GB2299909A

GB2299909A - Radar target detection

Info

Publication number: GB2299909A
Application number: GB9507448A
Authority: GB
Inventors: Peter Lintott; Angus Raymond Johnson; Paul George Wright; J M Wood
Original assignee: Commonwealth of Australia
Current assignee: Commonwealth of Australia
Priority date: 1995-04-11
Filing date: 1995-04-11
Publication date: 1996-10-16
Anticipated expiration: 2015-04-11
Also published as: GB9507448D0; GB2299909B

Abstract

Received radar signal data is arranged in cells eg by range, azimuth and doppler (ARD). Different categories of noise associated with respective cells being identified 6.White noise is applied to signal data associated with each respective cell 7-10, the type of white noise being selected in dependence on the category of noise associated with that cell, before the processed data is recombined 15. The invention is particularly applicable to over the horizon radars where different, clearly distinguishable noise sources exist, eg surface clutter, radio frequency interference and meteor streaks.

Description

IMPROVED RADAR TARGET DETECTION This invention relates to a method and apparatus for improved radar target detection and particularly to a method and apparatus for improved skywave over the horizon radar target detection.

In systems which operate by sending out energy, reflecting it from a target and detecting the reflected return the detection of the return is one of the most critical parts of the system.

This detection of return signals is a particular problem in long range radars because of the very low level of the return signal compared to general background noise and the capability to detect return signals only slightly above background noise level is one of the limits on long range radar performance. Signals from other sources further complicate the environment.

In a skywave over the horizon radar system the return signals are formed into a three dimensional array of cells, each cell having a unique range, azimuth and doppler value, the doppler value corresponding to the velocity of the target relative to the radar system and an associated received signal level. These cells are known as azimuth, range, doppler cells (ARD cells).

The standard method of detecting targets in radar systems employing ARD cells is to calculate the average noise level over all of the ARD cells and then to set a threshold value corresponding to a signal level having some predetermined relationship to the average noise level. Generally the threshold is set at some fixed margin above the averaged noise level.

A method of calculating the averaged noise level is to use noise whitening, also known as local averaging. In this technique each ARD cell is examined in turn and the average noise level of all of the ARD cells in three dimensional window centered on the ARD cell of interest is calculated. This local average noise level is then used as the basis for the threshold operation to detect targets.

This technique allows for noise levels varying gradually across the area scanned by the radar and so provides more reliable results, but it still has a number of problems, particularly when used in skywave over the horizon radars. Other problems particular to skywave over the horizon radars are caused by the types of noise encountered in such radars due to clutter, meteor trails and radio frequency interference.

Clutter is caused by scattering of the radar signals by surface objects. These are stationary relative to the radar receivers and transmitters and so generate returns at any range and azimuth values but having zero or very low doppler values. Clutter is a problem in all radar systems, not just skywave over the horizon radar.

The effect of meteors on the upper atmosphere is to generate meteor trails. The meteor trails interact with the radar signals refracted in the upper atmosphere, which skywave over the horizon radars rely on to operate to cause noise to be generated at a small number of range and azimuth locations, but at a very high range of doppler values these are known as streaks. Radio frequency interference (RFI) is caused by radio transmitters, generally radio broadcast transmitters, within the region scanned by the radar and their frequency characteristics cause noise having a small range of azimuth and doppler values but a large spread in range. These two phenomenons are a particular problem in skywave over the horizon radars.

In the past these sources of noise have been dealt with by specialist one dimensional noise whitening procedures which only average over adjacent cells in one dimension.

The disadvantage of this conventional approach is that the consecutive use of the multiple different noise whitening procedures required reduces the sensitivity of the radar system and so makes it harder to detect small targets, this reduction in sensitivity is often referred to as processing loss.

This invention was intended to produce a skywave over the horizon radar system overcoming these problems, at least in part.

In one aspect this invention provides a method of processing radar signal data arranged in cells and comprising the steps of; i) analysing the data and deducing which areas of the radar signal data are suffering from a type of noise; ii) carrying out separate noise whitening operations on the areas suffering from said type of noise and the remainder of the radar signal data using a noise whitening technique matched to the type of noise; and iii) recombining the processed data from the separate noise whitening operations.

In another aspect the invention provides apparatus for use in the method.

This invention is based upon the realisation that where specialist clutter, streak and RFI suppression noise whitening techniques have been used in the past to operate on all ARD cells in a skywave over the horizon radar output, most of the ARD cells they have operated on have been free of the type of noise suppressed by the specialist noise whitening technique. As a result, if it were possible to identify those parts of the radar output containing a particular type of noise the specialist noise whitening technique could be used on those parts of the output only.

