CN109166132B - Side-scan sonar image target identification method with variable initial distance symbolic function - Google Patents

Side-scan sonar image target identification method with variable initial distance symbolic function Download PDF

Info

Publication number
CN109166132B
CN109166132B CN201810779433.XA CN201810779433A CN109166132B CN 109166132 B CN109166132 B CN 109166132B CN 201810779433 A CN201810779433 A CN 201810779433A CN 109166132 B CN109166132 B CN 109166132B
Authority
CN
China
Prior art keywords
target
function
circle
curve
sonar image
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.)
Active
Application number
CN201810779433.XA
Other languages
Chinese (zh)
Other versions
CN109166132A (en
Inventor
杜雪
龚秋婷
严浙平
徐健
李娟�
周佳加
张耕实
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Engineering University
Original Assignee
Harbin Engineering University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin Engineering University filed Critical Harbin Engineering University
Priority to CN201810779433.XA priority Critical patent/CN109166132B/en
Publication of CN109166132A publication Critical patent/CN109166132A/en
Application granted granted Critical
Publication of CN109166132B publication Critical patent/CN109166132B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)

Abstract

A side scan sonar image target identification method with a variable initial distance symbolic function belongs to the field of sonar image target identification. The invention aims to solve the problems that a CV model is not sensitive to the position and the shape of an initial symbol distance function, noise is easy to be regarded as a target in the curve evolution process, and the target is easy to lose by calculating the mean value of the target gray scale through the noise gray scale. The method has the advantages that the position of the target is accurately found when the target deviates from the middle position of the picture; extracting a clear and complete target contour under the condition of large noise in the picture; the method is simple and reliable, easy to implement, small in calculated amount and good in accuracy, improves the feasibility and the practicability of side-scan sonar target identification, and has positive significance for the development of unmanned underwater vehicles in underwater operation tasks and the like in future.

