Communications: directive radio wave systems and devices (e.g. – Return signal controls radar system – Receiver
Reexamination Certificate
2003-01-23
2004-10-05
Sotomayor, John B. (Department: 3662)
Communications: directive radio wave systems and devices (e.g.,
Return signal controls radar system
Receiver
C342S098000, C342S099000, C342S189000, C342S192000, C342S194000, C342S196000
Reexamination Certificate
active
06801155
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a method of and an apparatus for recognising a radar target.
2. Discussion of Prior Art
Moving target identification (MTI) radars are known. Such radars can provide all-weather, day and night surveillance capability and have a wide-area search capability as a result of rapid scanning of the radar beam. An MTI radar may be used to provide location information relating to a moving target and may also be used to identify a target as belonging to a particular class of targets. For example an MTI radar may be used to classify a target as being a person, a wheeled vehicle or a tracked vehicle. Presently, the task of target recognition using an MTI radar is carried out by a human operator. In one known MTI radar, echo signals from targets are converted to audio signals which are output to an audio speaker. To identify a particular target, the radar's antenna is directed to a target for several seconds during which time an operator listens to the audio speaker output. An operator requires training to develop the ability to classify targets from their corresponding audio frequencies. However, if a scene under observation contains many different types of targets, a human operator, however trained, cannot classify all types in the scene and cannot provide up-to-date information on a rapidly changing scene.
In the case of imaging radars, to perform target recognition on radar signals algorithms have been developed which rely on the availability of 2-dimensional high-resolution imagery of targets. They exploit the differences between spatial structures of radar images in order to classify targets. They are unfortunately unsuitable for use with MTI radars which are mainly low resolution sensors: MTI radars show target images as single bright point objects with no spatial structure in either range or azimuth.
There are certain circumstances in which an MTI radar can achieve better spatial resolution. If the radar operates at higher bandwidth then the radar signature will have a higher slant range resolution and the target will be resolved in range, e.g. a bandwidth of 500 MHz will give a slant range resolution of 0.3 m. However, it is not possible to get higher azimuth resolution. This is because imaging radars use a synthetic aperture radar (SAR) processing, which provides an azimuth resolution that is many times smaller than the real beam azimuth aperture.
SAR processing assumes a static scene and moving objects will not be focused in azimuth using it. Thus at best an MTI radar can only obtain 1-dimensional high range resolution profiles of moving targets. Limited success has been achieved in developing recognition algorithms that exploit the 1-dimensional high range resolution profile of targets for classification. A technique based on 1-dimensional range profile template matching is described in the Proceedings of the International Radar Symposium IRS-98. There are, however, two main drawbacks with a range profile classifier. Firstly, high bandwidth radars are more expensive and it is more difficult to extract high range resolution profiles because moving targets can migrate through a succession of range cells and causing considerable difficulties. Secondly, target range profiles are very sensitive to changes in target orientation to the radar and the target physical build characteristics.
Since a target can have any number of articulations, a range of build variations and an almost infinite number of external fitting configurations, the number of potential independent realisations of range profiles of a particular target is large. This presents a major problem for classifiers using template matching, because they carry out recognition by comparing the range profile of an unknown target with a set of templates and choosing the class that yields the best match: this relies on having a template set that contain every single independent realisation of target signature for a class of targets to be recognised.
Another option open to an MTI radar is to use the Doppler characteristics of moving targets in determining the target classification. Doppler is the phenomenon by which the radar return from an object is shifted in frequency due to the object's radial motion relative to a radar system. A Doppler return from a target can be observed by looking at a received echo in the frequency domain. Raw radar data from a target is recorded as a series of temporal samples. Using a n-point Fast Fourier Transform (FFT), temporal samples from a target are transformed into a spectrum comprising n frequency samples or bins. The n frequency samples form a Doppler profile where the maximum unambiguous frequency is given by the inverse of the radar pulse repetition interval and the Doppler resolution per frequency bin is 1
th of the unambiguous Doppler.
A target classifier might be based on template matching similar to a range profile classifier but using templates of independent realisations of the Doppler profiles of moving targets. Although Doppler profiles are comparatively less sensitive to target articulations, they still exhibit fluctuations with changing target orientation and imaging geometry. Furthermore, Doppler profiles vary as a function of target speed and a profile's shape is modulated as a result of a target's vibrational and rotational motion. A target classifier based on template matching would require a prohibitively large number of stored Doppler profiles.
SUMMARY OF THE INVENTION
The present invention provides a method of recognising a radar target comprising receiving radar returns from a scene and processing the returns to produce a Doppler spectrum characterised in that the method also includes processing the Doppler spectrum to produce a Doppler feature vector and using hidden Markov modelling (HMM) to identify the Doppler feature vector as indicating a member of a particular class of targets.
The invention provides the advantage that moving targets may be classified without recourse to a large database of radar signature data.
In a preferred embodiment, the method includes arranging for targets to be encompassed within a single radar range cell. It may involve processing radar returns to obtain a sequence of Doppler spectra for each target and producing therefrom a sequence of Doppler feature vectors, and using HMM to identify the sequence of Doppler feature vectors as indicating a member of a particular class of targets. This enables classification to use linked Doppler profiles as part of a structural observation sequence: it exploits the fact that radar data for a target provides a series of Doppler profiles each offset slightly in time. Each Doppler profile is different, but over a sequence of observations the shape of the profile varies accordingly to some deterministic process. This embodiment exploits useful information in the variation of the Doppler profile with time. HMM algorithms have evolved to make use of sequences of speech signals in this way.
The method may include using HMM to identify the sequence of Doppler feature vectors by assigning to each feature vector an occurrence probability by selecting a probability distribution or state from a set thereof associated with a class of targets, multiplying the occurrence probabilities together to obtain an overall probability, repeating for other probability distributions in the set to determine a combination of probability distributions giving highest overall probability for that class of targets, then—repeating for at least one other class of targets and selecting the target class as being that which yields the highest overall occurrence probability. Probability distributions for successive feature vectors may be selected on the basis of some transitions between distributions allocated to successive feature vectors being allowed for the class of targets and others being forbidden.
A preliminary HMM training procedure may be implemented in which parameters for the states or probability distributions and transition probabi
Jahangir Mohammed
Ponting Keith M
Nixon & Vanderhye P.C.
QinetiQ Limited
Sotomayor John B.
LandOfFree
Method and apparatus for recognising a radar target does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for recognising a radar target, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for recognising a radar target will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3279771