CN106067172A - A kind of underwater topography image based on suitability analysis slightly mates and mates, with essence, the method combined - Google Patents
A kind of underwater topography image based on suitability analysis slightly mates and mates, with essence, the method combined Download PDFInfo
- Publication number
- CN106067172A CN106067172A CN201610363682.1A CN201610363682A CN106067172A CN 106067172 A CN106067172 A CN 106067172A CN 201610363682 A CN201610363682 A CN 201610363682A CN 106067172 A CN106067172 A CN 106067172A
- Authority
- CN
- China
- Prior art keywords
- image
- underwater topography
- essence
- suitability
- sigma
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
Landscapes
- Image Processing (AREA)
Abstract
The present invention provides a kind of underwater topography image based on suitability analysis slightly to mate the method mating combination with essence, the underwater topography altitude data detected for sonar system, obtain its real time imaging, first the suitability in template area is analyzed, if being suitable for coupling, select Different matching mode by course angle departure degree again, district to be measured is mated.If course angle deviation is relatively big, directly use essence matching way;If course angle deviation is less, use by slightly to the layering and matching mode of essence.Underwater topography image is slightly mated by the absolute difference algorithm wherein using gray scale;Essence coupling step is to choose the maximum correlation coefficient of gray level co-occurrence matrixes, the average of Gray level-gradient co-occurrence matrix and 7 not bending moment, totally 9 characteristic parameter constitutive characteristic vectors, uses these characteristic vectors that image carries out essence coupling.When equal external interference, use same position matching algorithm, it is possible to accomplish to judge well to underwater topography suitability, improve judgment accuracy.
Description
Technical field
The present invention relates to a kind of Digital Image Processing skill, particularly relate to a kind of underwater topography image analyzed based on suitability
Thick coupling mates, with essence, the method combined.
Background technology
Underwater topography suitability, is i.e. to judge that one piece of underwater topography is if appropriate for mating the one of location in topographic database
Plant and analyze.The complexity of underwater topography determines its characteristic parameter intensity of variation size, thus, the uniqueness of each piece of underwater topography
Property be to determine its key factor of precision in coupling position fixing process.For each underwater topography, secondary navigation system has
Different ability to work, such as underwater topography secondary navigation system, because relying on multibeam sounding system, thus high in landform
The region that journey amplitude of variation is big, its changing features amplitude is big, it is possible to performance is good in the matching process, and landform altitude change width
Spending little region, its changing features amplitude is also the least, and matching effect is the most relatively poor.For this situation, need under water
The suitability of shape image does certain constraint, it is judged that whether the real-time figure that underwater hiding-machine detection obtains can carry out location matches.Suitable
Join the research of sex chromosome mosaicism, it is possible to a large amount of underwater hiding-machine time loss in the matching process of position of saving, minimizing is not because registrating
And the operation time scanned for and unnecessary workload.
The present invention mainly propose a kind of new underwater topography suitability Rule of judgment for judge underwater topography if appropriate for
Coupling.By analyzing the suitability scheming in real time between reference map, Combining with terrain suitability tradition criterion, design produces base
In the underwater topography suitability Rule of judgment that thick coupling and the essence coupling of image texture characteristic direction character parameter combine;
Summary of the invention
The invention aims to judge that underwater topography provides a kind of based on suitability analysis if appropriate for coupling
Underwater topography image slightly mates and mates, with essence, the method combined.
The object of the present invention is achieved like this: comprises the steps:
The first step: utilize sonar system detection to obtain underwater topography altitude data;
Second step: the altitude data of acquisition is changed into figure in real time;
3rd step: analyze the suitability in real time between figure and reference map, Combining with terrain suitability tradition criterion, draw
Produce underwater topography suitability Rule of judgment based on image texture characteristic direction character parameter:
{ R/ σ > 0.15 ∩ UNImax-UNImin> 0.4 ∩ CONmax-CONmin> 0.6 ∩ CORmax-CORmin> 0.2}
In formula: R is gray scale roughness;σ is gray standard deviation;UNI is 4 obtained by the gray level co-occurrence matrixes of figure extraction in real time
The angle second moment in individual direction, and do normalized, obtain UNI by comparingmax、UNImin;CON is contrast, and COR is phase
Close, the discriminant value CON of CON Yu CORmax、CONminWith CORmax、CORminAcquisition methods is identical with UNI;
If meeting the formula of above-mentioned underwater topography suitability Rule of judgment, i.e. regarding as this region and being suitable for coupling, entering
Row the 4th step, otherwise re-starts the first step;
4th step: determine matching way according to actual heading angle deviation situation: if course angle deviation is less, carries out the successively
Five steps and the 6th step provide by thick to smart layering and matching mode;If course angle deviation is relatively big, directly carrying out the 6th step provides
Described essence matching way;
5th step: underwater topography image is slightly mated by absolute difference algorithm based on underwater picture gray scale;
6th step: choose and extracted the maximum correlation coefficient MCC of gray level co-occurrence matrixes, Gray Level-Gradient symbiosis square by figure in real time
The mean μ of battle array1With 7 not bending moment φ1—φ7Totally 9 characteristic parameter constitutive characteristic vectors, carry out based on image features
Essence coupling;
7th step: actual measurement underwater topography algorithm simulating, in different suitability Rule of judgment contrast simulation results.
