CN101187579A - Multi-wave band submarine tunnel effect elimination method - Google Patents

Abstract

The invention provides a method for eliminating multi wave beam measuring channel effect in seafloor, which comprises collecting an array original signal of a multi-beam depth sounder, processing the original signal to preformed multi wave beams, determining whether a strong mirror image echo exists or not, choosing an inference beam as a reference input and choosing a beam which is interrupted as an original input after determining an inference beam date existing, doing an operation of a self-adaptive side-lobe cancellation, determining a filter parameter and a recursive number M, a time initialization, an order initialization, a lattice type anticipator, joint process estimation filter of a lattice algorithm, and obtaining a beam output result em+1 after the side-lobe cancellation through M recursive operation. The invention effectively eliminates the side-lobe inference, increases seafloor measuring accuracy, and greatly reduces computing amount, which is applied in real time occasion. The invention can effectively control a recursive process, thereby the computing amount is further reduced, which is particularly applied to realize programmable devices such as FPGA and the like, and is beneficial for the application of a real-time occasion.

Description

Multi-wave band submarine tunnel effect elimination method

(1) technical field

What the present invention relates to is a kind of method that improves measuring accuracy, specifically a kind of method of utilizing self-adaptation secondary lobe cancellation algorithms to eliminate multi-wave band submarine tunnel effect.

(2) background technology

Multibeam sounding system has been brought into play enormous function with its higher work efficiency and total accuracy of sounding as novel seabed mapping equipment in the research work of submarine topography, be the high precision sounding device that is widely used on the boats and ships, it has greatly improved the understanding to submarine structure.Its principle is to utilize the estimation that the backscattering echoed signal in each intersection wave beam is carried out time of arrival and arrived angle, under the known situation of the velocity of sound, calculate seabed depth, yet through repeatedly observing the experimental data of multibeam sounding system, some illusions in the discovery system output data, these illusions can be given the explanation that makes mistake to submarine topography.By relatively intersecting depth measurement band and analyze digitizing original acoustic data and obtained confirmation, the cause of these illusions is not because there is mistake in the sea-bottom contour algorithm that uses to these illusions, but because the echo that uses detects and disposal route improper due to.Rule of thumb such detection mistake and character of the bottom have much relations (such as the unexpected variation on geology or slope) or come from the interference (such as the seismic origin) of other sound source.

The multiple sum of errors that exists in the actual depth measurement process disturbs and causes measurement data inaccurate, thereby influences the habitata precision, causes the distortion of seabed form.Wherein velocity of sound section is by the crooked habitata precision that directly influences of sound ray, when velocity of sound section modified value during less than the value of reality, the tip-tilted spill landform in both sides will occur, and the depth of water of its core is also corresponding to shoal; When velocity of sound section modified value during greater than the value of reality, both sides will appear to the convex landform of sinking, and the depth of water of its core is also corresponding to have deepened.The far-reaching factor of another one is that secondary lobe disturbs, very strong mirror is released into the main lobe of other wave beam to the secondary lobe of wave beam, when measured seabed relatively flat and when having very strong reflection, the secondary lobe interference phenomenon is more obvious, multibeam sounding system can be drawn the both sides concave upright submarine topography of tilting upward, and so-called " tunnel effect " occur.

Though find the existence of tunnel effect in traditional multi-wave band submarine, up to the present do not further investigate and propose effectively to eliminate means, and the research in more concern sound ray correction field.Yet effectively eliminate the precision that tunnel effect can further improve multi-wave band submarine.

(3) summary of the invention

The object of the present invention is to provide a kind of multi-wave band submarine tunnel effect elimination method that can improve mapping precision.

The object of the present invention is achieved like this:

(1) gathers multibeam echosounder basic matrix original signal;

(2) original signal is carried out the preformation multi-beam and handle, determined whether that strong mirror image echo exists;

(3) there is the beam data of disturbing to determining, gets interfering beam, choose disturbed wave beam, carry out the self-adaptation secondary lobe and offset operation as original input as with reference to input;

(4) determine the recurrence number of times M of filter parameter and lattice type algorithm;

(5) time initialization (n=0), m=0,1, L M, wherein M is the final exponent number of least square lattice type fallout predictor

${\mathrm{\δ}}_m(-1)={\mathrm{\δ}}_m^d(-1)=0,$

${\mathrm{\ϵ}}_m^f(-1)={\mathrm{\ϵ}}_m^b(-1)=\mathrm{\δ}>0,$ δ is little positive constant

γ _m(-1)＝1，

$e_m^b(-1)=0.;$

(6) exponent number initialization (m=0), n 〉=0

$e_0^f\left(n\right)=e_0^b\left(n\right)=x\left(n\right),$

${\mathrm{\ϵ}}_0^b\left(n\right)={\mathrm{\ϵ}}_0^f\left(n\right)=\mathrm{\λ}{\mathrm{\ϵ}}_0^f(n-1)+{\left|x\left(n\right)\right|}^2,$

γ ₀(n)＝1，

e ₀(n)＝d(n).；

(7) lattice type fallout predictor, beginning n=0, to m=0,1, L M calculates the renewal on each rank

${\mathrm{\δ}}_m\left(n\right)=\mathrm{\λ}{\mathrm{\δ}}_m(n-1)+e_m^{b*}(n-1)e_m^f\left(n\right)/{\mathrm{\γ}}_m(n-1),$

${\mathrm{\γ}}_{m+1}\left(n\right)={\mathrm{\γ}}_m\left(n\right)-{\left|e_m^b\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^b\left(n\right),$

$k_m^b\left(n\right)={\mathrm{\δ}}_m\left(n\right)/{\mathrm{\ϵ}}_m^f\left(n\right),$

$k_m^f\left(n\right)={\mathrm{\δ}}_m^{*}\left(n\right)/{\mathrm{\ϵ}}_m^b(n-1),$

$e_{m+1}^b\left(n\right)=e_m^b(n-1)-k_m^b\left(n\right)e_m^f\left(n\right),$

$e_{m+1}^f\left(n\right)=e_m^f\left(n\right)-k_m^f\left(n\right)e_m^b(n-1),$

${\mathrm{\ϵ}}_{m+1}^b\left(n\right)={\mathrm{\ϵ}}_m^b(n-1)-{\left|{\mathrm{\δ}}_m\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^f\left(n\right),$

${\mathrm{\ϵ}}_{m+1}^f\left(n\right)={\mathrm{\ϵ}}_m^f\left(n\right)-{\left|{\mathrm{\δ}}_m\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^b(n-1).;$

(8) combined process estimation filter, beginning n=0, to m=0,1, L M calculates the renewal on each rank

${\mathrm{\δ}}_m^d\left(n\right)=\mathrm{\λ}{\mathrm{\δ}}_m^d(n-1)+e_m^{b*}\left(n\right)e_m\left(n\right)/{\mathrm{\γ}}_m\left(n\right),$

$v_m\left(n\right)={\mathrm{\δ}}_m^d\left(n\right)/{\mathrm{\ϵ}}_m^b\left(n\right),$

$e_{m+1}\left(n\right)=e_m\left(n\right)-v_m\left(n\right)e_m^b\left(n\right).;$

(9) obtain secondary lobe by M recursive operation and offset back wave beam output e as a result _M+1

The present invention can also comprise:

1, the described definite method that determines whether that strong mirror image echo exists is promptly to determine to disturb to exist less than decision threshold by all direction main lobes in the more same time and side-lobe energy ratio.

2, described get interfering beam as with reference to input and choose disturbed wave beam as the method for original input for based on self-adaptation secondary lobe principle of cancellation, with disturbed wave beam s+n ₀As original input, with interfering beam n ₁As with reference to input, utilize interference secondary lobe and input with reference to relevant, estimate secondary lobe interference output y with the hypothesis that disturbed wave beam main lobe is irrelevant; Sef-adapting filter is to n ₁Filter and produce approximate secondary lobe interference output y, should export from disturbed wave beam s+n ₀In deduct the output s+n of the system that obtains ₀-y is real seabed involuting wave signal.

Fig. 1 has provided self-adaptation secondary lobe principle of cancellation, is example for ease of explanation with the larboard wave beam.Original input channel s+n ₀Get wave beam No. 2, with reference to input n ₁Get wave beam No. 0, n ₀Being No. 0 wave beam disturbs the secondary lobe of No. 2 wave beams, and s is the main lobe of No. 2 wave beams.Suppose that secondary lobe disturbs n ₀With reference input n ₁The main lobe height correlation, the main lobe of the main lobe of No. 0 wave beam and No. 2 wave beams is irrelevant, this supposition tallies with the actual situation.Sef-adapting filter is to n ₁Filter and produce approximate secondary lobe interference n ₀Output y, should export from original input s+n ₀In deduct the output s+n of the system that obtains ₀-y is real seabed involuting wave signal.Suppose s, n ₀And n ₁All be to add up steady and have zero-mean, all sides are got on both sides, consider s and n ₀And y is uncorrelated, obtains:

