CN105205817A - Underwater terrain matching method based on sonar image edge angular point histogram - Google Patents

Abstract

The invention relates to the field of digital image processing, in particular to an underwater terrain matching method which is based on a sonar image edge angular point histogram and applied to an underwater submersible vehicle navigation system. The method comprises the steps that 1, normalization and interpolation processing is conducted on original multi-beam data, and a depth value is converted to range from 0 to 255 through linear transformation; 2, a three-dimensional model is projected to an XY plane, and if gray values of all projection points are set to be height values of the projection points in the three-dimensional model, the projection serves as a gray image of an under water terrain; 3, traversing is conducted on all gray levels, the number of pixels on the same gray level is counted, a gray histogram is formed, and the total number of all the pixels on all the gray levels is obtained. According to the underwater terrain matching method based on the sonar image edge angular point histogram, a histogram theory is utilized, analyzing is conducted on all the gray levels of the image, and the defect that a traditional gradation histogram is poor in resolution is overcome; the edge angular point histogram is utilized to serve as an image feature, and positioning is conducted on the position of an underwater vehicle according to a similarity calculation result. On the condition that noise, directional errors and scale variations exist, the reliability of the result can still be kept.

Description

A kind of based on the histogrammic underwater terrain matching method of sonar image edge angle point

Technical field

What the present invention relates to is a kind of digital image processing field, be specifically related to a kind of apply to underwater hiding-machine navigational system based on the histogrammic underwater terrain matching method of sonar image edge angle point.

Background technology

In recent years, world technology power is unmanned device technology of diving under primary study and developing water just, and plan it can be used as the main tool of exploratory and exploitative ocean.With regard to current research emphasis, the underwater navigation technology of device of diving is one of topmost technological challenge in this field.

Conventional underwater navigation localization method mainly contains two classes: one is the air navigation aid based on acoustics, and another kind of is air navigation aid based on dead reckoning.Although the two all can realize positioning function, but due to its inherent shortcoming, as cumulative errors increases in time, uses region area less and maintenance cost costliness etc., they are all not suitable for the Underwater Navigation task of long-time, the long distance of complete independently, and need to carry out periodicity correction by assisting navigation technology to it.As a kind of important geophysical navigation method, namely Terrain-aided Navigation produces under this background and is developed gradually.

From the angle of principle of work, Terrain-aided Navigation can be divided into two classes: scene matching aided navigation is mated with Terrain Elevation.Terrain Elevation coupling uses a kind of terrain auxiliary navigation method comparatively widely, it passes through survey the topography elevation statics in real time and measured data and terrain elevation data storehouse are carried out the matching analysis in Matching band, thus obtain most relevant position, the i.e. physical location of carrier.The terrain auxiliary navigation method that the present invention studies belongs to Terrain Elevation coupling, utilize image processing techniques to analyze actual measurement terrain elevation data, reach by the similarity asked between actual measurement region and the edge angle point histogram of reference area the object determining latent device position.

Summary of the invention

The object of the present invention is to provide a kind of utilize underwater topography depth data to latent device carry out real-time location based on the histogrammic underwater terrain matching method of sonar image edge angle point.

The object of the present invention is achieved like this:

A kind of based on the histogrammic underwater terrain matching method of sonar image edge angle point, comprise the steps:

(1) Multibeam Data is standardized and interpolation processing, for depth value, by linear transformation, be transformed into 0-255 scope;

(2) projected in XY plane by three-dimensional model, if the gray-scale value of each subpoint to be set to this some height value in the three-dimensional model, then this projection is the gray level image of underwater topography;

(3) travel through each gray level, add up the number of pixels under same gray level, form grey level histogram, obtain the sum of pixel under each gray level;

(4) extract the marginal point of each gray level, and record its position and number, wherein marginal point is defined as:

In all pixels of certain gray level, if at least one four neighborhood pixel has different gray-scale value to certain point from it, then this point is the marginal point of this gray level;

(5) travel through the marginal point of each gray level, extract the edge angle point of each gray level, and record its number, wherein edge angle point is defined as follows:

In all marginal points of certain gray level, if the gray-scale value of certain point is different from the consecutive point of its at least one vertical direction, and the gray-scale value of this point is different from the consecutive point of its at least one horizontal direction, then this point is the edge angle point of this gray level;

A bit depth be L, resolution is in the gray level image of N × M, L is positive integer, and N, M are respectively image length and width, makes g (i, j) represent the gray-scale value of pixel P (i, j), i=1,2 ... N; J=1,2 ... M; K=1,2...2 in this image kth gray level ^l, the set of its marginal point is E _k, the set of edge angle point is C _k, C _ke _ksubset:

