CN105427342B - A kind of underwater Small object sonar image target detection tracking method and system - Google Patents
A kind of underwater Small object sonar image target detection tracking method and system Download PDFInfo
- Publication number
- CN105427342B CN105427342B CN201510789128.5A CN201510789128A CN105427342B CN 105427342 B CN105427342 B CN 105427342B CN 201510789128 A CN201510789128 A CN 201510789128A CN 105427342 B CN105427342 B CN 105427342B
- Authority
- CN
- China
- Prior art keywords
- target
- image
- template
- area
- current
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000001514 detection method Methods 0.000 title claims abstract description 33
- 238000005259 measurement Methods 0.000 claims abstract description 12
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 230000008569 process Effects 0.000 claims description 19
- 230000011218 segmentation Effects 0.000 claims description 14
- 238000013213 extrapolation Methods 0.000 claims description 8
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 claims 1
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 claims 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims 1
- 238000004364 calculation method Methods 0.000 description 9
- 238000003709 image segmentation Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 230000007123 defense Effects 0.000 description 2
- 238000010191 image analysis Methods 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 241000251468 Actinopterygii Species 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
Landscapes
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
- Image Analysis (AREA)
Abstract
The present invention provide a kind of underwater Small object sonar image target detection tracking method and with the corresponding system of the method, this method includes：Obtain the normalized template of goal-orientation point；Obtain the image for currently including target；It calculates and whether similar to template according to likeness coefficient or image of the related coefficient judgement comprising target, if similar, measurement point of the current image center comprising target of judgement for current goal track；Starting flight path is obtained, the future position of current goal track is obtained using Kalman filtering, measurement point and future position are subjected to data correlation, determine the target trajectory point of present image；Target point trace information is exported according to the target trajectory point of starting flight path and present image.The method that underwater Small object sonar image target detection tracking method disclosed in this invention and system are combined using template matches and likeness coefficient, can greatly save operation time, meet the requirement of real-time of detection system.
Description
Technical Field
The invention relates to the field of image processing and underwater target detection, in particular to an underwater small target sonar image target detection tracking method and system.
Background
At present, underwater defense for targets such as ports, coasts, ships and warships and the like in China is in a weak stage, and particularly defense for small targets such as frogmans, frogman carriers, small AUVs and the like is provided. In the imaging process of the sonar, a large amount of noise and interference sources can be generated in an image due to submarine reverberation, fish schools, reefs and the like, and the existence of the noise and the interference sources makes small underwater targets which are weak originally more difficult to detect.
The object detection of the sonar image aims to extract an object region from a complex seabed reverberation background, and the object region is a key step of image analysis. Only on the basis of accurate detection of sonar image targets, feature extraction and parameter measurement can be carried out on underwater targets, so that higher-level sonar image analysis and identification become possible. However, due to the complexity of the underwater sound field environment and the nonlinearity of imaging of sonar equipment, the acquired underwater sonar images have the characteristics of low contrast, poor imaging quality, serious noise pollution and the like. Because the targets of underwater small targets such as frogmans are small, the signal-to-noise ratio is very low, the targets are often submerged in background noise, the brightness of the targets changes greatly at different distances, and most filtering methods and detection methods have unsatisfactory effects.
The traditional image detection method based on the edge information or the statistical information is difficult to obtain a target detection result with high precision and strong robustness. The target can be well detected under low signal-to-noise ratio by adopting template matching, but because the matched filter needs to carry out related operation, and the related operation belongs to an exhaustive search method, a large amount of time is consumed for operation, so that the time for detecting the sonar image target is too long, and the requirement of the system on real-time property cannot be met.
Disclosure of Invention
The invention aims to provide a method and a system for detecting and tracking an underwater small target sonar image target, which aim to solve the problems.
The underwater small target sonar image target detection and tracking method comprises the following steps:
obtaining a normalized template with the target as a central point;
obtaining an image currently containing a target;
calculating a similarity coefficient or a correlation coefficient between the current image containing the target and the template;
judging whether the current image containing the target is similar to the template or not according to the similarity coefficient or the correlation coefficient,
if the image of the current object is similar to the template, the central point of the image of the current object is judged to be the measuring point of the current object track, and the similarity coefficient or the correlation coefficient of other images containing the object and the template is continuously calculated,
if the current image containing the target is not similar to the template, directly calculating similarity coefficients or correlation coefficients of other images containing the target and the template;
obtaining an initial track;
obtaining a predicted point of the current target track by using Kalman filtering;
determining a target track point of a current image;
and outputting target point track information according to the starting track and the target track point of the current image.
