Object Recognition (OBJ)
The department Object Recognition (Objekterkennung - OBJ) develops and evaluates algorithms for automatic object detection and object tracking in sensor networks.
The department´s activities range from the evaluation of video streams in the infrared and visual spectral band and the analysis of laser sensor data to the description of a three-dimensional, dynamic environment via multi-sensory data acquisition and automatic alerting in case of specifically defined occurrences. In addition, the possible real-time implementation of the algorithms is evaluated on the basis of heterogeneous hardware structures.
In detail, the department focuses on the following fields of work:
Object Recognition in Sensor Networks
The research work is focused on the detection and presentation of objects in imagery data streams of interconnected mobile sensors. In this context the technologies investigated include the aspect-independent description of objects, the registration of sensor-generated images with three-dimensional context data, and bandwidth-economical transfer of object information.
Video Content Analysis
Video Content Analysis combines methods for the detection and tracking of objects in video streams with algorithms for the conceptual description and analysis of the extracted, quantitative information. The studies aim at coming up with systems for the semantic analysis of videos, which means that videos are not only analyzed quantitatively, but that the extracted information is associated to conceptual background knowledge in order to draw conclusions from the visually perceived environment.
Machine vision algorithms extend from simple filtering functions up to complex analysis methods. Currently available hardware also varies with respect to computing performance, programming paradigms, architectures, and power consumption. The field of Heterogeneous Hardware Structures
deals with the specification and combination of hardware structures suitable for complex real-time vision systems.
Tracking and Tracking Assessment
Especially with regard to military tracking systems, performance evaluation is an essential theme. The topics “Tracking and Tracker Assessment
” are dealt with on the basis of years of experience in the development and design of evaluation schemes that interrelate and evaluate both the performance ability of tracking algorithms and the risk analysis, while taking into account possible counter-measures.
Object Recognition in 3D Data
The acquisition and analysis of 3D data is of increasing importance in those application areas that require a high degree of automation and reliability of object recognition. The department's research activities in this field are concerned with the development, optimization and evaluation of methods for 3D data analysis for use with established sensor techniques as well as prototypical hardware. In addition to object recognition, data acquired by these sensors are used for area-wide monitoring (e.g., 3D change detection) and to provide context information for image exploitation.
Publications of the department