$C_k\⊆E_k$

C _krepresent:

$C_k=\left\{P\left(i,j\right)|\begin{array}{c}P\left(i,j\right)\⊆E_k\\ g\left(i,j\right)\≠g\left(i-1,j\right)\left|\right|g\left(i,j\right)\≠g\left(i+1,j\right)\\ g\left(i,j\right)\≠g\left(i,j-1\right)\left|\right|g\left(i,j\right)\≠g\left(i,j+1\right)\end{array}\right\};$

(6) the edge angle point complexity of each gray level is calculated:

The edge angle point complexity γ of certain gray level _kedge angle for this gray level is counted the ratio of counting with edge, wherein λ (E _k) and λ (C _k) edge that is respectively this gray level counts and to count with edge angle

${\mathrm{\γ}}_k=\frac{\mathrm{\λ}\left(C_k\right)}{\mathrm{\λ}\left(E_k\right)};$

(7) using the edge angle point complexity of each gray level as the value of this gray level in histogram, thus obtain the edge angle point histogram of this image;

(8) utilize histogram as template, reference map is scanned; The edge angle point histogram of each scanning position in statistics reference map, and the similarity calculating itself and THE TEMPLATE HYSTOGRAM, similarity measurement adopts maximum Mean Square Error;

(9), after completing the scanning of view picture reference map, the scanning position coordinate with maximum similarity is exported as matching result.

The present invention utilizes histogram theoretical, analyzes respectively each gray level of image, overcomes the shortcoming that traditional grey level histogram differentiates rate variance.Utilize edge angle point histogram as characteristics of image, according to Similarity measures result, underwater carrier position is positioned.Having noise, deflection error and dimensional variation to deposit in case, the reliability of result still can be kept.Because the distribution of gray-level pixels same in underwater topography image disperses very much, therefore the pixel major part of same gray level is all judged as marginal point, thus makes often to have similar grey level histogram between different images and edge angle is counted.The present invention is directed to this problem, by definition edge angle point, effectively can extract the space distribution information between such image, construct the edge angle point histogram with enough discriminations, thus obtain good matching result.

Accompanying drawing explanation

Fig. 1 is the continuous three-dimensional underwater topography model obtained by interpolation.

Fig. 2 is the two-dimensional projection of underwater topography model.

Fig. 3 is underwater topography image.

Fig. 4 is institute's likely form of edge angle point.

Fig. 5 is the matching result figure under different noise background.

Fig. 6 is the matching result figure utilizing the real-time figure of different size to obtain.

The matching result figure under different rotary angle is there is in real time in Fig. 7 between figure and reference map.

Embodiment

Below in conjunction with accompanying drawing citing, the present invention is described in more detail:

A kind of based on the histogrammic underwater terrain matching method of sonar image edge angle point, comprise multi-beam Bathymetric Data normalization and be interpolated to figure, Extraction of Topographic Patterns, reference map scanning and location, multi-beam Bathymetric Data forms the data dot matrix with proportional spacing through normalization and interpolation processing, using the gray scale of this data dot matrix as image, form underwater topography image, submit to Extraction of Terrain Features analysis; Extraction of Topographic Patterns comprises edge angle point complexity corresponding to each gray level of calculating input image and asks for corresponding edge angle point histogram, and it can be used as the feature of landform; Scan reference map, calculate the histogrammic similarity of edge angle point of in real time figure and reference map correspondence position, the maximum position of similarity exports as matching result.Described Extraction of Terrain Features comprises according to gradation of image order from low to high, calculates edge angle point complexity and the structure edge angle point histogram of each gray-level pixels.Wherein each gray-level pixels edge angle point complexity will by calculating its edge pixel number successively, edge angle point pixel count obtains; Edge angle point histogram is obtained by the edge angle point complexity of adding up all gray-level pixels.Described reference map scans and locates to comprise and utilizes the edge angle point histogram of real-time figure as template, and scanning reference map obtains the edge angle point histogram feature of each local location, calculates the similarity between itself and template; Similarity measurement adopts mean square deviation criterion; The current location of carrier is thought in the maximum position of similarity, using this position coordinates as output.The edge angle point histogram feature of each local location of reference map can precalculate and store, and only needs the edge angle point histogram calculating real-time figure in actual applications, effectively can shorten match time thus.

