WO2004097452A1

WO2004097452A1 - Detection of small objects in bodies of water

Info

Publication number: WO2004097452A1
Application number: PCT/GB2004/001868
Authority: WO
Inventors: Wieslaw Jerzy Szajnowski; Simon Potter; John Benjamin Wynne
Original assignee: Qinetiq Limited
Priority date: 2003-04-30
Filing date: 2004-04-29
Publication date: 2004-11-11

Abstract

An object in a body of liquid such as seawater is detected by transmitting a signal, receiving a reflection of the signal, repeatedly determining the phase of the reflection, and detecting whether the distribution of the differences between phases of the reflection occurring at intervals which are no less than a predetermined value is significantly non-uniform. The influence of the phases on the detection of non-uniform distribution is weighted in accordance with the power of the reflection at the time of occurrence of the respective phases. A signal indicative of the presence of an object is provided in dependence on the result of the detection step.

Description

DETECTION OF SMALL OBJECTS IN BODIES OF WATER

FIELD OF THE INVENTION

This invention relates to a method and apparatus for detecting objects in bodies of liquid and particularly, but not exclusively, to arrangements for detecting a microwave signal reflected from a small floating object in the presence of interfering signals backscattered by a disturbed sea surface.

BACKGROUND OF THE INVENTION Radars operating in a maritime environment are expected to reliably detect various small objects of potential interest in the presence of unwanted signals reflected from the sea surface. The small objects to be detected include boats and rafts, buoys, swimmers, various debris and small fragments of icebergs. Some of those objects may pose a significant threat to safe ship navigation, whereas other objects are of interest in search-and-rescue missions, coastal surveillance etc.

For example, small fragments of icebergs protruding only about 1 m above the water line, yet weighing up to 50 tonnes, can be extremely hazardous to shipping. Targets of this type are very difficult to detect with conventional noncoherent marine radars, due to the fact that they produce low-level signals in substantial background noise and other interference.

It is known that sea clutter significantly degrades the detection performance of radars operating in a maritime environment. Although noncoherent radars cannot reliably detect small objects in spiky sea clutter, it has been shown experimentally that the coherent representation of backscattered signals may contain sufficient information for reliable detection of objects of interest.

In coherent radar systems, targets can be discriminated from clutter if they exhibit reflectivity (i.e., radar cross section, RCS) and phase characteristics different from those exhibited by the clutter. The interaction between the target and the interrogating signal can be examined in terms of amplitude modulation signatures and/or angle modulation signatures.

Traditionally, both types of modulation impressed by a target on the interrogating signal are analysed in the time domain and the frequency domain. Unfortunately, because a sinusoidal signal with purely amplitude modulation and a sinusoidal signal with purely phase modulation can both have the same power spectra, the spectral analysis is of limited use in systems intended for detecting small objects in spiky sea clutter. Also, spectral analysis becomes more complicated when the signals are sampled in a nonuniform manner, for example, as a result of utilizing a staggered pulse train as an mterrogating radar signal.

Fig. 1 is a simplified functional block diagram of a typical state-of-the-art radar system utilizing coherent pulses of microwave energy with no intrapulse modulation.

The system comprises a stable oscillator OSC producing a sinusoidal carrier signal, a power amplifier PAM whose control input CI is driven by a pulse generator PGR, a transmit antenna TAN, a receive antenna RAN connected to a low-noise amplifier LNA, a 90°-phase shifter PHS, two mixers, MXI and MXQ, two low-pass filters, LFI and LFQ, a delay unit DEL, two sample-and-hold circuits, SHI and SHQ, followed by two analogue-to-digital converters, ADI and ADQ, and a suitable digital signal processor DSP.