The invention also discloses an underwater small target sonar image target detection and tracking system, which comprises:
the template acquisition module is used for acquiring a normalized template with a target as a central point;
an image segmentation module for obtaining an image currently containing a target;
a similarity judging module, connected to the template obtaining module and the image segmentation module, respectively, for calculating a similarity coefficient or a correlation coefficient between the current image containing the target and the template, and judging whether the current image containing the target is similar to the template according to the similarity coefficient or the correlation coefficient,
if the image of the current object is similar to the template, the central point of the image of the current object is judged to be the measuring point of the current object track, and the similarity coefficient or the correlation coefficient of other images containing the object and the template is continuously calculated,
if the current image containing the target is not similar to the template, directly calculating similarity coefficients or correlation coefficients of other images containing the target and the template;
the starting track acquiring module is used for acquiring a starting track;
the device comprises a predicted point acquisition module, a target trajectory prediction module and a target trajectory prediction module, wherein the predicted point acquisition module is used for acquiring a predicted point of a current target trajectory by using Kalman filtering;
the target track point acquisition module is respectively connected with the similarity judgment module and the prediction point acquisition module and is used for determining a target track point of the current image;
and the output module is respectively connected with the target track point acquisition module and the starting track acquisition module and is used for outputting target point track information according to the starting track and the target track points of the current image.
The invention discloses a method for detecting and tracking targets of underwater small target sonar images. Then obtaining a plurality of images containing the target at present, solving similarity coefficients (or correlation coefficients) between the images containing the target at present and the template by utilizing the template and each image containing the target at present (possibly a target area) in the images, detecting whether pixel points in the images containing the target at present are possible target points according to the size of the similarity coefficients (or correlation coefficients), and taking the pixel points which are possible target points in the images containing the target at present as measuring points. And performing data association on the prediction point of the current target track obtained by Kalman filtering and the measurement point, and determining the target track point of the current image by taking the similarity coefficient (or correlation coefficient) as a weighting coefficient of a distance measurement function during data association. And finally, obtaining the motion trail of the target according to the initial track obtained by adopting a two-point extrapolation method and the target track point of the current image, and outputting the motion trail. The underwater small-target sonar image target detection and tracking method disclosed by the invention adopts a method combining template matching and similarity coefficients, can greatly save the operation time, and meets the real-time requirement of the system.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments of the present invention or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flowchart of an underwater small target sonar image target detection and tracking method disclosed by the embodiment of the invention;
fig. 2 is a schematic view of an underwater small target sonar image target detection tracking system disclosed by the embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses an underwater small target sonar image target detection and tracking method, which comprises the following steps of:
a normalized template centered at the target is obtained.
An image currently containing the target is obtained.
And calculating a similarity coefficient or a correlation coefficient of the image currently containing the target and the template.
And judging whether the current image containing the target is similar to the template or not according to the similarity coefficient or the correlation coefficient, if so, judging that the central point of the current image containing the target is a measuring point of the current target track, continuously calculating the similarity coefficient or the correlation coefficient of the other images containing the target and the template, and if not, directly calculating the similarity coefficient or the correlation coefficient of the other images containing the target and the template.
And obtaining a starting track.
And obtaining a predicted point of the current target track by using Kalman filtering.
And performing data association on the measuring points and the predicting points to determine target track points of the current image.
And outputting target point track information according to the starting track and the target track point of the current image.
The invention discloses a method for detecting and tracking targets of underwater small target sonar images. Then obtaining a plurality of images containing the target at present, solving similarity coefficients (or correlation coefficients) between the images containing the target at present and the template by utilizing the template and each image containing the target at present (possibly a target area) in the images, detecting whether pixel points in the images containing the target at present are possible target points according to the size of the similarity coefficients (or correlation coefficients), and taking the pixel points which are possible target points in the images containing the target at present as measuring points. And performing data association on the prediction point of the current target track obtained by Kalman filtering and the measurement point, and determining the target track point of the current image by taking the similarity coefficient (or correlation coefficient) as a weighting coefficient of a distance measurement function during data association. And finally, obtaining the motion trail of the target according to the initial track obtained by adopting a two-point extrapolation method and the target track point of the current image, and outputting the motion trail. The underwater small-target sonar image target detection and tracking method disclosed by the invention adopts a method combining template matching and similarity coefficients, can greatly save the operation time, and meets the real-time requirement of the system.
The normalized template with the target as a central point is obtained by utilizing brightness information of the target at different distances, and the specific process comprises the following steps:
and constructing a rectangular window with the size of n multiplied by n, wherein the rectangular window divides the sonar image into a plurality of areas, so that subsequent template establishment is facilitated.
Extracting areas A of the targets at different distances in the sonar image by using the window { area ═ area }1,area2,…,areatAnd f, wherein t is the number of the target areas.
For target area a ═ area1,area2,…,areatAny pixel area ini(θ, d) normalized to [0, 1%]The normalized result is:
area‘i(θ,d)＝areai(θ,d)/[max(areai)-min(areai)]wherein i ∈ [1, t ]]，max(areai) Representing and obtaining areaiMaximum pixel value of min (area)i) Representing and obtaining areaiThe minimum pixel value of (a) is θ, which is the angle of the target with respect to the sonar, as the abscissa of the sonar image, and d, which is the distance between the target and the sonar, as the ordinate of the sonar image.
Further, a normalized target region a ' ═ area ' is obtained '1,area’2,…,area‘t}。
Constructing a template T with the size of n multiplied by n, and enabling the normalized target area A 'to be { area'1,area’2,…,area‘tLocating the center of the template T, and determining the pixel value of the template T:
the template T is used as a reference standard of the similarity of the current sonar images.
In the embodiment, the segmentation threshold is obtained by using a method of solving a local mean value, the current sonar image is segmented, the current image containing the target is obtained, and the situation that the remote target is submerged in the background due to the adoption of a global threshold is avoided. The process of obtaining an image currently containing a target includes:
dividing pixels in a current sonar image f (theta, d) with width w and height h into m sections according to the distance from the bottom of the image (corresponding to the position where the sonar is located), wherein each distance section is as follows:
wherein k is 0,1,2, m;
calculating the distance from each pixel in the current sonar image f (theta, d) to the bottom of the image:
dis(θ,d)＝d；
dividing a distance interval where each pixel is located according to the distance from each pixel in the current sonar image to the bottom of the image;
calculating and obtaining the mean value mu and the variance sigma of the pixels in each distance interval2；
The segmentation threshold th is then:
th＝μ+pσ2wherein, p is a proportionality coefficient p ═ dis (θ, d)/max { dis (θ, d) }, p ∈ (0, 1);
and utilizing the segmentation threshold th to segment the distance section corresponding to the segmentation threshold th, and obtaining an image f' (theta, d) containing the target at present.
Since there is one segmentation threshold for each of the m distance sections, m images f' (θ, d) currently containing the target are obtained.
A process of calculating a similarity coefficient or correlation coefficient between an image currently containing an object and a template, comprising:
and carrying out binarization processing on the current image containing the target, and acquiring the central point of each non-zero region in the current image containing the target after binarization processing. And the process comprises:
performing binarization processing on the current image f '(theta, d) containing the target by using the segmentation threshold th to obtain a binarized image f'b(θ,d)：
Labeling binarized image f'bThe non-zero region B in (θ, d) corresponds to m non-zero regions, i.e., B ═ B1,b2,…,bmIn which b isiIs the ith binarized image f'bIn the non-zero region (theta, d), i is greater than or equal to 1 and less than or equal to m.
Calculating a binarized image f'bCenter point c of non-zero region in (theta, d)i(θ,d)：
Wherein M is an image f 'after binarization processing'b(theta, d) the number of pixels in the ith non-zero region, j is an integer, and j is more than or equal to 1 and less than or equal to M, and the center point ci(theta, d) is the jth binarized image f'bA center point of the non-zero region in (θ, d).
And assigning the central point pixel value of the non-zero area to the central point of the template, and assigning values to the template in proportion. The assignment is to associate the value of the template with the position of the region, because different regions are at different distances, and the pixel values of the same object at different distances are different due to the attenuation of the signal, so that two values are obtainedValue-processed image f'bAnd (theta, d), after the central point of the non-zero area in the template is obtained, assigning the central point pixel value of the non-zero area to the central point of the template, and assigning values to the template in proportion, so that the template has brightness characteristics similar to the areas at different distances for subsequent calculation. Since the pixel value of each pixel in the template is [0,1 ]]Wherein the pixel value of the central point is 1, the pixel values of the other pixels can be obtained through a proportional relationship.
And calculating a similarity coefficient or a correlation coefficient between the current image containing the target and the template according to the assignment result of the template. The process comprises the following steps:
calculating the similarity coefficient R between the image containing the target and the templatefT：
Or, calculating a correlation coefficient gamma (theta, d) of the image currently containing the target and the template:
the process of judging whether the current image is similar to the template or not according to the similarity coefficient or the correlation coefficient comprises the following steps:
comparing similarity coefficients RfTAnd a preset similarity threshold th0 if RfT>th0, judging that the image containing the target is similar to the template, and at the moment, judging that the central point of the image containing the target is the measuring point of the current target track; if R isfTAnd h0, judging that the image currently containing the target is not similar to the template, and discarding the image currently containing the target.
Or, taking an absolute value | γ (θ, d) | of the correlation coefficient γ (θ, d);
comparing the absolute value | gamma (theta, d) | of the correlation coefficient with a preset similarity threshold th0, if | gamma (theta, d) | > th0, determining that the image currently containing the target is similar to the template, and at the moment, determining that the central point of the image currently containing the target is a measuring point of the current target track; if | γ (θ, d) | ≦ th0, it is determined that the image currently containing the target is not similar to the template, and the image currently containing the target is discarded without use.
In the present invention, theoretically, the point of the target with the largest pixel value in the current image containing the target is the largest, but when the target area in the current image containing the target is small, the effect of taking the area center point and the area pixel value maximum point is the same, and there may be a plurality of points with the largest pixel value in this area, so it is reasonable to take the area center point as the measurement point of the current target track in this embodiment.
Each track has a start, and the start track is the start. The initial track is obtained by adopting a two-point extrapolation method, the two-point extrapolation method is used for extrapolating a third point by using the measured values of the first two points which are successfully associated, the third point can be used as a predicted value and the measured point obtained by the third frame image is associated with the second point, and if no effective measured point is obtained, the third point obtained by extrapolation is directly used as a track point to calculate the fourth frame image. And the subsequent predicted values are obtained by Kalman filtering.
In this embodiment, the process of obtaining the starting track includes:
measuring coordinate points (theta) according to the first two times of the current ith target1,d1) And (theta)2,d2) And predicting the coordinates of the current ith target by adopting a two-point extrapolation method:
the process of performing data association on the measuring points and the predicting points and determining the target track points of the current image comprises the following steps:
calculating the Euclidean distance Dis between the measuring point and the predicting point:
and in the correlation gate, taking the measured value with the minimum Euclidean distance Dis as a target track point in the current image.
And in the correlation gate, taking the measured value with the minimum Euclidean distance Dis as a target track point in the current image.
It should be noted that a plurality of suspicious target regions (i.e., the image f' (θ, d) currently containing the target) are included in one frame image, and each region corresponds to one similarity coefficient (or correlation coefficient) and euclidean distance. For example, if there is a target track and a plurality of suspicious regions are obtained from the current image, the corresponding similarity coefficient (or correlation coefficient) and euclidean distance are calculated between the suspicious regions and the target track. And calculating the Euclidean distance according to the similarity coefficient (or the correlation coefficient), and selecting the target point in the current image f' (theta, d) containing the target with the minimum Euclidean distance as the track of the current target.
The associated gate is determined according to the motion characteristics of the targets, and a certain type of target has the maximum moving distance within a limited time, and the maximum moving distance is the associated gate. There is no need for calculation outside the correlation gate, and the amount of calculation can be greatly reduced within the correlation.
And finally, outputting target point track information according to the starting track and the target track point of the current image so as to further analyze.
The underwater small target sonar image target detection tracking method disclosed by the invention adopts a template matching method to carry out target detection on sonar images, and can well detect targets under low signal-to-noise ratio. And further, a method of combining template matching and similarity coefficients is adopted, so that the operation time can be greatly saved, and the real-time requirement of the detection system is met.
Another embodiment of the present invention further discloses an underwater small target sonar image target detection tracking system, as shown in fig. 2, including:
the template acquisition module is used for acquiring a normalized template with a target as a central point;
an image segmentation module for obtaining an image currently containing a target;
a similarity judging module, connected to the template obtaining module and the image segmentation module, respectively, for calculating a similarity coefficient or a correlation coefficient between the current image containing the target and the template, and judging whether the current image containing the target is similar to the template according to the similarity coefficient or the correlation coefficient,
if the image of the current object is similar to the template, the central point of the image of the current object is judged to be the measuring point of the current object track, and the similarity coefficient or the correlation coefficient of other images containing the object and the template is continuously calculated,
if the current image containing the target is not similar to the template, directly calculating similarity coefficients or correlation coefficients of other images containing the target and the template;
the starting track acquiring module is used for acquiring a starting track;
the device comprises a predicted point acquisition module, a target trajectory prediction module and a target trajectory prediction module, wherein the predicted point acquisition module is used for acquiring a predicted point of a current target trajectory by using Kalman filtering;
the target track point acquisition module is respectively connected with the similarity judgment module and the prediction point acquisition module and is used for performing data association on the measurement points and the prediction points and determining target track points of the current image;
and the output module is respectively connected with the target track point acquisition module and the starting track acquisition module and outputs target point track information according to the starting track and the target track points of the current image.
Wherein, the template acquisition module comprises:
the window construction unit is used for constructing a rectangular window with the size of n multiplied by n, and the rectangular window divides the sonar image into a plurality of areas so as to facilitate subsequent template establishment.
An image extraction unit connected with the window construction unit and used for extracting areas A ═ area of the targets at different distances in the sonar image by using the window1,area2,…,areatAnd f, wherein t is the number of the target areas.