E[e ²]＝E[s] ²+E[(n ₀-y)] ²+2E[s(n ₀-y)]＝E[s ²]+E[(n ₀-y) ²]

Make E[e when regulating wave filter ²] hour, signal power is ground E[s ²] with unaffected, E[(n ₀-y) ²] reaching minimum, corresponding minimum output power is E _Min[e ²]=E[s ²]+E _Min[(n ₀-y) ²], wave filter output this moment y is secondary lobe and disturbs n ₀The best all just estimate.When output power reaches minimum possible value is E _Min[e ²]=E[s ²], this moment E[(n ₀-y) ²]=0, so y=n ₀, e=s, this moment, the secondary lobe interference signals was removed in wave filter output fully.If original input, that is to say that No. 0 wave beam does not produce secondary lobe to No. 2 wave beams and disturbs with uncorrelated fully with reference to input, the output y of wave filter is uncorrelated with original input in this case, and output power is

E[e ²]＝E[(s+n ₀) ²]-2E[y(s+n ₀)]+E[y ²]＝E[(s+n ₀) ²]+E[y ²]

Make the output power minimum, require E[y ²] minimum, make that all coefficients of wave filter are zero, wave filter output E[y ²] be zero, original input signal is impregnable by system, and promptly No. 2 wave beams are almost less than the wave filter that passes through of decaying.

Among the present invention, when all just estimating, the best adopted combined process recurrence least square lattice types (JPLSL) algorithm finding the solution, as shown in Figure 2, and the parameter k of lattice filter _m ^f(n) and k _m ^b(n) be forward reflection coefficient and retroreflection coefficient, e _m ^f(n) and e _m ^b(n) be forward prediction sum of errors back forecast error, ε _m ^f(n) and ε _m ^b(n) be forward prediction residual sum back forecast residual error, δ _m(n) be e _m ^f(n) and e _m ^b(n-1) partial correlation coefficient, γ _m(n) be conversion factor between prior uncertainty and the posteriority error, we adopt the posteriority evaluated error at this, because in the adaptive updates process, wave filter is always become better and better.v _m(n) be regression coefficient, e _m(n) be the output error that combined process is estimated, δ _m ^d(n) be e _m ^b(n) and e _m(n) partial correlation coefficient.The back forecast error of the continuous exponent number of lattice filter is incoherent, λ is forgetting factor and 0＜λ≤1, its effect is to adding bigger error from n near more error of the moment, and error far away is more added smaller error, because the information in past is concerning coefficient update, it ignores degree is ever-increasing.

Major advantage of the present invention is:

(1) utilizes self-adaptation secondary lobe principle of cancellation, disturb elimination because secondary lobe disturbs the tunnel effect that causes, improved the subsea survey precision thereby eliminated secondary lobe effectively.

(2) utilize combined process recurrence least square lattice types (JPLSL) algorithm to find the solution all square estimation problem of the best in the sef-adapting filter, the algorithm high modularization reduces operand greatly, is fit to real-time occasion.

(3) thus utilize that combined process estimates in the middle of the output result can effectively control recursive process and further reduce operand.

(4) high modularization of lattice structure is realized in hardware effectively, especially is fit to programming device realizations such as FPGA, thereby helps real-time applications.

(4) description of drawings

Fig. 1 is a self-adaptation secondary lobe principle of cancellation block diagram of the present invention;

Fig. 2 is a self-adaptation recurrence least square lattice filter;

Fig. 3 does not carry out the multiple-beam system lakebed data beam formation result that secondary lobe is offset;

Fig. 4 is that the multiple-beam system lakebed data beam that carries out after secondary lobe is offset forms the result.

(5) embodiment

For example the present invention is done description in more detail below in conjunction with accompanying drawing:

Carry out secondary lobe at the true lakebed data of certain multibeam echosounder and offset processing.This type multibeam echosounder adopts horizontal basic matrix, and frequency of operation is 180kHz, and the depth measurement scope is a 1-200 rice, 21 wave beams of 8 passages, laterally wide 6.67 degree of received beam, wide 8 degree of fore-and-aft direction, mirror has 1 wave beam to direction, each 10 wave beam of port and starboard, sample frequency is 30kHz, and the time interval between two sampled points (timeslice) is approximately 33 microseconds, because the beam angle broad, therefore the secondary lobe interference ratio is more serious, its certain measure the wave beam time series as shown in Figure 3.Method is carried out secondary treating to data among employing the present invention, on filter parameter is chosen, gets forgetting factor λ=0.99, and littler λ will cause restraining faster, but increase imbalance; Initiation parameter δ=0.01, along with the increase of iterations, the deviation of the sef-adapting filter estimation coefficient that initialization causes goes to zero rapidly when λ＜1; Filter order M gets M=8, when secondary lobe disturb very strong and filter order can be got when occurring continuously in time more greatly, the JPLSL algorithm is revised at every turn needs about 20M computing, computational complexity is proportional to M, big 1 M is complete acceptable.