A sinusoidal carrier signal, supplied by the oscillator OSC, is amplified and modulated in an on-off fashion in the power amplifier PAM, and transmitted as a burst of pulses of microwave energy by the transmit antenna TAN. The transmitted microwave pulses are frequency-shifted replicas of the pulses provided by the pulse generator PGR. A reflected signal received at the receive antenna RAN is amplified in the amplifier LNA and applied to the signal inputs, IS and QS, of the two mixers, MXI and MXQ. The reference inputs, IR and QR, of the mixers are driven, respectively, by a sinusoidal carrier, supplied by the oscillator OSC, and its replica shifted by 90° in the phase shifter PHS. The output

signals of the mixers, after low-pass filtering in filters LFI and LFQ, are applied to the signal inputs, IH and QH, of the two sample-and-hold circuits, SHI and SHQ. Pulses supplied by the pulse generator PGR are suitably delayed in the delay unit DEL and then applied to the control inputs, IP and QP, of the sample-and-hold circuits SHI and SHQ to determine the time instants at which the input signals, zι(t) and Z_Q(t), appearing respectively at inputs IH and QH, are to be sampled. The output discrete-time signals of the circuits SHI and SHQ, zι(t_k) and Z_Q(t_k), are converted into digital formats, ZI and ZQ, by, respectively, converters ADI and ADQ. The digital word^'s, ZI and ZQ, and also the trigger pulse TP produced by ^■ the delay unit DEL, are applied to the corresponding inputs of the digital signal processor DSP for further processing. The main objective of the digital signal processor DSP is to process its input signals in an optimum manner in order to make a reliable decision as to whether or not an object is present at a selected range. The exact value of the range under examination is determined by the value of the delay Ts introduced by the delay unit DEL. For example, a delay of 10 μs corresponds to a range of 1500 m. Usually, in order to examine different ranges a plurality of values of the delay Ts is used.

The binary decision regarding the presence or absence of an object is provided at

the output DD of the digital signal processor DSP.

Fig. 2 shows schematically the time relationship between transmitted pulses of duration τ ₀, interpulse interval T_R, and sampling time instants, i and t₂, delayed by Ts in the delay unit DEL with respect to the corresponding transmitted pulses. The value of the interval TR may either remain the same for all transmitted pulses, or it may vary in some suitable manner. The range resolution, determined from the pulse duration τ o, is usually referred to as the extent of a 'range cell'. For example, a pulse of duration 20 ns corresponds to a resolution (or a range cell) of 3 m. Both the interpulse interval TR, which may vary from pulse to pulse, and the pulse duration τ ₀ are defined by the modulating pulse train supplied by the pulse generator PGR.

It is known that a microwave signal backscattered by a given small patch of the sea surface can be modelled as the product of two independent random processes, often referred to as the speckle component and the texture component. The slow- varying texture component assumes only positive values and it can be regarded as the local mean level of the fast-varying complex Gaussian speckle component.

A complex signal z(t) representing the baseband form of sea clutter can be expressed as z(t) = zj(t) + j z_Q(t) = g(t)[x(t) + j y(t)], g(t)X) where zι(t) and Z_Q(t) are, respectively, in-phase and quadrature components of z(t), g(t) is the texture component process, and x(t) and y(t) are two independent Gaussian component processes. The in-phase and quadrature components, zι(t) and Z_Q(t), of clutter can be observed, respectively, at the inputs IH and QH of the sample-and-hold circuits SHI and SHQ of the system shown in Fig. 1.

The envelope |z(t)| of sea clutter can be determined from

If the texture component process g(t) had a constant value, then the envelope |z(t)| of sea clutter would have a Rayleigh distribution. The non-Gaussian 'spiky' nature of sea clutter results from the variability inherent in the texture process g(t) which modulates simultaneously the two quadrature Gaussian components, x(t) and y(t), of a speckle process. For example, when the square of the texture process, g (t), has a gamma distribution, the envelope |z(t)| of sea clutter exhibits a K- distribution, commonly used for spiky sea clutter characterisation.

The multiplicative form of the compound Gaussian model of sea clutter makes the clutter phase φ(t) totally independent of the texture component g(t), since

Therefore, statistical properties of the phase of sea clutter do not depend on the clutter 'spikiness'.

The apparatus of Fig. 1 operates by sampling the received radar signal in such a way as to obtain a substantial difference between:

- the properties of the phase of a signal reflected from an object of interest, and

- the properties of the phase of a signal reflected from background clutter. More specifically, if the time interval between consecutive signal samples is sufficiently large, then the phase of a signal backscattered by sea clutter will have a substantially uniform distribution, whereas the phase of a signal reflected by a small floating object and its sea clutter background will exhibit a non-uniform distribution. The digital signal processor DSP operates to distinguish between uniform and non-uniform phase distributions.