Description

Side-scan sonar image target identification method with variable initial distance symbolic function
Technical Field
The invention belongs to the field of target identification of sonar images, and particularly relates to a side-scan sonar image target identification method with a variable initial distance symbolic function.
Background
Sonar is an effective tool for surveying submarine topography to detect underwater targets. Side scan sonar is a typical active sonar, also known as side view sonar or undersea geomorphology side-detecting instrument. The basic working principle of the side scan sonar is that ultrasonic waves with directivity, wide vertical beam angle and narrow horizontal beam angle are transmitted to the sea bottom by the ultrasonic array, reflected waves and scattered waves of the sea bottom are received by the receiving array, and finally the reflected waves and the scattered waves are processed into acoustic images by the system. The method can display the topography of the sea bottom and determine the approximate position and height of a target, and is a technical tool widely applied to the field of sea bottom imaging. The sonar receives signals with target echoes, reverberation and noise. The target echo is a signal required for forming a sonar image, reverberation and noise can affect the quality of the sonar image, and relevant measures need to be taken to weaken the sonar image by knowing a generation mechanism of the sonar image.
In the aspect of extracting the outline of a side-scan sonar image, scholars have developed related researches by using a threshold method, a clustering method, a Markov random field simulation method and the like. However, when the background of the acoustic image is complex and the noise pollution is serious, ideal contour information is difficult to obtain by various threshold methods and clustering methods. The Markov random field model has a large amount of calculation in the actual calculation process, and has certain difficulty in actual application. In recent years, some scholars develop research on acoustic images by adopting a level set concept, embed an evolved curved surface or a curve into a curved surface represented by a high-dimensional function, and describe image information by using topological changes between the curve and the plane, so as to realize extraction of a target contour. The Chan-Vese model is the most famous model based on the region activity contour based on the level set theory, and is called C-V model for short. The model selects the gray of a certain region of the target to replace the gray of the target, namely the gray of the target is regarded as a mean value region, the mean value of the gray of the regions except the target is solved to be used as a background gray value, then evolution of a level set equation is carried out, and a zero level set is solved to obtain a target contour curve. The method utilizes the gray level mean value of the target and the background, can well filter out noise with irregular shape and gray level close to the background mean value, and has good robustness.
In the process of level set evolution, in order to avoid severe oscillation during the evolution of the level set function, the level set function is generally initialized to be a symbol distance function before the evolution, and then the level set function is periodically reinitialized to be the symbol distance function during the evolution. However, the CV model is insensitive to the position and the shape of the symbol distance function, and the noise is easily regarded as a target in the curve evolution process, and the target gray level mean value is easily calculated by the noise gray level, so that the target loss is easily caused. Therefore, on the basis of the C-V model, a method for changing the position and the shape of the distance symbolic function is added, the proper shape and position of the symbolic distance function are selected to fit the target characteristic gray value and the range, the target contour can be completely extracted, and noise can be well filtered to obtain the clear target contour.
Disclosure of Invention
The invention aims to provide a side-scan sonar image target identification method with a variable initial distance symbolic function. In order to solve the problems that a CV model is not sensitive to the position and the shape of an initial symbol distance function, noise is easily regarded as a target in the curve evolution process, and the target is easily lost due to the fact that the mean value of the target gray scale is calculated by the noise gray scale, a method for changing the position and the shape of the initial symbol distance function is added, and the shape and the position of the initial symbol distance function are selected to fit a target characteristic gray scale value and a target characteristic range, so that the target contour is completely extracted and the noise is well filtered.
The purpose of the invention is realized as follows:
a side-scan sonar image target identification method with a variable initial distance symbolic function comprises the following steps:
(1.1) according to the characteristic that the level set function is periodically reinitialized into a symbol distance function during the evolution of the level set function and whether a point to be measured is in the symbol distance function or not, selecting a circle with the minimum judgment calculation amount as the symbol distance function;
(1.2) according to the selected symbol distance function circle, when the side scan sonar image is input, the gray level characteristics of the target to be extracted are enveloped into the circle center position and the radius of the symbol distance function circle to the maximum extent;
(1.3) establishing an evolution level set function of the C-V model;
and (1.4) solving the active contour model of the C-V model by using a gradient descent method to obtain a parameter evolution equation of the active contour model.
The sign distance function needs to satisfy two conditions of | (d | ═ 1 and d (x, y) ═ 0
The position of the symbol distance function circle requires that when the noise gray value is lighter than the target, a region with the deepest gray level in the target and the largest difference between the noise gray level is selected, the center of the region is used as the center of the symbol distance function circle, and the size of the region is used as the radius of the symbol distance function circle; when the noise gray scale is deeper than the target, selecting a region with the lighter target gray scale and the largest noise gray scale difference, wherein the center of the region is used as the center of a symbol distance function circle, and the size of the region is used as the radius of the symbol distance function circle.