According to the method described in the present invention, choose central beam (MIRROR SITE) as a reference, respectively other wave beams are carried out secondary lobe and offset operation, concrete steps are as follows:

(1) time initialization (n=0), m=0,1, L M (M is the final exponent number of least square lattice type fallout predictor)

${\mathrm{\δ}}_m(-1)={\mathrm{\δ}}_m^d(-1)=0,$

${\mathrm{\ϵ}}_m^f(-1)={\mathrm{\ϵ}}_m^b(-1)=\mathrm{\δ}>0,$ δ is little positive constant

γ _m(-1)＝1，

$e_m^b(-1)=0.$

(2) exponent number initialization (m=0), n 〉=0

$e_0^f\left(n\right)=e_0^b\left(n\right)=x\left(n\right),$

${\mathrm{\ϵ}}_0^b\left(n\right)={\mathrm{\ϵ}}_0^f\left(n\right)=\mathrm{\λ}{\mathrm{\ϵ}}_0^f(n-1)+{\left|x\left(n\right)\right|}^2,$

γ ₀(n)＝1，

e ₀(n)＝d(n).

(3) lattice type fallout predictor, beginning n=0, to m=0,1, L M calculates the renewal on each rank

${\mathrm{\δ}}_m\left(n\right)=\mathrm{\λ}{\mathrm{\δ}}_m(n-1)+e_m^{b*}(n-1)e_m^f\left(n\right)/{\mathrm{\γ}}_m(n-1),$

${\mathrm{\γ}}_{m+1}\left(n\right)={\mathrm{\γ}}_m\left(n\right)-{\left|e_m^b\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^b\left(n\right),$

$k_m^b\left(n\right)={\mathrm{\δ}}_m\left(n\right)/{\mathrm{\ϵ}}_m^f\left(n\right),$

$k_m^f\left(n\right)={\mathrm{\δ}}_m^{*}\left(n\right)/{\mathrm{\ϵ}}_m^b(n-1),$

$e_{m+1}^b\left(n\right)=e_m^b(n-1)-k_m^b\left(n\right)e_m^f\left(n\right),$

$e_{m+1}^f\left(n\right)=e_m^f\left(n\right)-k_m^f\left(n\right)e_m^b(n-1),$

${\mathrm{\ϵ}}_{m+1}^b\left(n\right)={\mathrm{\ϵ}}_m^b(n-1)-{\left|{\mathrm{\δ}}_m\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^f\left(n\right),$

${\mathrm{\ϵ}}_{m+1}^f\left(n\right)={\mathrm{\ϵ}}_m^f\left(n\right)-{\left|{\mathrm{\δ}}_m\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^b(n-1).$

(4) combined process estimation filter, beginning n=0, to m=0,1, L M calculates the renewal on each rank

${\mathrm{\δ}}_m^d\left(n\right)=\mathrm{\λ}{\mathrm{\δ}}_m^d(n-1)+e_m^{b*}\left(n\right)e_m\left(n\right)/{\mathrm{\γ}}_m\left(n\right),$

$v_m\left(n\right)={\mathrm{\δ}}_m^d\left(n\right)/{\mathrm{\ϵ}}_m^b\left(n\right),$

$e_{m+1}\left(n\right)=e_m\left(n\right)-v_m\left(n\right)e_m^b\left(n\right).$

After 8 rank recursive operation, the beam pattern of output as shown in Figure 4, as seen this algorithm secondary lobe neutralization effect is very obvious, can effectively eliminate because the measuring error brought of tunnel effect.

Claims (3)

1. multi-wave band submarine tunnel effect elimination method is characterized in that:

(1) gathers multibeam echosounder basic matrix original signal;

(2) original signal is carried out the preformation multi-beam and handle and determined whether strong mirror image echo existence;

(3) there is the beam data of disturbing to determining, gets interfering beam, choose disturbed wave beam, carry out the self-adaptation secondary lobe and offset operation as original input as with reference to input;

(4) determine the recurrence number of times M of filter parameter and lattice type algorithm;

(5) time initialization (n=0), m=0,1, L M, wherein M is the final exponent number of least square lattice type fallout predictor

${\mathrm{\δ}}_m(-1)={\mathrm{\δ}}_m^d(-1)=0,$

${\mathrm{\ϵ}}_m^f(-1)={\mathrm{\ϵ}}_m^b(-1)=\mathrm{\δ}>0,$ δ is little positive constant

γ _m(-1)＝1，

$e_m^b(-1)=0.;$

(6) exponent number initialization (m=0), n 〉=0

$e_0^f\left(n\right)=e_0^b\left(n\right)=x\left(n\right),$

${\mathrm{\ϵ}}_0^b\left(n\right)={\mathrm{\ϵ}}_0^f\left(n\right)=\mathrm{\λ}{\mathrm{\ϵ}}_0^f(n-1)+{\left|x\left(n\right)\right|}^2,$

γ ₀(n)＝1，

e ₀(n)＝d(n).；

(7) lattice type fallout predictor, beginning n=0, to m=0,1, L M calculates the renewal on each rank

${\mathrm{\δ}}_m\left(n\right)=\mathrm{\λ}{\mathrm{\δ}}_m(n-1)+e_m^{b*}(n-1)e_m^f\left(n\right)/{\mathrm{\γ}}_m(n-1),$

${\mathrm{\γ}}_{m+1}\left(n\right)={\mathrm{\γ}}_m\left(n\right)-{\left|e_m^b\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^b\left(n\right),$

$k_m^b\left(n\right)={\mathrm{\δ}}_m\left(n\right)/{\mathrm{\ϵ}}_m^f\left(n\right),$

$k_m^f\left(n\right)={\mathrm{\δ}}_m^{*}\left(n\right)/{\mathrm{\ϵ}}_m^b(n-1),$

$e_{m+1}^b\left(n\right)=e_m^b(n-1)-k_m^b\left(n\right)e_m^f\left(n\right),$

$e_{m+1}^f\left(n\right)=e_m^f\left(n\right)-k_m^f\left(n\right)e_m^b(n-1),$

${\mathrm{\ϵ}}_{m+1}^b\left(n\right)={\mathrm{\ϵ}}_m^b(n-1)-{\left|{\mathrm{\δ}}_m\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^f\left(n\right),$

${\mathrm{\ϵ}}_{m+1}^f\left(n\right)={\mathrm{\ϵ}}_m^f\left(n\right)-{\left|{\mathrm{\δ}}_m\left(n\right)\right|}^2/{\mathrm{\ϵ}}_m^b(n-1).;$

(8) combined process estimation filter, beginning n=0, to m=0,1, L M calculates the renewal on each rank

${\mathrm{\δ}}_m^d\left(n\right)={\mathrm{\λ\δ}}_m^d(n-1)+e_m^{b*}\left(n\right)e_m\left(n\right)/{\mathrm{\γ}}_m\left(n\right),$

$v_m\left(n\right)={\mathrm{\δ}}_m^d\left(n\right)/{\mathrm{\ϵ}}_m^b\left(n\right),$

$e_{m+1}\left(n\right)=e_m\left(n\right)-v_m\left(n\right)e_m^b\left(n\right).;$

(9) obtain secondary lobe by M recursive operation and offset back wave beam output e as a result _M+1

2. multi-wave band submarine tunnel effect elimination method according to claim 1 is characterized in that: the described definite method that determines whether that strong mirror image echo exists is promptly to determine to disturb to exist less than decision threshold by all direction main lobes in the more same time and side-lobe energy ratio.

3. multi-wave band submarine tunnel effect elimination method according to claim 1 and 2, it is characterized in that: described get interfering beam as with reference to input and choose disturbed wave beam as the method for original input for based on self-adaptation secondary lobe principle of cancellation, with disturbed wave beam s+n ₀As original input, with interfering beam n ₁As with reference to input, utilize interference secondary lobe and input with reference to relevant, estimate secondary lobe interference output y with the hypothesis that disturbed wave beam main lobe is irrelevant; Sef-adapting filter is to n ₁Filter and produce approximate secondary lobe interference output y, should export from disturbed wave beam s+n ₀In deduct the output s+n of the system that obtains ₀-y is real seabed involuting wave signal.

Images