Radar target classification by micro-Doppler contributions
Molchanov, Pavlo (2014)
Molchanov, Pavlo
Tampere University of Technology
2014
Teknis-taloudellinen tiedekunta - Faculty of Business and Technology Management
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:ISBN:978-952-15-3398-3
https://urn.fi/URN:ISBN:978-952-15-3398-3
Tiivistelmä
This thesis studies non-cooperative automatic radar target classification. Recent developments in silicon-germanium and monolithic microwave integrated circuit technologies allows to build cheap and powerful continuous wave radars. Availability of radars opens new applications in different areas. One of these applications is security. Radars could be used for surveillance of huge areas and detect unwanted moving objects. Determination of the type of the target is essential for such systems. Microwave radars use high frequencies that reflect from objects of millimetre size.
The micro-Doppler signature of a target is a time-varying frequency modulated contribution that arose in radar backscattering and caused by the relative movement of separate parts of the target. The micro-Doppler phenomenon allows to classify non-rigid moving objects by analysing their signatures. This thesis is focused on designing of automatic target classification systems based on analysis of micro-Doppler signatures.
Analysis of micro-Doppler radar signatures is usually performed by second-order statistics, i.e. common energy-based power spectra and spectrogram. However, the information about phase coupling content in backscattering is totally lost in these energy-based statistics. This useful phase coupling content can be extracted by higher-order spectral techniques. We show that this content is useful for radar target classification in terms of improved robustness to various corruption factors.
A problem of unmanned aerial vehicle (UAV) classification using continuous wave radar is covered in the thesis. All steps of processing required to make a decision out of the raw radar data are considered. A novel feature extraction method is introduced. It is based on eigenpairs extracted from the correlation matrix of the signature. Different classes of UAVs are successfully separated in feature space by support vector machine. Within experiments or real radar data, achieved high classification accuracy proves the efficiency of the proposed solutions.
Thesis also covers several applications of the automotive radar due to very high growth in technologies for intelligent vehicle radar systems. Such radars are already build-in in the vehicle and ready for new applications. We consider two novel applications. First application is a multi-sensor fusion of video camera and radar for more efficient vehicle-to-vehicle video transmission. Second application is a frequency band invariant pedestrian classification by an automotive radar. This system allows us to use the same signal processing hardware/software for different countries where regulations vary and radars with different operating frequency are required.
We consider different radar applications: ground moving target classification, aerial target classification, unmanned aerial vehicles classification, pedestrian classification. The highest priority is given to verification of proposed methods on real radar data collected with frequencies equal to 9.5, 10, 16.8, 24 and 33 GHz.
The micro-Doppler signature of a target is a time-varying frequency modulated contribution that arose in radar backscattering and caused by the relative movement of separate parts of the target. The micro-Doppler phenomenon allows to classify non-rigid moving objects by analysing their signatures. This thesis is focused on designing of automatic target classification systems based on analysis of micro-Doppler signatures.
Analysis of micro-Doppler radar signatures is usually performed by second-order statistics, i.e. common energy-based power spectra and spectrogram. However, the information about phase coupling content in backscattering is totally lost in these energy-based statistics. This useful phase coupling content can be extracted by higher-order spectral techniques. We show that this content is useful for radar target classification in terms of improved robustness to various corruption factors.
A problem of unmanned aerial vehicle (UAV) classification using continuous wave radar is covered in the thesis. All steps of processing required to make a decision out of the raw radar data are considered. A novel feature extraction method is introduced. It is based on eigenpairs extracted from the correlation matrix of the signature. Different classes of UAVs are successfully separated in feature space by support vector machine. Within experiments or real radar data, achieved high classification accuracy proves the efficiency of the proposed solutions.
Thesis also covers several applications of the automotive radar due to very high growth in technologies for intelligent vehicle radar systems. Such radars are already build-in in the vehicle and ready for new applications. We consider two novel applications. First application is a multi-sensor fusion of video camera and radar for more efficient vehicle-to-vehicle video transmission. Second application is a frequency band invariant pedestrian classification by an automotive radar. This system allows us to use the same signal processing hardware/software for different countries where regulations vary and radars with different operating frequency are required.
We consider different radar applications: ground moving target classification, aerial target classification, unmanned aerial vehicles classification, pedestrian classification. The highest priority is given to verification of proposed methods on real radar data collected with frequencies equal to 9.5, 10, 16.8, 24 and 33 GHz.
Kokoelmat
- Väitöskirjat [4864]