Present invention additionally comprises so some architectural features:
1. the thick coupling in the 5th step is: known S (x, y) be size be m*n by coupling image, (x is y) that size is to T
The template image of M*N, in traveling through image S to be matched, takes so that (i is j) upper left corner, the subgraph of M*N size, calculates itself and mould
Plate image similarity, can obtain in subgraph all, finds the subgraph most like with template image finally to export as algorithm
As a result, and mean absolute difference is the least, illustrates that subgraph is the most similar to template image, therefore find minimum mean absolute difference D (i, j)
Just can determine that the position of subgraph:
In formula, 1≤i≤n-M+1,1≤j≤n-N+1.
2. the 6th step specifically includes: first, extracts the thick feature mating 5 approximate regions of gained in characteristic vector data storehouse
Vector also calculates the characteristic vector of real-time figure;Secondly, approximation district immediate with required position is tried to achieve by Similar measure function
Territory;Again, point on the basis of its coordinate, does the traversal of the most each 5 pixels, finds FsThe coordinate of minima:
Defined feature vector is c, it may be assumed that
C=W (MCC, μ1,φ1,φ2,φ3,φ4,φ5,φ6,φ7)T
In formula, W is weight matrix, ωMCC、Respectively MCC,
μ1、φ1—φ7Corresponding weights;
Similar measure function FsFor:
Fs=| c-c'|
In formula, c' is the set of eigenvectors of real-time figure, and c is the reference map subregion carrying out in each search procedure mating
Set of eigenvectors, makes FsThe region obtaining minima is institute's refinement matching area.
Compared with prior art, the invention has the beneficial effects as follows: 1, the present invention is under circular matching template, underwater topographic map
The maximum correlation coefficient of gray level co-occurrence matrixes, the average of Gray Level-Gradient Co-occurrence Matrix and characteristic parameter as being extracted can have good
Good rotational invariance, also has preferably the difference of the terrain graph resolution caused because of multibeam sounding system and detection range
Resistivity.2, by slightly to essence layering and matching mode, i.e. based on underwater topography image gray scale thick coupling, in conjunction with based under water
The essence coupling way of search of terrain graph characteristic vector, under conditions of allowable error is 3 pixels, mates locating accuracy
More traditional based on underwater topography elevation coupling TERCOM algorithmic method greatly improve, time depletion few.3, in order to solve tradition water
Lower Approach of Terrain Matching is used alone elevation information analysis makes underwater topography suitability judge not accurate enough limitation, the present invention
Design is based on underwater topography image grain direction feature adaptation determination methods, and both is effectively combined, complements each other, and concludes
Summary makes new advances underwater topography suitability criterion, equal external interference, use same position matching algorithm time, it is possible to water
Lower landform suitability accomplishes to judge well, improves judgment accuracy.And by certain lake ripple actual measurement underwater terrain matching, checking
The correctness of algorithm and effectiveness.
Accompanying drawing explanation
Fig. 1 is the flow chart of the present invention;
Fig. 2 is altitude data pcolor under water;
Fig. 3 is seafloor data modeling and simulating gray scale elevation map;
Fig. 4 (a) is not add process figure in real time, and Fig. 4 (b) is to add to make an uproar and scheme in real time after half-twist;
Fig. 5 (a) be real-time figure image size be the thick coupling of 15*15, Fig. 5 (b) be real-time figure image size be 17*17
Thick coupling, Fig. 5 (c) be real-time figure image size be the thick coupling of 19*19, Fig. 5 (d) be real-time figure image size be 21*21
Thick coupling;
Fig. 6 (a) be step-size in search be the thick coupling of 1 pixel, Fig. 6 (b) be step-size in search be thick of 2 pixels
Join, Fig. 6 (c) be step-size in search be the thick coupling of 3 pixels, Fig. 6 (d) be step-size in search be the thick coupling of 5 pixels;
Fig. 7 is that feature based parameter essence mates simulation result;
Fig. 8 (a) is terrain data under certain lake ripple actual water, and Fig. 8 (b) is region to be matched, and Fig. 8 (c) is actual measurement region;
Fig. 9 is terrain graph matching result under actual water;
Figure 10 is characteristic vector weight table;
Figure 11 is different suitability Rule of judgment contrast simulation result tables.
Detailed description of the invention
With detailed description of the invention, the present invention is described in further detail below in conjunction with the accompanying drawings.
The present invention is for the real-time positioning of underwater hiding-machine.The underwater topography altitude data detected for sonar system, obtains
Take its real time imaging.Under circular matching template, first analyzing the suitability in template area, if being suitable for coupling, then passing through
Course angle departure degree selects Different matching mode, mates district to be measured.If course angle deviation is relatively big, directly use essence
Formula formula;If course angle deviation is less, use more quickly by slightly to the layering and matching mode of essence.Wherein use the exhausted of gray scale
Underwater topography image is slightly mated by difference algorithm;Essence coupling step be choose the maximum correlation coefficient of gray level co-occurrence matrixes,
The average of Gray Level-Gradient Co-occurrence Matrix and 7 not bending moment, totally 9 characteristic parameter constitutive characteristics vectors, use these characteristic vectors
Image is carried out essence coupling.When equal external interference, use same position matching algorithm, it is possible to underwater topography suitability is done
To judging well, improve judgment accuracy.