Preferably, instead of determining the uniformity of the distribution of the phase measurements, the digital signal processor operates on the differences between suitably selected phase measurements. A non-uniform distribution of phase measurements will also result in a non-uniform distribution of phase difference measurements. Furthermore, a moving object will be more readily detected, because although the phase of reflections from the object may alter in a regular manner, thus resulting in a uniform distribution, the phase differences (for a constant velocity) remain constant. The probability of detection will depend on the statistical properties of the phase difference; hence, it will depend on the radar cross section RCS of the object and also on the object's relative motion. In the case of high signal-to-clutter ratio . (SCR), i.e., when the power of a signal reflected from the obj ect is large with respect to that of clutter, the phase difference will be dominated by that resulting from the object's relative movement between the consecutive sampling time instants.

Typical signal-plus-clutter cases giving rise to different phase characteristics can be summarised as follows:

When the signal-to-clutter ratio (SCR) is high and the object is substantially stationary, the observed phase values will be clustered around a. single dominant phase value. Since the observed values of the phase difference will also be clustered around a single (zero) value, this will produce a phase difference distribution which is significantly non-uniform.

When the SCR is high and the obj ect is moving in such a manner that the change in its radial distance from the radar remains almost the same from sample to sample, the observed values of the phase difference will be clustered around a single value determined by the obj ect radial velocity. Consequently, the phase difference distribution will again be significantly non-uniform.

When the SCR is moderate or low and the object is either stationary or in motion, the observed phase values can be perturbed significantly at the time instants coinciding with the occurrence of clutter spikes. At some time instants, the object will be unobservable at all, and its presence will be perceived as intermittent. Such cases will generate gross errors, both for phase and phase- difference measurements.

Depending on the object geometry, and the sea state, some other effects, such as fluctuating radar cross section (RCS) of the object, multipath, shadowing and submersion etc., will decrease the accuracy of phase measurement thereby reducing the probability of object detection.

It would therefore be desirable to provide a method and an apparatus for detecting small objects in spiky sea clutter in a more efficient way than that provided by the prior art techniques .

DESCRIPTION OF THE INVENTION

Aspects of the present invention are set out in the accompanying claims.

In accordance with a further aspect of the invention, the influence of a measured phase on the determination of uniformity of phase distribution is dependent upon (at least) the power of the reflection at the time of the measurement. Thus, weak signals, which are likely to result in phase inaccuracies, have a smaller influence. When the signal reflected by an object of interest is very small, larger excursions of the envelope of signal-plus-clutter returns are more likely to result from clutter spikes rather than from the signal itself. In this case, the overall effect of low- accuracy phase measurements on the detection probability is, in accordance with a f rther aspect of the invention, reduced by assigning proportionately lower weights to those phase measurements which coincide with relatively large values of the envelope of a received signal, preferably using a non-linear function. In the preferred embodiment of the invention, the detection procedure utilizes phase differences rather than phases. In this case, the weight assigned to a phase

difference is a function of the envelope samples at the times of the individual phases from which the phase difference is derived.

The preferred embodiment operates by representing values of phase differences, which are circular variables, by points on a circle, and determining a measure of the dispersion of the points. The points may have different weights, depending on the accuracy or reliability of the measurements they represent. Consequently, the influence of accurate measurements will be enhanced, and that of less accurate measurements will be suppressed. More specifically, the dispersion of points is preferably represented by the mean resultant length (when the points have the same weights) or the modified mean resultant length (when the points have different weights). In particular, the weights are a function of the power of signal samples used for determining the phase differences.

The derived measure of dispersion is preferably compared to a predetermined threshold, which is so selected as to obtain a specified value of the false alarm probability. An object of interest is declared as being present when the threshold value has been exceeded by the derived measure of dispersion.

In the preferred embodiment of the invention, the time between successive phase measurements is always greater than a predetermined interval such that the phase measurements are substantially uncorrelated. However,- by suitable choice of threshold, the invention is also applicable to situations in which there is a degree of correlation, for example because the nature of the environment under investigation makes this unavoidable or because some or all the intervals are too short.

