CN104484891A - An underwater terrain matching method based on textural feature and terrain feature parameters - Google Patents

Abstract

The invention relates to the digital image processing field, and specifically relates to an underwater terrain matching method based on textural feature and terrain feature parameters. The method comprises the followings steps: carrying out downsampling on actually-measured underwater terrain data and reference data; obtaining an actually-measured data matrix and a reference data matrix which are same in resolution; obtaining angular second moment, contrast, correlation and deficit moments; obtaining a terrain elevation average value, terrain elevation standard deviation, terrain roughness, terrain correlation coefficient, terrain entropy and terrain difference entropy; endowing weight for each feature parameter; keeping a similarity value of each searching position; and selecting the positions, of which the similarities are largest, that is, the similarity values are the smallest, as matching positions for outputting. According to the method, through carrying out downsampling on the actually-measured underwater terrain data and the underwater terrain reference data of a special area, the calculation cost is reduced, underwater measured data precision is improved, and moreover, the underwater terrain matching effect based on the textural feature and terrain feature parameters is achieved.

Description

A kind of underwater terrain matching method based on textural characteristics and terrain feature parameter

Technical field

What the present invention relates to is a kind of digital image processing field.A kind of specifically underwater terrain matching method based on textural characteristics and terrain feature parameter.

Background technology

Along with the continuous progress of science and technology, underwater hiding-machine is towards the future development that can perform the large degree of depth, long distance task, and this has higher requirement to underwater navigation technology.As most widely used two kinds of navigation means, inertial navigation system and accommodation push away method and all have the shortcoming self being difficult to overcome, i.e. the positioning error feature of dispersing in time.Therefore, they and be not suitable for completing alone long distance, for a long time navigation task.

The great advantage of Models in Terrain Aided Navigation is that error is not accumulated with the increase of distance in time.In addition, Terrain-aided Navigation also has high precision, long-acting, hidden and completely autonomous advantage, and therefore, this technology is the good auxiliary of underwater hiding-machine navigational system, and in causing the attention of various countries in recent years.But China is at the early-stage for the research of topographical navigation technology, and mainly concentrates on aviation field, then few for its underwater applied research.

At present, it is also unknown whether external submarine is equipped with underwater terrain matching system, but carried out actual loading test on AUV.The BPOZAUV sea examination that in May, 2002 to June, NATO organized six units to carry out, platform is the latent device of HUG wound that Norway KONGSBERG develops, there are two weeks Special Survey navigational system, wherein just comprise the topographical navigation device developed by FFI (Norway Defence Research Establishment), carry out test of many times again after this.The OBERON device of diving of Australia is also equipped with Models in Terrain Aided Navigation.U.S.'s " America and Japan's defence " report on January 6th, 2006, Lockheed Martin Corporation and US military have signed the revision contract of 1,060 ten thousand dollars, sensor array is integrated in " the unmanned device of diving of development in advance " of naval, makes it have three-dimensional obstacle detection and communicate with identification, VHF and three-dimensional habitata ability.This kind of so unmanned device of diving just has possessed the hardware condition of topographical navigation.

About underwater terrain matching method, what mostly adopt at present is the TERCOM method and the SITAN method that come from aviation field.As described in document [1] and document [2], methodical ultimate principle is all to survey the topography elevation along navigation route, using it and relevant numerical map as the input of matching process, obtains the optimal estimation value of position or inertial navigation system error.But, because the data precision of underwater survey is lower than airborne survey, and underwater topography feature is different from land landform, therefore, the matching process of simple reference other field can not reach good locating effect, and the underwater terrain matching method that research is applicable to underwater environment is significant.

The domestic research about underwater terrain matching method is not also very ripe, and the pertinent literature of underwater terrain matching method is not also a lot, from existing document, its to the research emphasis of underwater terrain matching method mainly based on the matching process that landform altitude is relevant.

List of references related to the present invention comprises:

Tian Minfeng. the underwater hiding-machine terrain auxiliary navigation method based on priori topographic data processing studies [J]. Harbin Engineering University Ph.D. Dissertation, 2007.

Liu Chengxiang. the Terrain Matching Assistant Positioning Technology research of underwater hiding-machine. Harbin Engineering University Ph.D. Dissertation, 2003.

Summary of the invention

The object of this invention is to provide a kind of underwater terrain matching method based on underwater topography textural characteristics and terrain feature parameter.

The object of the present invention is achieved like this:

(1) actual measurement underwater topography data and reference data are carried out down-sampled;

(2) interpolation is carried out to the data after down-sampled, obtain measured data matrix and the reference data matrix of equal resolution;

(3) according to gray level co-occurrence matrixes, calculate the textural characteristics of measured data matrix, obtain angle second order distance, contrast, be correlated with and 4 textural characteristics parameters such as unfavourable balance square;

(4) calculate the terrain feature parameter of measured data matrix, obtain landform altitude average, landform altitude standard deviation, terrain roughness, landform related coefficient, terrain entropy and terrain variance entropy 6 terrain feature parameters;

(5) by vectorial for 10 characteristic parameter composition characteristics of step (3) and step (4), the feature of corresponding shaped area is characterized; According to the priori of landform characteristic parameter, for each characteristic parameter gives weights;

(6) by step length searching reference data matrix, for each searching position, proper vector is calculated, compare with the proper vector of measured data matrix, calculate similarity, similarity measure adopts least-mean-square-error criterion, retains the similarity figure of each searching position;

(7) the maximum position that namely similarity figure is minimum of wherein similarity is chosen, as the matched position exported.

Gray level co-occurrence matrixes

p(m,n,d,θ)＝{(i,j),(i+Δi,j+Δj)|f(i,j)＝m,f(i+Δi,j+Δj)＝n}，

It is m=f (i, j) that landform altitude image I mono-coordinate points (i, j) of gray level to be L size be M × N puts gray-scale value, another coordinate points (the i+ Δ i departed from, j+ Δ j) gray-scale value n=f (i+ Δ i, j+ Δ j), putting right gray-scale value is (m, n), gray level co-occurrence matrixes P is the square formation of L × L, and i, j are respectively transverse and longitudinal coordinate a little, i=0,1,2 ... M-1, j=0,1,2 ..., N-1; M, n=0,1 ..., L-1; Δ i, Δ j are the side-play amounts of pixel position; D is the generation step-length of gray level co-occurrence matrixes; θ is the generation direction of gray level co-occurrence matrixes;

Described angle second moment:

$E=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}p{(m,n)}^2,$

Contrast:

$I=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}{(m-n)}^{2}p(m,n),$

Relevant:

$C=[\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}\left(\left(\mathrm{mn}\right)p(m,n)\right)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y]/{\mathrm{\δ}}_x{\mathrm{\δ}}_y,$

Wherein μ _x, μ _y, meet

${\mathrm{\μ}}_x=\underset{m}{\mathrm{\Σ}}m\underset{n}{\mathrm{\Σ}}p(m,n),$

${\mathrm{\μ}}_y=\underset{n}{\mathrm{\Σ}}n\underset{m}{\mathrm{\Σ}}p(m,n),$

${\mathrm{\δ}}_x^2=\underset{m}{\mathrm{\Σ}}{(m-{\mathrm{\μ}}_x)}^2\underset{n}{\mathrm{\Σ}}p(m,n),$

${\mathrm{\δ}}_y^2=\underset{n}{\mathrm{\Σ}}{(n-{\mathrm{\μ}}_y)}^2\underset{m}{\mathrm{\Σ}}p(m,n),$

Unfavourable balance square:

$L=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}\frac{1}{1+{(m-n)}^2}p(m,n).$

Landform altitude average

$\stackrel{\‾}{z}=\frac{1}{\mathrm{mn}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}z(i,j),$

Landform altitude standard deviation sigma,

$D\left(z\right)=\frac{1}{m(n-1)}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(z(i,j)-\stackrel{\‾}{z})}^2,$

$\mathrm{\σ}=\sqrt{D\left(z\right)},$

Terrain roughness r,

$r=\frac{S_S}{S},$

S is regional feature surface area, S _sfor the projected area of this regional feature,

Landform coefficient R,

$R_{\mathrm{\λ}}=\frac{1}{(m-1)n{\mathrm{\σ}}^2}\underset{i=1}{\overset{m-1}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}[z(i,j)-\stackrel{\‾}{z}][z(i+1,j)-\stackrel{\‾}{z}],$

$R_{\mathrm{\φ}}=\frac{1}{m(n-1){\mathrm{\σ}}^2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n-1}{\mathrm{\Σ}}}[z(i,j)-\stackrel{\‾}{z}][z(i,j+1)-\stackrel{\‾}{z}],$

$R=\frac{R_{\mathrm{\λ}}+R_{\mathrm{\φ}}}{2},$

R _λ, R _φbe respectively longitude, latitudinal related coefficient;

Terrain entropy H _f,

$H_f=-\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{p}_{\mathrm{ij}}\lg p_{\mathrm{ij}},$

Wherein p _ijit is the normalization height value at topographic(al) point coordinate place;

Terrain variance entropy H _e,

$H_e=-\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{p}_{i,j}\left(z\right)\lg p_{i,j}\left(z\right),$

Wherein D _z(i, j) is terrain differences value:

$D_z(i,j)=\frac{|z(i,j)-\stackrel{\‾}{z}|}{\stackrel{\‾}{z}}$

Utilize terrain differences value shape disparity probability p calculably _i,j():

$p_{i,j}\left(z\right)=\frac{D_z(i,j)}{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{D}_z(i,j)}.$

Described proper vector:

$\stackrel{\→}{c}=W{(E,I,C,L,\stackrel{\‾}{z},\mathrm{\σ},r,R,H_f,H_e)}^T,$

$W=\begin{array}{cccccccccc}{\mathrm{\ω}}_E& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& {\mathrm{\ω}}_I& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& {\mathrm{\ω}}_C& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& {\mathrm{\ω}}_L& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& {\mathrm{\ω}}_{\stackrel{\‾}{z}}& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& {\mathrm{\ω}}_{\mathrm{\σ}}& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_r& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_R& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& {{\mathrm{\ω}}_H}_f& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& {{\mathrm{\ω}}_H}_e\end{array}$

Wherein, W is weight matrix, ω _e, ω _i, ω _c, ω _l, ω _σ, ω _r, ω _r, be respectively corresponding weights.

Described similarity

$F_S=|\stackrel{\→}{c}-\stackrel{{\→}^{\′}}{c}|$

Wherein, the eigenvector in actual measurement region, it is the eigenvector carrying out the reference area subregion mated in each search.

Beneficial effect of the present invention is: the underwater topography reference data of the present invention to the underwater topography data measured in real time and specific region is carried out down-sampled, assess the cost to reduce, improve the data of underwater survey, reach the underwater terrain matching effect based on underwater topography textural characteristics and terrain feature parameter further.

Accompanying drawing explanation

Fig. 1 is the underwater topography 3-D view after down-sampled and interpolation, and the lower zone that red solid line is determined is actual measurement region, and view picture figure is reference area;

Fig. 2 is the actual measurement zone location result with rotating deviation, and dotted yellow line region is actual position, and blue rectangle region is maximum 5 positions of similarity;

Fig. 3 is the actual measurement zone location result under different noise background, and dotted yellow line region is actual position, and blue rectangle region is maximum 5 positions of similarity;

Fig. 3 a is signal to noise ratio (S/N ratio) positioning result figure when being 10dB;

Fig. 3 b is signal to noise ratio (S/N ratio) positioning result figure when being 2dB;

Fig. 4 is the positioning result under different actual measurement region area, and dotted yellow line region is actual position, and blue rectangle region is maximum 5 positions of similarity;

Fig. 4 a is positioning result figure when actual measurement region is 50 × 50 meters;

Fig. 4 b is positioning result figure when actual measurement region is 100 × 100 meters;

Fig. 5 is this method process flow diagram.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described further.

Its concrete methods of realizing is as follows:

(1) carry out down-sampled to actual measurement underwater topography data and reference data;

(2) interpolation is carried out to the data after down-sampled, obtain measured data matrix and the reference data matrix of equal resolution;

(3) calculate the gray level co-occurrence matrixes P of measured data, obtain textural characteristics parameter;

For the landform altitude image I (M is image length, and N is picture traverse) that a width size is M × N, if its gray level is L.Get arbitrary coordinate points (i, j) in I, this gray-scale value is m=f (i, j), gray-scale value n=f (the i+ Δ i of its another coordinate points (i+ Δ i, j+ Δ j) is departed from, j+ Δ j), then the gray-scale value that this point is right is (m, n).The number of times that every bit occurs (m, n) is counted to whole matrix I, is arranged in the square formation P of a L × L.Occurrence number normalization is obtained each point to the Probability p (m, n, d, θ) occurred, namely obtain the gray level co-occurrence matrixes P of this landform altitude image.

p(m,n,d,θ)＝{(i,j),(i+Δi,j+Δj)|f(i,j)＝m,f(i+Δi,j+Δj)＝n}

In formula, i, j are respectively the transverse and longitudinal coordinate of this point, i=0, and 1,2 ..., M-1, j=0,1,2 ..., N-1; M, n=0,1 ..., L-1; Δ i, Δ j are the side-play amounts of pixel position; D is the generation step-length of gray level co-occurrence matrixes; θ is the generation direction of gray level co-occurrence matrixes.In practical application, often suitably choose d, and θ gets 0 °, 45 °, 90 °, 135 °.

On this basis, the present invention calculates following 4 textural characteristics, and wherein p (m, n, d, θ) is abbreviated as p (m, n):

Angle second moment (energy) E

$E=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}p{(m,n)}^2$

Contrast I

$I=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}{(m-n)}^{2}p(m,n)$

Relevant C

$C=[\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}\left(\left(\mathrm{mn}\right)p(m,n)\right)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y]/{\mathrm{\δ}}_x{\mathrm{\δ}}_y$

Wherein μ _x, μ _y, meet

${\mathrm{\μ}}_x=\underset{m}{\mathrm{\Σ}}m\underset{n}{\mathrm{\Σ}}p(m,n)$

${\mathrm{\μ}}_y=\underset{n}{\mathrm{\Σ}}n\underset{m}{\mathrm{\Σ}}p(m,n)$

${\mathrm{\δ}}_x^2=\underset{m}{\mathrm{\Σ}}{(m-{\mathrm{\μ}}_x)}^2\underset{n}{\mathrm{\Σ}}p(m,n)$

${\mathrm{\δ}}_y^2=\underset{n}{\mathrm{\Σ}}{(n-{\mathrm{\μ}}_y)}^2\underset{m}{\mathrm{\Σ}}p(m,n)$

Local stationary (unfavourable balance square) L

$L=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}\frac{1}{1+{(m-n)}^2}p(m,n)$

(4) calculate the terrain feature parameter of measured data matrix, obtain terrain feature parameter;

If the longitude and latitude span of certain shaped area is m × n point, m is latitude direction span, and n is longitudinal span, and z (i, j) is the height value at point coordinate (i, j) place, present invention employs following 6 terrain feature parameters:

Landform altitude average

$\stackrel{\‾}{z}=\frac{1}{\mathrm{mn}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}z(i,j)$

Landform altitude standard deviation sigma

$D\left(z\right)=\frac{1}{m(n-1)}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(z(i,j)-\stackrel{\‾}{z})}^2$

$\mathrm{\σ}=\sqrt{D\left(z\right)}$

Terrain roughness r

$r=\frac{S_S}{S}$

S is regional feature surface area, S _sfor the projected area of this regional feature

Landform coefficient R

$R_{\mathrm{\λ}}=\frac{1}{(m-1)n{\mathrm{\σ}}^2}\underset{i=1}{\overset{m-1}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}[z(i,j)-\stackrel{\‾}{z}][z(i+1,j)-\stackrel{\‾}{z}]$

$R_{\mathrm{\φ}}=\frac{1}{m(n-1){\mathrm{\σ}}^2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n-1}{\mathrm{\Σ}}}[z(i,j)-\stackrel{\‾}{z}][z(i,j+1)-\stackrel{\‾}{z}]$

$R=\frac{R_{\mathrm{\λ}}+R_{\mathrm{\φ}}}{2}$

R _λ, R _φbe respectively longitude, latitudinal related coefficient

Terrain information entropy H _f

$H_f=-\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{p}_{\mathrm{ij}}\lg p_{\mathrm{ij}}$

Wherein p _ijit is the normalization height value at topographic(al) point coordinate place.

Terrain variance entropy H _e

$H_e=-\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{p}_{i,j}\left(z\right)\lg p_{i,j}\left(z\right)$

Wherein D _z(i, j) is terrain differences value, can be calculated by following formula:

$D_z(i,j)=\frac{|z(i,j)-\stackrel{\‾}{z}|}{\stackrel{\‾}{z}}$

Utilize terrain differences value shape disparity probability p calculably _i,j():

$p_{i,j}\left(z\right)=\frac{D_z(i,j)}{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{D}_z(i,j)}$

(5) by (3) and the vector of the characteristic parameter composition characteristic described in (4) be used for characterizing the feature of corresponding shaped area; According to the priori of different topographic characteristics parameter, for each characteristic parameter gives specific weights, make different terrain characteristic of correspondence vector significant difference;

$\stackrel{\→}{c}=W{(E,I,C,L,\stackrel{\‾}{z},\mathrm{\σ},r,R,H_f,H_e)}^T$

$W=\begin{array}{cccccccccc}{\mathrm{\ω}}_E& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& {\mathrm{\ω}}_I& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& {\mathrm{\ω}}_C& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& {\mathrm{\ω}}_L& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& {\mathrm{\ω}}_{\stackrel{\‾}{z}}& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& {\mathrm{\ω}}_{\mathrm{\σ}}& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_r& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_R& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& {{\mathrm{\ω}}_H}_f& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& {{\mathrm{\ω}}_H}_e\end{array}$

Wherein, W is weight matrix, ω _e, ω _i, ω _c, ω _l, ω _σ, ω _r, ω _r, be respectively corresponding weights.

(6) by certain step length searching reference data matrix, for each searching position, the proper vector that it is above-mentioned is calculated, and compare with the proper vector of measured data matrix, calculate similarity, similarity measure adopts least-mean-square-error criterion, retains the similarity figure of each searching position;

Similarity F _s:

$F_S=|\stackrel{\→}{c}-\stackrel{{\→}^{\′}}{c}|$

Wherein, the eigenvector in actual measurement region, it is the eigenvector carrying out the reference area subregion mated in each search.

(7) after completing the search of reference data matrix, the position of wherein similarity maximum (namely similarity function value is minimum) is chosen, as the matched position exported.

Embodiment

(1) carry out down-sampled to terrain data under initial condition, obtain Fig. 1;

(2) gray level co-occurrence matrixes in actual measurement region is calculated, texture feature extraction parameter;

(3) the terrain feature parameter in actual measurement region is calculated;

(4) the textural characteristics parameter in actual measurement region and terrain feature parameter are multiplied by corresponding weight value respectively, are combined as proper vector;

(5) search for reference data matrix, for each searching position, calculate the proper vector that it is above-mentioned, and compare with the proper vector of measured data matrix, calculate similarity, similarity measure adopts least-mean-square-error criterion, retains the similarity figure of each searching position;

(6) after completing the search of reference area, choose the position of wherein similarity maximum (namely similarity function value is minimum), as the matched position exported, according to diverse location and the magnitude relationship of surveying region and reference area, Fig. 2,3,4 can be obtained.

Claims (5)

1., based on a underwater terrain matching method for textural characteristics and terrain feature parameter, it is characterized in that:

(1) actual measurement underwater topography data and reference data are carried out down-sampled;

(2) interpolation is carried out to the data after down-sampled, obtain measured data matrix and the reference data matrix of equal resolution;

(3) according to gray level co-occurrence matrixes, calculate the textural characteristics of measured data matrix, obtain angle second order distance, contrast, relevant and unfavourable balance square 4 textural characteristics parameters;

(4) calculate the terrain feature parameter of measured data matrix, obtain landform altitude average, landform altitude standard deviation, terrain roughness, landform related coefficient, terrain entropy and terrain variance entropy 6 terrain feature parameters;

(5) by vectorial for 10 characteristic parameter composition characteristics of step (3) and step (4), the feature of corresponding shaped area is characterized; According to the priori of landform characteristic parameter, for each characteristic parameter gives weights;

(6) by step length searching reference data matrix, for each searching position, proper vector is calculated, compare with the proper vector of measured data matrix, calculate similarity, similarity measure adopts least-mean-square-error criterion, retains the similarity figure of each searching position;

(7) the maximum position that namely similarity figure is minimum of wherein similarity is chosen, as the matched position exported.

2. a kind of underwater terrain matching method based on textural characteristics and terrain feature parameter according to claim 1, is characterized in that: described gray level co-occurrence matrixes

p(m,n,d,θ)＝{(i,j),(i+Δi,j+Δj)|f(i,j)＝m,f(i+Δi,j+Δj)＝n}，

It is m=f (i, j) that landform altitude image I mono-coordinate points (i, j) of gray level to be L size be M × N puts gray-scale value, another coordinate points (the i+ Δ i departed from, j+ Δ j) gray-scale value n=f (i+ Δ i, j+ Δ j), putting right gray-scale value is (m, n), gray level co-occurrence matrixes P is the square formation of L × L, and i, j are respectively transverse and longitudinal coordinate a little, i=0,1,2 ... M-1, j=0,1,2 ..., N-1; M, n=0,1 ..., L-1; Δ i, Δ j are the side-play amounts of pixel position; D is the generation step-length of gray level co-occurrence matrixes; θ is the generation direction of gray level co-occurrence matrixes;

Described angle second moment:

$E=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}p{(m,n)}^2,$

Contrast:

$I=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}{(m-n)}^{2}p(m,n),$

Relevant:

$C=[\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}\left((m,n)p(m,n)\right)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y]/{\mathrm{\δ}}_x{\mathrm{\δ}}_y,$

Wherein μ _x, μ _y, meet

${\mathrm{\μ}}_x=\underset{m}{\mathrm{\Σ}}m\underset{n}{\mathrm{\Σ}}p(m,n),$

${\mathrm{\μ}}_y=\underset{n}{\mathrm{\Σ}}n\underset{m}{\mathrm{\Σ}}p(m,n),$

${\mathrm{\δ}}_x^2=\underset{m}{\mathrm{\Σ}}{(m-{\mathrm{\μ}}_x)}^2\underset{n}{\mathrm{\Σ}}p(m,n),$

${\mathrm{\δ}}_y^2=\underset{n}{\mathrm{\Σ}}{(n-{\mathrm{\μ}}_y)}^2\underset{m}{\mathrm{\Σ}}p(m,n),$

Unfavourable balance square:

$L=\underset{m}{\mathrm{\Σ}}\underset{n}{\mathrm{\Σ}}\frac{1}{1+{(m-n)}^2}p(m,n).$

3. a kind of underwater terrain matching method based on textural characteristics and terrain feature parameter according to claim 1, is characterized in that: described landform altitude average ,

$\stackrel{\‾}{z}=\frac{1}{\mathrm{mn}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}z(i,j),$

Landform altitude standard deviation sigma,

$D\left(z\right)=\frac{1}{m(n-1)}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(z(i,j)-\stackrel{\‾}{z})}^2,$

$\mathrm{\σ}=\sqrt{D\left(z\right)},$

Terrain roughness r,

$r=\frac{S_S}{S},$

S is regional feature surface area, S _sfor the projected area of this regional feature,

Landform coefficient R,

$R_{\mathrm{\λ}}=\frac{1}{(m-1)n{\mathrm{\σ}}^2}\underset{i=1}{\overset{m-1}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}[z(i,j)-\stackrel{\‾}{z}][z(i+1,j)-\stackrel{\‾}{z}],$

$R_{\mathrm{\φ}}=\frac{1}{m(n-1){\mathrm{\σ}}^2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n-1}{\mathrm{\Σ}}}[z(i,j)-\stackrel{\‾}{z}][z(i,j+1)-\stackrel{\‾}{z}],$

$R=\frac{R_{\mathrm{\λ}}+R_{\mathrm{\φ}}}{2},$

R _λ, R _φbe respectively longitude, latitudinal related coefficient;

Terrain entropy H _f,

$H_f=-\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{p}_{\mathrm{ij}}\lg p_{\mathrm{ij}},$

Wherein p _ijit is the normalization height value at topographic(al) point coordinate place;

Terrain variance entropy H _e,

$H_e=-\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{p}_{i,j}\left(z\right)\lg p_{i,j}\left(z\right),$

Wherein D _z(i, j) is terrain differences value:

$D_z(i,j)=\frac{|z(i,j)-\stackrel{\‾}{z}|}{\stackrel{\‾}{z}}$

Utilize terrain differences value shape disparity probability p calculably _i,j():

$p_{i,j}\left(z\right)=\frac{D_z(i,j)}{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{D}_z(i,j)}.$

4. a kind of underwater terrain matching method based on textural characteristics and terrain feature parameter according to claim 1, is characterized in that:

Described proper vector:

$\stackrel{\→}{c}=W{(E,I,C,L,\stackrel{\text{-}}{z},\mathrm{\σ},r,R,H_f,H_e)}^T,$

$W=\begin{array}{cccccccccc}{\mathrm{\ω}}_E& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& {\mathrm{\ω}}_I& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& {\mathrm{\ω}}_C& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& {\mathrm{\ω}}_L& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& {\mathrm{\ω}}_{\stackrel{\‾}{z}}& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& {\mathrm{\ω}}_{\mathrm{\σ}}& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_r& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_R& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_{H_f}& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& {\mathrm{\ω}}_{H_e}\end{array}$

Wherein, W is weight matrix, ω _e, ω _i, ω _c, ω _l, , ω _σ, ω _r, ω _r, , be respectively E, I, C, L, σ, r, R, H _f, H _ecorresponding weights.

5. a kind of underwater terrain matching method based on textural characteristics and terrain feature parameter according to claim 1, is characterized in that: described similarity

$F_S=|\stackrel{\→}{c}-{\stackrel{\→}{c}}^{\′}|$

Wherein, the eigenvector in actual measurement region, it is the eigenvector carrying out the reference area subregion mated in each search.