A method and system for detecting the overlap width of a body based on sonar images

Abstract

The application provides a kind of based on echo sounder image's body joint width detection method and system, wherein the method comprises: obtaining the first echo sounder image containing target body, the first echo sounder image is input into preset echo sounder image processing neural network model, and the outline and the flap rotating frame corresponding to target body are determined, wherein the long side of the flap is perpendicular to the laying direction of target body;According to the outline corresponding to target body, the joint point and the boundary line of target body are determined, and the joint width of laying is determined according to joint point and boundary line.Because the flap has a significant feature in the echo sounder image, and the long side of the flap is perpendicular to the laying direction of the target body, the image information of the flap can better determine the boundary line of the target body, thereby improving the recognition accuracy of the echo sounder image and improving the generalization ability of the body joint width detection based on the echo sounder image.

Description

Technical Field

[0001] This invention relates to the field of underwater sonar detection and identification, specifically to a method and system for detecting the overlap width of panels based on sonar images. Background Technology

[0002] Traditional methods for detecting the overlap width of rafts mainly involve divers diving into the water to visually inspect or use underwater cameras to check if pre-placed distance warning lines between adjacent rafts are visible. If visible, it indicates that the overlap distance is too small, and the rafting operation is substandard. Manual underwater operations have drawbacks such as high safety risks and high labor costs. Furthermore, the erosion of sediment by water currents can obscure visual or underwater images, leading to inaccurate results. The current mainstream method involves using sonar to scan the underwater work area during rafting. Inspectors then interpret the images based on experience, marking the boundary line between the upper raft and the riverbed, as well as the overlap line (point) between the two rafts. A perpendicular line is drawn from this point to the boundary line, and the length of this perpendicular line, converted to a scale, is used as the actual overlap width detected.

[0003] However, existing methods for detecting the overlap width of objects in sonar images still suffer from insufficient accuracy and poor generalization. Traditional image morphology algorithms rely too heavily on morphological features that exist only in parts of the image, such as sharpness, shadows, and lighting. Therefore, the morphological transformation parameters used are only effective under specific conditions. Other general-purpose sonar image target detection algorithms aim to identify a stationary or moving tangible object, while objects in rows often cover the entire sonar image, making it impossible to distinguish individual objects, and the overlapping areas lack precise object features. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide a method for detecting the overlap width of a row of structures based on sonar images, and a corresponding device for detecting the overlap width of a row of structures based on sonar images, in order to overcome or at least partially solve the above problems.

[0005] To address the aforementioned problems, this invention discloses, in one aspect, a method for detecting the overlap width of rows based on sonar images, comprising:

[0006] Acquire the first sonar image containing the target swarm;

[0007] The first sonar image is input into a preset sonar image processing neural network model, and the contour line corresponding to the target row and the flip-plate rotation frame corresponding to the target row are determined according to the sonar image processing neural network model, wherein the long side of the flip-plate is perpendicular to the laying direction of the target row;

[0008] The overlapping point of the target row and the boundary line of the target row are determined based on the outline of the target row and the flip plate.

[0009] The overlap width of the paving is determined based on the overlap point and the boundary line.

[0010] Optionally, determining the contour line corresponding to the target column based on the sonar image processing neural network model includes:

[0011] The first sonar image is input into a preset sonar image processing neural network model, and a second sonar image is obtained according to the sonar image processing neural network model. The second sonar image includes the outline corresponding to the target structure.

[0012] Optionally, several rafts are connected and set on the riverbed in sequence. The rafts include laid rafts and rafts to be laid. The boundary line of the target raft is the dividing line formed between the side of the laid raft closest to the raft to be laid and the riverbed.

[0013] Optionally, determining the overlap point of the target row body based on the outline corresponding to the target row body and the flap includes:

[0014] The boundary line between the target raft and the riverbed area in the first sonar image is determined based on the outline line corresponding to the target raft.

[0015] The overlap point of the target row is determined according to the boundary line. The overlap point is located in the overlap area of ​​the target row and the row adjacent to the target row.

[0016] Optionally, determining the boundary line of the target row based on the outline corresponding to the target row and the flip plate includes:

[0017] Based on the outline corresponding to the target array, identify the marker in the first sonar image that has a definite angle with the laying direction of the target array;

[0018] Calculate the slope of the boundary line based on the markers;

[0019] Select a point close to the overlapping area and located on the boundary line as the anchor point;

[0020] The boundary line is determined based on the slope of the boundary line and the anchor point;

[0021] Optionally, determining the overlap width of the paving based on the overlap point and the boundary line includes:

[0022] Calculate a first distance value from the overlap point to the boundary line, where the first distance value is the pixel distance in the first sonar image;

[0023] The first distance value is converted into a second distance value according to a preset scale. The second distance value is the overlap width of the row.

[0024] Optionally, the process of training the preset sonar image processing neural network model includes:

[0025] Acquire training data, which includes a third sonar image containing the target swarm;

[0026] The third sonar image is annotated, and the outline of the target platoon and the corresponding flap rotation frame are determined.

[0027] The preset sonar image processing neural network model is trained based on the third sonar image, the outline of the target column, and the flip-top rotating frame corresponding to the target column.

[0028] On the other hand, embodiments of the present invention disclose a system for detecting the overlap width of rows based on sonar images, comprising:

[0029] A data acquisition component, wherein the data acquisition component is used to acquire a first sonar image containing the target swarm;

[0030] A first data processing component is used to input the first sonar image into a preset sonar image processing neural network model, and determine the outline line corresponding to the target row and the flip-plate rotation frame corresponding to the target row according to the sonar image processing neural network model, wherein the long side of the flip-plate is perpendicular to the laying direction of the target row;

[0031] The second data processing component determines the overlap point of the target row and the boundary line of the target row based on the outline line corresponding to the target row and the flip plate.

[0032] A third data processing component determines the overlap width of the paving based on the overlap point and the boundary line;

[0033] A drawing component provides an interactive drawing interface and outputs the boundary line corresponding to the target swarm in the first sonar image, the perpendicular line segment from the overlap point to the boundary line, and the numerical value of the overlap width.

[0034] On the other hand, embodiments of the present invention also provide an electronic device, which includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the sonar image-based method for detecting the overlap width of row structures.

[0035] On the other hand, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the sonar image-based method for detecting the overlap width of row structures.

[0036] This invention acquires a first sonar image containing a target row of structures, inputs the first sonar image into a preset sonar image processing neural network model, and determines the contour line and flap rotation frame corresponding to the target row of structures, wherein the long side of the flap is perpendicular to the laying direction of the target row of structures. The overlap point and boundary line of the target row of structures are determined based on the contour line, and the overlap width is determined based on the overlap point and boundary line. Since the flap has significant features in the sonar image, and the long side of the flap is perpendicular to the laying direction of the target row of structures, combining the image information of the flap can better determine the boundary line of the target row of structures, thereby improving the recognition accuracy of the sonar image and enhancing the generalization ability of row of structures overlap width detection based on sonar images. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of a sonar image-based system for detecting the overlap width of rows according to an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of the physical deployment of a sonar image array overlap width detection system based on a combined model according to an embodiment of the present invention.

[0039] Figure 3 This is a schematic diagram of a sonar image provided in an embodiment of this application;

[0040] Figure 4 These are examples of sonar images and manual interpretation annotations according to embodiments of the present invention;

[0041] Figure 5 This is an example diagram of the input for detecting the overlap width of sonar image arrays according to an embodiment of the present invention;

[0042] Figure 6 This is a flowchart of boundary line detection reasoning according to an embodiment of the present invention;

[0043] Figure 7 This is a flowchart of the focus transformation of the overlapping area according to an embodiment of the present invention;

[0044] Figure 8 This is a visual schematic diagram of the focus change in the overlapping area according to an embodiment of the present invention;

[0045] Figure 9 This is a flowchart of the overlap point detection reasoning process according to an embodiment of the present invention;

[0046] Figure 10 These are visualized sonar images representing the output results of each detection step according to an embodiment of the present invention. Detailed Implementation

[0047] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0048] River regulation and riverbed protection are engineering projects implemented to ensure the stability of navigation channels and slow down the rate of riverbed evolution. Flexible revetments, due to their characteristics of filtration, isolation, scour prevention, and good integrity, are widely used in scour prevention after the construction of dikes and diversion structures, as well as in beach stabilization and bank protection projects. This prevents direct scouring of the riverbed and local deformation and damage caused by water seepage. Currently used flexible revetments (hereinafter also referred to as revetments) are mostly formed by compacting polypropylene woven fabric and other materials with stones or concrete. These revetments are then laid on the riverbed by construction vessels to prevent scouring of the riverbed bottom and to prevent the collapse of dikes or underwater structures. During the laying process, it is necessary to ensure a certain overlap width between adjacent revetments. If the overlap width is too small, water flow will scour the overlap area, causing bottom sand leakage and failure to protect the riverbed. This also increases the difficulty of subsequent construction, bringing greater construction risks and higher construction costs. Therefore, the quality inspection of the revetment laying is very important. First, the sonar image array overlap width detection system based on a combined model, according to an embodiment of the present invention, will be described with reference to the accompanying drawings.

[0049] Figure 1 This is a schematic diagram of a component of a row overlap width detection system based on sonar images, according to an embodiment of the present invention. Figure 1As shown, the system includes: a data acquisition component 100, a data annotation component 200, an imaging analysis component 300, and a drawing component 400. The data acquisition component 100 contains software that controls the acquisition, processing, transmission, and storage of sonar images, used to acquire sonar images and convert them into the input data format required for the combined model. The data annotation component 200 provides data annotation functionality; it receives the original sonar images and user annotation instructions, performs data annotation, and generates annotation data that can be used for training. The imaging analysis component 300 receives the sonar images, uses the combined model sonar image overlap width detection method of this patent for intelligent detection, and outputs the coordinates of the overlap points, the boundary line formula, and the perpendicular coordinates from the overlap points to the layout edge lines. The drawing component 400 provides an interactive drawing interface. By default, it draws the layout edge lines output by the imaging analysis component and the perpendicular line segments from the overlap points to the edge lines, as well as the actual overlap distance after scale conversion, on the original sonar images. Users can also manually adjust and correct errors by dragging the line segments or their endpoints, thereby more accurately determining the overlap width of the layout.

[0050] Figure 2 This diagram illustrates the physical deployment of a row body overlap width detection system, which consists of a sonar device, a server, a display, and necessary communication equipment. The server is responsible for the operation of all system components and can transmit and interact with the sonar device and the display. Specifically, this example uses a two-dimensional mechanical side-scan sonar device. When the data acquisition component 100 in the server receives a corresponding data acquisition request or operation, it controls the sonar device to perform detection. After the imaging analysis component 300 outputs the row body overlap width detection results, the drawing component 400 sends the drawn image to the display for the user to view and modify.

[0051] Figure 3 This is a schematic diagram of a sonar image provided in an embodiment of this application. To determine the overlap width of rows based on the aforementioned sonar image row overlap width detection system using a combined model, this embodiment of the invention provides a row overlap width detection method based on sonar images, comprising:

[0052] A first sonar image containing the target row is acquired; the first sonar image is input into a preset sonar image processing neural network model, and the contour line corresponding to the target row and the flip-plate rotation frame corresponding to the target row are determined according to the sonar image processing neural network model, wherein the long side of the flip-plate is perpendicular to the laying direction of the target row; the overlap point of the target row and the boundary line of the target row are determined according to the contour line corresponding to the target row and the flip-plate; the overlap width of the laying is determined according to the overlap point and the boundary line.

[0053] Since the flap has significant features in sonar images, and the long side of the flap is perpendicular to the laying direction of the target row, combining the image information of the flap can better determine the boundary line of the target row, thereby improving the recognition accuracy of sonar images and enhancing the generalization ability of row overlap width detection based on sonar images.

[0054] In some embodiments, a plurality of rafts are sequentially connected and arranged on a riverbed. The rafts include laid rafts and rafts to be laid. The boundary line of the target raft is a dividing line formed by the laid rafts near the side of the raft to be laid. Determining the contour line corresponding to the target raft according to a sonar image processing neural network model includes: inputting a first sonar image into a preset sonar image processing neural network model, and obtaining a second sonar image according to the sonar image processing neural network model. The second sonar image includes the contour line corresponding to the target raft. A morphological algorithm is used to determine the overlap point and the boundary line in the second sonar image. The morphological algorithm includes dilation, erosion, and edge detection of the binary image.

[0055] For reference only. Figure 4 These are example images of sonar images acquired and manually interpreted and annotated according to an embodiment of the present invention. Figure 5 This is an example input image (original image, unlabeled) of the overlap width detection of the pavement according to an embodiment of the present invention. In this embodiment, a pavement boundary line detection method is provided. This method is executed by the imaging analysis component 300 and is used to detect expressions in the image coordinate system. The pavement boundary line detection method used in this embodiment can employ a neural network model, morphological algorithms, and expert rules accumulated by the user based on their own experience. An exemplary deep learning model can include a layout semantic segmentation model and a flip-board rotating bounding box target detection model. The semantic segmentation model uses UNet, and the rotating bounding box target detection model uses CSL-RetinaNet. The morphological algorithm is a binary mask edge (contour) detection algorithm. The expert rules are: the long side of the flip-board is perpendicular to the paving direction, and the flip-board is located between the image center and the overlap position. The expert rules are set not only based on the fundamental facts of the paving project but also considering that the flip-board has significant features in the sonar image, while the boundary line may be blurred due to noise. Therefore, identifying the angle of significant objects and converting it into the paving direction will have higher accuracy and generalization ability compared to directly identifying the boundary line. The methods listed here are implemented using specific algorithms selected according to the requirements of this embodiment. This invention does not limit the specific algorithms and rules used.

[0056] The method for detecting the boundary line of the row structure in this embodiment can be generally divided into two stages: training and inference. In the training stage, after the data acquisition component 100 receives instructions and acquires sufficient data, the data annotation component 200 provides annotation functionality. A human annotator marks the outline of the area enclosed by the row structure (using polygonal dots) and the quadrilateral presented by the flap area in the sonar image. Then, the imaging analysis component 300 performs model training. In the inference stage, as the first step in the overall process of detecting the overlap width of the row structure, this embodiment is responsible for the boundary line detection process.

[0057] according to Figure 5 As shown, the boundary line detection inference process of this embodiment includes the following steps (all steps are executed by the imaging analysis component 300):

[0058] Step S101: Crop the center of the sonar image into a square (remove invalid black borders around it), with a side length of l. s This is the diameter of the circle in a two-dimensional mechanical side-scan sonar image.

[0059] Step S102: Use the CSL-RetinaNet flip-box object detection model to identify the flip-box expression [x] in the cropped image. f ,y f ,w f ,h f ,θ f ],θ f ∈[-90,0) (This example uses the 90-degree rotated rectangle representation in OpenCV).

[0060] Step S103, calculate the slope k′ of the straight line containing the long side of the flap:

[0061]

[0062] Step S104: Calculate the tiling direction slope as k = -1 / k′.

[0063] Step S105: Use the UNet layout semantic segmentation model to identify the binary mask image of the layout region.

[0064] Step S106: Extract the coordinates of the contour polygon using a binary mask edge detection algorithm. Note that since a binary mask image may consist of multiple connected components, only the outline of the connected component with the largest area is preserved here.

[0065] Step S107: Calculate the slope of the line segment (hereinafter referred to as the contour line segment) between two consecutive adjacent points in the polygon of the mask arrangement.

[0066] Step S108: Filter the contour segments using an angle error threshold δ = 5°, so that the retained contour segments...

[0067] Step S109: Calculate the center point of all retained contour segments.

[0068] Step S110: Select the point among all the center points of the contour line segments that is closest to the center point of the flip plate as the anchor point:

[0069]

[0070] Step S111, determine the boundary line of the target row:

[0071] After obtaining the boundary line slope k of the target slab and the parameter b of the anchor point, the boundary line equation f can be obtained. b (x) = kx + b, b = y a -kx a .

[0072] It should be noted that the overlap point is obtained by sampling the importance of the overlap area, and is determined by the point coordinates p in the image coordinate system. l =(x l ,y l Description. Due to environmental complexity and sonar imaging noise, varying degrees of shadowing, blurring, and distortion often exist around the overlap points, making it difficult to identify specific overlap points. However, some highly probable points can be identified, forming a cluster of overlap points C. p ={p i},p i =(x i ,y i ), i = 1, 2, ..., n. Then p is obtained by the importance sampling method S(·). l =S(C p ).

[0073] The overlap width is the distance from the overlap point to the boundary line, and the overlap point is p. l =(x l ,y l ) to the boundary line f b Distance (x) = kx + b This distance is the pixel distance of the overlap width on the sonar image, which is then converted to the actual overlap distance using a known scale.

[0074] In some embodiments, the overlap point detection method includes a deep learning model and two expert rules. The deep learning model is an overlap region semantic segmentation algorithm, specifically, UNet can be used. The expert rules include: the flap is located between the image center and the overlap region; if the entire image is rotated so that the flap is in a horizontal or vertical state, the flap and the overlap region are always in the same quadrant of the coordinate system with the image center as the origin. The expert rules are still based on the fundamental facts of the laying project. In addition, considering that professional inspectors, when manually interpreting and finding overlap regions, will always focus their attention on a small area near the image center of the flap to carefully observe the overlap shadow between the two arrangements to determine the overlap point, rather than looking for similar shadows on the entire image. This "attention" mechanism can be used as prior knowledge to constrain the scope of the algorithm. Therefore, the method used in this embodiment focuses on a quarter rectangle near the origin in a certain quadrant of the square image (corresponding to a quarter sector of the circle), and transforms the image of this focusing operation into an overlap region focus transformation.

[0075] The overlap point detection method in this embodiment can be generally divided into two stages: training and inference. In the training stage, after the data acquisition component 100 receives instructions and acquires sufficient data, the data annotation component 200 provides annotation functionality. A human annotator marks the overlap area (using polygonal dots) and the quadrilateral formed by the flap area in the sonar image. Then, the imaging analysis component 300 performs model training. In the inference stage, as the second step in the overall process of detecting the overlap width of the row, this embodiment is responsible for the overlap point detection process.

[0076] Note that, as Figure 4 and Figure 5 As shown, overlapping areas are often very blurry and difficult to describe with a single point or line. This invention employs the "importance sampling" method to sample overlapping points from areas of high probability; therefore, annotation only requires marking the approximate area using polygonal dots.

[0077] The focus transformation of the overlapping area is based on the current sonar image and the flap rotation box expression in the image [x] f ,y f ,w f ,h f ,θ f ],θ f The input must be ∈[-90,0) (using OpenCV's 90-degree rotated rectangle representation as an example). During training, other annotation information (such as overlapping region polygons or masks) can also be included. During the training phase, the flip-top rotated box expression is derived from the output of the morphological minimum bounding rectangle algorithm on the manually annotated flip-top quadrilateral region; during the inference phase, the flip-top rotated box expression is derived from the intermediate results of boundary line detection.

[0078] according to Figure 7 As shown, the process of focus transformation in the overlapping area includes the following steps (all steps are performed by the imaging analysis component 300):

[0079] Step S201: Crop the center of the sonar image into a square (remove invalid black borders around it), with a side length of l. s This is the diameter of the circle formed by the two-dimensional mechanical side-scan sonar imaging. Simultaneously, the coordinate offset x in the flip-frame rotation expression, which is affected by image cropping, is adjusted. f and y f If there are other annotations affected by the clipping transformation, adjust those annotations as well.

[0080] Step S202, rotate the image by -θ f Adjust the angle (i.e., clockwise rotation) and simultaneously adjust the expression of the flip-top rotation frame affected by the image rotation. At this point, the flip-top rotation angle should be -90°, i.e., completely horizontal or vertical. If there are other annotations affected by the rotation transformation, adjust those annotations as well.

[0081] Step S203: Determine the quadrant where the flapper is located in the temporary coordinate system with the image center as the origin. The quadrant is determined by the left and right position descriptors I. w and up / down position descriptor I h Decision, I w ,I h ∈{-1,1}. Taking the image coordinate system with the upper left corner as the origin as an example, the formula for calculating the position descriptor is as follows:

[0082] I w =sign(x) f -l s / 2),I h =sign(y f -l s / 2)

[0083]

[0084] Step S204, crop out the image from the rotated image with (l s / 2,l s / 2),(l s / 2+I w (l s / 4),l s / 2+I h (l s / 4)) is the square region at the diagonal vertices, referred to below as the focal image. If there are other annotations affected by the cropping transformation, adjust those annotations as well.

[0085] Figure 8A visualization of the focus transformation in the overlapping region is shown, illustrating the focus image and its position within the original image. In this embodiment, the training of the semantic segmentation model for the overlapping region will be performed only on the focus image. As the second step in detecting the overlap width of the rows, the inference process for detecting the overlap point in this embodiment follows the aforementioned boundary line detection inference process. Figure 9 As shown, the overlap point detection inference process in this embodiment includes the following steps (all steps are executed by the imaging analysis component 300):

[0086] Step S301: Using the overlapping area focus transformation method, the square image cropped in step S101 and the flip-board rotation frame expression [x] detected in step S102 are used. f ,y f ,w f ,h f ,θ f [] is the input, and the focused image is obtained.

[0087] Step S302: Run the forward propagation of the overlapping region semantic segmentation model on the focal image to obtain a binary mask image of the overlapping region. Pixels with a value of 1 are selected as candidate overlapping point clusters C. p ={p i},p i =(x i ,y i ), i = 1, 2, ..., n.

[0088] Step S303: Run the importance sampling method to sample overlap points from the candidate overlap point cluster. Here, the importance sampling function S(·) is defined as the centroid coordinates of the point cluster:

[0089] p l =S(C p ) = mean(p i )

[0090] Step S304: Run the reverse process of the overlapping area focus transformation method to convert the coordinates of the overlapping points in the focus image into the coordinates of the overlapping points in the original image.

[0091] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0092] On the other hand, embodiments of the present invention also provide an electronic device, which includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the sonar image-based method for detecting the overlap width of row structures.

[0093] On the other hand, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the sonar image-based method for detecting the overlap width of row structures.

[0094] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0095] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0100] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0101] The above provides a detailed description of a method and system for detecting the overlap width of row structures based on sonar images, as provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for detecting the overlap width of rows of structures based on sonar images, characterized in that, include: Acquire the first sonar image containing the target swarm; The first sonar image is input into a preset sonar image processing neural network model, and the contour line corresponding to the target row and the flip-plate rotation frame corresponding to the target row are determined according to the sonar image processing neural network model, wherein the long side of the flip-plate is perpendicular to the laying direction of the target row; The overlapping point of the target row and the boundary line of the target row are determined based on the outline of the target row and the flip plate. The overlap width of the paving is determined based on the overlap point and the boundary line; The step of determining the overlap point of the target row body based on the outline corresponding to the target row body and the flip plate includes: The boundary line between the target raft and the riverbed area in the first sonar image is determined based on the outline line corresponding to the target raft. The overlap point of the target row is determined according to the boundary line. The overlap point is located in the overlap area of ​​the target row and the row adjacent to the target row. The step of determining the boundary line of the target row body based on the outline line corresponding to the target row body and the flip plate includes: Based on the outline corresponding to the target array, identify the marker in the first sonar image that has a definite angle with the laying direction of the target array; Calculate the slope of the boundary line based on the markers; Select a point close to the overlapping area and located on the boundary line as the anchor point; The boundary line is determined based on the slope of the boundary line and the anchor point.

2. The method according to claim 1, characterized in that, The step of determining the contour line corresponding to the target column based on the sonar image processing neural network model includes: The first sonar image is input into a preset sonar image processing neural network model, and a second sonar image is obtained according to the sonar image processing neural network model. The second sonar image includes the outline corresponding to the target structure.

3. The method according to claim 1, characterized in that, Several rafts are connected in sequence and placed on the riverbed. The rafts include laid rafts and rafts to be laid. The boundary line of the target raft is the dividing line formed between the side of the laid raft closest to the raft to be laid and the riverbed.

4. The method according to claim 1, characterized in that, Determining the overlap width of the paving based on the overlap point and the boundary line includes: Calculate a first distance value from the overlap point to the boundary line, where the first distance value is the pixel distance in the first sonar image; The first distance value is converted into a second distance value according to a preset scale. The second distance value is the overlap width of the row.

5. The method according to claim 1, characterized in that, The process of training the preset sonar image processing neural network model includes: Acquire training data, which includes a third sonar image containing the target swarm; The third sonar image is annotated, and the outline of the target platoon and the corresponding flap rotation frame are determined. The preset sonar image processing neural network model is trained based on the third sonar image, the outline of the target column, and the flip-top rotating frame corresponding to the target column.

6. A system for detecting the overlap width of row structures based on sonar images, the system being used to implement the method for detecting the overlap width of row structures based on sonar images according to any one of claims 1-5, characterized in that, include: A data acquisition component, wherein the data acquisition component is used to acquire a first sonar image containing the target swarm; An imaging analysis component is used to input the first sonar image into a preset sonar image processing neural network model, and determine, based on the sonar image processing neural network model, the contour line corresponding to the target row and the flip-plate rotation frame corresponding to the target row, wherein the long side of the flip-plate is perpendicular to the laying direction of the target row; determine the overlap point of the target row and the boundary line of the target row based on the contour line corresponding to the target row and the flip-plate; and determine the overlap width of the laying based on the overlap point and the boundary line. A drawing component provides an interactive drawing interface and outputs the boundary line corresponding to the target swarm in the first sonar image, the perpendicular line segment from the overlap point to the boundary line, and the numerical value of the overlap width.

7. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the method as described in any one of claims 1-5.

Images