An embodiment of the invention will now be described by way of example with reference to Figure 1 which shows schematically a radar target detection system. In Figure 1, output data from a skywave over the horizon radar is supplied along a line 2 as a series of ARD cell identities and associated received signal levels. This ARD cell data is supplied to a clutter detector 3, a radio frequency interference (RFI) detector 4 and a streak detector 5. Each of the detectors 3, 4 and 5 contains a memory into which the ARD cell data is fed until a whole radar field of view is stored. The detectors 3, 4 and 5 then begin operation.

The clutter detector 3 applies a fixed threshold to all ARD cells having a doppler value below a set limit, generally this limit is set very close to zero. The clutter detector 3 then identifies all ARD cells which have received signal levels above the threshold and are in a contiguous group as including clutter and define an area containing these clutter containing cells.

The location and size of this clutter area are passed to a conflict resolving unit 6.

The RFI detector 4 noise whitens the ARD cell data using a noise whitening window which is narrow in the range dimension to reduce RFI effects. The RFI detector 4 then applies a fixed threshold, having a fixed value over the locally calculated noise level, which need not be the same as the clutter detector 3 fixed threshold, to the ARD cells and then checks all of the ARD cells which exceed the threshold value for RFI interference. This checking is done by looking at a window of ARD cells centred on each ARD cell containing a threshold crossing.

The RFI detection window takes in N contiguous ARD cells having different range values but identical Azimuth and doppler values. If M or more of the N ARD cells have an associated power level above the threshold, known as a threshold crossing, the RFI detector 4 defines them as being subject to RFI noise and defines an area or areas containing all of the RFI affected ARD cells. The RFI detector 4 then passes on the location and size of the RFI affected area or areas to the conflict resolving unit 6.

The streak detector 5 operates in a similar manner to the RFI detector 4. The streak detector 5 noise whitens the ARD cell data using a noise whitening window which is narrow in the doppler dimension. The streak detector 5 then applies a fixed threshold having a fixed value over the locally calculated noise level for each ARD cell to the ARD cells. This fixed threshold need not be the same as the fixed thresholds used in the clutter and RFI detectors 3 and 4. All thresholds can be uniquely set. The streak detector 5 then checks all of the ARD cells which exceed the threshold value for streak interference by looking at a window of X contiguous ARD cells having different doppler values and the same azimuth and range values centred on each ARD cell containing a threshold crossing.If Y or more of the X ARD cells within the window contain a threshold crossing the streak detector defines them as being subject to streak noise and defines an area or areas containing all of the streak affected ARD cells. The streak detector 5 then passes on the location and size of the streak affected area or areas to the conflict resolving unit 6.

The conflict resolving unit 6 compares the areas of the region scanned by the radar (the map) which are defined as containing clutter, RFI or streaks by the clutter, RFI or streak detectors 3, 4 and 5. Where parts of the map are defined as containing two or three of these types of noise the conflict resolving unit 6 decides which of the types of noise these overlapping parts of the map are to be treated. This decision can be based on a simple ranking process where a fixed set of instructions as to how each type of multiple interference is treated, the instructions being determined by the characteristics of the radar system.

Alternatively the clutter, RFI and streak detectors 3, 4 and 5 can produce local noise estimates for each of the clutter, RFI or streak affected areas and supply these to the conflict resolving unit 6. The conflict resolving unit then treats areas defined as containing several types of interference as areas affected by the type of interference having the highest local noise estimate in the overlapping area.

Thus the conflict resolving unit 6 defines all parts of the map as containing clutter or RFI or streaks, or none of them, and supplies this information to four noise whitening units 7 to 10.

The first noise whitening unit 7 uses a conventional noise whitening algorithm on only those areas of the map defined as not containing clutter, RFI or streaks by the conflict resolving unit 6.

The second noise whitening unit 8 uses a noise whitening algorithm optimised for clutter suppression on only those parts of the map defined as containing clutter by the conflict resolving unit 6.

The third noise whitening unit 9 uses a one dimensional noise whitening algorithm optimised for RFI suppression on only those parts of the map defined as containing RFI by the conflict resolving unit 6.

The fourth noise whitening unit 10 uses a one dimensional noise whitening algorithm optimised for streak suppression on only those parts of the map defined as containing streaks by the conflict resolving unit 6.

All of the noise whitening units 7 to 10 are supplied with the ARD data along the line 2.

