Dynamic sea state ship draft automatic identification method, device, medium and product based on relative wave rest algorithm
By processing time-series images of the ship's draft marking area using a relative wave static algorithm, and combining a dynamic leveling reference system and stability determination, the accuracy and robustness issues of automatic ship draft identification under dynamic sea conditions are solved. This achieves high-precision, low-cost, and real-time automatic draft identification, meeting international trade standards.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI MARITIME UNIVERSITY
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-16
AI Technical Summary
Existing technologies are insufficient for achieving high-precision, robust, and low-cost automatic identification of ship draft under dynamic sea conditions, and cannot meet real-time requirements and international trade measurement standards.
A method based on the relative wave static algorithm is adopted. By acquiring the time-series image sequence of the ship's draft indicator area, the instantaneous draft value is extracted. The instantaneous draft value sequence is then processed based on the relative wave static algorithm. Combined with the dynamic leveling reference coordinate system and stability judgment threshold, the final draft value is selected, outliers are removed, and an adaptive feature perception layer and a high-level decision smoothing layer are constructed to achieve accurate identification.
It achieves high-precision, low-cost automatic ship draft identification under dynamic sea conditions, with good robustness and real-time performance. It can meet international trade measurement standards, reduce computational redundancy, and improve the stability and accuracy of the algorithm.
Smart Images
Figure CN122211535A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of marine technology, specifically to a method, equipment, medium, and product for automatic identification of ship draft in dynamic sea conditions based on a relative wave static algorithm. Background Technology
[0002] Draft readings are crucial parameters for reflecting a ship's cargo status, ensuring navigational safety, and facilitating port measurement and settlement. They play a key role in ship stability assessment, loading and unloading operation control, and port supervision. With the rapid development of intelligent shipping and autonomous vessel technologies, achieving automated, accurate, and real-time acquisition of ship draft has become a significant technical requirement for reducing bulk cargo measurement errors and improving port operation efficiency and intelligence.
[0003] Currently, some research has been conducted both domestically and internationally on ship draft identification. The main technical methods include manual visual inspection, hull-mounted robotic inspection, infrared recognition, ultrasonic measurement, fiber optic sensing, and machine vision-based image recognition. Among these, manual visual inspection is heavily influenced by human experience, resulting in low efficiency and difficulty in achieving continuous monitoring. Hull-mounted robotic and sensing methods are costly, complex to maintain, and require strict operating environments, limiting their large-scale application. Visual recognition technology is susceptible to factors such as changes in lighting, water surface reflection, dirt obstruction, and complex background interference in practical applications, and its stability and robustness still need further improvement.
[0004] Current mainstream research on automated recognition mainly focuses on the field of computer vision, and its technical paths can be summarized into the following three categories:
[0005] The first category is recognition frameworks based on object detection. This type of technology mainly uses deep learning models (such as YOLO, SSD, PPOCR, etc.) to locate draft markers and local water surface areas in the surveillance video stream. By recognizing specific draft characters and combining them with image processing algorithms, the relative position of the water surface on the draft gauge is calculated, thereby obtaining the instantaneous draft value.
[0006] The second category is waterline extraction methods based on semantic segmentation. Some existing techniques use U2-NetP or segmentation networks with attention mechanisms to perform pixel-level segmentation of the water body and ship hull boundaries in the image to obtain continuous waterline coordinates.
[0007] The third category is multi-task learning models. Frameworks such as SDRNet attempt to simultaneously perform draft character recognition and waterline coordinate localization. To cope with reading fluctuations caused by waves, such techniques typically introduce multi-frame time averaging or use the median to handle outliers in the backend, and use Z-score for quantitative evaluation, that is, to offset the interference of fluctuations by taking the arithmetic mean or median of the observations over a period of time.