Specifically, in conjunction with Fig. 1, the present invention comprises the steps:
(1) underwater topography altitude data is obtained;
Altitude data pcolor is as shown in Figure 2 under water.Fig. 3 is a certain water depth number obtained by multibeam echosounding sonar
According to, it is converted into gray level image, referred to as reference map.
(2) altitude data is converted into real-time figure image;
In real time figure is that underwater hiding-machine carries bathymetric data that multibeam sounding system obtains in real time under circular shuttering, conversion
The gray level image become, as shown in Figure 4.
(3) real-time figure suitability condition value is analyzed;Analyzing the suitability in real time between figure and reference map, Combining with terrain is adaptive
Property tradition criterion, design produce underwater topography suitability Rule of judgment based on image texture characteristic direction character parameter;
If pixel along the x-axis direction is N altogetherx, pixel along the y-axis direction is N altogethery, G is matrix number of grey levels, is designated as:
Lx=1,2 ..., Nx}
Ly=1,2 ..., Ny}
G={1,2 ..., Ng}
First definition gray level co-occurrence matrixes is:
Pc=p (i, j, d, θ)
PcFor newly-generated gray level co-occurrence matrixes, p (i, j, d, θ) is matrix PcIn the i-th row, jth row element;I, j are former
Gray level in matrix;D is the distance of gray scale inter-stage, and the value of d is usually 1, for the coarse journey of post analysis image texture
Degree, for open grain, the newly-generated value on gray level co-occurrence matrixes leading diagonal is generally large, and is distributed along leading diagonal, right
In close grain, the newly-generated value on gray level co-occurrence matrixes leading diagonal is the most little, and is distributed in leading diagonal both sides;θ is
Value direction, if without particular/special requirement, the usual value of θ is 0 °, 45 °, 90 °, 135 °.
It is initial with X-axis, starts in the counterclockwise direction, with 0 °, 45 °, 90 °, 135 ° of four directions, entry of a matrix element is entered
Row definition:
P (i, j, d, 0 °)=π { (k, l) (m, n) ∈ (Lx*Ly)*(Lx*Ly) | k-m=0, | l-n |=d;F (k, l)=i, f
(m, n)=j}
P (i, j, d, 45 °)=π { (k, l) (m, n) ∈ (Lx*Ly)*(Lx*Ly) | (k-m=d, l-n=d) or (k-m=-d,
L-n=-d);F (k, l)=i, f (m, n)=j}
P (i, j, d, 90 °)=π { (k, l) (m, n) ∈ (Lx*Ly)*(Lx*Ly) | | k-m |=d, l-n=0;F (k, l)=i,
F (m, n)=j}
P (i, j, d, 135 °)=π { (k, l) (m, n) ∈ (Lx*Ly)*(Lx*Ly) | (k-m=d, l-n=-d) or (k-m=-
D, l-n=d);F (k, l)=i, f (m, n)=j}
In above formula, { x} represents the element number contained by set x, the i-th row in newly-generated matrix, jth column element to π
Be expressed as in original matrix along θ direction, neighbor distance be d pixel in gray value be i, another gray value be the unit of j
Element is to quantity.
Afterwards, to Pc=p (i, j, d, θ) gray level co-occurrence matrixes carries out normalization:
In formula, R is regularization parameter, and it is to make to be calculated by gray level co-occurrence matrixes that gray level co-occurrence matrixes carries out normalization
Eigenvalue there is more higher leveled texture divide example.The R=2N as d=1, θ=0 °y(Nx-1), the R=as d=1, θ=45 °
2(Ny-1)(Nx-1), the R=2N as d=1, θ=90 °x(Ny-1), the R=2 (N as d=1, θ=45 °x-1)(Ny-1)。
The gray level co-occurrence matrixes of the four direction obtained also cannot directly use, and needs to calculate respectively textural characteristics
Value, and the analysis of textural characteristics is carried out with the eigenvalue obtained.
Owing to the real-time figure of present invention research obtains for real-time detection in underwater hiding-machine traveling process, so turning
When being melted into after real-time figure compared with reference map, direction has certain angle and rotates, therefore, studying above-mentioned gray scale symbiosis square
During battle array, main utilization has rotational invariance and the relatively small characteristic parameter of amount of calculation, concrete is respectively angle second moment
(UNI), contrast (CON), relevant (COR), entropy (ENT) and entropy (SENT), difference entropy (DE), mutual information tolerance (IMC),
Big correlation coefficient (MCC), totally 8 kinds.
Angle second moment (UNI):
UNI=∑ ∑ p (i, j) }2
Angle second moment mainly reflects that in image, intensity profile is the most uniform, also referred to as energy.For open grain, angle second moment
Value is relatively big, and for close grain, angle second moment is less.
Contrast (CON):
| i-j |=n in formula.
The meaning of contrast is the readability of image, namely the readability of texture.Contrast is the biggest, then texture ditch
Stricture of vagina is the most obvious, and degree is the deepest, and image effect is the best.For open grain, contrast value is less, and for close grain, contrast value is relatively
Greatly.
Relevant (COR):
In formula, μxIt is pxAverage, σxIt is pxMean square deviation, μyIt is pyAverage, σyIt is pyMean square deviation.