The overall ('global') detection performance of the method is preferably improved^' by examining a selected number of consecutive local decisions, and declaring that an object of interest is present if the ratio of 'local' detection decisions to non- detection decisions exceeds a predetermined value. This is achieved in the preferred embodiment by combining a sequence of decisions according to a scheme usually referred to as 'binary integration' or K-of-M detection.

In binary integration, a number M of consecutive decisions is examined: if the total number of detection decisions exceeds K, then an object is declared as being present. The statistical properties of binary integration are known in the prior art.

A method according to the invention can be implemented either in a hardware form of an application specific integrated circuit (ASIC), or it can be implemented in software as a suitable computer program; it is also possible to suitably combine the two approaches to obtain the required functional equivalence.

The invention extends to a detection method, a detection apparatus and also a signal processor (preferably a single integrated circuit) which can be used in ^' detection apparatus to carry out the method of the invention. DESCRIPTION OF THE DRAWINGS

An arrangement embodying the invention will now be described by way of example with reference to the accompanying drawings, in which:

Fig. 1 is a functional block diagram of a radar system utilizing coherent pulses of microwave energy;

Fig. 2 shows the time relationship between transmitted pulses of duration τ₀, interpulse interval T_R, and sampling time instants, ti and t₂, delayed by Ts with respect to the corresponding transmitted pulses;

Fig. 3 is an example of a phasor representation of a complex clutter sample z(t_k) observed at some arbitrary time instant t^;

Figs. 4 (a) and (b) respectively show examples of representations of widely dispersed angles and clustered angles together with their corresponding mean resultant lengths R;

Fig. 5 shows an example of combining varioμs clutter phasors with a phasor a t) representing a low-level constant signal;

Fig. 6 depicts the shape of a suitable weight function W_k which depends on the circular average U_k of the magnitudes of two consecutive samples; and

Fig. 7 is a functional block diagram of a digital signal processor DSP configured in accordance with the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT The embodiment of the invention to be described below has the structure shown in Fig. 1, the difference with respect to the prior art comprising the way in which the digital signal processor DSP operates.

The in-phase and quadrature clutter components, zι(t) and Z_Q(_), are sampled at the time instants t_k by the sample-and-hold circuits SHI and SHQ of the system shown in Fig. 1, to produce a discrete-time representation, z(t_k) = zι(tk) + jzQ(t ), of clutter. It is convenient to view each complex clutter sample z(t_k) as a phasor, characterised by either its two orthogonal components, zι(t_k) and Z(j(tk), or by its magnitude |z(t_k)| and phase φ (t_k).

Fig. 3 shows an example of a phasor representation of a complex clutter sample z(t_k) observed at some arbitrary time instant t .

The representation z(t) is sampled at time instants

tlj t-2, ■■■ , t_n_ι, t_n, t_n+l selected in such a way that the resulting clutter samples z(t , z(t₂), ... , z(t_n-ι), z(t_n), z(t_n+ι) are uncorrelated. Therefore, the corresponding phase angles

q> l, φ ₂, ... , φ n-l, φn, φ n+l will be mutually independent and distributed uiiiformly in the (-π, π) interval. It has been found experimentally that sampling, intervals TR of duration equal to or greater than 10 ms will generally produce uncorrelated samples of sea clutter.

A primary set of (n+1) observed phase angles φ i, φ₂, ... , φ_n-1, φ_n, φ _n+ι s utilized to construct a corresponding set of n phase differences

θi, θ₂, ... , θ_n-ι, θ_n where θ = φ ₊i - φk or k = 1, 2, ... , n.

From the fact that the phase angle is a circular variable, it follows that if two independent phase angles are distributed uniformly in the (-π, π) interval, then their difference will also have a uniform distribution in the (-π, π) interval.

For the purpose of statistical phase processing, it is convenient to represent observed phase differences θ as points on the unit circle. By construction, the angles θ and (θ +2 π ) will be represented by the same point on the circle.

A measure of dispersion R, called the mean resultant length, of the phase difference θ can be determined as follows. First, the coordinates, C and S, of the 'centre of mass' are calculated from

C = -∑cosθ_k S = -∑sinθ_k n _k=1 n _k=1

where θ _k are observed phase differences. Then, the mean resultant length R is determined as

If the points representing observed phase differences θ i, ... , θ _n are widely . dispersed on the circle, then R will assume a value close to zero. On the other hand, if θ 1 , ... , θ _n are tightly clustered, then R will be almost 1.