[0008] While the aforementioned computer vision-based methods perform well in laboratory environments or ideal calm sea conditions, they still suffer from the following serious limitations in the dynamic sea conditions of actual port operations: First, existing technologies have blind spots in observing the physical characteristics of waves. Most existing automated algorithms treat draft identification as a purely static image recognition task, while waves in sea conditions are not simple periodic movements but are composed of superimposed wavelets of multiple frequencies, amplitudes, and random phases. Existing multi-frame averaging methods are based on linear assumptions and do not consider the superposition characteristics of waves, failing to effectively cope with the randomness and nonlinearity of waves. This results in a significant deviation between the calculated average value and the actual calm water draft when swells are large. Second, existing technologies lack consistency checks on temporal logic. Existing algorithms often process each frame of the image independently or simply perform numerical stacking, lacking in-depth utilization of the continuity of object motion. When waves fluctuate violently, instantaneous observations contain a large amount of random noise, such as spray obstruction and abrupt changes in the waterline. Existing systems struggle to accurately distinguish the true floating and sinking state of the vessel from this high-frequency noise, leading to frequent jumps in output results and extremely poor robustness. Secondly, existing outlier removal mechanisms lack industrial adaptability. Traditional methods for processing outlier observation data often employ statistical methods based on the assumption of normal distribution (such as Z-score) or density-based clustering methods (such as DBSCAN). However, field observation data is affected by ship movement, changes in light and shadow, and wave impact, often exhibiting significant skewed distributions or containing a large number of extreme outliers. These traditional methods have high computational complexity and low identification accuracy, making it difficult to provide reliable data support in industrial environments while ensuring real-time performance. Finally, the accuracy of existing technologies falls short of international trade standards. According to relevant maritime standards, high-precision draft measurement errors should be controlled within three per thousand. However, due to the aforementioned physical and algorithmic limitations, existing technologies often exceed this standard under dynamic sea conditions, making them unsuitable as a legal basis for cargo volume calculation and trade settlement. They still rely on manual visual inspection, significantly hindering the progress of port digitization.
[0009] Therefore, how to achieve high-precision, robust, and low-cost automatic identification of ship draft under dynamic sea conditions, while meeting real-time requirements and international trade measurement standards, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention
[0010] The embodiments of this application provide a method for automatically identifying the draft of a ship in dynamic sea conditions based on a relative wave static algorithm, which is used to automatically identify the draft value of a ship.
[0011] Firstly, this application provides a dynamic sea state ship draft automatic identification method based on a relative wave stillness algorithm. This method acquires a time-series image sequence containing ship draft indicator regions. Each frame in the time-series image sequence is processed to extract draft indicator features from each frame. The instantaneous draft value corresponding to the image with the draft indicator features is calculated based on these features, forming an instantaneous draft value sequence. The instantaneous draft value sequence is then processed using the relative wave stillness algorithm to obtain the final draft value. The calculation of the final draft value includes the following steps: a preset stability threshold for water surface dwell time is established; the instantaneous draft value sequence is traversed, and based on the stability threshold, the difference between the instantaneous draft values of adjacent frames in the instantaneous draft value sequence is determined to satisfy the stability threshold. If so, the adjacent frames are considered continuous stable frames, and the continuous stable frame count is incremented by 1; otherwise, the continuous stable frame count is reset to 1; the maximum continuous stable frame count is selected from all continuous stable frame counts; and the instantaneous draft value of the frame corresponding to the maximum continuous stable frame count is taken as the final draft value.
[0012] Optionally, the instantaneous draft calculation method includes: extracting at least two feature point pairs from the draft scale line in each frame of the time-series image sequence, each feature point pair including a first feature point and a second feature point located on the same draft scale line; constructing a dynamic leveling reference coordinate system based on the at least two feature point pairs for tilt correction of the image; obtaining multiple contact points between the hull and the water surface in the image as waterline feature points; determining the waterline position based on the tilt-corrected image and the waterline feature points, and calculating the instantaneous draft value in conjunction with the draft scale line.
[0013] Optionally, constructing a dynamic leveling reference coordinate system includes: determining the reference baseline equation based on at least two pairs of feature points through fitting, wherein the fitting method is the least squares method.
[0014] Optionally, determining the waterline location includes: performing linear fitting on multiple waterline feature points to obtain the waterline equation.
[0015] Optionally, the first feature point is the center of the ship's draft scale, and the second feature point is the center of the number corresponding to the center of the draft scale.
[0016] Optionally, the stability determination threshold is dynamically adjusted based on real-time acquired environmental parameters, including sea state level and wind speed.
[0017] Optionally, the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm also includes: obtaining the instantaneous draft value of the frame corresponding to the maximum continuous stable frame count, removing outliers, and taking the instantaneous draft value without outliers as the final draft value.
[0018] Secondly, this application also provides an electronic device, including a memory and a processor; the memory stores a computer program, and the processor runs the computer program in the memory to perform operations in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm provided in the first aspect.
[0019] Thirdly, this application also provides a storage medium storing multiple instructions adapted for loading by a processor to execute the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm provided in the first aspect.
[0020] Fourthly, this application also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm provided in the first aspect.
[0021] The method, equipment, medium, and product for automatic ship draft identification in dynamic sea conditions based on the relative wave stillness algorithm provided in this application achieve accurate identification of effective observation points by capturing the "quasi-static" features that momentarily approach equilibrium during wave superposition and interference. Secondly, by capturing the "metastable moment" in the dynamic environment, this application can effectively filter out nonlinear interference caused by wave splash, light and shadow fluctuations, and random sea state noise. When waves are violently fluctuating, existing algorithms struggle to accurately distinguish the ship's true floating state from high-frequency noise. This application, however, actively captures the optimal observation moment when the water surface is relatively still during random fluctuations by tracking the continuity of draft values between frames, achieving "selective acquisition" rather than full-time averaging, thus exhibiting better robustness. Furthermore, this application employs a computationally efficient algorithm architecture, reducing computational redundancy through a decoupling framework, ensuring real-time synchronous output of data under high-frequency dynamic sampling, and possessing real-time processing advantages. In addition, this application constructs a coupled processing architecture combining deep learning semantic features and improved statistical filtering, using an interquartile range (IQR) method instead of the traditional Z-score method. Field observation data is often affected by ship movement, changes in light and shadow, and the impact of waves, resulting in a significant skewed distribution or a large number of extreme outliers. The IQR method used in this application exhibits stronger robustness when processing industrial field data containing extreme abrupt changes, providing reliable data support in industrial environments while ensuring real-time performance. Finally, this application employs a highly decoupled modular design, making the lower-level feature perception layer and the higher-level decision smoothing layer completely independent, without relying on a specific visual detection model, and enabling generalized processing of multi-source instantaneous feature data streams. The feature perception layer not only supports convolutional neural network models such as YOLO series, SSD, and PPOCR, but is also compatible with attention mechanism models such as ViT and Swin-Transformer, as well as semantic segmentation networks. It can also establish an adaptive reference system by utilizing the inherent geometric features of the hull. By automatically identifying and connecting the center point of the M scale with the corresponding draft number center point, it calculates the average slope of multiple sets of feature line segments, anchors the ship's physical structure in real time, and automatically compensates for image rotation caused by ship roll, pitch, and camera shooting angle deviations. This ensures the consistency of the reference benchmark under dynamic sea conditions and achieves accurate tilt correction without manual intervention. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1This is a flowchart illustrating the automatic identification method for dynamic sea state ship draft based on a relative wave static algorithm provided in an embodiment of this application.
[0024] Figure 2 This is a schematic diagram of the calculation steps for the final draft value in some embodiments of this application;
[0025] Figure 3 These are schematic diagrams of dynamic leveling reference systems in some embodiments of this application;
[0026] Figure 4 This is a schematic diagram of Test Example 1 of this application;
[0027] Figure 5 This is a schematic diagram of Test Example 2 of this application;
[0028] Figure 6 This is a schematic diagram of Test Example 3 of this application. Detailed Implementation
[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0030] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0031] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.
[0032] The use of "applies to" or "configured to" in this application implies open and inclusive language, which does not preclude applicability to or configuration to devices performing additional tasks or steps. Furthermore, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more conditions or values may in practice be based on additional conditions or values beyond those conditions.
[0033] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.
[0034] The following describes, with reference to the accompanying drawings, the method, equipment, medium, and product for automatic identification of ship draft in dynamic sea conditions based on a relative wave static algorithm provided in the embodiments of this application.
[0035] Figure 1 This is a flowchart illustrating the automatic identification method for dynamic sea state ship draft based on a relative wave static algorithm according to an embodiment of this application.
[0036] like Figure 1 As shown in the embodiment of this application, a method for automatic identification of ship draft in dynamic sea conditions based on a relative wave static algorithm is provided to automatically identify the ship's draft value, including:
[0037] Step 102: Obtain a time-series image sequence containing the ship's draft indicator area;
[0038] Step 104: Process each frame of the time-series image sequence, extract the draft marker features in each frame, calculate the instantaneous draft value corresponding to the image with the draft marker features based on the draft marker features, and form an instantaneous draft value sequence.
[0039] In some embodiments of this application, feature extraction is performed through a feature-aware layer. This layer is configured with a replaceable feature extraction engine capable of extracting the coordinates of key points on the water level gauge or semantic features. The feature extraction engine may include: Convolutional Neural Network (CNN) series: such as YOLOv5-v11, SSD, PPOCR, and other object detection models; Transformer series: such as ViT, Swin-Transformer, and other attention mechanism models; and semantic segmentation networks: capable of pixel-level segmentation of the water level gauge reading region.
[0040] Those skilled in the art will understand that while some embodiments of this application list feature extraction engines, these are merely exemplary and not intended to limit the scope of this application. In other embodiments of this application, other suitable architectures may also be used, such as one-stage detectors: including but not limited to SSD, FCOS, EfficientDet, or PP-series networks optimized for edge computing, to achieve higher inference speeds. Another example is two-stage detectors: such as Faster R-CNN and its variants, which further improve localization accuracy in complex backgrounds through pre-selection mechanisms. Yet another example is dedicated text / keypoint recognition networks: these can also integrate text recognition frameworks such as PP-OCR to directly locate and semantically parse characters, serving as auxiliary verification for coordinate extraction.
[0041] In some embodiments of this application, the method for calculating instantaneous draft includes:
[0042] Step 1: Extract at least two feature point pairs from the draft scale line in each frame of the time-series image sequence. Each feature point pair includes a first feature point and a second feature point located on the same draft scale line.
[0043] Step 2: Based on at least two feature point pairs, construct a dynamic leveling reference coordinate system for image tilt correction.
[0044] Step 3: Obtain multiple contact points between the ship's hull and the water surface in the image as waterline feature points;
[0045] Step 4: Based on the tilt-corrected image and waterline feature points, determine the waterline position and calculate the instantaneous draft value by combining the draft scale line.
[0046] Figure 3 This is a schematic diagram of a dynamic leveling reference system in some embodiments of this application.
[0047] like Figure 3 As shown, in some embodiments of this application, constructing a dynamic leveling reference coordinate system includes: determining the reference baseline equation based on at least two feature point pairs by fitting, wherein the fitting method is the least squares method. Each feature point pair includes a first feature point and a second feature point, the first feature point being the center of the ship's draft scale (i.e., the M scale center), and the second feature point being the digital center corresponding to the scale center.
[0048] In some embodiments of this application, determining the waterline position includes: performing linear fitting on multiple waterline feature points to obtain the waterline equation.
[0049] like Figure 3 As shown, in some embodiments of this application, multiple parallel draft lines are formed by automatically identifying and connecting the M-markers with their corresponding digital centers. By calculating the average slope of these lines, a dynamic leveling reference system anchored to the ship's physical structure is established, thereby achieving automatic correction of image deflection angles and ensuring the consistency of the reference benchmark under dynamic sea conditions. Key points of the waterline and the M-line of the draft gauge are extracted from the automatically corrected image, and linear regression processing is performed. By minimizing the sum of squared vertical distances from the observation point to the fitted line, accurate waterline equations and benchmark M-line equations are constructed, respectively.
[0050] In some embodiments of this application, the intersection of the dynamic waterline equation and the water gauge scale lines is located based on the fitted geometric relationship. Using linear interpolation, the actual physical draft corresponding to the pixel coordinates is calculated based on the vertical distance between the intersection point and adjacent scale lines, transforming the visual detection results into a physically meaningful instantaneous draft observation sequence.
[0051] Step 106: Process the instantaneous draft value sequence based on the relative wave static algorithm to obtain the final draft value.
[0052] Figure 2 This is a flowchart 200 showing the steps for calculating the final draft in some embodiments of this application.
[0053] like Figure 2 As shown, in some embodiments of this application, the calculation of the final draft includes the following steps:
[0054] Step 202: Set a stability threshold for the water surface dwell time;
[0055] Step 204: Traverse the instantaneous draft value sequence and determine whether the difference between the instantaneous draft values of adjacent frames in the instantaneous draft value sequence meets the stability determination threshold based on the stability determination threshold. If yes, the adjacent frames are determined to be continuous stable frames, and the continuous stable frame count is incremented by 1; otherwise, the continuous stable frame count is reset to 1.
[0056] Step 206: Select the maximum consecutive stable frame count from all consecutive stable frame counts;
[0057] Step 208: Use the instantaneous draft value of the frame corresponding to the maximum continuous stable frame count as the final draft value.
[0058] In some embodiments of this application, the algorithm identifies the number of frames with the longest water surface dwell time through frame sequence analysis, i.e., finding the maximum continuous stable interval, and extracts the instantaneous draft value Di for each frame in the N frames of the video stream. In some embodiments of the application, the determination logic is to set a continuous stable frame count Ci. If the draft difference between adjacent frames |Di+1 - Di|=0, then the counter Ci →Ci + 1; otherwise, it is reset to 1. All counts are iterated, and the draft value Dfinal corresponding to the maximum number of continuous frames Cmax is selected as the final result.
[0059] Cmax = max{C1, C2, . . . , CN};
[0060] Dfinal = Dk where Ck = Cmax;
[0061] Where Cmax is the maximum number of consecutive frames, and Dfinal is the final dynamic draft recognition result.
[0062] In some embodiments of this application, the stability determination threshold is dynamically adjusted based on real-time acquired environmental parameters, including sea state level and wind speed.
[0063] In some other embodiments of this application, the method for automatic identification of ship draft in dynamic sea conditions based on the relative wave stillness algorithm further includes: obtaining the instantaneous draft value of the frame corresponding to the maximum continuous stable frame count, removing outliers, and using the instantaneous draft value without outliers as the final draft value.
[0064] This application uses three test cases as examples to illustrate the automatic identification method of ship draft in dynamic sea conditions based on the relative wave static algorithm. Those skilled in the art will understand that the three test cases are examples and are not intended as a limitation not to apply.
[0065] Figure 4 This is a schematic diagram of Test Example 1 of this application.
[0066] Test Example 1
[0067] Location: Yangjiang Wharf
[0068] Device: iPhone 14 Pro
[0069] Sea state: Light waves (Force 1-3 winds)
[0070] Dataset: 150 labeled images
[0071] Model training:
[0072] Using pre-trained models such as YOLO11
[0073] Hyperparameter settings: epoch=100, batch_size=auto, optimizer=SGD, lr=0.001, momentum=0.9
[0074] Hardware: NVIDIA RTX 4070, Intel i5-12800HX, 16GB RAM
[0075] Recognition process:
[0076] Frames are extracted from the video and input into the YOLO model to detect watermarks and characters.
[0077] A dynamic water reference system is constructed, and the waterline equation is fitted.
[0078] Calculate the draft value Di for each frame and count the number of consecutive stable frames Ci.
[0079] Choose Dfinal=4.85m corresponding to Cmax=14.
[0080] Use IQR to remove outliers and output the final draft.
[0081] like Figure 4 As shown, the results are: recognition error: 0%, median processing time: 32.0ms.
[0082] Figure 5 This is a schematic diagram of Test Example 2 of this application.
[0083] Test Example 2: Draft Recognition under Medium Wave Conditions
[0084] Experimental environment:
[0085] Location: Vietnam Port
[0086] Device: Huawei Mate 60
[0087] Sea state: Moderate waves (Force 5 winds)
[0088] Dataset: Same as test case 1
[0089] Recognition process: Same as test example 1, finally select Dfinal=5.71m corresponding to Cmax=14.
[0090] like Figure 2 As shown, the results are: recognition error: 0.17%, median processing time: 36.0 ms.
[0091] Figure 6 This is a schematic diagram of Test Example 3 of this application.
[0092] Test Example 3: Draft Recognition under High Wave Conditions
[0093] Experimental environment:
[0094] Same as in Example 2, but with large waves.
[0095] Recognition process:
[0096] Similar to Example 1, the final value of Dfinal=13.50m corresponding to Cmax=11 was selected.
[0097] like Figure 6 As shown, the results are: recognition error: 0.07%, median processing time: 37.0 ms.
[0098] The method, equipment, medium, and product for automatic ship draft identification in dynamic sea conditions based on the relative wave stillness algorithm provided in this application achieve accurate identification of effective observation points by capturing the "quasi-static" features that momentarily approach equilibrium during wave superposition and interference. Secondly, by capturing the "metastable moment" in the dynamic environment, this application can effectively filter out nonlinear interference caused by wave splash, light and shadow fluctuations, and random sea state noise. When waves are violently fluctuating, existing algorithms struggle to accurately distinguish the ship's true floating state from high-frequency noise. This application, however, actively captures the optimal observation moment when the water surface is relatively still during random fluctuations by tracking the continuity of draft values between frames, achieving "selective acquisition" rather than full-time averaging, thus exhibiting better robustness. Furthermore, this application employs a computationally efficient algorithm architecture, reducing computational redundancy through a decoupling framework, ensuring real-time synchronous output of data under high-frequency dynamic sampling, and possessing real-time processing advantages. In addition, this application constructs a coupled processing architecture combining deep learning semantic features and improved statistical filtering, using an interquartile range (IQR) method instead of the traditional Z-score method. Field observation data is often affected by ship movement, changes in light and shadow, and the impact of waves, resulting in a significant skewed distribution or a large number of extreme outliers. The IQR method used in this application exhibits stronger robustness when processing industrial field data containing extreme abrupt changes, providing reliable data support in industrial environments while ensuring real-time performance. Finally, this application employs a highly decoupled modular design, making the lower-level feature perception layer and the higher-level decision smoothing layer completely independent, without relying on a specific visual detection model, and enabling generalized processing of multi-source instantaneous feature data streams. The feature perception layer not only supports convolutional neural network models such as YOLO series, SSD, and PPOCR, but is also compatible with attention mechanism models such as ViT and Swin-Transformer, as well as semantic segmentation networks. It can also establish an adaptive reference system by utilizing the inherent geometric features of the hull. By automatically identifying and connecting the center point of the M scale with the corresponding draft number center point, it calculates the average slope of multiple sets of feature line segments, anchors the ship's physical structure in real time, and automatically compensates for image rotation caused by ship roll, pitch, and camera shooting angle deviations. This ensures the consistency of the reference benchmark under dynamic sea conditions and achieves accurate tilt correction without manual intervention.
[0099] On the other hand, this application also provides an electronic device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform the operations in the dynamic sea state ship draft automatic identification method based on the relative wave static algorithm provided in the first aspect.
[0100] On the other hand, this application also provides a storage medium storing multiple instructions adapted for loading by a processor to execute the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm provided in the first aspect.
[0101] On the other hand, this application also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm provided in the first aspect.
[0102] The method, equipment, medium, and product for automatic ship draft identification based on a relative wave stillness algorithm provided in this application achieve accurate identification of effective observation points by capturing the "quasi-static" characteristics that momentarily approach equilibrium during wave superposition and interference. Secondly, by capturing "metastable moments" in the dynamic environment, this application effectively filters out nonlinear interference caused by wave splash, light and shadow fluctuations, and random sea state noise. When waves are violently fluctuating, existing algorithms struggle to accurately distinguish the ship's true floating or sinking state from high-frequency noise. This application, however, actively captures the optimal observation moment when the water surface is relatively still during random fluctuations by tracking the continuity of draft values between frames, achieving "selective acquisition" rather than full-time averaging, thus exhibiting better robustness. Furthermore, this application employs a computationally efficient algorithm architecture, reducing computational redundancy through a decoupling framework, ensuring real-time synchronous output of data under high-frequency dynamic sampling, and possessing real-time processing advantages. In addition, this application constructs a coupled processing architecture combining deep learning semantic features and improved statistical filtering, using an interquartile range (IQR) method instead of the traditional Z-score method. Field observation data is often affected by ship movement, changes in light and shadow, and the impact of waves, resulting in a significant skewed distribution or a large number of extreme outliers. The IQR method used in this application exhibits stronger robustness when processing industrial field data containing extreme abrupt changes, providing reliable data support in industrial environments while ensuring real-time performance. Finally, this application employs a highly decoupled modular design, making the lower-level feature perception layer and the higher-level decision smoothing layer completely independent, without relying on a specific visual detection model, and enabling generalized processing of multi-source instantaneous feature data streams. The feature perception layer not only supports convolutional neural network models such as YOLO series, SSD, and PPOCR, but is also compatible with attention mechanism models such as ViT and Swin-Transformer, as well as semantic segmentation networks. It can also establish an adaptive reference system by utilizing the inherent geometric features of the hull. By automatically identifying and connecting the center point of the M scale with the corresponding draft number center point, it calculates the average slope of multiple sets of feature line segments, anchors the ship's physical structure in real time, and automatically compensates for image rotation caused by ship roll, pitch, and camera shooting angle deviations. This ensures the consistency of the reference benchmark under dynamic sea conditions and achieves accurate tilt correction without manual intervention.
[0103] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0104] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.
[0105] The foregoing has provided a detailed description of a method, apparatus, device, medium, and product for automatic identification of ship draft in dynamic sea conditions based on a relative wave static algorithm, as provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. 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 this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for automatic identification of ship draft in dynamic sea conditions based on a relative wave static algorithm, used to automatically identify the ship's draft value, characterized in that, include: Obtain a time-series image sequence containing the ship's draft indicator region; Each frame of the time-series image sequence is processed to extract the draft marker features in each frame of the image, and the instantaneous draft value corresponding to the image corresponding to the draft marker features is calculated based on the draft marker features, forming an instantaneous draft value sequence. The instantaneous draft sequence is processed using a relative wave static algorithm to obtain the final draft value. The calculation of the final draft includes the following steps: A preset threshold for determining the stability of the water surface dwell time; Traverse the instantaneous draft value sequence, and determine whether the difference between the instantaneous draft values of adjacent frames in the instantaneous draft value sequence meets the stability determination threshold based on the stability determination threshold. If yes, the adjacent frames are determined to be continuous stable frames, and the continuous stable frame count is incremented by 1; otherwise, the continuous stable frame count is reset to 1. Select the largest consecutive stable frame count from all the aforementioned consecutive stable frame counts; The instantaneous draft of the frame corresponding to the maximum continuous stable frame count is taken as the final draft.
2. The method for automatic identification of ship draft in dynamic sea conditions based on a relative wave stationary algorithm according to claim 1, characterized in that, The method for calculating the instantaneous draft includes: Extract at least two feature point pairs from the draft scale line in each frame of the time-series image sequence, wherein each feature point pair includes a first feature point and a second feature point located on the same draft scale line; A dynamic leveling reference coordinate system is constructed based on at least two of the feature point pairs to perform tilt correction on the image; Multiple contact points between the ship's hull and the water surface in the image are obtained as waterline feature points; Based on the tilt-corrected image and the waterline feature points, the waterline position is determined, and the instantaneous draft value is calculated by combining the draft scale lines.
3. The method for automatic identification of ship draft in dynamic sea conditions based on a relative wave stationary algorithm according to claim 2, characterized in that, Constructing the dynamic leveling reference coordinate system includes: Based on at least two of the aforementioned feature point pairs, the equation of the reference baseline is determined through fitting. The fitting method used is the least squares method.
4. The method for automatic identification of ship draft in dynamic sea conditions based on a relative wave stationary algorithm according to claim 2, characterized in that, Determining the waterline position includes: Linear fitting is performed on the multiple waterline feature points to obtain the waterline equation.
5. The method for automatic identification of ship draft in dynamic sea conditions based on a relative wave stationary algorithm according to claim 2, characterized in that, The first feature point is the center of the ship's draft scale, and the second feature point is the center of the number corresponding to the center of the draft scale.
6. The method for automatic identification of ship draft in dynamic sea conditions based on a relative wave stationary algorithm according to claim 1, characterized in that, The stability determination threshold is dynamically adjusted based on real-time acquired environmental parameters, including sea state level and wind speed.
7. The method for automatic identification of ship draft in dynamic sea conditions based on a relative wave stationary algorithm according to claim 1, characterized in that, Also includes: Obtain the instantaneous draft value of the frame corresponding to the maximum continuous stable frame count, remove outliers, and use the instantaneous draft value without outliers as the final draft value.
8. An electronic device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor runs the computer program in the memory to perform the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm according to any one of claims 1 to 7.
9. A storage medium, characterized in that, The storage medium stores multiple instructions, which are adapted for loading by a processor to execute the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm as described in any one of claims 1 to 7.
10. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a processor, implement the steps in the dynamic sea state ship draft automatic identification method based on the relative wave stillness algorithm as described in any one of claims 1 to 7.