CN113642650B - Multi-beam sonar sunken ship detection method based on multi-scale template matching and adaptive color screening - Google Patents
Multi-beam sonar sunken ship detection method based on multi-scale template matching and adaptive color screening Download PDFInfo
- Publication number
- CN113642650B CN113642650B CN202110935841.1A CN202110935841A CN113642650B CN 113642650 B CN113642650 B CN 113642650B CN 202110935841 A CN202110935841 A CN 202110935841A CN 113642650 B CN113642650 B CN 113642650B
- Authority
- CN
- China
- Prior art keywords
- image
- algorithm
- hsv
- sonar
- detection frame
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 110
- 238000012216 screening Methods 0.000 title claims abstract description 25
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 11
- 238000010606 normalization Methods 0.000 claims abstract description 4
- 238000000034 method Methods 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000003384 imaging method Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 description 4
- 238000012876 topography Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of sunken ship detection methods, and particularly discloses a sunken ship detection method based on multi-scale template matching and adaptive color screening, which mainly comprises the following steps: acquiring topographic data of the submarine image by utilizing the multi-beam sonar; establishing a template matching database by using the marked sunken ship area in the multibeam sonar image; calculating the matching degree of a specific region of the sonar image to be detected and the template image by adopting a normalization correlation matching algorithm, and further generating a detection frame and a confidence coefficient corresponding to the optimal matching result; obtaining a detection frame and confidence coefficient thereof under an HSV color space algorithm through self-adaptive HSV threshold screening; based on the detection results of the two algorithms under the same scene and view angle, the position of the detection frame is adjusted by calculating the self-adaptive weight, so that a sunken ship detection result in a final multi-beam sonar image is obtained; the invention provides a sunken ship detection method which is labor-saving and high in detection efficiency.
Description
Technical Field
The invention belongs to the technical field of sunken ship detection methods, and particularly relates to a sunken ship detection method based on multi-scale template matching and adaptive color screening.
Background
The working principle of the multi-beam sonar is that the transmitting transducer array is used for transmitting sound waves covered by a wide sector to the seabed, the receiving transducer array is used for receiving the sound waves in a narrow beam mode, the irradiation footprints of the seabed topography are formed through the orthogonalization of the transmitting and receiving sectors, the footprints are properly processed, and the water depth values of hundreds or even more seabed measured points in a vertical plane perpendicular to the heading can be given out by one detection, so that the size, shape and height change of a target under water in a certain width along a navigation line can be accurately and rapidly measured, and the three-dimensional characteristics of the seabed topography can be reliably depicted (figure I).
In the field of target detection of sonar images, traditional target object identification is usually a manual verification method, but in the underwater detection process of sonar, a large amount of invalid data can be generated, if a sunken ship needs to be found in a large amount of picture data, the manual verification method is time-consuming and laborious. In addition, in order to accurately find that the sunken ship exists in the sonar image, an expert who is quite aware of the field needs to adopt own expertise to judge and identify (for example, the imaging characteristics of the underwater sunken ship under the detection of the multi-beam sonar need experience accumulation to identify, otherwise, the sunken ship is very likely to be filtered out as a background), and if the expert spends a great deal of time in the sonar image to find the sunken ship, the expert also wastes great resources.
At present, in the field of target detection of sonar images, some people try to extract features of the sonar images by using a convolutional neural network through a marked data set, so that the content of the images is learned, and the network can identify objects in the images in the subsequent detection process. However, the difficulty of the method is that the data set of the sonar image is extremely difficult to acquire, the number of single-class samples in the data set is usually thousands or even tens of thousands by the neural network, however, in practical situations, the method is extremely difficult to acquire a high-quality sonar image. Therefore, at the current stage, the adoption of a deep learning method for performing high-accuracy target detection on the sonar image is not realistic.
Up to now, there is no detection method capable of automatically detecting the position of an object in a multi-beam sonar image.
Disclosure of Invention
The invention aims to provide a multi-beam sonar sunken ship detection method which is labor-saving and high in detection efficiency and is based on multi-scale template matching and self-adaptive color screening.
Based on the above purpose, the invention adopts the following technical scheme:
a multi-beam sonar sunken ship detection method based on multi-scale template matching and self-adaptive color screening, which comprises the following steps,
step 1, capturing a sunken ship image marked in a sonar image, and performing rotation, overturning and scaling treatment on the captured sunken ship image; taking the original sunken ship image and the processed sunken ship image as template images, and establishing a template matching database which contains a plurality of different template images;
step 2, acquiring a sonar image to be detected of an area of the seabed by adopting multi-beam sonar; calculating the matching degree of the sonar image to be detected and the template image through a template matching algorithm, and generating a matching algorithm detection frame and a matching algorithm confidence coefficient corresponding to the optimal matching result;
step 3, acquiring a gray level image of a sonar image to be detected through a self-adaptive HSV color threshold screening algorithm, and generating a corresponding HSV algorithm detection frame and an HSV algorithm confidence coefficient;
step 4, setting a threshold value of the cross-correlation ratio IOU, and calculating the cross-correlation ratio of the matching algorithm detection frame and the HSV algorithm detection frame; if the cross ratio is lower than the threshold value, acquiring sonar images to be detected of other areas again, and repeating the steps 2-4; if the cross-correlation ratio is higher than or equal to the threshold value, calculating the weight alpha of the matching algorithm detection frame and the weight beta of the HSV algorithm detection frame; the location of the final detection box and the final confidence are then calculated.
Further, the method for generating the confidence coefficient of the matching algorithm in the step 2 is as follows: sequentially selecting template images from a template matching database, and adopting a normalization correlation matching algorithm CV_TM_CCORR_NORMED in a sliding window mode to obtain the similarity between the template images and the sonar images to be detected, wherein the similarity calculation formula is as follows:
wherein (x ', y') is the position coordinate of the pixel point in the template image T, and (x, y) is the position coordinate of the pixel point in the sonar image I to be detected;
measuring the similarity degree of the template image T and the covered area of the sonar image I to be detected by moving the T one pixel at a time and sliding the T one pixel at each pixel point, and storing the measurement result calculated at each position (x, y) in a corresponding position R (x, y) in a result matrix R; match algorithm confidence match_c=r (x, y) max Wherein R (x, y) max Is the maximum in the result matrix R.
Further, in step 2, the matching algorithm detection frame is a detection frame obtained by using (x) in the sonar image I to be detected 1 ，y 1 ) Is the upper left corner and the width is w 1 The height is h 1 X is the region of (x) 1 For R (x, y) max The value of x, y 1 For R (x, y) max The value of y, w 1 Is the width of the template image, h 1 Is the high of the template image.
Further, in step 3, the method for generating the HSV algorithm detection box includes: acquiring a height value h of the highest position of the seabed in an image to be detected from the sonar image max The method comprises the steps of carrying out a first treatment on the surface of the The self-adaptive HSV color threshold value screening algorithm is based on h max Selecting a corresponding imaging color, and performing inrange () processing on a sonar image to be detected according to the set chromaticity, saturation and brightness threshold of the color to obtain a gray level image corresponding to the sonar image to be detected; performing Loc () processing on the gray level image to obtain an HSV algorithm detection frame on the sonar image to be detected; the upper left and lower right coordinates of the HSV algorithm detection frame are the upper left and lower right coordinates of the white region on the gray scale map,the pixel value of the white area is 255.
Further, in step 3, the calculation formula of the HSV algorithm confidence level HSV-c is:
wherein w is 2 The width h of the HSV algorithm detection frame 2 For HSV algorithm detection box high, v (i, j) is the pixel value of the (i, j) point in the gray scale image.
Further, in step 3, a threshold value of the confidence coefficient of the HSV algorithm is set, if the confidence coefficient of the HSV algorithm is lower than the threshold value, the sonar images to be detected in other areas are acquired again, and step 2-3 is repeated; if the HSV algorithm confidence level is greater than or equal to the threshold, continuing to step 4.
Further, in step 4, the cross-over ratio
Wherein area (template) is a region of a matching algorithm detection frame, and area (HSV) is a region of an HSV algorithm detection frame;
the calculation formula of the final detection frame is as follows:；
the final detection frame is that the upper left corner coordinate on the sonar image to be detected is (x) result ，y result ) A region of width w and height h;
confidence＝α·Match_c+β·HSV_c
further, in step 3, the threshold of the confidence coefficient of the HSV algorithm is set to be 0.4, and in step 4, the threshold of the cross ratio is set to be 0.4.
Compared with the prior art, the invention has the following beneficial effects:
the method has the advantages that the marked sunken ship image in the sonar image of the sunken ship is intercepted, and the data enhancement operations such as rotation, overturning, scaling and the like are carried out on the sunken ship image, so that the purposes of enriching data set samples and improving the robustness of the matching model algorithm can be achieved. And comparing the matching degree of the template image and the sonar image to be detected of the region to be detected by using a template matching library by adopting a normalized correlation matching algorithm, obtaining the result of the optimal matching region, and generating the position information and the confidence result of a matching algorithm detection frame so as to achieve the aim of preliminarily obtaining the detection result.
According to the imaging characteristics of the underwater topography height and the characteristics of the sunken ship which is generally protruded on the seabed, the approximate position of the sunken ship on the seabed can be judged through different imaging colors. The gray level map corresponding to the specific color can be obtained through the self-adaptive HSV threshold value screening algorithm, the gray level map of the color corresponding to the highest position is the position where the sunken ship is possible, the possible position of the sunken ship can be obtained through the imaging position of the white area in the gray level map, and then the confidence coefficient result of the algorithm can be obtained.
The confidence coefficient of the matching algorithm and the confidence coefficient of the HSV algorithm obtained through the two algorithms can obtain respective self-adaptive weights; the adjusted final detection frame position and final confidence coefficient can be obtained through the self-adaptive weight, and algorithm fusion is completed, so that a detection result with high positioning accuracy and high confidence coefficient is achieved.
According to the method, when the sunken ship in the multi-beam sonar image is detected in the underwater archaeological process, the heavy and low-efficiency manual operation of related experts is not needed, and the sunken ship with high efficiency, high positioning accuracy, high confidence and high robustness can be directly detected on a large number of obtained sonar images. According to the method, for automatic detection of the sunken ship in the multi-beam sonar image, on the basis of greatly saving labor and financial cost, the detection accuracy is improved, and the method has a very high use value for underwater archaeology.
Drawings
FIG. 1 is a sonar image of the present invention with a sunken ship;
FIG. 2 is a schematic diagram of a template matching database of the present invention;
FIG. 3 is a sonar image to be detected according to the present invention;
FIG. 4 is a gray scale view of the process of FIG. 3;
fig. 5 is a schematic diagram of the algorithm fusion of the present invention.
Detailed Description
Example 1
A multi-beam sonar sunken ship detection method based on multi-scale template matching and self-adaptive color screening comprises the following steps:
step 1, establishing a template matching database. Firstly, manually selecting one sonar image containing a sunken ship from the previous multibeam sonar images, and taking out an area containing the sunken ship section in the image as a sunken ship image; the intercepted original sunken ship image is respectively processed by rotation, overturning, scaling and the like so as to achieve the purpose of data enhancement; storing the original sunken ship image and the enhanced sunken ship image (the sunken ship image obtained after the sunken ship image is processed) into a template matching database; as shown in fig. 2, a template matching database is established by taking a sunken ship image and a plurality of images processed by the sunken ship image as a template image T.
And 2, a template matching algorithm. Acquiring a sonar image I to be detected of an area of the seabed by adopting multi-beam sonar; and calculating the matching degree of the sonar image to be detected and the template image through a template matching algorithm, and generating a matching algorithm detection frame corresponding to the best matching result and a matching algorithm confidence coefficient. According to the established template matching database, sequentially selecting template images from the template matching database, and sequentially adopting a normalized correlation matching algorithm CV_TM_CCORR_NORMED in a sliding window mode to obtain the similarity between the template images and the sonar images to be detected in the detection area, wherein the similarity calculation formula is as follows:
wherein (x ', y') is the position coordinate of the pixel point in the template image T, and (x, y) is the position coordinate of the pixel point in the sonar image I to be detected; by moving T one pixel at a time (sliding from left to right and from top to bottom), a calculation is performed once at each position on the sonar image to be detected to measure the similarity of the template image T and the covered area of the sonar image I to be detected. And saving the measurement result calculated at each position (x, y) of the sonar image to be detected in a corresponding position R (x, y) in a result matrix R, namely, each position in the result matrix R contains a matching measurement value.
The width and height of the template image are w respectively 1 、h 1 In the obtained result matrix R, the value of R (x, y) represents that in the sonar image I to be detected, the (x, y) is taken as the upper left corner and the width and the height are taken as w 1 、h 1 Matching degree of the region of (c) with the template image. Therefore, only the maximum value in the result matrix needs to be selected to obtain the best matching area. Correspondingly, the result of the confidence coefficient Match-c of the matching algorithm can be directly obtained as the normalization processing is carried out on the related matching algorithm:
Match_c＝R(x，y) max (2)
wherein R (x, y) max Is the maximum in the result matrix R. The matching algorithm detection frame is formed by using (x) in the sonar image I to be detected 1 ,y 1 ) Is the upper left corner and the width is w 1 The height is h 1 X is the region of (x) 1 For R (x, y) max The value of x, y 1 For R (x, y) max The value of y, w 1 Is the width of the template image, h 1 Is the high of the template image.
And 3, a self-adaptive HSV color threshold value screening algorithm. And acquiring a gray level image of the sonar image to be detected through a self-adaptive HSV color threshold screening algorithm, and generating a corresponding HSV algorithm detection frame and an HSV algorithm confidence coefficient. According to the characteristics of underwater archaeology, the height of a sunken ship on the seabed is generally higher than that of the periphery seabed. As shown in fig. 3 and 4, a height value h of the highest position of the seabed in the image to be detected is obtained from the sonar image max The method comprises the steps of carrying out a first treatment on the surface of the The self-adaptive HSV color threshold value screening algorithm automatically carries out the color filtering according to h max Selecting the imaging color h corresponding to the right side of the figure 1 max color H in the present embodiment max color Yellow, and doing an inrange () on the sonar image to be detected according to the set H (chromaticity), S (saturation) and V (brightness) threshold values of the colorObtaining a gray level image (figure 4) corresponding to the sonar image to be detected; the white area (pixel value 255) presented in the gray level image obtained after the processing corresponds to the yellow area in the sonar image to be detected, and is also the area with the highest position in the sonar image to be detected, and the area is the place where sinking ship is most likely to exist in the image. The position of the sunken ship can be characterized by rectangular frames as long as the upper left and lower right corner coordinates of the white area on the gray scale map are obtained.
Performing Loc () processing on the gray level image to obtain an HSV algorithm detection frame on the sonar image to be detected, wherein the HSV algorithm detection frame comprises the following formula:
wherein src represents the original image, and the image processing unit,h max colorlow is a H, S, V threshold value corresponding to the color (imaging color corresponding to the highest point), dst is a gray image returned after the inrange () process, and Loc () is used to obtain the upper left and lower right coordinates of a white region in the gray image. The left upper left corner coordinate and the left lower right corner of the HSV algorithm detection frame are the left upper left coordinate and the right lower right coordinate of the white area respectively.
The confidence level is a visual reflection of the positioning accuracy of the detection frame, and the high-quality detection frame can be selected by setting a threshold value (HSV_confidence_thr > 0.4) for the confidence level so as to prevent false detection. The calculation formula of the HSV algorithm confidence HSV-c is as follows:
wherein w is 2 The width h of the HSV algorithm detection frame 2 For HSV algorithm detection box high, v (i, j) is the pixel value of the (i, j) point in the gray scale image, 255 is the pixel value of the white area. If HSV-c is lower than 0.4, deleting the HSV algorithm detection frame, and re-acquiring the sonar images to be detected of other areasRepeating the steps 2-3; if HSV-c is higher than or equal to 0.4, reserving an HSV algorithm detection frame, and continuing to carry out the step 4.
And 4, algorithm fusion. The algorithm fusion stage mainly has three works: calculating the area intersection ratio (IOU) of the detection frames of the two algorithms; corresponding weights alpha and beta of the two algorithms are adaptively generated according to the confidence coefficient Match-c and the HSV-c; and finally adjusting the position of a final detection frame on the sonar image to be detected by using weights generated by the two algorithms.
(1) Calculating an overlap ratio (IOU): in order to verify the consistency of the results of the two algorithms and thus to improve the accuracy of the final algorithm, it is necessary to compare the template matching algorithm with the detection frames of the HSV color space algorithm. The Intersection-over-Union (IOU) is a concept used in target detection, and here represents the overlapping rate of two algorithms on a detection frame in the same sonar image detection result to be detected, namely the ratio of the Intersection of a detection frame of a matching algorithm and a detection frame of an HSV algorithm to the Union, and the optimal condition is complete overlapping, namely the ratio is 1. The calculation formula of the IOU is as follows:
wherein area (template) is a region of a matching algorithm detection frame, and area (HSV) is a region of an HSV algorithm detection frame;
setting a threshold value IOU_thr to be more than 0.4 for improving the accuracy of the algorithm; the IOU is higher than or equal to 0.4, and all detection frames generated by the two algorithms are reserved and the weight alpha and the weight beta are continuously calculated; if the IOU is lower than 0.4, all detection frames generated by the two algorithms are abandoned, sonar images to be detected of other areas of the seabed are acquired again, and the steps 2-4 are repeated; i.e. to a certain extent, the consistency of the results of both algorithms is ensured.
(2) Generating an adaptive weight alpha and a weight beta: the detection frames which are reserved after the threshold value screening are provided with confidence coefficients representing the accuracy of respective algorithm results, and the formulas for generating the weight alpha and the weight beta are as follows according to the common method that the confidence coefficient self-adaptively distributes the corresponding weight to be the target detection field:
(3) Adjusting the position of a detection frame: and adjusting the position of the detection frame by obtaining the alpha and beta weights. As shown in fig. 5, the upper left corner coordinates of the matching algorithm detection frame area (template) (red rectangular frame in fig. 5) generated by the template matching algorithm is (x) 1 ，y 1 ) Width and height w 1 ，h 1 The upper left corner coordinate of an HSV algorithm detection frame area (HSV) (yellow rectangular frame in fig. 5) generated by the HSV color space algorithm is (x) 2 ，y 2 ) Width and height w 2 ，h 2 Then the following weighting formula is used:
the final detection frame (black rectangle in FIG. 5) is that the upper left corner coordinates on the sonar image to be detected are (x) result ，y result ) A region of width w and height h; the final confidence of the final detection box after weighting is also obtained.
The accurate sunken ship detection result can be obtained through the obtained position of the final detection frame.
Claims (8)
1. A multi-beam sonar sunken ship detection method based on multi-scale template matching and self-adaptive color screening is characterized by comprising the following steps,
step 1, capturing a sunken ship image marked in a sonar image, and performing rotation, overturning and scaling treatment on the captured sunken ship image; taking the original sunken ship image and the processed sunken ship image as template images, and establishing a template matching database;
step 2, acquiring a sonar image to be detected by adopting multi-beam sonar; calculating the matching degree of the sonar image to be detected and the template image through a template matching algorithm, and generating a matching algorithm detection frame and a matching algorithm confidence coefficient corresponding to the optimal matching result;
step 3, acquiring a gray level image of a sonar image to be detected through a self-adaptive HSV color threshold screening algorithm, and generating a corresponding HSV algorithm detection frame and an HSV algorithm confidence coefficient; the method for generating the HSV algorithm detection frame comprises the following steps: acquiring a height value h of the highest position of the seabed in an image to be detected from the sonar image max The method comprises the steps of carrying out a first treatment on the surface of the The self-adaptive HSV color threshold value screening algorithm is based on h max Selecting a corresponding imaging color, and performing inrange () processing on a sonar image to be detected according to the set chromaticity, saturation and brightness threshold of the color to obtain a gray level image corresponding to the sonar image to be detected; performing Loc () processing on the gray level image to obtain an HSV algorithm detection frame on the sonar image to be detected; the upper left corner coordinate and the lower right corner coordinate of the HSV algorithm detection frame are respectively the upper left coordinate and the lower right coordinate of a white area on the gray level diagram, and the pixel value of the white area is 255;
step 4, setting a threshold value of the cross-correlation ratio IOU, and calculating the cross-correlation ratio of the matching algorithm detection frame and the HSV algorithm detection frame; if the cross ratio is lower than the threshold value, acquiring sonar images to be detected of other areas again, and repeating the steps 2-4; if the cross-correlation ratio is higher than or equal to the threshold value, calculating the weight alpha of the matching algorithm detection frame and the weight beta of the HSV algorithm detection frame; and then calculating the position of the final detection frame and the final confidence.
2. The sunken ship detection method based on multi-scale template matching and adaptive color screening as set forth in claim 1, wherein in step 2, the method for generating the confidence level of the matching algorithm is as follows: sequentially selecting template images from a template matching database, and adopting a normalization correlation matching algorithm CV_TM_CCORR_NORMED in a sliding window mode to obtain the similarity between the template images and the sonar images to be detected, wherein the similarity calculation formula is as follows:
wherein (x ', y') is the position coordinate of the pixel point in the template image T, and (x, y) is the position coordinate of the pixel point in the sonar image I to be detected;
measuring the similarity degree of the template image T and the covered area of the sonar image I to be detected by moving the T one pixel at a time and sliding the T one pixel at each pixel point, and storing the measurement result calculated at each position (x, y) in a corresponding position R (x, y) in a result matrix R; match algorithm confidence match_c=r (x, y) max Wherein R (x, y) max Is the maximum in the result matrix R.
3. The sunken ship detection method based on multi-scale template matching and adaptive color screening as claimed in claim 2, wherein in step 2, the matching algorithm detection frame is a detection frame obtained by using (x) in the sonar image I to be detected 1 ，y 1 ) Is the upper left corner and the width is w 1 The height is h 1 X is the region of (x) 1 For R (x, y) max The value of x, y 1 For R (x, y) max The value of y, w 1 Is the width of the template image, h 1 Is the high of the template image.
4. The sunken ship detection method based on multi-scale template matching and adaptive color screening as set forth in claim 3, wherein in step 3, the calculation formula of the HSV algorithm confidence level hsv_c is as follows:
wherein w is 2 The width h of the HSV algorithm detection frame 2 For HSV algorithm detection box high, v (i, j) is the pixel value of the (i, j) point in the gray scale image.
5. The sunken ship detection method based on multi-scale template matching and self-adaptive color screening is characterized in that in the step 3, a threshold value of HSV algorithm confidence is set, if the HSV algorithm confidence is lower than the threshold value, a sonar image to be detected is acquired again, and the step 2-3 is repeated; if the HSV algorithm confidence level is greater than or equal to the threshold, continuing to step 4.
6. The method for detecting sunken ship based on multi-scale template matching and adaptive color screening multi-beam sonar as defined in claim 4, wherein in step 4, the cross-over ratio is
Wherein area (tenplate) is a region of a matching algorithm detection frame, and area (HSV) is a region of an HSV algorithm detection frame; weighting of
；
The calculation formula of the final detection frame is as follows: x is x result ＝αx 1 +βx 2
y result ＝αy 1 +βy 2
w＝αw 1 +βw 2
h＝αh 1 +βh 2
The final detection frame is that the upper left corner coordinate on the sonar image to be detected is (x) result ，y result ) A region of width w and height h;
confidence＝α·Match_c+β·HSV_c。
7. the multi-scale template matching and adaptive color screening-based sunken ship detection method based on multi-beam sonar of claim 5, wherein the threshold value of the confidence level of the HSV algorithm is set to be 0.4 in the step 3.
8. The multi-scale template matching and adaptive color screening-based multi-beam sonar sunken ship detection method as defined in claim 6, wherein the threshold value of the cross-over ratio is set to be 0.4 in the step 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110935841.1A CN113642650B (en) | 2021-08-16 | 2021-08-16 | Multi-beam sonar sunken ship detection method based on multi-scale template matching and adaptive color screening |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110935841.1A CN113642650B (en) | 2021-08-16 | 2021-08-16 | Multi-beam sonar sunken ship detection method based on multi-scale template matching and adaptive color screening |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113642650A CN113642650A (en) | 2021-11-12 |
| CN113642650B true CN113642650B (en) | 2024-02-20 |
Family
ID=78421975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110935841.1A Active CN113642650B (en) | 2021-08-16 | 2021-08-16 | Multi-beam sonar sunken ship detection method based on multi-scale template matching and adaptive color screening |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113642650B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103971127A (en) * | 2014-05-16 | 2014-08-06 | 华中科技大学 | Forward-looking radar imaging sea-surface target key point detection and recognition method |
| CN108520511A (en) * | 2018-03-19 | 2018-09-11 | 中国海洋大学 | A kind of underwater fish target detection and identification method based on fish finder |
| CN108921066A (en) * | 2018-06-22 | 2018-11-30 | 西安电子科技大学 | Remote sensing image Ship Detection based on Fusion Features convolutional network |
| CN110837870A (en) * | 2019-11-12 | 2020-02-25 | 东南大学 | Sonar image target identification method based on active learning |
| CN111767939A (en) * | 2020-05-11 | 2020-10-13 | 浙江大学 | Underwater sonar system target extraction method |
-
2021
- 2021-08-16 CN CN202110935841.1A patent/CN113642650B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103971127A (en) * | 2014-05-16 | 2014-08-06 | 华中科技大学 | Forward-looking radar imaging sea-surface target key point detection and recognition method |
| CN108520511A (en) * | 2018-03-19 | 2018-09-11 | 中国海洋大学 | A kind of underwater fish target detection and identification method based on fish finder |
| CN108921066A (en) * | 2018-06-22 | 2018-11-30 | 西安电子科技大学 | Remote sensing image Ship Detection based on Fusion Features convolutional network |
| CN110837870A (en) * | 2019-11-12 | 2020-02-25 | 东南大学 | Sonar image target identification method based on active learning |
| CN111767939A (en) * | 2020-05-11 | 2020-10-13 | 浙江大学 | Underwater sonar system target extraction method |
Non-Patent Citations (2)
| Title |
|---|
| Weighted boxes fusion: Ensembling boxes from different object detection models;Roman Solovyev et al.;《Image and Vision Computing》;第107卷(第2021期);第1-6页 * |
| 基于车载序列影像中路名牌的提取与识别研究;周江玲;《中国优秀硕士学位论文全文数据库 基础科学辑》(第03期);论文第8-34页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113642650A (en) | 2021-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109615611B (en) | Inspection image-based insulator self-explosion defect detection method | |
| CN105182350B (en) | A kind of multibeam sonar object detection method of application signature tracking | |
| CN106960449B (en) | Heterogeneous registration method based on multi-feature constraint | |
| CN111028154B (en) | Side-scan sonar image matching and stitching method for rugged seafloor | |
| CN110246151B (en) | Underwater robot target tracking method based on deep learning and monocular vision | |
| CN110210448B (en) | Intelligent face skin aging degree identification and evaluation method | |
| CN111985376A (en) | Remote sensing image ship contour extraction method based on deep learning | |
| CN112801074B (en) | Depth map estimation method based on traffic camera | |
| CN108520511A (en) | A kind of underwater fish target detection and identification method based on fish finder | |
| CN115761550A (en) | Water surface target detection method based on laser radar point cloud and camera image fusion | |
| CN109544694A (en) | A kind of augmented reality system actual situation hybrid modeling method based on deep learning | |
| Mahiddine et al. | Underwater image preprocessing for automated photogrammetry in high turbidity water: An application on the Arles-Rhone XIII roman wreck in the Rhodano river, France | |
| CN109961013A (en) | Recognition methods, device, equipment and the computer readable storage medium of lane line | |
| CN104318559A (en) | Quick feature point detecting method for video image matching | |
| CN111192194A (en) | Panoramic image splicing method for curtain wall building vertical face | |
| CN115272353A (en) | Image processing method suitable for crack detection | |
| CN116279592A (en) | Method for dividing travelable area of unmanned logistics vehicle | |
| CN107169412A (en) | Remote sensing image harbor-berthing ship detection method based on mixed model decision | |
| CN114066795A (en) | DF-SAS high-low frequency sonar image fine registration fusion method | |
| CN104268550A (en) | Feature extraction method and device | |
| CN112258536B (en) | Integrated positioning and segmentation method for calluses and cerebellum earthworm parts | |
| CN106204564A (en) | A kind of laser photocentre extracting method | |
| CN113642650B (en) | Multi-beam sonar sunken ship detection method based on multi-scale template matching and adaptive color screening | |
| Li et al. | Vision-based target detection and positioning approach for underwater robots | |
| CN111105390B (en) | Improved sea-sky-line detection and evaluation method |
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 |