Christian Nunn

A Decision Fusion and Reasoning Module for a Traffic Sign Recognition System

A novel approach for a decision fusion and reasoning system for vision-based traffic sign recognition is presented. This module consists of several steps. In the first stage, a track-based Bayesian fusion scheme is used to fuse the classification results from each frame to obtain a fusion result for each track to decide whether a sign is present, as well as to determine the sign type. In order to determine the sign type, the temporal fusion scheme has been combined with a decision tree. In the second stage, the system combines and fuses probable identical objects which help to further reduce failures in the recognition process. The decision is based on the fusion results, as well as a position cue. Finally, a reasoning module is used to decide which of the passed signs should be shown to the driver. In addition to these modules, a general evaluation method for multi-class tracking systems is shown. While some failures are observed from the evaluation on object level, the additional post processing steps improve the system in such a way that the finally presented signs are almost always correct on the test set.

A novel multi-lane detection and tracking system

In this paper a novel spline-based multi-lane detection and tracking system is proposed. Reliable lane detection and tracking is an important component of lane departure warning systems, lane keeping support systems or lane change assistance systems. The major novelty of the proposed approach is the usage of the so-called Catmull-Rom spline in combination with the extended Kalman filter tracking. The new spline-based model enables an accurate and flexible modeling of the lane markings. At the same time the application of the extended Kalman filter contributes significantly to the system robustness and stability. There is no assumption about the parallelism or the shapes of the lane markings in our method. The number of lane markings is also not restrained, instead each lane marking is separately modeled and tracked. The system runs on a standard PC in real time (i.e. 30 fps) with WVGA image resolution (752 × 480). The test vehicle has been driven on the roads with challenging scenarios, like worn out lane markings, construction sites, narrow corners, exits and entries of the highways, etc., and good performance has been demonstrated. The quantitative evaluation has been performed using manually annotated video sequences.

Time to contact estimation using interest points

This paper presents a novel approach to obtain reliable and robust time-to-contact estimates from a monocular moving camera observing various obstacles. The algorithm utilizes interest points to measure the relative scale change of an obstacle and applies robust estimation techniques to combine the different measurements into one of three possible motion models. These include a model with constant distance, with constant velocity and with constant acceleration. An interacting multiple model framework is used to select the appropriate model and finally to estimate the time-to-contact with the observed obstacle. The algorithm presented is evaluated utilizing a large set of recorded video sequences with radar ground truth. Due to its field of application the entire algorithm is designed to use as little computation time as possible and is thus realtime capable.

Performance evaluation of a real time traffic sign recognition system

Traffic sign recognition has been a very active research topic for many years now. However, during this time of intensive research, no common evaluation methodology has been established. This paper explains in detail, how we evaluated our real time video-based traffic sign recognition system and thus may serve as a building block towards establishing a commonly accepted evaluation methodology. The proposed evaluation methods are taken from the visual surveillance research community, which was very active in evaluation techniques during the recent years.

On Occlusion-Handling for People Detection Fusion in Multi-camera Networks

In this paper a system for people detection by means of Track-To-Track fusion of multiple cameras is presented. The main contribution of this paper is the evaluation of the fusion algorithm based on real image data. Before the fusion of the tracks an occlusion handling resolves implausible assignments.

Markov random field for image synthesis with an application to traffic sign recognition

In current state-of-the-art systems for object detection and classification a huge amount of data is needed. Even if large databases are available, some classes are typically underrepresented and therefore the classifier is not able to capture the variability in appearance. In this work we present a novel method to enrich the training database with natural looking synthetic images. The method can be used to transfer the object appearance from one image (template image) to another image (base image) containing a different object of the same or a similar category. In order to preserve natural appearance and avoid artifacts we only use the gray-level values of the base image for synthesis. The main contribution of this work is an extension of the shift-map approach [1]. An appropriate optimization criteria for the used Markov Random Field (MRF) is defined and the MRF is embedded into a general framework for training data synthesis, which is exemplary tailored to the generation of traffic signs. The influence of using synthetic images is evaluated using a convolutional neural network (CNN).

Improving light spot tracking for an Automatic Headlight Control Algorithm

Multi target tracking is an important task for an Automatic Headlight Control Algorithm(AHC). The task is challenging due to the presence of small, closely spaced light spots and the limited computing power of an embedded platform. Thus the tracking method has to be accurate and at the same time computationally efficient. This paper presents a novel method which achieves this task by combining the concepts of interest point tracking and position tracking. In interest point tracking, points are tracked using appearance based features while position tracking makes use of kinematic features. The interest point tracking method in this paper employs a feature set, which consists of well known appearance based features along with a novel light spot environmental feature. A genetic algorithm based search method was used to filter out this feature set from a bigger set. The information from these features is combined with the kinematic features using a computationally efficient method. This fused information is used to track light spots. The new method improved the system performance by reducing the tracking failures by 65% and showed better performance during worst cases like vehicle pitching.

Adaptation of the Mean Shift Tracking Algorithm to Monochrome Vision Systems for Pedestrian Tracking Based on HoG-Features

The mean shift tracking algorithm has become a standard in the field of visual object tracking, caused by its real time capability and robustness to object changes in pose, size, or illumination. The standard mean shift tracking approach is an iterative procedure that is based on kernel weighted color histograms for object modelling and the Bhattacharyyan coefficient as a similarity measure between target and candidate histogram model. The benefits of the approach could not been transferred to monochrome vision systems yet, because the loss of information from color to grey-scale histogram object models is too high and the system performance drops seriously. We propose a new framework that solves this problem by using histograms of HoG-features as object model and the SOAMST approach by Ning et al. for track estimation. Mean shift tracking requires a histogram for object modelling. In the proposed framework a set of high dimensional HoG-features is clustered via K-means and features inside the object area are matched to the clustercenters via a nearest neighbor search. This procedure is comparable to a Bag of Words algorithm. The proposed system is evaluated for advanced driver assistance systems and it is shown that the framework can be used as a reliable visual tracking system for a pedestrian recognition module.