Relevant effect is to weigh the similar journey on the direction that each element in four direction gray level co-occurrence matrixes is expert at
Degree.Such as, piece image has a texture in vertical direction, then the gray level co-occurrence matrixes of calculated θ=90 ° of this image
Correlation values typically can be bigger than the correlation of the gray level co-occurrence matrixes on θ=0 °, θ=45 °, direction, 3, θ=135 °.
Entropy (ENT):
Entropy represents a kind of tolerance of this amount of image information.For not having veined image, its gray level co-occurrence matrixes is almost
For null matrix, so entropy also levels off to zero;Element value approximation phase if image is covered with close grain, in its gray level co-occurrence matrixes
Deng, then the entropy of this image is the most maximum;If only having less texture in image, the element value difference of its gray level co-occurrence matrixes is relatively
Greatly, then the entropy of diagram picture is the least.
With entropy (SENT):
Difference entropy (DE):
Mutual information tolerance (IMC):
f12=(HXY-HXY1)/max{HX, HY}
f13=(1-exp{-2 (HXY2-HXY) })1/2
In formula, HX is pxEntropy, HY is pyEntropy, HXY is p (i, entropy j).
Maximum correlation coefficient (MCC):
MCC=(the maximum second order eigenvalue of Q)1/2
In formula:
Gray level co-occurrence matrixes owing to being worth to by calculating gradation of image analyzes textural characteristics value at image transform processes
Middle can there is certain difference, it is impossible to characterize picture characteristics completely, so on the basis of to obtaining gray level co-occurrence matrixes before, entering
One step calculates the Gray Level-Gradient Co-occurrence Matrix of image, because Gray Level-Gradient space is not only able to each pixel of good description
And the spatial relationship between this pixel adjacent domains pixel, describes image internal each pixel gray scale and gradient the most simultaneously
The regularity of distribution, the gray value information of image is combined with gradient information, the main limit considering grey scale pixel value and image
The statistical distribution of edge Grad, i.e. combines grey level histogram and edge gradient rectangular histogram.
If: Gray Level-Gradient Co-occurrence Matrix is H, its element be H (i, j);Gray matrix F after normalization, its element is F
(m,n);Gradient matrix G after normalization, its element be G (m, n).Definition Gray Level-Gradient Co-occurrence Matrix H (i, j) be F (m, n)
With G (m, n) in gray scale be i and the counting of total pixel that gradient is j.Such as, what H (5,2)=1 represented is pixel ash in image
Always the counting of pixel that degree is 5, gradient is 2 is 1.
The present invention uses Laplace operator to calculate the Grad in former gray level image:
G (m, n)=4f (m, n)-f (m+1, n)-f (m-1, n)-f (m, n+1)-f (m, n-1)
Gradient matrix calculated to above formula is normalized:
G (m, n)=INT (g (m, n) * Gmax/gmax)+1
In formula, INT represent to G (m, n) in each element round;GmaxRepresent maximum after gradient matrix normalization
Grad, G of the present inventionmaxValue is 16;gmaxRepresent the maximum of gradients in former gradient matrix.
In like manner, original image is carried out gray scale normalization process:
F (m, n)=INT (f (m, n) * Fmax/fmax)+1
In formula, FmaxRepresent maximum gradation value after gray matrix normalization, F of the present inventionmaxValue is 16;fmaxRepresent former
Gray scale maximum in gray matrix.
To after being normalized gray level image matrix F (m, n) and gradient image matrix G (m, n) statistics F (m, n)
(m, the some logarithm of the pixel of n)=j, (i, j) in the i-th row, jth to be newly-generated Gray Level-Gradient Co-occurrence Matrix H for=i and G
The element value h of rowij。
H (F (m, n), G (m, n))=H (F (m, n), G (m, n))+1
Obtain the probability of Gray Level-Gradient Co-occurrence Matrix further:
By calculating the eigenvalue of newly-generated Gray Level-Gradient Co-occurrence Matrix, eigenvalue is analyzed, obtains original image
Texture features.
The present invention is when studying underwater topography image feature, and for being preferably analyzed characteristics of image, minimizing need not
Ensure again while the amount of calculation wanted that picture characteristics well embodies, therefore select energy, gray average, gradient mean value, dependency, mix
Close entropy, difference square, unfavourable balance away from 7 characteristic parameters.
Energy:
Gray average:
Gradient mean value:
Dependency:
The entropy of mixing:
Difference square:
Unfavourable balance divides square:
Use the method that square represents stochastic variable distribution situation, extract the moment characteristics of gray level image.In the case of being located at continuously
Image function be f (x, y), then definable image p+q rank geometric moment (also known as standard square) is:
And define p+q rank centre-to-centre spacing and be:
In formula,m00It is image f (x, zeroth order geometric moment y), i.e. all pictures in image
The summation of element value, so m00Can regard as image quality,WithRepresent the barycenter of image, then can be by image
The heart is away from μpqReflect the gray scale distribution situation relative to gray scale barycenter of image.
For discrete digital picture, definition image p+q rank geometric moment is:
Definition p+q rank centre-to-centre spacing is:
In formula, N and M represents height and the width of image respectively.
