Report from a Partner: Polytechnical University of Timisoara
The activity of UPT ARTRAC team is carried out on the development of algorithms for road conditions classification and targets detection based on radar signal processing. This task assumes the use of several signal processing techniques, such as
- denoising of the received radar signal, which is affected by speckle noise;
- extraction of other useful signal parameters which can be used in the classification process;
- Classification of the scanned road surface based on the above estimate.
Our main activity in this period refers to denoising, a relatively new signal processing method which we propose to be used in ARTRAC project. Denoising is a wavelet-based noise removing method which does not distort the noiseless component of the acquired signal and can thus improve the target detection by automotive radar. A preliminary report presented in March 2013 at Timişoara ARTRAC meeting is an overview of method performances, highlighting its ability to improve the detection procedure used in ARTRAC project. Progress has been made towards a higher probability of targets detection by wavelet denoising. Using the wavelet denoising procedure for detecting a DC signal embedded in Gaussian noise, an important processing gain was obtained as compared with the usual detection procedure. Though, this gain is lower than the maximal value predicted by the information theory for DC signals, the investigation of denoising procedures should be continued on more complex signals, in order to implement it in target detection algorithms used in ARTRAC project.
In order to develop an algorithm for road conditions classification, a preliminary report on radar road detection techniques was presented in November 2012 at Braunschweig ARTRAC meeting. Preliminary road radar data obtained at VTT showed the way the radar data can be used to discriminate between different road surfaces, but more experimental results are needed in order to build a reliable road conditions database.
Another topic of interest was the implementation of new Kalman filter algorithms for tracking and identifying moving targets. Preliminary study undertaken on real automotive radar data were reported at Timişoara ARTRAC meeting and revealed that this approach combined with a data association algorithm is a true alternative to currently used target localization methods. These results will also be reported to an international technical conference.
Timişoara, July 3, 2013