The initial circle radius of the symbol distance function circle requires that the range defined by the symbol distance function circle does not exceed the target range and that the target subject is enveloped in the initial symbol distance function circle, but does not exceed the target range.
The C-V model evolution level set function is established by setting a target contour curve as C
Figure BDA0001732204550000021
Where φ is the level set form of C, λ1、λ2V and μ are normal numbers, c1And c2Mean value of the gray levels inside and outside the curve, Hε(phi) denotes the inner region of the curve, (1-H)ε(phi)) represents the outer region of the curve, and the third term and the fourth term represent the length of the curve and the area of the region enclosed by the curve, respectively.
The gradient descent method comprises the following steps:
(6.1)Hε(phi) and deltaε(φ) is the regularized Heaviside function and Dirac function, respectively, and when ε → 0, there is δε→δ,Hε→ H, defined as follows:
Figure BDA0001732204550000031
(6.2) according to the strength conditions:
Figure BDA0001732204550000032
Figure BDA0001732204550000033
calculating the average gray value by using a step function to obtain:
Figure BDA0001732204550000034
Figure BDA0001732204550000035
wherein c is1And c2The gray level mean values of the inside and the outside of the curve are respectively used as global quantities;
(6.3) minimizing the energy functional in the level set function, the corresponding Euler-Lagrange equation is evolved as:
Figure BDA0001732204550000036
and (6.4) integrating the above expression, solving a value obtained by solving phi at the time t to be 0, and solving to obtain a target profile curve.
The invention has the beneficial effects that:
(1) the side-scan sonar image target identification method with the variable initial distance symbolic function can accurately find the position of a target when the target deviates from the middle position of a picture;
(2) the side-scan sonar image target identification method with the variable initial distance symbolic function can extract clear and complete target contours under the condition that large noise exists in the image;
(3) the side-scan sonar image target identification method with the variable initial distance symbolic function, provided by the invention, is simple and reliable, is easy to realize, is small in calculated amount and good in accuracy, improves the feasibility and the practicability of side-scan sonar target identification, and has positive significance for development of unmanned underwater vehicles in underwater operation tasks and the like in future.
Drawings
FIG. 1 is a flowchart of a side-scan sonar image target identification method operation with a variable initial distance symbolic function;
FIG. 2 is a side scan sonar image containing an object to be identified;
FIG. 3 is a schematic diagram of the position of the original C-V model with a fixed initial symbol distance function;
FIG. 4 is a diagram illustrating the target contour extraction result of the original C-V model with the initial symbol distance function fixed;
FIG. 5 is a diagram illustrating an initial symbol distance function after changing its position and size according to the present invention;
FIG. 6 is a diagram illustrating the result of target contour extraction using the present invention.
Detailed Description
The following further describes embodiments of the present invention with reference to the accompanying drawings:
the invention discloses a sonar target extraction method with a variable initial distance symbolic function. The invention provides a method for improving an active contour model C-V model, a method for changing the position of an initial symbol distance function and a method for selecting the size of the coverage area of the initial symbol function. A side-scan sonar image target identification method based on an active contour variable initial distance symbolic function is characterized by comprising the following steps: for a side-scan sonar image, introducing a level set theory into an active contour model; dividing the sonar image into a target area and a background area according to the characteristics of the target sonar image area in the sonar image; taking the minimum difference between the gray average values of the target area and the background area as an evolution target, namely the minimum energy function of the target background area; according to the characteristic that an active contour model of a C-V model uses the average gray value inside and outside the contour to extract a target contour, most of noise points generated by reverberation and noise of a side scan sonar image are suppressed, and a contour with clear boundary and less noise is extracted; according to the characteristics that the level set function is periodically reinitialized into the symbol distance function and whether the point to be measured is in the symbol distance function or not during the evolution of the level set function, selecting a circle with the minimum judgment calculation amount as the symbol distance function, wherein the circle center position and the radius of the symbol distance function circle are variable; according to the selected symbol distance function circle, when the side scan sonar image is input, the gray level characteristics of the target to be extracted are selected to be enveloped into the circle center position and the radius of the symbol distance function circle to the maximum extent; and solving the active contour model of the C-V model by using a gradient descent method to obtain a parameter evolution equation of the active contour model.
(1) Circular initial distance sign function
When a traditional level set method is used for solving a target contour, a level set equation is initialized to be a symbol distance function, and is initialized to be a distance symbol function periodically in the process of evolution of the level set equation so as to avoid confusion or loss of the target contour caused by severe oscillation in the process of evolution of the level set equation. However, the C-V model is insensitive to the position and the shape of the initial symbol distance function, noise is easily regarded as a target in the curve evolution process, and the target gray level mean value is easily calculated by using the noise gray level, so that the target loss is easily caused.