A normalization unit connected with the image extraction unit and used for normalizing the target area A ═ area1,area2,…,areatAny pixel area ini(θ, d) normalized to [0, 1%]The normalized result is:
area‘i(θ,d)＝areai(θ,d)/[max(areai)-min(areai)]wherein i ∈ [1, t ]]，max(areai) Representing and obtaining areaiMaximum pixel value of min (area)i) Representing and obtaining areaiThe minimum pixel value of (a) is θ, which is the angle of the target with respect to the sonar, as the abscissa of the sonar image, and d, which is the distance between the target and the sonar, as the ordinate of the sonar image. And further obtaining a normalized target area A '═ area'1,area’2,…,area‘t}。
A template construction unit connected to the normalization unit for constructing a template T of size n × n and normalizing the target area a ═ area'1,area’2,…,area‘tLocating the center of the template T, and determining the pixel value of the template T:
the template T is used as a reference standard of the similarity of the current sonar images.
The image segmentation module comprises:
a section dividing unit, configured to divide pixels in a current sonar image f (θ, d) with a width w and a height h into m sections according to a distance from the bottom of the image (corresponding to a position where the sonar is located), where each distance section is:
wherein k is 0,1,2.
And the pixel induction unit is connected with the interval segmentation unit and used for calculating the distance from each pixel in the current sonar image f (theta, d) to the bottom of the image:
dis(θ,d)＝d；
and dividing a distance interval where each pixel is located according to the distance from each pixel in the current sonar image to the bottom of the image.
The threshold calculation unit is connected with the pixel induction unit and used for calculating and acquiring the mean value mu and the variance sigma of the pixels in each distance interval2；
The segmentation threshold th is then:
th＝μ+pσ2where p is a scaling factor, p ═ dis (θ, d)/max { dis (θ, d) }, and p ∈ (0, 1).
And the segmented image acquisition unit is connected with the threshold calculation unit and is used for segmenting the distance section corresponding to the segmented threshold th by using the segmented threshold th to obtain an image f' (theta, d) containing the target at present. Since there is one segmentation threshold for each of the m distance sections, m images f' (θ, d) currently containing the target are obtained.
The similarity determination module comprises:
and the binarization unit is used for carrying out binarization processing on the current image containing the target and acquiring the central point of each nonzero area in the current image containing the target after the binarization processing. The process comprises the following steps:
performing binarization processing on the current image f '(theta, d) containing the target by using the segmentation threshold th to obtain a binarized image f'b(θ,d)：
Labeling binarized image f'bThe non-zero region B in (θ, d) corresponds to m non-zero regions, i.e., B ═ B1,b2,…,bmIn which b isiIs the ith binarized image f'bIn the non-zero region in (theta, d), i is more than or equal to 1 and less than or equal to m;
calculating a binarized image f'bCenter point c of non-zero region in (theta, d)i(θ,d)：
Wherein M is the image after binarization processingf′b(theta, d) the number of pixels in the ith non-zero region, j is an integer, and j is more than or equal to 1 and less than or equal to M, and the center point ci(theta, d) is the jth binarized image f'bA center point of the non-zero region in (θ, d).
And the template assignment unit is connected with the binarization unit and is used for assigning the pixel value of the central point of the non-zero area to the central point of the template and assigning the value to the template in proportion. The assignment is to associate the value of the template with the position of the region, because different regions are at different distances, and the pixel values of the same object at different distances are different due to the attenuation effect of the signal, so that the image f 'after the binarization processing is obtained'bAnd (theta, d), after the central point of the non-zero area in the template is obtained, assigning the central point pixel value of the non-zero area to the central point of the template, and assigning values to the template in proportion, so that the template has brightness characteristics similar to the areas at different distances for subsequent calculation. Since the pixel value of each pixel in the template is [0,1 ]]Wherein the pixel value of the central point is 1, the pixel values of the other pixels can be obtained through a proportional relationship.
And the coefficient calculation unit is connected with the template assignment unit and used for calculating a similarity coefficient or a correlation coefficient between the image of the current contained target and the template according to the assignment result of the template. The process comprises the following steps:
calculating the similarity coefficient R between the image containing the target and the templatefT：
Or, calculating a correlation coefficient gamma (theta, d) of the image currently containing the target and the template:
a similarity judging unit connected to the coefficient calculating unit for judging whether the current image is similar to the template according to the similarity coefficient or the correlation coefficient, the process including:
comparing similarity coefficients RfTAnd a preset similarity threshold th0 if RfT>th0, judging that the image containing the target is similar to the template, and at the moment, judging that the central point of the image containing the target is the measuring point of the current target track; if R isfTAnd h0, judging that the image currently containing the target is not similar to the template, and discarding the image currently containing the target.