The noise whitening units 7 to 10 operate by averaging the signal strength over the adjacent ARD cells to the ARD cell of interest in one, two or three dimensions. The ARD cells averaged over may be immediately adjacent only or a window of any desired size centred on the ARD cell of interest. The average noise value produced is then either subtracted from the received signal power value in the ARD cell or used to normalise the power value in the ARD cell by dividing the received power value in the ARD cell by the average noise value.

The new power value produced is then used as the power value of the ARD cell for thresholding. The precise technique used is selected based upon the radar parameters of the specific radar system Each of the noise whitening units 7 to 10 generates a local average noise value for each of the ARD cells. The noise values are supplied to thresholding units 11 to 14 respectively and are supplied together with the threshold crossings directly to a data combiner 15.

The thresholding units 11 to 14 each compare the new signal power levels of the ARD cells with a threshold above the noise estimates.

The data combiner 15 brings together the threshold crossing outputs of the thresholding units 11 to 14 and the outputs of the noise whitening units 7 to 10 to provide two outputs, an output along a line 16 identifying all the ARD cells containing threshold crossings, and an output along a line 17 giving the noise whitened power value for each ARD cell. These outputs 1 are used in further, conventional processing in later stages of the radar system. In addition the raw input ARD data is available on line 18.

The processing loss is reduced because no part of the ARD cell data is noise whitened more than once. In addition, because the ARD cell data is retained it can be used as calibration values to allow received signal values from different ARD cells to be correctly scaled to allow comparison or combination, for example in interpolation to calculate the precise position of a target with an ARD cell or in peak detection where a peak must be a peak in all three dimensions.

Each of the noise whitening operations described above can be fixed or adaptive. An adaptive noise whitening algorithm is one where the type of noise whitening employed is varied in response to external parameters, variations in the noise whitening procedure include whether mean or order statistics are used or whether an arithmetric or geometric mean is used. Excision techniques may be employed, or the size and shape of the window averaged over may be varied.

These external parameters may be the operating parameters of the radar system or the statistics of the noise over the whole area scanned by the radar or just in the vicinity of the ARD cell of interest. Suitable noise statistics would be determined from the noise floor.

Although the technique above is described in the terms of a skywave over the horizon radar they could be used in any system where data is multiply noise whitened to reduce processing loss and where the detection method is matched to the noise. Although the invention has been described with reference to a radar system, the scope of the invention encompasses other similar systems which operate by detecting a return signal of some kind, for example sonar signals.

Claims

1. A method of processing radar signal data arranged in cells and comprising the steps of; i) analysing the data and deducing which areas of the radar signal data are suffering from a type of noise; ii) carrying out separate noise whitening operations on the areas suffering from said type of noise and the remainder of the radar signal data using different noise whitening techniques; and iii) recombining the processed data from the separate noise whitening operations.

2. A method as claimed in claim 1 including the further step of; iv) carrying out a thresholding operation on the processed data to identify potential targets.

3. A method as claimed in claim 2 in which step iv is carried out separately on processed data from the area suffering from said type of noise and the remainder of the processed radar signal data before step iii.

4. A method as claimed in any preceding claim in which; in step i the areas of the radar signal data suffering from a plurality of types of noise are deduced; in step ii separate noise whitening operations are carried out on the areas suffering from each type of noise and on the remainder of the radar signal data using different noise whitening techniques; and including an additional step v) identifying those parts of the radar signal data suffering from more than one of said plurality of types of noise and deciding which of the types of noise said parts are treated as suffering from between steps i and ii.

5. A method as claimed in claim 4 in which, in step v, said parts are treated as suffering from the type of noise having the highest local noise value in said part.

6. A method as claimed in claims 4 or 5 in which the types of noise are clutter, radio frequency interference and streaks.

7. A method as claimed in any preceding claim in which the noise whitening techniques used are altered in dependence upon the radar system parameters.

8. A method as claimed in any preceding claim in which the noise whitening techniques are altered in dependence upon the noise statistics for the area of the radar signal data being noise whitened.

9. A method as claimed in any preceding claim in which the radar data is skywave over the horizon radar data and the cells are azimuth range and doppler cells.

10. A method of radar data processing substantially as shown in or as described with reference to the accompanying figure.

11. Apparatus for processing radar signal data arranged in cells comprising; i) means for analysing the data and deducing which areas of the radar signal data are suffering from a type of noise; ii) means for carrying out separate noise whitening operations of the areas suffering from said type of noise and the remainder of the radar signal data using different noise whitening techniques; and iii) means for recombining the processed data from the separate noise whitening operations.

12. Appparatus for processing radar data substantially as shown in or as described with reference to the accompanying figure.