1, original multibeam bathymetric data is standardized and interpolation, depth value is limited within the scope of 0-255 by linear transformation, and according to concrete accuracy requirement, difference density is set, thus form the underwater topography image with non-uniform resolution;

2, travel through underwater topography image, add up each gray-level pixels number, by the sequence of gray level size, form the grey level histogram of underwater topography image;

3, travel through each gray level, extract the marginal point in pixel at different levels, record various point locations and sum;

4, define the edge angle point under same gray level, travel through the marginal point of each gray level, extract edge angle point wherein, record its number;

5, utilize the above-mentioned edge calculated to count and edge angle is counted and obtained the edge angle point complexity of each gray level, and construct the edge angle point histogram of image;

6, using this edge angle point histogram as template, traverse scanning is carried out to benchmark underwater topographic map, and calculates the similarity of each scanning position simultaneously;

7, after completing scanning, the scanning position coordinate with maximum similarity is positioning result.

The present invention is a kind of underwater terrain matching method based on sonar image range statistics feature.Content comprises carries out pre-service to multibeam bathymetric data, is become the underwater topography image with non-uniform resolution by normalized and interpolation; Marginal point, the edge angle point of each gray level are also extracted in definition successively, calculate the edge angle point complexity of each gray level; According to the order of each gray level, by angle point complexity composition edge, the edge angle point histogram of its correspondence, as the foundation of match search.The present invention utilizes multibeam bathymetric data to construct underwater topography image, by extracting the edge angle dot information of underwater topography in each degree of depth that image comprises, the exact position of underwater carrier can be provided, have good result for the multibeam bathymetric data with larger investigative range.

The present invention effectively can resist noise, and in real time figure is insensitive with the rotating deviation between reference map, can adjust real-time figure size according to actual needs, and detection time can not factor data amount to be increased and long.The more general matching process based on underwater topography provincial characteristics, calculated amount of the present invention is less, also can provide accurate positional information simultaneously, is conducive to like this being converted into practical application.

Different gray level image may have identical grey level histogram, therefore only utilizes the grey level histogram of image to carry out coupling and is difficult to obtain stable and accurate result.And the matching process utilizing image edge pixels improved, when process has the image of identical grey level histogram and same edge pixel count, effect same is not good.The present invention compares above two kinds of methods, further proposes edge angle point as characteristics of image, successfully solves the problems referred to above, and be especially applicable to the coupling of underwater topography image, positioning precision has a clear superiority in.

1. pair Multibeam Data is standardized and interpolation processing, and object is the underwater topography model obtaining continuously smooth.For depth value, by linear transformation, be transformed within the scope of 0-255 to meet the requirement of gray level image.Interpolation precision can be chosen as required, to obtain the result of different resolution.As shown in Figure 1, in figure, interpolation precision is 1 square metre/pixel in citing.

2. project in XY plane by the three-dimensional model shown in Fig. 1, citing as shown in Figure 2.If the gray-scale value of each subpoint to be set to this some height value in the three-dimensional model (i.e. water depth value), then this projection is the gray level image of underwater topography, and citing as shown in Figure 3.Fig. 2 and Fig. 3 precision is 1 square metre/pixel.

3. each gray level of traversing graph 3, adds up the number of pixels under same gray level, forms grey level histogram, obtains the sum of pixel under each gray level.

4. on the basis of previous step, extract the marginal point of each gray level, and record its position and number, wherein marginal point is defined as follows:

In all pixels of certain gray level, if at least one four neighborhood pixel (its four neighborhoods pixel can not belong to this gray level) have different gray-scale value to certain point from it, then this point is the marginal point of this gray level.The frontier point of image is marginal point of gray level belonging to it also.

5. travel through the marginal point of each gray level, extract the edge angle point of each gray level, and record its number, wherein edge angle point is defined as follows:

In all marginal points of certain gray level, if the gray-scale value of certain point is different from the consecutive point of its at least one vertical direction, and the gray-scale value of this point is different from the consecutive point of its at least one horizontal direction, then this point is the edge angle point of this gray level.

A bit depth be L, resolution is (L is positive integer, and N, M are respectively image length and width) in the gray level image of N × M, makes g (i, j) represent the gray-scale value of pixel P (i, j), i=1,2 ... N; J=1,2 ... M.(k=1,2...2 in this image kth gray level ^l), the set of its marginal point is E _k, the set of edge angle point is C _k, so C _kmust be E _ksubset:

$C_k\⊆E_k---\left(1\right)$

Thus, C _kcan be represented by (2) formula:

$C_k=\left\{P\left(i,j\right)|\begin{array}{c}P\left(i,j\right)\⊆E_k\\ g\left(i,j\right)\≠g\left(i-1,j\right)\left|\right|g\left(i,j\right)\≠g\left(i+1,j\right)\\ g\left(i,j\right)\≠g\left(i,j-1\right)\left|\right|g\left(i,j\right)\≠g\left(i,j+1\right)\end{array}\right\}---\left(2\right)$

Fig. 4 lists according to (1) and (2) formula, the form of all edges angle point that may exist in certain gray level, i.e. independent point, line segment end points and line segment flex point.