Fig. 4a and Fig. 4b show examples of representations of, respectively, widely dispersed angles and clustered angles, together with their corresponding mean resultant lengths R.

Since a sufficiently large sampling interval T_R is employed, the observed phase differences for clutter-alone backscatter have a circularly uniform distribution; hence, the value of the corresponding mean resultant length R will be almost 0, irrespective of the 'spikiness' of the sea clutter environment.

A constant-false-alarm rate (CFAR) detection procedure is based on the following decision rule: an object is declared as being present, if

R_N > γ(P_FA,N) = JVN

where RN is the value of the mean resultant length determined from N independent phase difference measurements, Y(PFA,N) is a predetermined decision threshold, and PPA is a specified value of false alarm probability.

For example, in the case of independent phase differences, when N = 64: γ = 0.38

for P_FA = 10^"4, γ = 0.42 for.P_FA = 10^"5, and γ = 0.46 for P_FA = 10^"6.

As indicated above, the invention is applicable also to situations in which there may be a degree of correlation between successive phase differences even in the absence of an object. In the case of dependent phase differences, the value of the decision threshold required for a specified value of false alarm probability P_FA will differ from the value of Y(P_FA,N) given above. Suitable values of the decision threshold can be determined experimentally, either from real data or from the results of computer simulations.

A microwave signal backscattered by a region of disturbed sea surface, which also contains a reflective object, can be expressed in the baseband form as

z(t) = [a, (t) + g(t)x(t)]+ j|a_Q (t) + g(t)y(t)J

where aι(t) and aς(t) are, respectively, the in-phase and quadrature components of a complex signal a(t) reflected from the object. Therefore;, the observed phase angle can be determined from

Ja_Q(t) + g(t)y(t)| (t)/g(t) + y(t)] φ(t) = arctan< — > = arctani — > t) + g(t)x(t)J tø/gtø + xtøJ

In this case, the phase φ(t) depends -on the texture component g(t); hence the clutter 'spikiness' will affect the angle of the signal-plus-clutter phasor. The texture component g(t) can be viewed as a process which, depending on its instantaneous value, either 'attenuates' or 'amplifies' the signal a(t).

Fig. 5 shows examples of combining various clutter phasors with a phasor a(t) representing a low-level constant signal. Even for a small signal, the phase distribution of the resultant signal-plus-clutter phasor can be distinguishable from a uniform distribution.

As indicated above, to avoid errors due to low-accuracy phase measurements, the influence of the phase measurements on the detection procedure can be adjusted by a weighting technique, whereby phase measurements which coincide with relatively large values of the envelope of a received signal are more influential.

In the preferred embodiment of the invention, the detection procedure utilizes phase differences rather than phases. In this case, the weight W_k assigned to a phase difference θ = φ_k+i - φk should be some function of the two envelope samples, |z(t_k+ι)| and |z(t_k)| at the times of occurrence of the individual phase measurements.

Preferably, the two arguments, |z(t_k+ι)| and |z(t_k)|, of the corresponding weight function W_k are replaced by a single argument U_k which is preferably dependent on the product of the envelope samples, and more preferably is proportional to the product and inversely proportional^'to the square root of the sum of the squares of the envelope samples. In the preferred embodiment, U_k is defined by

In the following, the above operation will be referred to as the circular average.

Preferably, a suitable non-linear function is applied to the circular average in order to obtain a weight function W in which particularly large envelope values give rise to proportionately lower weights. Fig. 6 depicts the shape of a suitable weight function W = fh(u_k) which depends on the circular average U of |z(tk₊ι)| and |z(t_k)|. Generally, the shape of the weight function should resemble that of a 'soft limiter'. Preferably, a modified CFAR detection procedure is based on the weighted mean resultant length determined from

where n 11

∑w_kcosθ_k ∑w_ksinθ_k c_w _ k=l c _ k=l n ' ° n

∑w_k k=l Σ k=l ^Wk

and the weights W_k are obtained from a suitable weight function. The n denominator ∑ w_k serves to normalise the weighted mean resultant length with k=l respect to the accumulated envelope samples, so that R_w lies between 0 and 1.

In this case, an object is declared as being present, if

where R_WN is the value of the weighted mean resultant length determined from N phase difference measurements and Y_W(P_FA_.N) is a predetermined modified decision threshold so selected as to obtain a specified value of false alarm probability PFA-

Fig. 7 is a functional block diagram of the digital signal processor DSP configured in accordance with the present invention. Although the digital signal processor DSP is in this case designed for use in the coherent pulse radar system shown in Fig. 1, it could alternatively be used in other systems, for example systems using different types of modulation, or different types of signals (e.g. electromagnetic waves in different wavebands or acoustic waves).

The digital signal processor DSP, which may be implemented as a single integrated circuit, comprises a magnitude extractor MEX, a phase extractor PEX, an auxiliary delay ADL, two magnitude storage registers, ZNE and ZOL, two phase storage registers, PNE and POL, a magnitude converter MAG, a phase subtracter DIF, a weight calculating unit WGT, two look-up tables WCO and WCI, three accumulators, AWE, AGO and ASI, a control and timing unit CTU, three storage registers, BWE, BGO, BSI, a threshold calculating unit THR, a polygonal approximator PAP, a comparator CMP, a serial-in-parallel-out binary

shift register SIPO, and a logic unit LGU.

Digital representations ZI and ZQ of the in-phase and quadrature components of a received signal, are applied simultaneously to the magnitude extractor MEX and the phase extractor PEX which determine digital representations, ZZ and PH, of respectively, the magnitude and the phase of a received signal. The functions of extractors MEX and PEX can also be performed by a single processor, known in the prior art as a Pythagoras processor.

At time instants determined by suitable clock pulses CP supplied by the auxiliary delay ADL, the current (new) values of ZZ and PH are stored, respectively, in the registers ZNE and PNE. At the same time, previous (old) values of ZZ and PH are transferred to the registers ZOL and POL. Such an arrangement makes two consecutive 'running' values of ZZ and PH available continually to magnitude converter MAG and phase subtracter DIF in a moving window fashion. The phase subtracter DIF calculates the difference PD between the phases P2 and PI supplied by the registers PNE and POL, and the magnitude converter MAG determines a digital representation the circular average CA of the two magnitudes, Z2 and Z 1 , supplied by the registers ZNE and ZOL.

The circular average CA is applied to one input of the weight calculating unit WGT whose other input is driven by a binary mode select input MS. If MS = 0, the weight calculating unit WGT supplies all weights WW of the same unit value, irrespective of the values of the circular average CA; otherwise, weights WW are digital representations of a specified, preferably non-linear, weight function, such as the function shown in Fig. 6. The weight calculating unit WGT can be implemented in the form of a suitable code converter.

The operation of the two look-up tables, WCO and WSI, can be summarised as follows. If MS = 0, then the look-up tables WCO and WSI supply, respectively, values CW and SW which are digital representations of the cosine and the sine of the input value PD, irrespective of the circular average CA value. However, if MS = 1, the outputs CW and SW become, respectively, digital representations of the product of the cosine, or the sine, of the input value PD, and a suitable weight depending on the value of the circular average CA and the shape of the weight function. The operations performed by the look-up tables WCO and WSI can also be provided by suitably configured digital multipliers.

The main function of the control and timing unit CTU is to use the clock pulses CP provided by the auxiliary delay ADL, and also the externally supplied information about the total number of pulses LP, to generate the following signals:

1. The data transfer pulses DT which are replicas of the clock pulses CP delayed by the time needed by the circuits MEX, PEX, MAG, DIF, WGT, WCO and WSI to complete their respective operations. The DT pulses follow the clock pulses, starting from pulse number 2 and including the last pulse number LP of the set. Therefore, data transfer operations occur at (LP-1) time instants,

2. The reset pulse RS pulse generated just before the occurrence of the first data transfer pulse DT. In order to perform this function, the CTU comprises a suitable pulse counter.

3. The data transfer pulse CT which is a replica of the data transfer pulse DT delayed by the time needed by the accumulators AWE, ACO and ASI to complete their respective operations. 4. The clock pulse CK used by the shift register SIPO to shift serially binary values LD. The pulse CK is delayed with respect to the data transfer pulse CT by the time needed by the circuits PAP, THR and CMP to complete their respective operations.

The weights WW and the products CW and SW determined for a predetermined number LP of samples are accumulated in the three accumulators, AWE, ACO and ASI.

Since two consecutive samples must always be used for the determination of the circular average CA and the phase difference PD, LP primary values of the magnitude ZZ and the phase PH, will result in (LP-1) values of the weight WW and products CW and SW.

In order to process information contained in a set of LP samples, the accumulators AWE, ACO and ASI are all reset to their initial state 'zero' by a pulse provided by the control and timing unit CTU, and applied to the common reset input RS. The input data, WW, CW and SW, are transferred at the (LP-1) time instants, determined by the pulses supplied by the control and timing unit CTU and appearing at the common input DT. The data transfer pulses are suitably delayed replicas of the clock pulses CP, starting from pulse number 2 and including pulse number LP (the last sample).

The operations performed jointly by the look-up table WCO and accumulator

ACO (and also those performed jointly by look-up table WSI and accumulator ASI) can also be performed by multipliers-accumulators known in the prior art.

When all LP samples have been processed, the contents of the accumulators AWE, ACO and ASI are transferred to the respective buffers BWE, BCO and BSI. A suitably timed data transfer pulse, provided by the control and timing unit CTU, is applied to the common input CT.

The output value AW of the buffer BWE is applied to the threshold calculating unit THR whose other input assumes a predetermined value PT of a decision threshold. The threshold calculating unit THR processes jointly (multiplies) values AW and PT to produce the operational value TH of the decision threshold. The output values AC and AS of the buffers BCO and BSI, are used by the polygonal approximator PAP to determine the value RL dependent on the mean resultant length (when MS = 0) or the modified mean resultant length (when MS ^: 1).

In order to simplify the required calculations, the polygonal approximator PAP utilizes the following formula if |AC| > |AS|j then RL = |AC| + |AS|/2; otherwise, RL = |AS| + |AC|/2

There are also other suitable 'polygonal approximations' known in the prior art.

The comparator CMP compares the value RL dependent on the (modified) mean resultant length supplied by the polygonal approximator PAP to the value TH of the decision threshold provided by the threshold calculating unit THR. If the decision threshold has been exceeded, then the binary output LD of the comparator CMP assumes value 1; otherwise, the value of LD is set to 0. Each decision LD is based on processing all LP sample pairs (ZI, ZQ) applied to the digital signal processor DSP.

It will be noted that this operation involves comparing RL with TH, where n

TH=PT x AW. AW is the sum of the weights (i.e. ∑ w_k for varying weights, or k=l n when MS=0 and all weights are set to unity). Accordingly the comparator operation is mathematically equivalent to comparing a normalised (weighted) mean resultant length R (or R_w )= RL/AW with the decision threshold PT. Binary integration is performed jointly by the serial-in-parallel-out binary shift register SIPO, comprising M binary cells, and the logic unit LGU; the process of binary integration is executed as follows.

The output values LD, regarded as local decisions, are shifted into the shift register SIPO at the time instants determined by the clock pulses CK supplied by the control and timing unit CTU. Each clock pulse CK is delayed with respect to the data transfer pulse CT by the time needed by the circuits PAP, THR and CMP to complete their respective operations.

The M binary outputs of the shift register SIPO are connected to the M inputs of the logic unit LGU which also receives a predetermined number K. If the total number of 'ones' appearing at the M outputs of the shift register SIPO exceeds K, then a 'global' decision DD is "object present", and DD is set to T; otherwise, DD = 0 which means "no object detected".

The foregoing description of the preferred embodiment of the invention has been presented for the purpose of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise foπn disclosed. In light of the foregoing description, it is evident that many alterations, modifications, and variations will enable those skilled in the art to utilize the invention in various embodiments suited to the particular use contemplated.

One possible modification would involve transmitting, instead of monotone pulses, pulses in which the phase or frequency is modulated in a specific manner to provide better range resolution (as is known per se in the art). Furthermore, a continuous signal, rather than pulses, could be transmitted.

The arrangement described above operates by determining the uniformity of distribution of the difference between consecutive phase measurements. However, this is not essential; the phase differences may be between any suitably- chosen pair of successive (not necessarily consecutive) measurements.

In the arrangement of Figure 7, the weights wj- are summed in order to normalise the weighted mean resultant length, so that the decision threshold can be predetermined, irrespective of the accumulated powers of the reflection. However, this is not essential; normalisation could be carried out in some other way, e.g. by measuring the average power of the reflection, or could in certain circumstances be omitted.

The arrangement described above could form part of a larger system which also uses conventional object-detection techniques.

Claims

CLAIMS:

1. A method of detecting an object in a body of liquid by receiving a reflection of a transmitted signal, the method comprising: repeatedly determining the phase of the reflection; detecting whether the distribution of successive phases of the reflection is

significantly non-uniform; and providing a signal indicative of the presence of an object in dependence on the result of the detection step; the method further including the step of weighting the influence of the phases on the detection of non-uniform distribution in accordance with the power of the reflection at the times of occurrence of the respective phases.

2. A method as claimed in claim 1, wherein the detection step comprises calculating the difference between successive phases and determining whether the distribution of said differences is significantly non-uniform.

3. A method as claimed in claim 2, including the step of calculating a value dependent on the powers of the reflection at the times of the individual phases from which a phase difference is derived in order to weight the influence of the phase difference on the detection of non-uniform distribution.

4. A method as claimed in claim 3, wherein the value is proportional to the product of the envelopes of the reflection at said times.

5. A method as claimed in claim 3 or 4, wherein the value is inversely proportional to the square root of the sum of the squares of the envelopes of the reflection at said times.

6. A method as claimed in any preceding claim, including the step of deriving weighting values from a non-linear function of the power of the reflection at the times of occurrence of the respective phases.

7. A method as claimed in claim 6, including the step of normalising a value representing the uniformity of distribution of the phases in accordance with the sum of the weighting values.

8. - A method as claimed in any preceding claim, including the step of making successive determinations of whether the distribution of phases is significantly non-uniform, and providing said signal indicative of the presence of an obj ect in dependence upon the ratio of different determinations .

,

9. A method as claimed in any preceding claim, when the transmitted signal comprises discrete coherent pulses.

-.

10. A method as claimed in any preceding claim, including the step of frequency or phase modulating the transmitted signal.

11. ^' A method as claimed in any preceding claim, wherein the intervals between successive phase measurements are no greater than a predetermined value such that the successive phase measurements are substantially uncorrelated.

12. A method as claimed in any preceding Claim, wherein the liquid is water.

13. Apparatus for detecting an object in a body of liquid, the apparatus being arranged to operate in accordance with a method as claimed in any preceding claim.

14. A signal processor for apparatus as claimed in claim 13, the signal processor being responsive to a reflected signal for ^"detecting whether the distribution of the phases of the reflection at intervals which are no less than a predetermined value is significantly non-uniform, the processor being operable to weight the influence of the phases on the detection of non-uniform distribution in accordance with the power of the reflection at the time of occurrence of the respective phases.

15. An apparatus for^' detecting an object in a body of liquid, the apparatus comprising a receiver operable to receive a reflected signal; a phase extractor operable to determine the phase of the reflected signal; a phase detector operable to determine the distribution of successive phases of the reflected signal having a first output; a phase magnitude detector operable to determine the power of each said successive phases of the reflected signal having a second output; and a weighting unit operable to weight said first output in response to said second output to provide a weighted input signal to a logic unit, said logic unit being operable to generate a signal indicative of the presence of an object based on said weighted input signal.

16. A computer program product, in a computer readable medium, for detecting an object in a body of liquid by receiving a reflection of a transmitted signal, the product comprising: instructions for repeatedly determining the phase of the reflection; instructions for detecting whether the distribution of successive phases of the reflection is significantly non-uniform; and instructions for providing a signal indicative of the presence of an object in dependence on the result of the detection step; the product further including instructions for weighting the influence of the phases on the detection of non-uniform distribution in accordance with the power of the reflection at the times of occurrence of the respective phases.