Skip to main content

Cognitive and Cooperative Signal Processing Technologies for Remote Sensing Applications

Final Report Summary - COGSENSE (Cognitive and Cooperative Signal Processing Technologies for Remote Sensing Applications)

Cognitive and cooperative sensing is a novel concept in signal processing that promises to revolutionize the way we interpret and understand our world. The definition of what exactly makes a system cognitive has been the subject of debate over the past few years. While some authors view cognition as being an extension of adaptivity to include the exploitation of knowledge into the decision making processing governing the sensor, others view cognition as being the first step of incorporating artificial intelligence into the sensor itself. This Project reverses the way cognition is traditionally applied, turning the sensor itself into a robot, capable of making autonomous decisions in response to changing situations.

Clearly, this is a very ambitious task. Under COGSENSE, a variety of sensors have been considered: in particular, radar, magnetic sensors, and various types of airborne and satellite imagery (e.g. optical, SAR, hyperspectral). Research is conducted under two main thrusts: 1) Detection, estimation and tracking algorithms for cognitive, distributed sensor networks, and 2) Multi-sensor intelligent information fusion. Under these headings, our research has focused on four specific applications:

• Biomedical Engineering (Assistive, Rehabilitation, and Quality of Life Technologies)
• Vehicle Detection and Localization
• Intelligent Satellite Data Processing for Disaster Relief and Agriculture
• Multi-Sensor Signal Integration on Airborne Platforms

More specifically, extensive progress has been made under COGSENSE in developing machine learning and pattern recognition algorithms that can adapt their processing based on environmental factors, such as noise level, observation duration and direction of motion, and thus lay the groundwork for a cognitive radar-based activity recognition system for elderly health monitoring. Furthermore, research has also been conducted on pushing the sensing limitations of passive sensors, such as magnetic sensors. A novel algorithm based on orthogonal matching pursuit was developed to enable the estimation of which quadrant the vehicle is located at from the data of a single magnetic sensor. Measurement and analysis of techniques for low-error radiometric calibration, pan-sharpening, multi-sensor exploitation for earthquake damage assessment have been developed. Signal level integration of optical and radar imagery has been explored by the estimation of a radar clutter map from simultaneously obtained optical imagery.

Additionally, COGSENSE has been successful in ensuring integration of the researcher not just to the host institution, TUBITAK UZAY, but also in forging ties with academia, industry, and other European institutions. Significant research collaborations have been nurtured with research groups at the University of Strathclyde (UK), Cranfield University (UK), University College London (UK) and Linkoping University (Finland). The researcher has been a co-advisor to graduate students and the Middle East Technical University and TOBB Univ. of Economics and Technology, resulting in six (6) M.S. Theses in related topics; has taught classes relating to electronics, remote sensing and signal processing at TOBB University and given seminars and lectures on cognitive remote sensing and activity recognition at Ankara University, Bilkent University, and the University College London. The researcher has also authored two book chapters on topics relating to COGSENSE, and participated in other FP7 and Horizon 2020 projects.