6. utilize the above results, calculate the edge angle point complexity of each gray level, it is defined as follows:

The edge angle point complexity γ of certain gray level _kedge angle for this gray level is counted the ratio of counting with edge, as shown in (3) formula, and wherein λ (E _k) and λ (C _k) edge that is respectively this gray level counts and to count with edge angle.

${\mathrm{\γ}}_k=\frac{\mathrm{\λ}\left(C_k\right)}{\mathrm{\λ}\left(E_k\right)}---\left(3\right)$

7. using the edge angle point complexity of each gray level as the value of this gray level in histogram, thus obtain the edge angle point histogram of this image.

8. utilize this histogram as template, reference map is scanned.The edge angle point histogram of each scanning position in statistics reference map, and the similarity calculating itself and THE TEMPLATE HYSTOGRAM, similarity measurement adopts maximum Mean Square Error.

After completing the scanning of view picture reference map, the scanning position coordinate with maximum similarity is exported as matching result.Fig. 5, Fig. 6 and Fig. 7 illustrate scanning result in contour map, and wherein dotted line institute region is carrier actual position (namely scheming in real time), and five rectangles represent the highest scanning position of similarity.Can see, along with the change of the real-time reduction of figure signal to noise ratio (S/N ratio), the increase of rotating deviation or real-time figure yardstick, best match position all correctly can contain actual position, has showed the robustness that the present invention is good.

Claims (1)

1., based on the histogrammic underwater terrain matching method of sonar image edge angle point, it is characterized in that, comprise the steps:

(1) Multibeam Data is standardized and interpolation processing, for depth value, by linear transformation, be transformed into 0-255 scope;

(2) projected in XY plane by three-dimensional model, if the gray-scale value of each subpoint to be set to this some height value in the three-dimensional model, then this projection is the gray level image of underwater topography;

(3) travel through each gray level, add up the number of pixels under same gray level, form grey level histogram, obtain the sum of pixel under each gray level;

(4) extract the marginal point of each gray level, and record its position and number, wherein marginal point is defined as:

In all pixels of certain gray level, if at least one four neighborhood pixel has different gray-scale value to certain point from it, then this point is the marginal point of this gray level;

(5) travel through the marginal point of each gray level, extract the edge angle point of each gray level, and record its number, wherein edge angle point is defined as follows:

In all marginal points of certain gray level, if the gray-scale value of certain point is different from the consecutive point of its at least one vertical direction, and the gray-scale value of this point is different from the consecutive point of its at least one horizontal direction, then this point is the edge angle point of this gray level;

A bit depth be L, resolution is in the gray level image of N × M, L is positive integer, and N, M are respectively image length and width, makes g (i, j) represent the gray-scale value of pixel P (i, j), i=1,2 ... N; J=1,2 ... M; K=1,2...2 in this image kth gray level ^l, the set of its marginal point is E _k, the set of edge angle point is C _k, C _ke _ksubset:

$C_k\⊆E_k$

C _krepresent:

$C_k=\left\{P\left(i,j\right)|\begin{array}{c}P\left(i,j\right)\⊆E_k\\ g\left(i,j\right)\≠g\left(i-1,j\right)\left|\right|g\left(i,j\right)\≠g\left(i+1,j\right)\\ g\left(i,j\right)\≠g\left(i,j-1\right)\left|\right|g\left(i,j\right)\≠g\left(i,j+1\right)\end{array}\right\};$

(6) the edge angle point complexity of each gray level is calculated:

The edge angle point complexity γ of certain gray level _kedge angle for this gray level is counted the ratio of counting with edge, wherein λ (E _k) and λ (C _k) edge that is respectively this gray level counts and to count with edge angle

${\mathrm{\γ}}_k=\frac{\mathrm{\λ}\left(C_k\right)}{\mathrm{\λ}\left(E_k\right)};$

(7) using the edge angle point complexity of each gray level as the value of this gray level in histogram, thus obtain the edge angle point histogram of this image;

(8) utilize histogram as template, reference map is scanned; The edge angle point histogram of each scanning position in statistics reference map, and the similarity calculating itself and THE TEMPLATE HYSTOGRAM, similarity measurement adopts maximum Mean Square Error;

(9), after completing the scanning of view picture reference map, the scanning position coordinate with maximum similarity is exported as matching result.