Generally, 0~3 rank centre-to-centre spacing of image are represented by the geometric moment of image:
In image processing process, in order to eliminate the impact that image scaled change produces, it is generally required in definition normalization
The heart away from for:
Centre-to-centre spacing is unrelated with the starting point of image, can solve translation invariance;And normalization centre-to-centre spacing eliminates image ratio
The impact that example is brought, has translation and scale invariance, but can not effectively solve rotational invariance.Utilize second order and three rank
7 of normalization central moment derivation not bending moment, can solve these problems, be shown below, and they have translation, rotate and chi
Degree invariance.
φ1=η20+η02
φ3=(η30-3η12)2+3(η21-η03)2
φ4=(η30+η12)2+(η21+η03)3
φ5=(η30+3η12)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η21-η03)(η21+η03)
[3(η30+η12)2-(η21+η03)2]
φ6=(η20-η02)[(η30+η12)2-(η21+η03)2]+4η11(η30+η12)(η21+η03)
φ7=(3 η21-η03)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η12-η30)(η21+η03)
[3(η30+η12)2-(η21+η03)2]
The present invention extracts 7 invariant moment features of underwater topography image, is primarily directed to underwater hiding-machine and carries multi-beam survey
The image caused due to sounding system resolution difference that the underwater picture that deep system is detected during advancing there will be
Scaling, the rotation of image that causes due to direct of travel difference, and bending moment does not has necessarily for the noise of underwater picture
Anti-interference.
Analyze and obtain suitability Rule of judgment:
{ R/ σ > 0.15 ∩ UNImax-UNImin> 0.4 ∩ CONmax-CONmin> 0.6 ∩ CORmax-CORmin> 0.2}
In formula, R is gray scale roughness;σ is gray standard deviation;UNI is 4 obtained by the gray level co-occurrence matrixes of figure extraction in real time
The angle second moment in individual direction, and do normalized, obtain UNI by comparingmax、UNImin;CON is contrast, and COR is phase
Closing, its discriminant value acquisition methods is identical with UNI.If i.e. meeting above formula, this region can be regarded as and be suitable for mating, on the contrary, or
It is unsatisfactory for obtaining new real-time figure from new, and re-starts suitability analysis.
(4) matching way is determined according to actual heading angle deviation situation.If course angle deviation is relatively big, directly use (6) institute
State essence matching way;If course angle deviation is less, use (5) and (6) described by slightly to the layering and matching mode of essence;
(5) underwater topography image is slightly mated by absolute difference algorithm based on underwater picture gray scale;
MAD algorithm (Mean Absolute Differences) is that one is widely used in images match
In matching algorithm.
If S (x, y) be size be m*n by coupling image, T (x, y) be size be the template image of M*N, i.e. by
Figure picture finds template image.In traveling through image S to be matched, take so that (i, j) is the upper left corner, the subgraph of M*N size, calculates
It can obtain in subgraph all with template image similarity, finds the subgraph most like with template image as algorithm
Output result eventually.Obviously, mean absolute difference is the least, illustrates that subgraph is the most similar to template image, therefore has only to find minimum
Mean absolute difference is assured that the position of subgraph.
In formula, 1≤i≤n-M+1,1≤j≤n-N+1.
Randomly select at one underwater topography and change into figure in real time, image being added the noise of 40dB, rotates 15 °, analyze not
With the impact on matching accuracy of the real-time figure of size, result is as shown in Figure 5.In Figure 5, blue solid lines is for being matched actual bit
Putting, red dotted line is, through algorithm, underwater hiding-machine detection is obtained 5 positions that terrain graph calculates that its probability is the highest.By scheming
5 it is found that scheme the least in real time, and device of i.e. diving moves ahead apart from too short, bigger on matching accuracy impact;Figure is too big in real time, makes again
The latent device traveling time becoming unnecessary calculates time waste with algorithm.Can be drawn by the comparison of 4 subgraphs, for this water
The underwater topography feature in territory, is considering the size of region of search, i.e. reference map information contained and pixel number, 19*19 is big in design
Little real-time figure mates, and both can guarantee that the effectiveness of coupling, is avoided that again time loss is long.
Randomly select underwater topography and change into figure in real time, image being added the noise of 40dB, rotates 15 °, analyze difference and search
The Suo Buchang impact on matching accuracy, result is as shown in Figure 6.In Fig. 6, blue solid lines is for being matched physical location, red empty
Line is, through algorithm, underwater hiding-machine detection is obtained 5 positions that terrain graph calculates that its probability is the highest.Can by Fig. 6 result
See, identical in matching area size, picture noise is identical, the anglec of rotation identical under conditions of, different step-size in search are to figure in real time
Coupling can produce certain impact, step-size in search is the longest, deviates the biggest to the physical location of images match, but in physical location
Near zone is included in algorithm in multiple similar area positions that figure speculates in real time, it is possible to for the essence of the vector of feature based afterwards
Coupling provides strong support.
(6) maximum correlation coefficient (MCC) of gray level co-occurrence matrixes, the average (μ of Gray Level-Gradient Co-occurrence Matrix are chosen1) and 7
Individual not bending moment (φ1—φ7), totally 9 characteristic parameter constitutive characteristic vectors, carry out essence coupling based on image features;
Specific practice is: first, extracts the characteristic vector of slightly coupling 5 approximate regions of gained in characteristic vector data storehouse also
Calculate the characteristic vector of real-time figure;Secondly, approximate region immediate with required position is tried to achieve by Similar measure function;Again
Secondary, point on the basis of its coordinate, does the traversal in the most each 5 pixels (i.e. 11*11 region), finds FsThe seat of minima
Mark, is latent device position, as shown in Figure 7.
