AS
IS
re
he
nic
lle
yer
lla
0”,
nic
Handmann, Uwe
ANALYSIS OF DYNAMIC SCENES
Uwe HANDMANN, Iris LEEFKEN, and Christian GOERICK
Institut für Neuroinformatik, Lehrstuhl für Theoretische Biologie
Ruhr-Universitát Bochum, 44780 Bochum, Germany
E-mail: Uwe.Handmann@neuroinformatik.ruhr-uni-bochum.de
KEY WORDS: Dynamic Scenes, Computer Vision, Driver Assistance
ABSTRACT
In this paper the proposed architecture for a dynamic scene analysis is illustrated by a driver assistance system. To
reduce the number of traffic accidents and to increase the drivers comfort, the thought of designing driver assistance
systems rose in the past years. Principal problems are caused by having a moving observer (ego motion) in predominantly
natural surroundings. In this paper we present a solution for a flexible architecture for a driver assistance system. The
architecture can be subdivided into four different parts: the object-related analysis, the knowledge base, the behavior-
based scene interpretation, and the behavior planning unit. The object-related analysis is fed with data by the sensors
(vision, radar). The sensor data are preprocessed (flexible sensor fusion) and evaluated (saliency map) searching for
object-related information (positions, types of objects, etc.). The knowledge base is represented by static and dynamic
knowledge. It consists of a set of rules (traffic rules, physical laws), additional information (GPS, lane-information) and
it is implicitly used by algorithms in the system. The scene interpretation combines the information extracted by the
object-related analysis and inspects the information for contradictions. It is strongly connected to the behavior planning
using only information needed for the actual task. In the scene interpretation consistent representations (i.e., bird's eye
view) are organized and interpreted as well as a scene analysis is performed. The results of the scene interpretation are
used for decision making in behavior planning, which is controlled by the actual task.
1 INTRODUCTION
Systems for automated image analysis are useful for a variety of tasks. Their importance is still growing due to tech-
nological advances and increased social acceptance. Especially driver assistance systems have reached a high level of
sophistication. Fully or partly autonomously guided vehicles, particularly for road traffic, require highly reliable algo-
rithms due to the conditions imposed by natural environments. At the /nstitut für Neuroinformatik, methods for analyzing
driving relevant scenes by computer vision are developed in cooperation with several partners from the automobile in-
dustry. We present a system extracting important information from an image taken by a CCD camera installed at the
rear-view mirror in a car.
The problems encountered in building a driver assistance system are numerous. The collection of information about
real environments by sensors is error-prone and incomplete. When the sensors are mounted on a moving observer, it
is difficult to find out whether a detected motion was caused by ego-motion or by an independent object moving. The
collected data can be analyzed by several algorithms with different features designed for different tasks. To gain the
demanded information, their results have to be integrated and interpreted. In order to achieve an increase in reliability
of information, a stabilization over time and knowledge about important features have to be applied. Different solutions
for driver assistance systems have been published. An approach published by (Rossi et al., 1996) showed an application
for a warning system for security distance and lane-keeping. An application being tested on highways and being based
on inverse perspective mapping has been presented by (Bertozzi and Broggi, 1998). (Dickmanns et al., 1997) presented
a driving assistance system based on a 4D-approach. Those systems were mainly designed for highway scenarios, while
the architecture presented by (Goerzig and Franke, 1998) has been tested in urban environment.
The content of this paper concentrates on a flexible, modular architecture of a driver assistance system working on eval-
uation and integration of the actual information gained from different sensors. The modules of the architecture are rep-
resented by the object-related analysis, the scene interpretation and the behavior planning. The accumulated knowledge
is organized in the knowledge base. The presented architecture is able to handle different tasks. New requirements to the
system can be integrated easily.
2 SYSTEM ARCHITECTURE
The proposed architecture is intended to produce different kinds of behavior according to given tasks. The structure is
shown in fig. 1. Information about the actual state of the environment is perceived by the system's sensors. The data
collected by each sensor have to be processed and interpreted to gain the desired information for the actual task. This is
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 347