Or, taking an absolute value | γ (θ, d) | of the correlation coefficient γ (θ, d);
comparing the absolute value | gamma (theta, d) | of the correlation coefficient with a preset similarity threshold th0, if | gamma (theta, d) | > th0, determining that the image currently containing the target is similar to the template, and at the moment, determining that the central point of the image currently containing the target is a measuring point of the current target track; if | γ (θ, d) | ≦ th0, it is determined that the image currently containing the target is not similar to the template, and the image currently containing the target is discarded without use.
The process of obtaining the initial track by the initial track obtaining module comprises the following steps:
measuring coordinate points (theta) according to the first two times of the current ith target1,d1) And (theta)2,d2) And predicting the coordinates of the current ith target by adopting a two-point extrapolation method:
the process of the target track point acquisition module performing data association on the measurement points and the prediction points and determining the target track points of the current image comprises the following steps:
calculating the Euclidean distance Dis between the measuring point and the predicting point:
and in the correlation gate, taking the measured value with the minimum Euclidean distance Dis as a target track point in the current image.
And in the correlation gate, taking the measured value with the minimum Euclidean distance Dis as a target track point in the current image.
It should be noted that a plurality of suspicious target regions (i.e., the image f' (θ, d) currently containing the target) are included in one frame image, and each region corresponds to one similarity coefficient (or correlation coefficient) and euclidean distance. For example, if there is a target track and a plurality of suspicious regions are obtained from the current image, the corresponding similarity coefficient (or correlation coefficient) and euclidean distance are calculated between the suspicious regions and the target track. And calculating the Euclidean distance according to the similarity coefficient (or the correlation coefficient), and selecting the target point in the current image f' (theta, d) containing the target with the minimum Euclidean distance as the track of the current target.
The associated gate is determined according to the motion characteristics of the targets, and a certain type of target has the maximum moving distance within a limited time, and the maximum moving distance is the associated gate. There is no need for calculation outside the correlation gate, and the amount of calculation can be greatly reduced within the correlation.
And finally, the output module outputs the target point track information according to the starting track and the target track point of the current image so as to further analyze.
The underwater small-target sonar image target detection tracking system disclosed by the invention adopts a template matching method to carry out target detection on sonar images, and can well detect targets under low signal-to-noise ratio. And further, a method of combining template matching and similarity coefficients is adopted, so that the operation time can be greatly saved, and the real-time requirement of the detection system is met.
The underwater small-target sonar image target detection and tracking method provided by the invention is described in detail above. The principles and embodiments of the present invention are explained herein using specific examples, which are presented only to assist in understanding the method and its core concepts. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
Claims (6)
1. The utility model provides a little target sonar image target detection tracking method under water which characterized in that includes:
obtaining a normalized template taking a target as a central point, and constructing a rectangular window with the size of n multiplied by n;
extracting areas A of the targets at different distances in the sonar image by using the window { area ═ area }1,area2,...,areatT is the number of the target areas;
for target area a ═ area1,area2,...,areatAny one ofPlain areai(θ, d) normalized to [0, 1%]The normalized result is:
area‘i(θ,d)＝areai(θ,d)/[max(areai)-min(areai)]wherein i ∈ [1, t ]]，max(areai) Representing and obtaining areaiMaximum pixel value of min (area)i) Representing and obtaining areaiTheta is the angle of the target relative to the sonar, and d is the distance between the target and the sonar;
obtaining a normalized target area A '{ area'1,area‘2,...,area‘t}；
Constructing a template T with the size of n multiplied by n, and enabling the normalized target area A 'to be { area'1,area‘2,...,area‘tLocating the center of the template T, and determining the pixel value of the template T:
obtaining a current image containing a target, dividing pixels in the current sonar image with the width of w and the height of h into m sections according to the distance from the bottom of the image, wherein each distance section is as follows:
wherein k is 0,1,2, m;
calculating the distance from each pixel in the current sonar image to the bottom of the image:
dis(θ,d)＝d；
dividing a distance interval where each pixel is located according to the distance from each pixel in the current sonar image to the bottom of the image;
calculating and obtaining the mean value mu and the variance sigma of the pixels in each distance interval2；
The segmentation threshold th is then:
th＝μ+pσ2wherein, p is a proportionality coefficient, and p belongs to (0, 1);
segmenting the distance interval corresponding to the segmentation threshold th by using the segmentation threshold th to obtain an image f' (theta, d) containing a target at present;
calculating a similarity coefficient or a correlation coefficient between the image currently containing the target and the template, wherein the similarity coefficient R between the image currently containing the target and the template is calculatedfT：
Wherein f' is the current image containing the target, s is the increment of the abscissa, and t is the increment of the ordinate;
calculating a correlation coefficient gamma (theta, d) of the image currently containing the target and the template:
wherein,is the average value of the pixels in the template,the average value of the pixels of the area corresponding to the template in the current image containing the target is obtained;
judging whether the current image containing the target is similar to the template or not according to the similarity coefficient or the correlation coefficient,
if the image of the current object is similar to the template, the central point of the image of the current object is judged to be the measuring point of the current object track, and the similarity coefficient or the correlation coefficient of other images containing the object and the template is continuously calculated,
if the current image containing the target is not similar to the template, directly calculating similarity coefficients or correlation coefficients of other images containing the target and the template;
obtaining an initial track;
obtaining a predicted point of the current target track by using Kalman filtering;
performing data association on the measuring points and the predicting points, and determining target track points of the current image;
and outputting target point track information according to the starting track and the target track point of the current image.
2. The underwater small-target sonar image target detection and tracking method according to claim 1, wherein the process of calculating a similarity coefficient or a correlation coefficient between a current target-containing image and a template includes:
carrying out binarization processing on the current image containing the target, and acquiring the central point of each non-zero region in the current image containing the target after binarization processing;
assigning the pixel value of the central point of the non-zero area to the central point of the template, and assigning values to the template in proportion;
and calculating a similarity coefficient or a correlation coefficient between the current image containing the target and the template according to the assignment result of the template.
3. The underwater small target sonar image target detection and tracking method according to claim 2, wherein the process of binarizing the current target-containing image and obtaining the center point of each non-zero region in the binarized current target-containing image includes:
performing binarization processing on the current image f '(theta, d) containing the target by using the segmentation threshold th to obtain a binarized image f'b(θ,d)：
Calculating a binarized image f'bCenter point c of non-zero region in (theta, d)i(θ,d)：
Wherein M is an image f 'after binarization processing'b(θ, d) the number of pixels in the ith non-zero region, j is an integer, and1≤j≤M。
4. the underwater small-target sonar image target detection and tracking method according to claim 3, wherein the process of judging whether the current image is similar to a template according to the similarity coefficient or the correlation coefficient includes:
comparing similarity coefficients RfTAnd a preset similarity threshold th0 if RfTIf the image containing the target is larger than th0, judging that the image containing the target is similar to the template;
or, taking absolute value | γ (θ, d) | of the correlation coefficient γ (θ, d);
and comparing the absolute value | gamma (theta, d) | of the correlation coefficient with a preset similarity threshold th0, and if | gamma (theta, d) | > th0, judging that the image currently containing the target is similar to the template.
5. The underwater small-target sonar image target detection and tracking method according to claim 4, wherein the process of obtaining the initial track comprises:
measuring coordinate points (theta) according to the first two times of the current ith target1,d1) And (theta)2,d2) And predicting the coordinates of the current ith target by adopting a two-point extrapolation method:
6. the underwater small-target sonar image target detection and tracking method according to claim 5, wherein the process of performing data association on the measurement points and the prediction points to determine the target track points of the current image includes:
calculating the Euclidean distance Dis between the measuring point and the predicting point:
in the correlation gate, taking the minimum measurement value of the Euclidean distance Dis as a target track point in the current image;
or,
and in the correlation gate, taking the measured value with the minimum Euclidean distance Dis as a target track point in the current image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510789128.5A CN105427342B (en) | 2015-11-17 | 2015-11-17 | A kind of underwater Small object sonar image target detection tracking method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510789128.5A CN105427342B (en) | 2015-11-17 | 2015-11-17 | A kind of underwater Small object sonar image target detection tracking method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105427342A CN105427342A (en) | 2016-03-23 |
| CN105427342B true CN105427342B (en) | 2018-06-08 |
Family
ID=55505522
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510789128.5A Active CN105427342B (en) | 2015-11-17 | 2015-11-17 | A kind of underwater Small object sonar image target detection tracking method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105427342B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107526087A (en) * | 2016-06-21 | 2017-12-29 | 北京臻迪科技股份有限公司 | A kind of method and system for obtaining underwater 3D faultage images |
| CN106114095B (en) * | 2016-06-27 | 2018-08-03 | 中国石油大学(华东) | A kind of amphibious sniffing robot |
| CN106338733B (en) * | 2016-09-09 | 2018-11-09 | 河海大学常州校区 | Forward-Looking Sonar method for tracking target based on frogeye visual characteristic |
| KR101720327B1 (en) * | 2016-10-28 | 2017-03-28 | 한국지질자원연구원 | Apparatus and method for localization of underwater anomalous body |
| CN106910204B (en) * | 2016-12-30 | 2018-04-27 | 中国人民解放军空军预警学院监控系统工程研究所 | A kind of method and system to the automatic Tracking Recognition of sea ship |
| CN107290731B (en) * | 2017-06-21 | 2019-12-10 | 哈尔滨工业大学 | Radar track starting method based on image domain template matching |