For improving efficiency, reducing time loss, reference map does the foursquare maximum inscribed circle of 19*19, and (round diameter is
The square length of side) step through to template, extract image features vector, set up underwater topography image characteristic vector data
Storehouse.
Because each characteristic parameter order of magnitude difference is excessive, so to set feature weight before characteristic vector so that it is vector
Weight is waited, in order to avoid affecting matching accuracy because a certain characteristic parameter is excessive between value.
Defined feature vector is c, it may be assumed that
C=W (MCC, μ1,φ1,φ2,φ3,φ4,φ5,φ6,φ7)T
In formula, W is weight matrix, ωMCC、Respectively MCC,
μ1、φ1—φ7Corresponding weights.
Design Similar measure function is as follows:
Fs=| c-c'|
In formula, c' is the set of eigenvectors of real-time figure, and c is the reference map subregion carrying out in each search procedure mating
Set of eigenvectors.Obviously, F is madesThe region obtaining minima is exactly required matching area.Weights are arranged as shown in Figure 10, finally
The threshold value such as Figure 10 is used to carry out smart matching area.
(7) actual measurement underwater topography algorithm simulating, the different suitability Rule of judgment contrast simulation result of contrast.
Algorithm simulating test is carried out based on terrain data under certain lake actual water, as shown in Fig. 8 (a), for convenience of computer
Computing, intercepts 220*220 pixel sized images thereon as region to be matched, as shown in Fig. 8 (b), real-time water-depth measurement
Data are obtained, as shown in Fig. 8 (c) by the 40dB-50dB noise that adds random in true depth of water sequence.Random-Rotation 5 °-75 °
Under the conditions of, by based on gradation of image the thick coupling that step-length is 5 pixels, locally 11*11 pixel traversal based on figure
As the essence coupling of characteristic vector distance, under conditions of allowable error is 3 pixels, underwater hiding-machine is positioned, result
As shown in Figure 8.Different suitability Rule of judgment comparing results are as shown in figure 11.
Compared with prior art, the invention has the beneficial effects as follows:
1, under circular matching template, the maximum correlation coefficient of the gray level co-occurrence matrixes that underwater topography image is extracted, ash
The average of degree-gradient co-occurrence matrix and 7 not bending moment totally 9 characteristic parameters can have good rotational invariance, maximum anti-noise energy
Power is 40dB, and the difference of the terrain graph resolution caused because of multibeam sounding system and detection range is also had preferably opposing
Ability.
2, tradition TERCOM algorithmic method based on underwater topography elevation coupling is simple, but computationally intensive, calculate time-consuming long,
And coupling correct localization is the highest;Present invention design is by slightly to essence layering and matching mode, i.e. global search step-length is 5 pixels
Underwater topography image is slightly mated, as shown in Figure 5 and Figure 6 by absolute difference algorithm based on underwater topography image gray scale.In conjunction with
11*11 region Local Search step-length is 1 pixel, chooses the maximum correlation coefficient (MCC) of gray level co-occurrence matrixes, gray scale-ladder
Average (the μ of degree co-occurrence matrix1) and 7 not bending moment (φ1—φ7), the essence coupling of totally 9 characteristic parameter constitutive characteristic vectors is searched
Rope mode, as shown in Figure 7.Under conditions of allowable error is 3 pixels, coupling locating accuracy reaches 91%, than tradition
TERCOM algorithmic match correct localization improves more than 10 percentage point, and depletion few 30% when calculating;And surveyed by certain lake ripple
Underwater terrain matching, the correctness of verification algorithm and effectiveness, as shown in Figure 8 and Figure 9.
3, conventional adaptation determination methods based on underwater topography depth displacement has certain limitation, is i.e. used alone height
It is not accurate enough that journey information analysis makes underwater topography suitability judge.Conventional adaptation Rule of judgment provides the result being suitable for coupling,
But the eigentransformation for coupling is inconspicuous, be unfavorable for coupling;Present invention design is special based on underwater topography image grain direction
Levy suitability determination methods, and both effectively combined, complements each other, induction and conclusion make new advances underwater topography suitability differentiate bar
Part, when equal external interference, use same position matching algorithm, it is possible to accomplish to judge well to underwater topography suitability,
Improve judgment accuracy, as shown in figure 11.
Claims (3)
1. a underwater topography image based on suitability analysis slightly mates and mates, with essence, the method combined, it is characterised in that: bag
Include following steps:
The first step: utilize sonar system detection to obtain underwater topography altitude data;
Second step: the altitude data of acquisition is changed into figure in real time;
3rd step: analyze the suitability in real time between figure and reference map, Combining with terrain suitability tradition criterion, draw generation
Underwater topography suitability Rule of judgment based on image texture characteristic direction character parameter:
{ R/ σ > 0.15 ∩ UNImax-UNImin> 0.4 ∩ CONmax-CONmin> 0.6 ∩ CORmax-CORmin> 0.2}
In formula: R is gray scale roughness;σ is gray standard deviation;UNI is to be extracted, by figure in real time, 4 sides that gray level co-occurrence matrixes obtains
To angle second moment, and do normalized, obtain UNI by comparingmax、UNImin;CON is contrast, and COR is relevant, CON
Discriminant value CON with CORmax、CONminWith CORmax、CORminAcquisition methods is identical with UNI;
If meeting the formula of above-mentioned underwater topography suitability Rule of judgment, i.e. regarding as this region and being suitable for coupling, carrying out the
Four steps, otherwise re-start the first step;
4th step: determine matching way according to actual heading angle deviation situation: if course angle deviation is less, carry out the 5th step successively
With the 6th step provide by thick to smart layering and matching mode;If course angle deviation is relatively big, directly carry out the institute that the 6th step provides
State essence matching way;
5th step: underwater topography image is slightly mated by absolute difference algorithm based on underwater picture gray scale;
6th step: choose and extracted the maximum correlation coefficient MCC of gray level co-occurrence matrixes, Gray Level-Gradient Co-occurrence Matrix by real-time figure
Mean μ1With 7 not bending moment φ1—φ7Totally 9 characteristic parameter constitutive characteristic vectors, carry out essence based on image features
Join;
7th step: actual measurement underwater topography algorithm simulating, in different suitability Rule of judgment contrast simulation results.
A kind of underwater topography image based on suitability analysis the most according to claim 1 slightly mates and mates combination with essence
Method, it is characterised in that: the thick coupling in the 5th step is: known S (x, y) be size be m*n by coupling image, (x y) is T
Size is the template image of M*N, in traveling through image S to be matched, takes so that (i, j) is the upper left corner, the subgraph of M*N size, calculates
It can obtain in subgraph all with template image similarity, finds the subgraph most like with template image as algorithm
Output result eventually, and mean absolute difference is the least, illustrates that subgraph is the most similar to template image, therefore finds the mean absolute difference of minimum
D (i, j) just can determine that the position of subgraph:
In formula, 1≤i≤n-M+1,1≤j≤n-N+1.
A kind of underwater topography image based on suitability analysis the most according to claim 1 and 2 slightly mates and mates knot with essence
The method closed, it is characterised in that: the 6th step specifically includes: first, extracts thick coupling gained 5 approximation in characteristic vector data storehouse
The characteristic vector in region also calculates the characteristic vector of real-time figure;Secondly, tried to achieve by Similar measure function and connect most with required position
Near approximate region;Again, point on the basis of its coordinate, does the traversal of the most each 5 pixels, finds FsMinima
Coordinate:
Defined feature vector is c, it may be assumed that
C=W (MCC, μ1,φ1,φ2,φ3,φ4,φ5,φ6,φ7)T
In formula, W is weight matrix, ωMCC、It is respectively MCC, μ1、
φ1—φ7Corresponding weights;
Similar measure function FsFor:
Fs=| c-c'|
In formula, c' is the set of eigenvectors of real-time figure, and c is the feature of reference map subregion carrying out in each search procedure mating
Vector set, makes FsThe region obtaining minima is institute's refinement matching area.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610363682.1A CN106067172B (en) | 2016-05-27 | 2016-05-27 | A method of slightly matching matches combination to the underwater topography image based on suitability analysis with essence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610363682.1A CN106067172B (en) | 2016-05-27 | 2016-05-27 | A method of slightly matching matches combination to the underwater topography image based on suitability analysis with essence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106067172A true CN106067172A (en) | 2016-11-02 |
CN106067172B CN106067172B (en) | 2018-10-26 |
Family
ID=57420854
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610363682.1A Active CN106067172B (en) | 2016-05-27 | 2016-05-27 | A method of slightly matching matches combination to the underwater topography image based on suitability analysis with essence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106067172B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180120472A1 (en) * | 2016-10-28 | 2018-05-03 | Korea Institute Of Geoscience And Mineral Resources | Apparatus and method for localizing underwater anomalous body |
CN108416168A (en) * | 2018-03-29 | 2018-08-17 | 北京航空航天大学 | Landform based on hierarchical decision making is adapted to area's Choice |
CN108921891A (en) * | 2018-06-21 | 2018-11-30 | 南通西塔自动化科技有限公司 | A kind of machine vision method for rapidly positioning that can arbitrarily rotate |
CN109313809A (en) * | 2017-12-26 | 2019-02-05 | 深圳配天智能技术研究院有限公司 | A kind of image matching method, device and storage medium |
CN110378425A (en) * | 2019-07-23 | 2019-10-25 | 北京隆普智能科技有限公司 | A kind of method and its system that intelligent image compares |
CN111486845A (en) * | 2020-04-27 | 2020-08-04 | 中国海洋大学 | AUV multi-strategy navigation method based on submarine topography matching |
CN108631788B (en) * | 2018-03-29 | 2020-12-18 | 北京航空航天大学 | Coding distortion optimization method for matching region adaptability analysis |
CN112950590A (en) * | 2021-03-03 | 2021-06-11 | 哈尔滨工程大学 | Terrain image adaptability analysis method and device and readable storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103761526A (en) * | 2014-01-26 | 2014-04-30 | 北京理工大学 | Urban area detecting method based on feature position optimization and integration |
CN104134209A (en) * | 2014-07-18 | 2014-11-05 | 北京国电富通科技发展有限责任公司 | Feature extraction and matching method and feature extraction and matching system in visual navigation |
CN105160665A (en) * | 2015-08-25 | 2015-12-16 | 东南大学 | Double-circle sub-template underwater terrain matching method |
CN105205817A (en) * | 2015-09-17 | 2015-12-30 | 哈尔滨工程大学 | Underwater terrain matching method based on sonar image edge angular point histogram |
-
2016
- 2016-05-27 CN CN201610363682.1A patent/CN106067172B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103761526A (en) * | 2014-01-26 | 2014-04-30 | 北京理工大学 | Urban area detecting method based on feature position optimization and integration |
CN104134209A (en) * | 2014-07-18 | 2014-11-05 | 北京国电富通科技发展有限责任公司 | Feature extraction and matching method and feature extraction and matching system in visual navigation |
CN105160665A (en) * | 2015-08-25 | 2015-12-16 | 东南大学 | Double-circle sub-template underwater terrain matching method |
CN105205817A (en) * | 2015-09-17 | 2015-12-30 | 哈尔滨工程大学 | Underwater terrain matching method based on sonar image edge angular point histogram |
Non-Patent Citations (1)
Title |
---|
《航天控制》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180120472A1 (en) * | 2016-10-28 | 2018-05-03 | Korea Institute Of Geoscience And Mineral Resources | Apparatus and method for localizing underwater anomalous body |
CN109313809A (en) * | 2017-12-26 | 2019-02-05 | 深圳配天智能技术研究院有限公司 | A kind of image matching method, device and storage medium |
CN109313809B (en) * | 2017-12-26 | 2022-05-31 | 深圳配天智能技术研究院有限公司 | Image matching method, device and storage medium |
CN108416168A (en) * | 2018-03-29 | 2018-08-17 | 北京航空航天大学 | Landform based on hierarchical decision making is adapted to area's Choice |
CN108631788B (en) * | 2018-03-29 | 2020-12-18 | 北京航空航天大学 | Coding distortion optimization method for matching region adaptability analysis |
CN108416168B (en) * | 2018-03-29 | 2021-11-02 | 北京航空航天大学 | Terrain adaptive area selection scheme based on layered decision |
CN108921891A (en) * | 2018-06-21 | 2018-11-30 | 南通西塔自动化科技有限公司 | A kind of machine vision method for rapidly positioning that can arbitrarily rotate |
CN110378425A (en) * | 2019-07-23 | 2019-10-25 | 北京隆普智能科技有限公司 | A kind of method and its system that intelligent image compares |
CN111486845A (en) * | 2020-04-27 | 2020-08-04 | 中国海洋大学 | AUV multi-strategy navigation method based on submarine topography matching |
CN111486845B (en) * | 2020-04-27 | 2022-02-11 | 中国海洋大学 | AUV multi-strategy navigation method based on submarine topography matching |
CN112950590A (en) * | 2021-03-03 | 2021-06-11 | 哈尔滨工程大学 | Terrain image adaptability analysis method and device and readable storage medium |
CN112950590B (en) * | 2021-03-03 | 2024-04-05 | 哈尔滨工程大学 | Terrain image suitability analysis method, equipment and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106067172B (en) | 2018-10-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106067172A (en) | A kind of underwater topography image based on suitability analysis slightly mates and mates, with essence, the method combined | |
US20200401842A1 (en) | Human Hairstyle Generation Method Based on Multi-Feature Retrieval and Deformation | |
CN111199214B (en) | Residual network multispectral image ground object classification method | |
CN104392426B (en) | A kind of no marks point three-dimensional point cloud method for automatically split-jointing of self adaptation | |
CN103871039B (en) | Generation method for difference chart in SAR (Synthetic Aperture Radar) image change detection | |
CN103778626B (en) | A kind of fast image registration method of view-based access control model marking area | |
CN113468968B (en) | Remote sensing image rotating target detection method based on non-anchor frame | |
CN103714148B (en) | SAR image search method based on sparse coding classification | |
CN114187255B (en) | Remote sensing image change detection method based on difference guidance | |
US20150131873A1 (en) | Exemplar-based feature weighting | |
CN102314610B (en) | Object-oriented image clustering method based on probabilistic latent semantic analysis (PLSA) model | |
Mao et al. | Uasnet: Uncertainty adaptive sampling network for deep stereo matching | |
CN106530321A (en) | Multi-graph image segmentation based on direction and scale descriptors | |
CN113343804B (en) | Integrated migration learning classification method and system for single-view fully-polarized SAR data | |
CN116563096B (en) | Method and device for determining deformation field for image registration and electronic equipment | |
CN110334599A (en) | Training method, device, equipment and the storage medium of deep learning network | |
CN106355607A (en) | Wide-baseline color image template matching method | |
CN115496720A (en) | Gastrointestinal cancer pathological image segmentation method based on ViT mechanism model and related equipment | |
CN108871342A (en) | Subaqueous gravity aided inertial navigation based on textural characteristics is adapted to area's choosing method | |
CN115809970A (en) | Deep learning cloud removing method based on SAR-optical remote sensing image combination | |
CN111709977A (en) | Binocular depth learning method based on adaptive unimodal stereo matching cost filtering | |
CN113011359B (en) | Method for simultaneously detecting plane structure and generating plane description based on image and application | |
CN109074643B (en) | Orientation-based object matching in images | |
CN113988198A (en) | Multi-scale city function classification method based on landmark constraint | |
CN102880869B (en) | Based on the fingerprint direction field under Morkov random field condition extracting method of priori |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |