Wave condition prediction system based on sea wave image remote sensing and artificial intelligence

Abstract

The invention relates to the technical field of ocean engineering, in particular to a wave condition prediction system based on sea wave image remote sensing and artificial intelligence, which comprises an image remote sensing unit, a data processing unit, a condition prediction unit, an analysis and inspection unit and a verification and adjustment unit, the image remote sensing unit performs preprocessing grid division on the sea wave image to generate a grid data set; the data processing unit comprises a data optimization module, a data segmentation module and a data analysis module and is used for evaluating and optimizing the grid data set; the condition prediction unit predicts the significant wave height of the sea waves, and the analysis and inspection unit performs mean square error inspection on prediction data and analyzes grid change conditions or adjusts prediction parameters in the condition prediction unit; and the verification adjustment unit judges the accuracy of prediction and analysis according to the prediction difference value. According to the method, the original data is optimized, and the prediction result is verified and adjusted, so that the prediction precision of the significant wave height of the sea waves is improved, and prediction errors are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of ocean engineering, and particularly to a wave condition prediction system based on remote sensing of sea wave images and artificial intelligence. Background Art

[0002] Wave forecasting plays an indispensable role in maintaining the safety of maritime activities and protecting the marine environment. For example, in terms of maritime safety, wave forecasting can help fishermen avoid bad sea conditions and ensure the safety of maritime operations; for coastal tourism, wave forecasting can warn tourists in advance and avoid accidents in dangerous sea areas; in the field of ocean energy development, it provides important references for the design and maintenance of offshore wind power and tidal energy facilities.

[0003] Chinese Patent Publication No.: CN107545250A discloses a real-time forecasting system for the movement of ocean floating bodies based on remote sensing of sea wave images and artificial intelligence; firstly, big data of the marine environment is obtained by a marine environment sensor module, and at the same time, big data of the dynamic response of the floating body is obtained by a floating body dynamic response sensor module, then the two types of data are transmitted to data storage and stored in a database, and then an artificial intelligence learning module is trained by using the data stored in the data storage to establish an artificial intelligence forecasting model that reflects the movement of the floating body based on the input of ocean images. Finally, based on the trained artificial intelligence forecasting model, and based on the big data of the marine environment and the big data of the dynamic response of the floating body, a real-time forecasting module for the movement response of the floating body is established, and the results are all transmitted to an auxiliary decision-making module and a computer. It can be seen that the real-time forecasting system for the movement of ocean floating bodies based on remote sensing of sea wave images and artificial intelligence has the following problems:

[0004] There are abnormal situations in the acquired data set due to uncollected, missing or irregular changes, and since the proportion of extreme situations in the training data set is very low, the training samples are unbalanced, resulting in extreme forecasting defects, and the intelligent forecasting model generally has a small bias for extreme value forecasting. Summary of the Invention

[0005] Therefore, the present invention provides a wave condition prediction system based on remote sensing of sea wave images and artificial intelligence to overcome the problems that the acquired data set in the prior art has abnormal situations and the intelligent forecasting model generally has a small bias for extreme value forecasting.

[0006] To achieve the above object, the present invention provides a wave condition prediction system based on remote sensing of sea wave images and artificial intelligence, including:

[0007] An image remote sensing unit, which preprocesses the original sea wave images acquired in the same area according to the initial sampling period to generate corrected sea wave images, divides the corrected sea wave images into grids according to the initial length to obtain grid sea wave parameters, and generates a grid data set for the grid sea wave parameters;

[0008] A data processing unit, which includes a data optimization module, a data segmentation module, and a data analysis module, for evaluating and optimizing the generated grid data set;

[0009] The data optimization module makes a conditional determination on the grid sea wave parameters corresponding to the image grid, judges the data state of the significant wave height in the grid data set according to the conditional determination result, the data state includes a first data state and a second data state, and takes corresponding optimization measures according to different determined conditions;

[0010] The data segmentation module, which is connected to the data optimization module, divides data windows for the missing data segments and extreme value data, removes the data contained in the abnormal windows from the grid data set, and reorders the grid data set in time;

[0011] The data analysis module, which is connected to the conditional prediction unit, analyzes the autocorrelation of the data, and tests the processing result of the grid data set according to the autocorrelation;

[0012] The conditional prediction unit is used to predict the significant wave height of the sea wave according to the test result of the data analysis module, according to the input grid data according to the initial number of steps and the step duration;

[0013] The analysis and test unit is used to perform a mean square error test on the predicted data, judge the prediction state of the predicted data, analyze the grid change situation according to the prediction state, or adjust the prediction parameters in the conditional prediction unit according to the specific error situation;

[0014] The verification and adjustment unit is used to judge the accuracy of the significant wave height prediction and the grid change situation analysis according to the prediction difference.

[0015] Furthermore, the image remote sensing unit generates an image training set according to the initial number of original sea wave images, and builds a sea wave image detection model;

[0016] Take the initial number of original sea wave images as the image test set, input them into the sea wave image detection model for detection according to the initial detection period, identify, crop and correct the sea waves in the original sea wave images to obtain corrected sea wave images;

[0017] Divide the corrected sea wave images into several image grids according to the initial length, and correspondingly obtain the grid sea wave parameters corresponding to the several image grids to generate a grid data set.

[0018] Furthermore, when the data in the grid data set meets the first determination condition, the data optimization module determines that the significant wave height is in the first data state, eliminates the data with abnormal sampling intervals, and correspondingly adjusts the sampling period of the data in the grid data set;

[0019] When the data in the grid dataset meets the second determination condition or the third determination condition, the data optimization module determines that the significant wave height is in the second data state and transmits the data to the data segmentation module.

[0020] Further, the first determination condition is that the data optimization module determines whether there is an abnormality in the sampling interval according to the data sampling period of the data in the grid dataset. If there is an abnormality in the sampling interval, the data optimization module determines that the data meets the first determination condition;

[0021] The second determination condition is that the data optimization module calculates the data proportion according to the number of data with the maximum value in each column of the grid dataset, and determines whether any corresponding numerical column is a missing data segment according to the data proportion. If any corresponding numerical column is a missing data segment, the data meets the second determination condition;

[0022] The third determination condition is that the data optimization module determines whether the data is an extreme value according to the difference between the data and the data average value. If the data is an extreme value, the data meets the third determination condition.

[0023] Further, the data segmentation module divides abnormal windows for missing data and extreme value data, removes the data included in the abnormal windows from the dataset. After removing the data, it reconstructs the column data object according to the time point information, and re-orders the dataset in time using the insertion sort method based on time.

[0024] Further, the data analysis module analyzes the autocorrelation of the data, calculates the autocovariance. If the first-order autocorrelation coefficient is close to the standard coefficient and gradually converges to zero as the lag order increases, the data analysis module determines that the significant wave height time series of the ocean waves has autocorrelation and inputs the grid dataset into the conditional prediction unit.

[0025] Further, the analysis and inspection unit conducts a mean square error test on the predicted data, calculates the mean square error between the actual value corresponding to the predicted data and the predicted value of the predicted data.

[0026] If any mean square error is less than or equal to the critical error, the analysis and inspection unit determines that the predicted data is in the first prediction state and analyzes the grid change situation.

[0027] If there is any mean square error greater than the critical error, the analysis and inspection unit determines that the predicted data is in the second prediction state and adjusts the prediction parameters in the conditional prediction unit.

[0028] Further, when the predicted data is in the first prediction state, the analysis and inspection unit compares the predicted grid significant wave height with the critical wave height to determine the wave conditions corresponding to the grid where the predicted grid significant wave height is located. When the grid is in the second wave condition, it outputs a warning signal including the warning duration.

[0029] The critical wave height is negatively correlated with the mean square error.

[0030] Further, when the prediction data is in the second prediction state, the analysis and inspection unit adjusts the step duration of the initial number of steps of the conditional prediction unit according to any step prediction corresponding to the mean square error greater than the critical error.

[0031] Further, the verification and adjustment module calculates the absolute value of the prediction difference between the predicted grid significant wave height and the actual significant wave height corresponding to the grid, compares it with the difference evaluation value, determines whether the accuracy of the analysis of the predicted significant wave height and the grid change situation can meet the prediction analysis requirements, and adjusts the initial length and initial detection period of the divided image grid and the initial number of acquired original sea wave images.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows.

[0033] Further, the significant wave height of the sea wave recorded at half-hour and one-hour sampling periods is an important reference for the size of the sea wave during this period. The sea wave image detection model of this system preprocesses the image with the model, making the sea waves contained in the image in a standard direction, and reducing the calculation amount and difficulty of obtaining sea wave parameters according to the sea wave image by homogenizing the image, increasing the accuracy of obtaining the significant wave height of the sea wave. By dividing the image into grids, the sea waves contained in the image are distinguished, laying an analysis foundation for subsequent analysis of the sea wave change situation.

[0034] Further, there are uncollected or missing situations in the acquired original sea wave images due to weather or unknown reasons, and the collection period also changes irregularly due to errors. Correspondingly, there are also abnormal situations in the data in the grid data set generated according to the grid sea wave parameters obtained from the sea wave images. At the same time, because the proportion of extreme situations in the training data set is very low, the imbalance of data extremes in the training samples will lead to extreme prediction defects. Training the intelligent prediction model according to the extremes in the training set will result in underestimated extreme predictions. The present invention analyzes and judges the data, takes corresponding optimization measures according to different judged conditions, judges whether it is a missing data segment according to the number of data with a value of 99.0 in each column, judges whether the data is an extreme value, judges the balance of the data, obtains the sampling period of the data, judges whether the sampling interval is abnormal, adjusts the sampling period, evaluates the state of the data in the grid data set through condition determination, provides an optimization basis for subsequent data optimization, improves the accuracy of subsequent optimization links, and increases the adaptability and flexibility of obtaining data for sea wave images.

[0035] Furthermore, the prediction experiment of the significant wave height of ocean waves requires a continuous sequence of significant wave heights of ocean waves of a certain length as input, and the significant wave height values at intervals of the prediction step length after the input sequence as labels; considering that the proportion of missing data is very small and evenly distributed, the existing missing data and extreme value data are removed from the data set, avoiding the adverse effects of abnormal data on the prediction of wave conditions, and increasing the accuracy and flexibility of predicting wave parameters.

[0036] Furthermore, autocorrelation refers to the degree of correlation of the time series of the same variable at different times. If the autocorrelation of the variable is zero, it means that the variable has no correlation at different time points, and using the historical information corresponding to the variable to predict future information is of no value and will have an adverse impact on the accuracy of the prediction result. Therefore, before performing statistical analysis on the significant wave height of ocean waves, this system conducts a correlation analysis to test the processing results of the grid data set, reducing the risk of adverse effects on the accuracy of the prediction result.

[0037] Furthermore, this system verifies whether the predicted data output is accurate by calculating the mean square error between the predicted significant wave height of ocean waves and the actual significant wave height of ocean waves. If the predicted data output is accurate, it analyzes the grid change situation based on the predicted data. If the predicted data output is inaccurate, it calls back the prediction parameters of the conditional prediction unit, increasing the accuracy of predicting wave parameters; at the same time, by analyzing the significant wave height of the grid, it determines the specific location where the set wave determination condition occurs, providing a basis for calculating early warning information for subsequent warning signals.

[0038] Furthermore, this system judges the accuracy of the predicted significant wave height and the analysis of the grid change situation through the predicted difference between the grid significant wave height and the actual significant wave height, and correspondingly adjusts the initial length and initial detection period of dividing the image grid and the initial number of obtaining the original ocean wave images. By increasing the initial number, it increases the computing ability of the image remote sensing unit to obtain the grid wave parameters corresponding to several image grids. By reducing the initial length of dividing the image grid, it reduces the grid unit size for analyzing the grid change situation, increasing the accuracy of the analysis. Description of the Drawings

[0039] Figure 1 It is a unit connection diagram of the wave condition prediction system based on ocean wave image remote sensing and artificial intelligence in the embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the wave crest and wave trough of ocean waves in the embodiment of the present invention;

[0041] Figure 3 It is a flowchart of the analysis and inspection unit judging the state of predicted data according to the mean square error in the embodiment of the present invention;

[0042] Figure 4 This is a flowchart for verifying the accuracy of prediction by the adjustment module according to the absolute value of the prediction difference in the embodiments of the present invention. Detailed implementation manners

[0043] In order to make the objectives and advantages of the present invention more clear, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] The preferred implementation manners of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these implementation manners are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0045] It should be noted that in the description of the present invention, the terms indicating directions or positional relationships such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0046] In addition, it should also be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0047] Please refer to Figures 1-4 as shown in Figure 1 This is a unit connection diagram of a wave condition prediction system based on sea wave image remote sensing and artificial intelligence in the embodiments of the present invention; Figure 2 This is a schematic diagram of sea wave crests and troughs in the embodiments of the present invention; Figure 3 This is a flowchart for the analysis and verification unit to judge the state of prediction data according to the mean square error in the embodiments of the present invention; Figure 4 This is a flowchart for verifying the accuracy of prediction by the adjustment module according to the absolute value of the prediction difference in the embodiments of the present invention.

[0048] The present invention provides a wave condition prediction system based on sea wave image remote sensing and artificial intelligence, including:

[0049] An image remote sensing unit that preprocesses the original ocean wave images obtained for the same area according to the initial sampling period to generate corrected ocean wave images, divides the corrected ocean wave images into grids according to the initial length, obtains grid ocean wave parameters, and generates a grid data set for the grid ocean wave parameters;

[0050] A data processing unit, which includes a data optimization module, a data segmentation module, and a data analysis module, and is used to evaluate and optimize the generated grid data set;

[0051] The data optimization module makes a conditional determination for the grid ocean wave parameters corresponding to the image grids, judges the data state of the significant wave height in the grid data set according to the conditional determination result, the data state includes a first data state and a second data state, and takes corresponding optimization measures according to different determined conditions;

[0052] The data segmentation module, which is connected to the data optimization module, divides data windows for missing data segments and extreme value data, removes the data included in the abnormal windows from the grid data set, and reorders the grid data set in time;

[0053] The data analysis module, which is connected to the conditional prediction unit, analyzes the autocorrelation of the data, and tests the processing result of the grid data set according to the autocorrelation;

[0054] The conditional prediction unit is used to predict the significant wave height of the ocean waves according to the test result of the data analysis module, with the input grid data according to the initial number of steps and step duration;

[0055] The analysis and test unit is used to perform a mean square error test on the predicted data, judge the prediction state of the predicted data, analyze the grid change situation according to the prediction state, or adjust the prediction parameters in the conditional prediction unit according to the specific error situation;

[0056] The verification and adjustment unit is used to judge the accuracy of the significant wave height prediction and the grid change situation analysis according to the prediction difference.

[0057] In this embodiment, the remotely sensed ocean wave images include, but are not limited to, unmanned aerial vehicle images and satellite images, and the ocean wave images can be obtained from any channel.

[0058] The image remote sensing unit performs filtering processing on the original ocean wave images obtained for the same area according to the initial sampling period, corrects the images to the standard coordinate system to generate corrected ocean wave images, and obtains a number of ocean wave parameters according to the corrected ocean wave images. The ocean wave parameters include the instantaneous height, wavelength, wave period, and propagation direction of the ocean waves;

[0059] The process of the image remote sensing unit performing image correction on the original ocean wave images includes,

[0060] The image remote sensing unit generates an image training set based on an initial number of original ocean wave images, uses the image training set to train the YOLO v8 object detection algorithm, and constructs an ocean wave image detection model;

[0061] Take the initial number of original ocean wave images as an image test set, input them into the ocean wave image detection model for detection according to the initial detection period, identify, crop, and correct the ocean waves in the original ocean wave images to obtain corrected ocean wave images.

[0062] Divide the corrected ocean wave images into several image grids according to the initial length. The several image grids include the first image grid, the second image grid, the third image grid,..., the nth image grid;

[0063] Correspondingly obtain the grid ocean wave parameters corresponding to the several image grids. The grid ocean wave parameters include the instantaneous height, wavelength, and wave period of the ocean waves within the grid;

[0064] The image remote sensing unit calculates the significant wave height based on the obtained instantaneous height and generates a grid data set for the calculated significant wave height.

[0065] The grid data set is in TXT format. The data set contains uncollected data and missing value data, where the uncollected data and missing value data are represented by 99.0, indicating that the data for the corresponding moment does not exist.

[0066] Specifically, the ocean wave height is obtained by calculating the difference between adjacent wave crests and wave troughs of continuous ocean waves. Sort the ocean wave heights collected within a period of time from largest to smallest, and take the average of the data ranked in the top one-third to obtain the significant wave height of the ocean waves within that period.

[0067] Specifically, the significant wave height of the ocean waves recorded with a sampling period of half an hour and one hour is an important reference for the size of the ocean waves within that period. The ocean wave image detection model built in this system preprocesses the images so that the ocean waves contained in the images are in a standard direction, and reduces the calculation amount and difficulty of obtaining ocean wave parameters based on the ocean wave images through homogenization processing of the images, increases the accuracy of obtaining the significant wave height of the ocean waves, and differentiates the ocean waves contained in the images by dividing the images into grids, laying an analysis foundation for subsequent analysis of the changes in ocean waves.

[0068] When the data in the grid data set meets the first determination condition, the data optimization module determines that the significant wave height is in the first data state, eliminates the data with abnormal sampling intervals, and correspondingly adjusts the sampling period of the data in the grid data set;

[0069] When the data in the grid data set meets the second determination condition, or meets the third determination condition, the data optimization module determines that the significant wave height is in the second data state and transmits the data to the data segmentation module;

[0070] The first determination condition is as follows: Obtain the data sampling period of the data in the grid dataset. If the data sampling period is greater than the standard period, the data optimization module determines that the sampling interval is abnormal, and the data meets the first determination condition;

[0071] If the data sampling period is less than or equal to the standard period, the data optimization module determines that the sampling interval is not abnormal, and the data does not meet the first determination condition.

[0072] The second determination condition is as follows: Obtain the number of data with the maximum value in each column of the grid dataset, calculate the data proportion based on the number of data. If the data proportion is greater than the critical proportion, the data optimization module determines that any corresponding numerical column is a missing data segment, and the data meets the second determination condition; if the number of data is less than or equal to the critical value, the data optimization module determines that any corresponding numerical column is not a missing data segment, and the data does not meet the second determination condition.

[0073] The third determination condition is as follows: Exclude the maximum value and calculate the data average value. If the difference between the data and the data average value is greater than the data evaluation value, the data optimization module determines that the data is an extreme value, and the data meets the third determination condition; if the difference between the value and the data average value is greater than the data evaluation value, the data optimization module determines that the data is not an extreme value, and the data does not meet the third determination condition.

[0074] Among them, the standard period is a preset value set according to the initial sampling period, the critical proportion is 90%, the data evaluation value is a preset value set according to the historical data of the significant wave height of the sea wave, and the maximum value is 99.0.

[0075] Specifically, this system uses the Pandas data processing library in Python to read txt data and split columns according to line breaks and spaces.

[0076] Specifically, there are cases where the acquired original ocean wave images are not collected or missing due to weather or unknown reasons, and the acquisition period also varies irregularly due to errors. Correspondingly, there are also anomalies in the data in the grid dataset generated based on the grid ocean wave parameters obtained from the ocean wave images. At the same time, since the proportion of extreme cases in the training dataset is very low, the imbalance of data extremes in the training samples will lead to extreme prediction defects, and training the intelligent prediction model based on the extremes in the training set will result in underestimated extreme predictions. The present invention analyzes and judges the data, takes corresponding optimization measures according to different judged conditions, judges whether it is a missing data segment based on the number of data with a value of 99.0 in each column, judges whether the data is an extreme value, judges the balance of the data, obtains the sampling period of the data, judges whether the sampling interval is abnormal, adjusts the sampling period, evaluates the state of the data in the grid dataset through condition determination, provides an optimization basis for subsequent data optimization, improves the accuracy of subsequent optimization links, and increases the adaptability and flexibility of obtaining data for ocean wave images.

[0077] The data segmentation module divides abnormal windows for missing data and extreme value data, removes the data included in the abnormal windows from the dataset. After removing the data, it reconstructs the column data object according to the time point information, and re - sorts the dataset in time using the insertion sort method based on time. Finally, a csv file is generated by Pandas.

[0078] Specifically, the prediction experiment of the significant wave height of ocean waves requires a continuous sequence of significant wave heights of a certain length as input, and the significant wave height values at intervals of the prediction step length after the input sequence as labels. Considering that the proportion of missing data is very small and evenly distributed, the existing missing data and extreme value data are removed from the dataset, avoiding the adverse effects of abnormal data on predicting wave conditions, and increasing the accuracy and flexibility of predicting ocean wave parameters.

[0079] The data analysis module analyzes the autocorrelation of the data and calculates the autocovariance.

[0080] Let \(X_t\) be a time - variable that takes values over time \(t\), then its \(k\) - th autocovariance is:

[0081] AF k =E[(X t -μ t )(X t-k -μ t-k )]=Cov(X t -X t-k )

[0082] In the formula, \(E\) represents the mathematical expectation, and \(\mu\) represents the mean.

[0083] The \(k\) - th autocorrelation coefficient of a random variable is:

[0084]

[0085] Among them, Var represents the variance of the variable.

[0086] If the first-order autocorrelation coefficient is close to the standard coefficient and gradually converges to zero as the lag order increases, the data analysis module determines that the time series of the significant wave height of the sea wave has autocorrelation and inputs the grid data set into the conditional prediction unit;

[0087] If the first-order autocorrelation coefficient is not close to the standard coefficient and does not gradually converge to zero as the lag order increases, the data analysis module determines that the time series of the significant wave height of the sea wave does not have autocorrelation;

[0088] Among them, the standard coefficient is 1.

[0089] Specifically, autocorrelation refers to the degree of correlation of the time series of the same variable at different times. If the autocorrelation of the variable is zero, it means that the variable does not have correlation at different time points, and using the historical information corresponding to the variable to predict future information has no value and will have an adverse impact on the accuracy of the prediction result. Therefore, before performing statistical analysis on the significant wave height of the sea wave, this system conducts a correlation analysis to test the processing result of the grid data set, reducing the risk of having an adverse impact on the accuracy of the prediction result.

[0090] When the time series of the significant wave height of the sea wave has autocorrelation, the grid data set is input into the conditional prediction unit, and the conditional prediction unit predicts the significant wave height of the sea wave according to the input grid data;

[0091] In this embodiment, the conditional prediction unit uses a convolutional residual neural network model to make predictions according to the initial number of steps and the step duration. The predictions are divided into 1-step prediction, 2-step prediction, 3-step prediction,..., 24-step prediction, and the step duration is 30 minutes.

[0092] The analysis and inspection unit conducts a mean square error test on the predicted data, calculates the mean square error between the actual value corresponding to the predicted data and the predicted value of the predicted data,

[0093] The calculation formula for the mean square error MSE between the predicted significant wave height of the sea wave and the actual significant wave height of the sea wave is as follows,

[0094]

[0095] Among them, yi is the actual value of the i-th data to be predicted, yi' is the predicted value of the i-th data to be predicted, and m is the total number of predicted data.

[0096] If any mean square error is less than or equal to the critical error, the analysis and inspection unit determines that the predicted data is in the first prediction state and analyzes the grid change situation;

[0097] If there is any mean square error greater than the critical error, the analysis and inspection unit determines that the predicted data is in the second prediction state and adjusts the prediction parameters in the conditional prediction unit;

[0098] Wherein, the critical error is a preset value set according to the mean square error of the historical prediction data that has passed the accuracy test.

[0099] When the predicted data is in the first prediction state, the analysis and inspection unit compares the predicted grid significant wave height with the critical wave height.

[0100] If any grid significant wave height is less than or equal to the critical wave height, the analysis and inspection unit determines that the grid corresponding to the predicted grid significant wave height is in the first wave condition and does not need to output a warning signal;

[0101] If any grid significant wave height is greater than the critical wave height, the analysis and inspection unit determines that the grid corresponding to the predicted grid significant wave height is in the second wave condition, calculates the wave speed by combining the wavelength and wave period, and predicts and outputs a warning signal including the warning duration according to the wave speed;

[0102] Wherein, the critical wave height can be set according to the application scenario of the prediction system, with the unit of m, and the critical wave height is negatively correlated with the mean square error.

[0103] The analysis and inspection unit adjusts the parameters of the conditional prediction unit according to the specific situation of the error.

[0104] When the predicted data is in the second prediction state, if the mean square error corresponding to the prediction result of any step of prediction is greater than the critical error, the analysis and inspection unit determines that the prediction result after any step of prediction is an invalid prediction.

[0105] Reduce the initial number of steps for the conditional prediction unit to predict the sea wave parameters according to the corresponding number of steps, and at the same time reduce the step duration according to the ratio of the reduced number of steps to the initial number of steps.

[0106] Specifically, the system verifies whether the output predicted data is accurate by calculating the mean square error between the predicted significant wave height of the sea wave and the actual significant wave height of the sea wave. If the output predicted data is accurate, it analyzes the grid change situation according to the predicted data. If the output predicted data is inaccurate, it calls back the prediction parameters of the conditional prediction unit, increasing the accuracy of predicting the sea wave parameters; at the same time, by analyzing the significant wave height of the grid, it determines the specific position where the set wave determination condition occurs, providing a basis for calculating the warning information for the subsequent warning signal.

[0107] The verification and adjustment module calculates the absolute value of the prediction difference between the predicted grid significant wave height and the actual significant wave height corresponding to the grid.

[0108] If the absolute value of the predicted difference is less than or equal to the difference evaluation value, the verification and adjustment module determines that the accuracy of the predicted significant wave height and the analysis of the grid change situation can meet the requirements of the prediction analysis, and there is no need to adjust the parameters;

[0109] If the absolute value of the predicted difference is greater than the difference evaluation value, the verification and adjustment module determines that the accuracy of the predicted significant wave height and the analysis of the grid change situation cannot meet the requirements of the prediction analysis, and adjusts the initial length of the divided image grid, the initial detection period, and the initial number of original sea wave images obtained;

[0110] Specifically, the initial number of original sea wave images obtained is increased according to the ratio of the absolute value of the predicted difference to the difference evaluation value, and the initial length of the divided image grid, the initial detection period, and the initial sampling period of obtaining the original sea wave images for the same area are decreased according to the ratio of the difference evaluation value to the absolute value of the predicted difference.

[0111] Specifically, the system determines the accuracy of the predicted significant wave height and the analysis of the grid change situation through the predicted difference between the predicted grid significant wave height and the actual significant wave height, and correspondingly adjusts the initial length of the divided image grid, the initial detection period, and the initial number of original sea wave images obtained. By increasing the initial number, the calculation ability of the image remote sensing unit to obtain the grid sea wave parameters corresponding to several image grids is increased, and by decreasing the initial length of the divided image grid, the grid unit size for analyzing the grid change situation is reduced, increasing the accuracy of the analysis.

[0112] So far, the technical solution of the present invention has been described in combination with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0113] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A wave condition prediction system based on ocean wave image remote sensing and artificial intelligence, characterized in that: include: An image remote sensing unit, which pre-processes the original wave image obtained in the same area according to the initial sampling period to generate a corrected wave image, divides the corrected wave image into grids according to the initial length, obtains grid wave parameters, and generates a grid data set for the grid wave parameters; A data processing unit, which includes a data optimization module, a data segmentation module and a data analysis module, is used to evaluate and optimize the generated grid data set; The data optimization module performs conditional judgment on the grid wave parameters corresponding to the image grid, and judges the data state of the significant wave height in the grid data set according to the conditional judgment result. The data state includes the first data state and the second data state, and takes corresponding optimization measures according to different judgment conditions; The data segmentation module is connected to the data optimization module, which divides the data window according to the missing data segment and the extreme value data, removes the data contained in the abnormal window from the grid data set, and reorders the grid data set in time; A data analysis module, which is connected to the conditional prediction unit, analyzes the autocorrelation of the data and verifies the processing results of the grid data set based on the autocorrelation; The condition prediction unit is used to predict the effective wave height of the sea wave according to the test results of the data analysis module and the initial step number and step time through the input grid data; An analysis and inspection unit is used to perform a mean square error inspection on the predicted data, determine the prediction state of the predicted data, analyze the grid change according to the prediction state, or adjust the prediction parameters in the conditional prediction unit according to the specific error situation; The verification and adjustment unit is used to determine the accuracy of the significant wave height prediction and grid change analysis based on the prediction difference.

2. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 1, characterized in that: The image remote sensing unit generates an image training set based on an initial number of original wave images and builds a wave image detection model; An initial number of original ocean wave images are used as image test sets, and are input into an ocean wave image detection model for detection according to an initial detection cycle, and the ocean waves in the original ocean wave images are identified, cropped, and corrected to obtain corrected ocean wave images; The corrected wave image is divided into several image grids according to the initial length, and the grid wave parameters corresponding to the several image grids are obtained accordingly to generate a grid data set.

3. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 2 is characterized in that: When the data in the grid data set meets the first determination condition, the data optimization module determines that the effective wave height is in the first data state, removes the data with abnormal sampling interval, and adjusts the sampling period of the data in the grid data set accordingly; When the data of the grid data set meets the second determination condition or the third determination condition, the data optimization module determines that the effective wave height is in the second data state and transmits the data to the data segmentation module.

4. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 3 is characterized in that: The first determination condition is that the data optimization module determines whether there is an abnormality in the sampling interval according to the data sampling period of the data in the grid data set, and if there is an abnormality in the sampling interval, the data optimization module determines that the data meets the first determination condition; The second judgment condition is that the data optimization module calculates the data proportion according to the number of data with the maximum value in each column of the grid data set, and determines whether any corresponding value column is a missing data segment according to the data proportion. If any corresponding value column is a missing data segment, the data meets the second judgment condition; The third judgment condition is that the data optimization module determines whether the data is an extreme value based on the difference between the data and the data average value. If the data is an extreme value, the data meets the third judgment condition.

5. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 1, characterized in that: The data segmentation module divides abnormal windows according to missing data and extreme value data, removes the data contained in the abnormal windows from the data set, and after removing the data, rebuilds the column data object according to the time point information, and reorders the data set in time based on time by using the insertion sort method.

6. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 1, characterized in that: The data analysis module analyzes the autocorrelation of the data and calculates the autocovariance. If the first-order autocorrelation coefficient is close to the standard coefficient and gradually converges to zero with the increase of the lag order, the data analysis module determines that the time series of the significant wave height of the ocean wave has autocorrelation and inputs the grid data set into the conditional prediction unit.

7. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 1, characterized in that: The analysis and inspection unit performs a mean square error inspection on the predicted data, and calculates the mean square error between the actual value corresponding to the predicted data and the predicted value of the predicted data. If any mean square error is less than or equal to the critical error, the analysis and verification unit determines that the predicted data is in the first prediction state and analyzes the grid change situation; If any mean square error exists that is greater than the critical error, the analysis and verification unit determines that the prediction data is in the second prediction state and adjusts the prediction parameters in the conditional prediction unit.

8. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 7, characterized in that: When the forecast data is in the first forecast state, the analysis and verification unit compares the predicted grid effective wave height with the critical wave height to determine the wave condition of the grid corresponding to the predicted grid effective wave height, and outputs a warning signal including the warning duration when the grid is in the second wave condition; The critical wave height is negatively correlated with the mean square error.

9. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 7, characterized in that: When the prediction data is in the second prediction state, the analysis and verification unit adjusts the step duration of the initial step number of the condition prediction unit according to any step prediction corresponding to any mean square error greater than the critical error.

10. The wave condition prediction system based on ocean wave image remote sensing and artificial intelligence according to claim 1, characterized in that: The verification and adjustment module calculates the absolute value of the predicted difference between the predicted grid effective wave height and the actual effective wave height corresponding to the grid, compares it with the difference evaluation value, determines whether the accuracy of the predicted effective wave height and grid change analysis can meet the prediction and analysis requirements, adjusts the initial length and initial detection period of the divided image grid, and obtains the initial number of original wave images.

Images