| CN109242019B (en) * | 2018-09-01 | 2022-05-17 | 哈尔滨工程大学 | Rapid detection and tracking method for optical small target on water surface |
| CN109559321A (en) * | 2018-11-28 | 2019-04-02 | 清华大学 | A kind of sonar image dividing method and equipment |
| CN111426693A (en) * | 2020-04-26 | 2020-07-17 | 湖南恒岳重钢钢结构工程有限公司 | Quality defect detection system and detection method thereof |
| CN113052872B (en) * | 2021-03-12 | 2022-06-10 | 浙江大学 | Underwater moving object tracking method based on sonar image |
| CN116310769A (en) * | 2023-01-03 | 2023-06-23 | 深海技术科学太湖实验室 | Deep sea target detection tracking method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101673403A (en) * | 2009-10-10 | 2010-03-17 | 安防制造（中国）有限公司 | Target following method in complex interference scene |
| CN102215377A (en) * | 2010-04-07 | 2011-10-12 | 北京智安邦科技有限公司 | Device and method for PTZ (planning, tilting and zooming) single-target automatic tracking |
| CN102915545A (en) * | 2012-09-20 | 2013-02-06 | 华东师范大学 | OpenCV(open source computer vision library)-based video target tracking algorithm |
| CN103996208A (en) * | 2014-05-21 | 2014-08-20 | 国家电网公司 | Method for conducting automatic tracking of PTZ single target in video image |
| CN105046258A (en) * | 2015-09-08 | 2015-11-11 | 中国电子科技集团公司第三研究所 | Target detection method and target detection device for small target detection sonar images |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101279561B1 (en) * | 2012-01-19 | 2013-06-28 | 광운대학교 산학협력단 | A fast and accurate face detection and tracking method by using depth information |
-
2015
- 2015-11-17 CN CN201510789128.5A patent/CN105427342B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101673403A (en) * | 2009-10-10 | 2010-03-17 | 安防制造（中国）有限公司 | Target following method in complex interference scene |
| CN102215377A (en) * | 2010-04-07 | 2011-10-12 | 北京智安邦科技有限公司 | Device and method for PTZ (planning, tilting and zooming) single-target automatic tracking |
| CN102915545A (en) * | 2012-09-20 | 2013-02-06 | 华东师范大学 | OpenCV(open source computer vision library)-based video target tracking algorithm |
| CN103996208A (en) * | 2014-05-21 | 2014-08-20 | 国家电网公司 | Method for conducting automatic tracking of PTZ single target in video image |
| CN105046258A (en) * | 2015-09-08 | 2015-11-11 | 中国电子科技集团公司第三研究所 | Target detection method and target detection device for small target detection sonar images |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105427342A (en) | 2016-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105427342B (en) | A kind of underwater Small object sonar image target detection tracking method and system | |
| CN108805904B (en) | Moving ship detection and tracking method based on satellite sequence image | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN102879786B (en) | Detecting and positioning method and system for aiming at underwater obstacles | |
| CN105279772B (en) | A kind of trackability method of discrimination of infrared sequence image | |
| CN107167810B (en) | Method for rapidly extracting underwater target by side-scan sonar imaging | |
| CN102722891A (en) | Method for detecting image significance | |
| CN101739686A (en) | Moving object tracking method and system thereof | |
| CN109919002B (en) | Yellow stop line identification method and device, computer equipment and storage medium | |
| CN105654516B (en) | Satellite image based on target conspicuousness is to ground weak moving target detection method | |
| Lipschutz et al. | New methods for horizon line detection in infrared and visible sea images | |
| CN110415208A (en) | A kind of adaptive targets detection method and its device, equipment, storage medium | |
| CN107942329A (en) | Motor platform single-channel SAR is to surface vessel object detection method | |
| CN106910204A (en) | A kind of method and system to the automatic Tracking Recognition of sea ship | |
| CN111369570A (en) | Multi-target detection tracking method for video image | |
| CN107369163B (en) | Rapid SAR image target detection method based on optimal entropy dual-threshold segmentation | |
| CN109100697A (en) | A kind of target condensing method based on ground surveillance radar system | |
| Li et al. | Outlier-robust superpixel-level CFAR detector with truncated clutter for single look complex SAR images | |
| CN111161308A (en) | Dual-band fusion target extraction method based on key point matching | |
| CN108983194B (en) | Target extraction and condensation method based on ground monitoring radar system | |
| CN111105390A (en) | Improved sea-sky-line detection and evaluation method | |
| CN111332306A (en) | Traffic road perception auxiliary driving early warning device based on machine vision | |
| CN109544574B (en) | Target extraction method based on all-solid-state VTS radar | |
| CN108959355B (en) | Ship classification method and device and electronic equipment | |
| WO2019165626A1 (en) | Methods and apparatus to match images using semantic features |
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 |