown that knowledge fusion largely improves the robustness and reliability of the detection system.

Multi-modal speed limit assistants: Combining camera and GPS maps

We propose a method for fusing two modalities of information for speed limit assistants: (i) camera based speed sign recognition and (ii) digitized speed limit maps combined with a GPS sensor. The fusion is based on a Bayesian framework. Here, we rely on two modeling assumptions: (i) the speed sign recognizerpsilas score being probabilistic and (ii) a model describing speed limit sign probabilities conditioned on the map information. Speed limit assistants incorporating the proposed fusion can particularly benefit over uni-modal solutions in situations, where a solution based on a single modality is ill-posed, that is, adverse lighting or weather conditions in case of camera based speed sign recognition, and dynamic traffic guidance systems, construction zones, or incomplete maps in case of GPS maps. We give exemplary evidence of the proposed solutionpsilas effectiveness.

Passing vehicle detection from dynamic background using robust information fusion

This work presents a robust method of passing vehicle detection. Obstacle detection algorithms that rely on motion estimation tend to be sensitive to image outliers caused by structured noise and shadows. To achieve a reliable vision system, we have developed two important techniques, motion estimation with robust information fusion and dynamic scene modeling. By exploiting the uncertainty of flow estimates, our information fusion scheme gives robust estimation of image motion. In addition, we also model the background and foreground dynamics of road scenes and impose coherency constraints to eliminate outliers. The proposed detection scheme is used by a single-camera vision system developed for driver assistance. Our test results have shown superior performance achieved by the new detection method.