In order to solve the problem, a method for changing the sign distance function is provided, and finally an ideal target contour is obtained. The size and the position of the circle can be determined by only determining the position and the radius of the circle center, the selection is convenient, and when points inside and outside the target are judged through iterative calculation, the calculated amount is small only by selecting the Euclidean distance from any point on the circle center calculation image to the circle center and subtracting the radius from the Euclidean distance, so that the target extraction time can be greatly shortened, and the circle is selected as a symbol distance function. According to the side-scan sonar image target identification method based on the variable initial distance symbolic function of the active contour, a circle is selected as the distance symbolic function. And during initialization, the position and the radius of the initial SDF circle are selected by a mouse so as to completely extract the outline and well filter noise. The selection requires that the range defined by the SDF circle does not exceed the target range. According to a side-scan sonar image target identification method based on an active contour variable initial distance symbolic function, an initial SDF circle position is selected. When the noise gray value is lighter than the target, selecting an area with the deepest gray in the target and the largest difference from the noise gray; when the noise gray scale is deeper than the target, the region where the target gray scale is lighter and the noise gray scale is most different is selected. According to a side-scan sonar image target identification method based on an active contour variable initial distance symbolic function, an initial SDF circle radius is selected. The range circumscribed by the SDF circle is required not to exceed the target range and to envelope the target subject into the initial SDF circle. But not beyond the target range to be immune to background and noise gray scale effects.
When the circle is selected as the symbolic distance function, the equation is as follows:
φ(x,y)＝x2+y2-1
where | # | ═ 1 and Φ (x, y) ═ 0. This is a implicit functional representation of a circle, which represents the division of a plane Ω into two regions by a unit circle on a two-dimensional plane space Ω, Ω-An inner region of unit satisfying a small phi (x, y) < 0; omega+Is the outer region of the unit circle, satisfies phi (x, y) > 0. The unit circle curve can be expressed as Ω, and satisfies the condition of the symbol distance function: |. Φ | ═ 1 and Φ (x, y) ═ 0, and the three area divisions are specifically: phi (x, y) < 0, representing the inner region of the symbol distance function circle; phi (x, y) > 0, represents the outer region of the symbolic distance function circle; when Φ (x, y) is 0, the symbol distance function circular curve is represented. When the side scan sonar image is input, the gray level characteristics of the target to be extracted are selected to be enveloped into the circle center position and the radius of the symbol distance function circle to the maximum extent.
(2) Energy function model of C-V model
The energy function of the C-V model represents the gray level variance inside and outside the contour curve respectively, and the final goal of evolution is to minimize the internal and external variance of the curve simultaneously. The energy function expression is as follows:
Figure BDA0001732204550000051
in the formula, λ1、λ2V and μ are normal numbers, c1And c2Mean value of the gray levels inside and outside the curve, respectivelyε(phi) denotes the inner region of the curve, (1-H)ε(phi)) represents the outer region of the curve, and the third term and the fourth term represent the length of the curve and the area of the region enclosed by the curve, respectively.
(3) Solution of C-V active contour model
In the equation Hε(phi) and deltaε(φ) is the regularized Heaviside function and Dirac function, respectively, and when ε → 0, there is δε→δ,Hε→ H, defined as follows:
Figure BDA0001732204550000061
the minimization is realized by solving the Euler-Lagrange equation corresponding to the energy functional, and then the corresponding level set function evolution equation is expressed as:
Figure BDA0001732204550000062
in the formula:
Figure BDA0001732204550000063
Figure BDA0001732204550000064
c1and c2The mean values of the gray levels inside and outside the curve are global quantities. After the image to be measured is input, the time t can be introduced by determining the size and the position of the initial symbol distance function, and the partial differential equation is solved by utilizing a gradient descent method.
(4) Adjusting symbol distance function during image input
The operation platform is Matlab 2014a, and the image to be detected after filtering and denoising is input.
And stopping program operation by using a Pause function in a Matlab function library during image initialization, preventing iteration of a level set equation, selecting the circle center position of the SDF circle by using a mouse, and reading a keyboard input numerical value as the radius of the SDF circle.
The selection principle is as follows:
and when the noise gray value is lighter than the target, selecting the center of the area with the deepest gray value in the target and the largest difference between the noise gray value and the target as the center of the SDF circle, and taking the size of the area as the radius.
Similarly, when the noise gray scale is deeper than the target, the center of the area with the lighter target gray scale and the largest noise gray scale difference is selected as the center of the SDF, and the size of the area is taken as the radius.
The selection requires that the range defined by the SDF circle does not exceed the target range.
(5) Results of the experiment
And verifying the text method by using Matlab language programming, and giving a fixed evolution result of the initial symbol distance function and a variable evolution experiment result of the initial symbol distance function for comparison.
Fig. 1 is a side scan sonar image containing an object to be identified. FIG. 2 shows the position and size of the original C-V model initial distance symbolic function, and FIG. 3 shows the result of extracting the sonar image target contour by the original C-V model. The result shows that the original C-V model initial distance symbolic function is fixed at the central part of the image, so that flexible initial contour gray level acquisition cannot be flexibly carried out on targets appearing at different positions of the sonar image, and partial target loss and noise interference phenomena occur.
Fig. 4 is a diagram showing the position of the bit circle of the initial symbolic distance function and the set radius size found by the method described in the text when the level set equation is initialized after an image is input, and fig. 5 is a diagram showing the result of iterative evolution by the method of changing the distance symbolic function. The result shows that under the condition that the sonar image has noise, the method can accurately find the target to be identified, has good noise immunity, and extracts a clear and complete target contour.
(1) Selecting a circle as a distance sign function; the position and radius size of an initial Symbol Distance Function (SDF) circle are selected by a mouse during initialization so as to completely extract the contour and well filter noise. The selection requires that the range defined by the SDF circle does not exceed the target range.
Usually, a circle symbol distance function is selected, which has the advantages that the size and the position of a circle can be determined only by determining the position and the radius of the circle center, the selection is convenient, and when points inside and outside the target are judged through iterative calculation, the calculated amount is small only by selecting the Euclidean distance from any point on a circle center calculation graph to the circle center and subtracting the radius of the Euclidean distance from any point on the circle center calculation graph to the circle center, so that the target extraction time can be greatly shortened.
The sign distance function is selected to satisfy two conditions of | d | ═ 1 and d (x, y) ═ 0, and the certification process is as follows.
For a plane Ω, a symbol distance function φ is defined, and divides the plane Ω into two regions. It is calculated as:
φ(x,y,t＝0)＝Sign(x,y)·d(x,y) (1)
wherein Sign (·) represents a Sign of the distance, that is, Sign (·) · -1 if the point is inside the curve, and Sign (·) +1 if the point is outside the curve; the distance function d (x, y) is calculated as the shortest distance from the point (x, y) to the curve C (x, y) using the euclidean distance.
The distance sign function d (v) is expressed as:
d(v)＝min{|v-vi|},vi∈Ω (2)
for boundaries
Figure BDA0001732204550000072
The v of the point (v) satisfies d (v) ═ 0, otherwise, for any point (v) not on the boundary, if the point (v) nearest to v is found on the boundarycAnd d (v) ═ v-vcL. According to Euclidean distance, d (ν) is expressed as:
Figure BDA0001732204550000071
the gradient of d is then:
Figure BDA0001732204550000081
| (d | ═ 1) is known from the above formula. The re-initialization operation performed periodically during the curve evolution process is to re-calculate the symbol distance function from the closed curve C (x, y, t) expressed by the zero level set to replace the current level set function phi (x, y, t). Since the symbol distance function conforms to the zero-level set function, when t is 0, phi (x, y) is 0, that is, d (x, y) is 0. Therefore, it is necessary to satisfy two conditions of | d | ═ 1 and d (x, y) ═ 0 when selecting the symbol distance function.
The circle-symbol distance function meets the condition of the distance-symbol function, and is selected as the symbol distance function, and the verification process is as follows.
Let y ═ f (x) be an arbitrary curve in a two-dimensional plane space, and let the function represent a correspondence between an abscissa x and an ordinate y of an arbitrary point (x, y) on the curve, and this correspondence is expressed as y-f (x) being 0. If so:
φ(x,y)＝y-f(x,y) (5)
then phi (x, y) ═ 0 is an implicit expression for a two-dimensional planar space curve. Assume the level set function on the two-dimensional plane space Ω as:
φ(x,y)＝x2+y2-1 (6)
the function represents a unit circle in a two-dimensional plane space omega that divides the plane omega into two regions, omega-An inner region of unit satisfying a small phi (x, y) < 0; omega+Is the outer region of the unit circle, and satisfies phi (x, y) > 0; the unit circle curve is expressed as omega, and satisfies phi (x, y) 0
The equation of the circle thus satisfies the condition that the sign distance function | # | -, 1 and Φ (x, y) | 0.
(2) Selecting the position and the radius of an initial symbol distance circle; selecting an initial SDF circle position, wherein when the noise gray value is lighter than the target, selecting an area with the deepest gray value in the target and the largest difference between the noise gray value and the deepest gray value in the target, wherein the center of the area is used as the center of the SDF circle, and the size of the area is used as the radius; when the noise gray scale is deeper than the target, selecting a region with the lighter target gray scale and the largest noise gray scale difference, wherein the center of the region is used as the center of the SDF circle, and the size of the region is used as the radius; selecting the initial SDF circle radius requires that the range circumscribed by the SDF circle not exceed the target range and envelope the target subject into the initial SDF circle, but not exceed the target range to be immune to background and noise gray scale.
(3) Establishing an evolution level set function of the C-V model; if the target contour curve is C, then:
Figure BDA0001732204550000091
where φ is the level set form of C, λ1、λ2V and μ are normal numbers, c1And c2Mean value of the gray levels inside and outside the curve, Hε(phi) denotes the inner region of the curve, (1-H)ε(phi)) represents the outer region of the curve, and the third term and the fourth term represent the length of the curve and the area of the region enclosed by the curve, respectively.
(4) Solving the C-V active contour model, solving the active contour model of the C-V model by using a gradient descent method to obtain a parameter evolution equation of the active contour model, and comprising the following steps of:
the first step is as follows: h in equation (7)ε(phi) and deltaε(φ) is the regularized Heaviside function and Dirac function, respectively, and when ε → 0, there is δε→δ,Hε→ H, defined as follows:
Figure BDA0001732204550000092
the second step is that: according to the strength conditions:
Figure BDA0001732204550000093
Figure BDA0001732204550000094
calculating the average gray value by using a step function to obtain:
Figure BDA0001732204550000095
Figure BDA0001732204550000096
wherein c is1And c2The mean values of the gray levels inside and outside the curve are global quantities.
The third step: by the above-obtained energy function, we can minimize the energy functional in equation (7). The evolution of the corresponding Euler-Lagrange equation is as follows:
Figure BDA0001732204550000101
this is a partial differential equation derived from the level set function of equation (7) with respect to the time parameter.
The fourth step: and (4) integrating the expression (13), and solving a value obtained by solving phi at the time t to be 0 to obtain the target profile curve.

Claims (6)

1. A side scan sonar image target identification method of a variable initial distance symbolic function is characterized by comprising the following steps:
(1.1) according to the characteristic that the level set function is periodically reinitialized into a symbol distance function during the evolution of the level set function and whether a point to be measured is in the symbol distance function or not, selecting a circle with the minimum judgment calculation amount as the symbol distance function;
(1.2) according to the selected symbol distance function circle, when the side scan sonar image is input, the gray level characteristics of the target to be extracted are enveloped into the circle center position and the radius of the symbol distance function circle to the maximum extent;
(1.3) establishing an evolution level set function of the C-V model;
and (1.4) solving the active contour model of the C-V model by using a gradient descent method to obtain a parameter evolution equation of the active contour model.
2. The method for identifying the side-scan sonar image target of the variable initial distance symbolic function according to claim 1, wherein the symbolic distance function needs to satisfy
Figure FDA0003274753130000011
d (x, y) ═ 0.
3. The method for recognizing the side-scan sonar image target with the variable initial distance symbolic function according to claim 1, wherein the method comprises the following steps: the position of the symbol distance function circle requires that when the noise gray value is lighter than the target, a region with the deepest gray level in the target and the largest difference between the noise gray level is selected, the center of the region is used as the center of the symbol distance function circle, and the size of the region is used as the radius of the symbol distance function circle; when the noise gray scale is deeper than the target, selecting a region with the lighter target gray scale and the largest noise gray scale difference, wherein the center of the region is used as the center of a symbol distance function circle, and the size of the region is used as the radius of the symbol distance function circle.
4. The method for recognizing the side-scan sonar image target with the variable initial distance symbolic function according to claim 1, wherein the method comprises the following steps: the initial circle radius of the symbol distance function circle requires that the range defined by the symbol distance function circle does not exceed the target range and that the target subject is enveloped in the initial symbol distance function circle, but does not exceed the target range.
5. The method for recognizing the side-scan sonar image target with the variable initial distance symbolic function according to claim 1, wherein the method comprises the following steps: the C-V model evolution level set function is established by setting a target contour curve as C
Figure FDA0003274753130000012
Where φ is the level set form of C, λ1、λ2V and μ are all normal numbers, c1And c2Mean value of the gray levels inside and outside the curve, Hε(phi) denotes the inner region of the curve, (1-H)ε(phi)) represents the outer region of the curve, and the third term and the fourth term represent the length of the curve and the area of the region enclosed by the curve, respectively.
6. The method for recognizing the side-scan sonar image target with the variable initial distance symbolic function according to claim 1, wherein the method comprises the following steps: the gradient descent method comprises the following steps:
(6.1)Hε(phi) and deltaε(φ) is the regularized Heaviside function and Dirac function, respectively, and when ε → 0, there is δε→δ,Hε→ H, defined as follows:
Figure FDA0003274753130000021
(6.2) according to the strength conditions:
Figure FDA0003274753130000022
Figure FDA0003274753130000023
calculating the average gray value by using a step function to obtain:
Figure FDA0003274753130000024
Figure FDA0003274753130000025
wherein c is1And c2The gray level mean values of the inside and the outside of the curve are respectively used as global quantities;
(6.3) minimizing the energy functional in the level set function, the corresponding Euler-Lagrange equation is evolved as:
Figure FDA0003274753130000026
and (6.4) integrating the above expression, solving a value obtained by solving phi at the time t to be 0, and solving to obtain a target profile curve.
CN201810779433.XA 2018-07-16 2018-07-16 Side-scan sonar image target identification method with variable initial distance symbolic function Active CN109166132B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810779433.XA CN109166132B (en) 2018-07-16 2018-07-16 Side-scan sonar image target identification method with variable initial distance symbolic function

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810779433.XA CN109166132B (en) 2018-07-16 2018-07-16 Side-scan sonar image target identification method with variable initial distance symbolic function

Publications (2)

Publication Number Publication Date
CN109166132A CN109166132A (en) 2019-01-08
CN109166132B true CN109166132B (en) 2022-01-07

Family

ID=64897843

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810779433.XA Active CN109166132B (en) 2018-07-16 2018-07-16 Side-scan sonar image target identification method with variable initial distance symbolic function

Country Status (1)

Country Link
CN (1) CN109166132B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110048783B (en) * 2019-04-19 2021-01-01 美钻深海能源科技研发（上海）有限公司 Sound wave communication unit, system, base station, inspection device and sound wave communication method
CN111260674B (en) * 2020-01-14 2023-04-18 武汉理工大学 Method, system and storage medium for extracting target contour line from sonar image

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779346A (en) * 2012-07-05 2012-11-14 西安电子科技大学 SAR (storage address register) image changing detection method based on improved C-V model
CN103886599A (en) * 2014-03-26 2014-06-25 北京工业大学 Blood vessel ROI dividing method based on intravascular ultrasonic image
CN104217422A (en) * 2014-06-03 2014-12-17 哈尔滨工程大学 Sonar image detection method of self-adaption narrow-band level set
CN104463889A (en) * 2014-12-19 2015-03-25 中国人民解放军国防科学技术大学 Unmanned plane autonomous landing target extracting method based on CV model
CN104657736A (en) * 2013-11-19 2015-05-27 中国科学院沈阳自动化研究所 Active contour-based sonar image mine target recognition method
CN105469408A (en) * 2015-11-30 2016-04-06 东南大学 Building group segmentation method for SAR image
CN107240108A (en) * 2017-06-06 2017-10-10 衢州学院 Movable contour model image partition method based on local Gaussian fitting of distribution and the poor energy driving of local symbol

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779346A (en) * 2012-07-05 2012-11-14 西安电子科技大学 SAR (storage address register) image changing detection method based on improved C-V model
CN104657736A (en) * 2013-11-19 2015-05-27 中国科学院沈阳自动化研究所 Active contour-based sonar image mine target recognition method
CN103886599A (en) * 2014-03-26 2014-06-25 北京工业大学 Blood vessel ROI dividing method based on intravascular ultrasonic image
CN104217422A (en) * 2014-06-03 2014-12-17 哈尔滨工程大学 Sonar image detection method of self-adaption narrow-band level set
CN104463889A (en) * 2014-12-19 2015-03-25 中国人民解放军国防科学技术大学 Unmanned plane autonomous landing target extracting method based on CV model
CN105469408A (en) * 2015-11-30 2016-04-06 东南大学 Building group segmentation method for SAR image
CN107240108A (en) * 2017-06-06 2017-10-10 衢州学院 Movable contour model image partition method based on local Gaussian fitting of distribution and the poor energy driving of local symbol

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Distance Regularized Level Set Evolution and Its Application to Image Segmentation;Chunming Li等;《IEEE TRANSACTIONS ON IMAGE PROCESSING》;20101231;第19卷(第12期);全文 *
基于圆形约束CV-LIF模型的原木端面图像分割;官俊等;《计算机工程与应用》;20141231;全文 *
结合BV-L2分解的CV变分水平集模型;唐利明等;《重庆大学学报》;20180331;第41卷(第3期);全文 *

Also Published As

Publication number Publication date
CN109166132A (en) 2019-01-08

Similar Documents

Publication Publication Date Title
CN109903327B (en) Target size measurement method of sparse point cloud
CN112229404B (en) Method for improving interpolation accuracy of ocean gravity field based on submarine topography three-dimensional optimization principle
CN110663060B (en) Method, device, system and vehicle/robot for representing environmental elements
CN110807781B (en) Point cloud simplifying method for retaining details and boundary characteristics
Luo et al. Sediment classification of small-size seabed acoustic images using convolutional neural networks
CN108764027A (en) A kind of sea-surface target detection method calculated based on improved RBD conspicuousnesses
CN104217422B (en) A kind of sonar image detection method of adaptive narrow-band level set
EP2757529B1 (en) Systems and methods for 3D data based navigation using descriptor vectors
CN105787886A (en) Multi-beam image sonar-based real-time image processing method
EP2054835B1 (en) Target orientation
CN109166132B (en) Side-scan sonar image target identification method with variable initial distance symbolic function
Song et al. Application of acoustic image processing in underwater terrain aided navigation
Long et al. A comprehensive deep learning-based outlier removal method for multibeam bathymetric point cloud
Zhang et al. Object detection and tracking method of AUV based on acoustic vision
CN112907615A (en) Submarine landform unit contour and detail identification method based on region growing
EP2757526A1 (en) Systems and methods for 3D data based navigation using a watershed method
Padial et al. Correlation of imaging sonar acoustic shadows and bathymetry for ROV terrain-relative localization
Song et al. Underwater terrain-aided navigation based on multibeam bathymetric sonar images
US11874407B2 (en) Technologies for dynamic, real-time, four-dimensional volumetric multi-object underwater scene segmentation
Hammond et al. Automated point cloud correspondence detection for underwater mapping using AUVs
Li et al. A combinatorial registration method for forward-looking sonar image
Chen et al. Single Ping Filtering of Multi-Beam Sounding Data Based on Alpha Shapes
Palomer et al. Multi-beam terrain/object classification for underwater navigation correction
Luyuan et al. Sonar Image MRF Segmentation Algorithm Based on Texture Feature Vector
Du et al. Sidescan sonar image target extraction method based on variable initial signed distance function-based active contour CV model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant