Radar and ship swaying information fusion-based significant wave height inversion technology

Abstract

The invention discloses a technology for inverting significant wave height based on fusion of radar and ship swaying information. The technology comprises the following steps: step 1, constructing a wave measurement model; 2, determining a ship type to be simulated; step 3, determining the wave direction of the sea wave to be simulated; 4, determining a to-be-simulated working condition; step 5, working condition simulation, including time calendar data simulation and radar image simulation based on ship motion response; step 6, data extraction, including swaying time calendar data meaningful value feature extraction and signal-to-noise ratio feature extraction; step 7, carrying out data normalization; 8, training a weight coefficient vector; and step 9, performing inversion. According to the method, sea wave feature information obtained based on ship swaying motion is fused into traditional non-coherent radar wave measurement, and a multi-sensor information fusion wave measurement model is established; and the sea wave feature information from the oscillation motion of the ship and the signal-to-noise ratio feature of the radar image are combined to calculate the more accurate significant wave height. The measurement precision is higher, the error is smaller, and the actual situation of sea waves can be reflected more accurately.

Description

Technical Field

[0001] The invention relates to the field of ocean monitoring, and in particular to a technology for inverting effective wave height based on fusion of radar and ship swaying information. Background Art

[0002] Since any shipborne decision support system needs to take the wave height and other information of the relevant waves encountered as system input to make the system accurate and reliable, it is particularly important to estimate the sea conditions around the ship.

[0003] Marine operations, such as cargo replenishment, helicopter landings, and optimal maneuvering of surface ships, have high requirements for the prediction accuracy of sea waves. They are carried out in a specific marine environment to ensure safe and effective completion. According to the requirements of the sea wave parameter measurement performance verification test outline of radar wave measuring equipment, the range of sea wave effective wave height (main sea wave parameters) is usually 1.3 to 6 meters. When the sea wave effective wave height is 1.3 to 3.5 meters, the measurement accuracy must be ≤0.5 meters; when the sea wave effective wave height is 3.5 to 6 meters, the measurement accuracy must be ≤15%.

[0004] At present, the wave height is measured by buoys and ship-borne radars, etc. Among them, buoys are generally used for fixed-point measurement and cannot measure the waves encountered in real time and continuously during the voyage of the ship.

[0005] Shipborne radars, such as non-coherent X-band navigation radars that travel with ships, are often used for ship navigation. In recent years, sea clutter in navigation radar images has been widely used for remote sensing of sea waves. When non-coherent radars are used for sea surface remote sensing, they are usually calibrated using a wave spectrum based on a signal-to-noise ratio. However, since the wave spectrum analysis method based on a signal-to-noise ratio uses traditional three-dimensional Fourier transform to obtain a relative spectrum, the extracted signal-to-noise ratio requires external dedicated equipment to calibrate the effective wave height. Due to insufficient calibration accuracy and large errors, it cannot meet actual needs.

[0006] Since non-coherent X-band radar is the most widely equipped navigation radar at present, the improvement of its wave measurement accuracy is of great significance to the improvement of ship's operational capability. Summary of the invention

[0007] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned prior art and to provide a technology for inverting effective wave height based on the fusion of radar and ship swaying information. The technology for inverting effective wave height based on the fusion of radar and ship swaying information integrates the characteristic information obtained from the ship swaying motion into the traditional non-coherent radar wave measurement process, and proposes an effective wave height inversion method that combines the ship swaying characteristics with the signal-to-noise ratio, thereby improving the wave height measurement accuracy.

[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0009] A technology for inverting effective wave height based on the fusion of radar and ship sway information includes the following steps.

[0010] Step 1: Construct a wave measurement model, specifically:

[0011] Ax=b (1)

[0012] Among them, A is the observation matrix, x is the weight coefficient vector, and b is the effective wave height vector. The specific expressions are:

[0013]

[0014] In the formula, and 1 are the i-th group of observation row vectors in the observation matrix A; where 1≤i≤m; m is the total number of groups of observation row vectors.

[0015] and They are l hi , l pi and θ i The normalized value of .

[0016] snr i , l hi , l pi and θ i are the signal-to-noise ratio, heave 1 / 3 significant value, pitch 1 / 3 significant value and wave direction of the i-th group of observation row vectors in the observation matrix A respectively.

[0017] x1, x2, x3, x4 and x5 are the five column vectors in each set of observation vectors. and a weight coefficient of 1.

[0018] b1, b2, b i and b m are the effective wave height values ​​corresponding to the first group of observation row vectors, the second group of observation row vectors, the i-th group of observation row vectors and the m-th group of observation row vectors in the observation matrix A respectively.

[0019] Step 2: Determine the ship type to be simulated: Determine the ship type to be simulated by investigating the historical types of ships that have passed through the ocean to be monitored.

[0020] Step 3, determine the direction of the waves to be simulated: According to the ship type determined in step 2, and by investigating the historical data of the ocean waves to be monitored, determine the direction θ of all waves that need to be simulated for each type of ship.

[0021] Step 4, determine the working conditions to be simulated: According to the ship type determined in step 2, the wave direction θ determined in step 3, and by investigating the historical data of the ocean waves to be monitored, determine the number of working conditions to be simulated for each type of ship at each wave direction, as well as the effective wave height value b and wave period to be simulated for each working condition.

[0022] Step 5: Working condition simulation, specifically including the following steps.

[0023] Step 5-1, time-history data simulation based on ship motion response: For each type of ship in each working condition under each wave, a time-history data simulation based on ship motion response is performed for a set duration L, and a sway sensor is used to monitor the ship's sway data in real time.

[0024] Step 5-2, radar image simulation: using the non-coherent radar installed in the ocean to be monitored, a non-coherent radar image of a set duration L is simulated for each type of ship in each working condition of each wave downward.

[0025] Step 6: Data extraction, specifically including the following steps:

[0026] Step 6-1, Time history data extraction based on ship motion response: Starting from the starting time, according to the set time interval D, the swaying data obtained by the simulation of any ship type A in step 5-1 is subjected to the heave 1 / 3 significant value l h and pitch 1 / 3 meaningful value l p The extraction of the effective wave height b and the wave direction θ corresponding to the acquisition time is recorded, so as to obtain n groups l corresponding to the n effective wave height values ​​b and n wave directions θ of ship type A arranged in time sequence h and l p Value; where n>m.

[0027] Step 6-2, extracting the signal-to-noise ratio: starting from the starting time, according to the set time interval D, intercept the non-coherent radar image of the set time length L obtained by the simulation of any ship type A in step 5-2, and obtain a total of n non-coherent radar images arranged in chronological order, and record the effective wave height value b and the wave direction θ corresponding to the interception time; then, each non-coherent radar image is transformed into an image spectrum using Fourier transform, and the image spectrum is filtered and corrected to obtain the signal-to-noise ratio snr of the waves in the wave spectrum, and then obtain n groups of signal-to-noise ratio root mean squares corresponding to the n effective wave height values ​​b and n wave directions θ of ship type A arranged in chronological order.

[0028] Step 7: Data normalization: The root mean square of the signal-to-noise ratio of the m groups of ship type A obtained in step 6 is m group l h 、m group l pThe values ​​and the m wave directions θ are normalized to form m sets of dimensionless normalized data and

[0029] Step 8: Train the weight coefficient vector: normalize the m groups of data after step 7 and Substitute the observation matrix A constructed in step 1 into the observation matrix A, and then combine it with m groups of normalized data and The m corresponding effective wave height values ​​b are substituted into the effective wave height vector of the wave constructed in step 1 for training, thereby obtaining the weight coefficient vector x; the trained weight coefficient vector x is substituted into the wave measurement model in step 1, thereby obtaining the inversion wave measurement model.

[0030] Step 9, inversion, includes the following steps.

[0031] Step 9-1, obtain the actual signal-to-noise ratio of ship B: obtain the non-coherent radar image of the actual ship B at time K, and refer to the method of step 6-2 to extract the root mean square signal-to-noise ratio of the actual ship B at time K

[0032] Step 9-2, obtain the motion response time history data of the actual ship B: use the sway sensor installed on the actual ship B to obtain the wave direction θ at any time K k , 1 / 3 of the meaningful value l hk and pitch 1 / 3 meaningful value l pk .

[0033] Step 9-3: Normalize the actual data: Use the method in step 7 to normalize the data in step 9-1. l in step 9-2 hk , l pk ,θ k , are normalized to form the normalized and

[0034] Step 9-4, Inversion: and Substitute into the observation matrix A of the inversion wave measurement model in step 8 to obtain the actual effective wave height b of ship B at time K k .

[0035] In step 3, the values ​​of the wave directions θ of all waves are 0°, 20°, 40°, 60°, 80°, 100°, 120°, 140°, 160° and 180° respectively.

[0036] In step 4, each type of ship needs to simulate 7 working conditions at each wave level, and the 7 working conditions need to be selected from sea conditions of level 3 to 6 according to the type of ship; among them, the effective wave height of waves corresponding to level 3 sea conditions is 0.5-1.25m; the effective wave height of waves corresponding to level 4 sea conditions is 1.25-2.5m; the effective wave height of waves corresponding to level 5 sea conditions is 2.5-4.0m; the effective wave height of waves corresponding to level 6 sea conditions is 4.0-6.0m.

[0037] In step 6-1, the swing 1 / 3 has a significant value l h and pitch 1 / 3 meaningful value l p The calculation formula is:

[0038]

[0039] Where Q is the number of large waveforms in the first 1 / 3 of the heave time history data; h q is the height of the corresponding single waveform; R is the number of large waveforms in the first 1 / 3 of the pitch history data; h r is the corresponding single waveform height.

[0040] In step 7, The normalized formula is:

[0041]

[0042] in:

[0043]

[0044] In the formula, is the root mean square signal-to-noise ratio of m groups The column mean of .

[0045] is the root mean square signal-to-noise ratio of m groups The column variance of .

[0046] In step 9-3, the actual signal-to-noise ratio of ship B at time K is When normalizing, use the column mean of the root mean square of the signal-to-noise ratio of the m groups in step 7 and column variance

[0047] In step 7, The normalized formula is:

[0048]

[0049] In the formula, For m group l h The column mean of the values.

[0050] var(l hi) is m group l h Column variance of the values.

[0051] In step 9-3, the actual ship B is at time K. hk When normalizing, use the m group l in step 7 h The column mean of and column variance var(l hi ).

[0052] In step 7, The normalized formula is:

[0053]

[0054] In the formula, For m group l p The column mean of the values.

[0055] var(l pi ) is m group l p Column variance of the values.

[0056] In step 9-3, the actual ship B is at time K pk When normalizing, use the m group l in step 7 p The column mean and column variance var(l pi ).

[0057] In step 7, The normalized formula is:

[0058]

[0059] In the formula, For m groups of θ i The column mean of the values.

[0060] var(θ i ) is m groups of θ i Column variance of the values.

[0061] In step 9-3, the actual θ of ship B at time K k When normalizing, use the m groups of θ in step 7 i The column mean of and column variance var(θ i ).

[0062] In step 7, the RMS signal-to-noise ratio of the remaining nm groups is l h , l p The values ​​and wave direction θ are normalized; the normalized nm group l h , l pand θ, together with the remaining nm groups of significant wave height values ​​b, form the validation data set; when in step 8, m groups of normalized data and After the wave measurement model training is completed, the inversion wave measurement model is verified using the validation data set. During the verification, the correlation coefficient r and the root mean square error of the effective wave height b are used for evaluation.

[0063] In step 4, the wave period ranges from 6.5 to 10.5 s.

[0064] The present invention has the following beneficial effects:

[0065] 1. The present invention integrates the wave characteristic information obtained based on the ship's swaying motion into the traditional non-coherent radar wave measurement, and establishes a new multi-sensor information fusion wave measurement model. The wave characteristic information from the ship's oscillation motion and the signal-to-noise ratio characteristics of the radar image are combined to calculate a more accurate effective wave height. The measurement accuracy is higher and the error is smaller, which can more accurately reflect the actual situation of the waves.

[0066] 2. The present invention compares the ship to a wave buoy, providing a reasonable alternative for shipborne wave assessment. Under the action of waves, the ship sways, and the motion response reflects the wave excitation encountered by the ship. Currently, there are a variety of measurement devices that can measure the swaying motion of the ship in real time and record the time history information of the ship's pitch, heave, etc. Therefore, no additional equipment and facilities are required, which greatly reduces the cost of marine environment monitoring and improves cost-effectiveness.

[0067] 3. The effective wave height measurement technology proposed in the present invention can accurately measure the waves encountered in real time and continuously during the navigation of the ship, and has the advantages of low cost and no need for additional equipment calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic diagram showing the principle of inverting effective wave height based on the fusion of radar and ship sway information of the present invention.

[0069] Figure 2 A schematic diagram of the training process of the weight coefficient vector in the present invention is shown.

[0070] Figure 3 The line graph of the 1 / 3 significant value of the S175 container ship in this embodiment is shown at a wave direction of 0°.

[0071] Figure 4 The line graph of the 1 / 3 significant value of the S175 container ship in this embodiment is shown when the wave direction is 40°.

[0072] Figure 5The line graph of the 1 / 3 significant value of the S175 container ship at a wave direction of 120° in this embodiment is shown.

[0073] Figure 6 The line graph of the 1 / 3 significant value of the S175 container ship in this embodiment under the same working condition and different wave directions is shown.

[0074] Figure 7 The line graph of the average wave period of the S175 container ship in this embodiment under the same working condition and different wave directions is shown.

[0075] Figure 8 The line graph of 1 / 3 significant value extracted based on heave time history data of S175 container ship at significant wave height of 1.5 in this embodiment is shown.

[0076] Fig. 9 The line graph of the 1 / 3 significant value extracted based on the heave time history data of the S175 container ship at a significant wave height of 5.5 in this embodiment is shown.

[0077] Fig.10 Schematic diagram of significant wave height estimated based on signal-to-noise ratio characteristics.

[0078] Fig.11 It is a scatter plot of the wave height estimated based on the signal-to-noise ratio feature and the true value.

[0079] Fig.12 Schematic diagram of significant wave height estimated based on combined characteristics.

[0080] Fig.13 It is a scatter plot of the effective wave height estimated based on the combined features and the true value. DETAILED DESCRIPTION

[0081] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific preferred embodiments.

[0082] like Figure 1 and Figure 2 As shown, a technology for inverting effective wave height based on the fusion of radar and ship sway information includes the following steps.

[0083] Step 1: Construct a wave measurement model, specifically:

[0084] Ax=b (1)

[0085] Among them, A is the observation matrix, x is the weight coefficient vector, and b is the effective wave height vector. The specific expressions are:

[0086]

[0087] In the formula, and 1 are the i-th group of observation row vectors in the observation matrix A; where 1≤i≤m; m is the total number of groups of observation row vectors.

[0088] and They are l hi , l pi and θ i The normalized value of .

[0089] snr i , l hi , l pi and θ i are the signal-to-noise ratio, heave 1 / 3 significant value, pitch 1 / 3 significant value and wave direction of the i-th group of observation row vectors in the observation matrix A respectively.

[0090] x1, x2, x3, x4 and x5 are the five column vectors in each set of observation vectors. and a weight coefficient of 1.

[0091] b1, b2, b i and b m are the effective wave height values ​​corresponding to the first group of observation row vectors, the second group of observation row vectors, the i-th group of observation row vectors and the m-th group of observation row vectors in the observation matrix A respectively.

[0092] To prevent the weight imbalance among features, before constructing the observation matrix, considering the differences among the meaningful value data of SNR root mean square, heave and pitch history data, the m groups of SNR root mean square, l h and l p The feature data is normalized to ensure that different features have a relatively balanced influence in the model.

[0093] The SNR root mean square extracted from the radar image has an approximately linear relationship with the significant wave height. The x1 and x4 in the weight coefficient vector are to consider the influence of this weak linear relationship. For the convenience of expression, the coefficient 1 is added to the observation matrix.

[0094] Step 2: Determine the ship type to be simulated: Determine the ship type to be simulated by investigating the historical types of ships that have passed through the ocean to be monitored.

[0095] There are many types of ships, which can be divided into many categories according to their functions, designs and uses. Common ship classifications: passenger ships, cargo ships (such as bulk carriers, container ships, tankers), warships, engineering ships, etc.

[0096] Ships can usually be divided into the following categories according to their weight and size: Small ships: ships with a displacement of less than 200 tons; Medium ships: ships with a displacement of between 200 tons and several thousand tons, which can carry more cargo or passengers, such as cargo ships and ferries; Large ships: ships with a displacement of more than 10,000 tons, such as tankers, super cargo ships, passenger ships, etc., which are usually used for ocean transportation and international trade. Ultra-large ships: ships with a displacement of more than 100 million tons, mainly used for oil transportation.

[0097] In this embodiment, it is preferred to select the main ship type S175 container ship which has most of the characteristics of a container ship.

[0098] Step 3, determine the direction of the waves to be simulated: According to the ship type determined in step 2, and by investigating the historical data of the ocean waves to be monitored, determine the direction θ of all waves that need to be simulated for each type of ship.

[0099] In this embodiment, the values ​​of all wave directions θ are preferably 0°, 20°, 40°, 60°, 80°, 100°, 120°, 140°, 160° and 180° respectively.

[0100] In actual use, appropriate adjustments can be made according to the type of ship.

[0101] Step 4, determine the working conditions to be simulated: According to the ship type determined in step 2, the wave direction θ determined in step 3, and by investigating the historical data of the ocean waves to be monitored, determine the number of working conditions to be simulated for each type of ship at each wave direction, as well as the effective wave height value b and wave period to be simulated for each working condition.

[0102] In this embodiment, each type of ship needs to simulate 7 working conditions at each wave level, and the 7 working conditions need to be selected from sea conditions of level 3 to 6 according to the type of ship. Among them:

[0103] The effective wave height corresponding to level 3 sea conditions is 0.5 to 1.25 m.

[0104] The effective wave height corresponding to level 4 sea conditions is 1.25 to 2.5 meters.

[0105] The effective wave height corresponding to level 5 sea conditions is 2.5 to 4.0 m.

[0106] The effective wave height corresponding to level 6 sea condition is 4.0~6.0m.

[0107] The value range of the above wave period is preferably 6.5 to 10.5 s.

[0108] For the specific values ​​of the significant wave height and wave period in each working condition, refer to the distribution characteristics table of wave periods under different significant wave heights given by IMO in SDC 1 / INF.6.

[0109] In this embodiment, the seven working conditions to be simulated for the S175 container ship are marked as case 1 to case 7, which are:

[0110] Case 1: effective wave height 0.5m, wave peak period 6.5s, wave direction 0°~180°, interval 20°.

[0111] Case 2: effective wave height 1.5m, wave peak period 7.5s, wave direction 0°~180°, interval 20°.

[0112] Case 3: effective wave height 2.5m, wave peak period 8.5s, wave direction 0°~180°, interval 20°.

[0113] Case 4: effective wave height 3.5m, wave peak period 8.5s, wave direction 0°~180°, interval 20°.

[0114] Case 5: effective wave height 4.5m, wave peak period 9.5s, wave direction 0°~180°, interval 20°.

[0115] Case 6: effective wave height 5.5m, wave peak period 9.5s, wave direction 0°~180°, interval 20°.

[0116] Case 7: effective wave height 6.5m, wave peak period 10.5s, wave direction 0°~180°, interval 20°.

[0117] Step 5: Working condition simulation, specifically including the following steps.

[0118] Step 5-1, time-history data simulation based on ship motion response: For each type of ship in each working condition under each wave, a time-history data simulation based on ship motion response is performed for a set duration L, and a sway sensor is used to monitor the ship's sway data in real time.

[0119] In this embodiment, the setting time L is preferably 3600 seconds, the resolution is preferably 0.1 seconds, and the simulation speed is preferably 0 knots. It is preferred to use a two-and-a-half-dimensional method to simulate the motion of the ship under typical sea conditions to obtain the time history data of the ship's pitch and heave.

[0120] Step 5-2, radar image simulation: using the non-coherent radar installed in the ocean to be monitored, a non-coherent radar image of a set duration L is simulated for each type of ship in each working condition of each wave downward.

[0121] In this embodiment, the simulation of non-coherent radar images is performed according to the actual working conditions and corresponding wave parameters when the ship motion response is carried out. Based on the input actual working conditions and the characteristics of the typical sea area, the wave spectrum and the sea surface elevation are first obtained, and then the installation height of the radar and the imaging mechanism of the radar image (shadow modulation and tilt modulation) are combined to generate a nearly real non-coherent radar image using the sea surface elevation image.

[0122] Step 6: Data extraction, specifically including the following steps:

[0123] Step 6-1, Time history data extraction based on ship motion response: Starting from the starting time, according to the set time interval D, the swaying data obtained by the simulation of any ship type A in step 5-1 is subjected to the heave 1 / 3 significant value l h and pitch 1 / 3 meaningful value l p The extraction of the effective wave height b and the wave direction θ corresponding to the acquisition time is recorded, so as to obtain n groups l corresponding to the n effective wave height values ​​b and n wave directions θ of ship type A arranged in time sequence h and l p Value; where n>m.

[0124] According to the actual ship model, the ship motion is simulated under typical sea conditions with different wave periods and wave heights to obtain the historical data of the ship's heave and pitch. When the ship sails on the waves, the action of wind and waves will produce 6-degree-of-freedom motion, among which the ship's roll, pitch and heave motion are more prominent. In addition, the reliability of simulated pitch and heave is higher than that of roll. Therefore, the present invention mainly considers the historical data of pitch and heave. Using the generated ship pitch and heave time history data under different wave periods and effective wave height sea conditions, the corresponding 1 / 3 significant values ​​are extracted as statistical features.

[0125] Extracting wave meaningful value features based on the ship's heave and pitch time history data: Based on the generated ship pitch and heave time history data, the corresponding maximum value, 1 / 10 meaningful value, 1 / 3 meaningful value, average value and other statistical features can be extracted respectively. Among them, the 1 / 3 meaningful value is often used to characterize the effective wave height of waves. Therefore, the present invention intends to integrate the 1 / 3 meaningful value feature information obtained based on the ship's swaying motion into the traditional non-coherent radar wave measurement process.

[0126] Based on the recorded historical data of heave and pitch of the ship, the height of each waveform can be calculated according to the peak extreme point and the zero axis crossing. According to the determined number and height of the heave waveforms, the individual height values ​​of the waveforms are arranged in descending order. Therefore, the above heave 1 / 3 meaningful value l h and pitch 1 / 3 meaningful value l p The calculation formula is preferably:

[0127]

[0128] Where Q is the number of large waveforms in the first 1 / 3 of the heave time history data; h q is the height of the corresponding single waveform; R is the number of large waveforms in the first 1 / 3 of the pitch history data; h r is the corresponding single waveform height.

[0129] In this embodiment, for the S175 container ship, 7 working conditions, case 1 to case 7, are selected. At 0° wave direction, 40° wave direction and 120° wave direction, a line graph is drawn using the 1 / 3 significant value extracted from the heave and pitch history data, as shown in the figure below. Figure 3 , Figure 4 and Figure 5 As shown. Figures 3 to 5 It can be seen that the 1 / 3 significant values ​​of heave and pitch are close to the true values ​​and have a linear relationship.

[0130] In this embodiment, the heave characteristic data is further analyzed, wherein: Figure 6 The line graph is a graph of the 1 / 3 significant value extracted from the heave time history data. It can be seen that when the wave direction is about 100°, the 1 / 3 significant value has a maximum value; when the wave direction is 0°~100°, the 1 / 3 significant value shows an increasing trend; when the wave direction is 100°~180°, the 1 / 3 significant value shows a decreasing trend.

[0131] Figure 7 This is a line graph of the average wave period extracted from the heave time data. It can be seen that when the wave direction is around 100°, the average wave period has a minimum value; when the wave direction is 0°~100°, the average wave period shows a decreasing trend; when the wave direction is 100°~180°, the average wave period shows an increasing trend.

[0132] Based on the simulated sway time history data, the heave 1 / 3 meaningful value, pitch 1 / 3 meaningful value, and true value extracted under the significant wave height of 1.5m and wave period of 7.5s are as follows: Figure 8 As shown in the figure, the black curve and the red curve are the 1 / 3 significant value of heave and the 1 / 3 significant value of pitch, respectively. The blue curve is the 1 / 3 significant value extracted from the input data of the simulated ship swaying time history. It can be found from the figure that the 1 / 3 significant value extracted from the input time history of the ship motion response changes slightly in different wave directions and is close to 1.5m, while the 1 / 3 significant values ​​of pitch and heave change with the wave direction. The 1 / 3 significant value of heave increases first and then decreases, reaching the maximum value near the 90° wave direction; the 1 / 3 significant value of pitch reaches the maximum near 60° and 120°, and the significant value near 90° is the smallest.

[0133] Based on the simulated sway time history data, the heave 1 / 3 meaningful value, pitch 1 / 3 meaningful value, and true value extracted under the significant wave height of 5.5m and wave period of 9.5s are as follows: Fig. 9 As shown in the figure, it can be found that the 1 / 3 significant value extracted from the input time history of the ship motion response changes slightly in different wave directions, but is slightly lower than 5.5m, while the 1 / 3 significant values ​​of heave and pitch change with the wave direction. The characteristics of the 1 / 3 significant values ​​of pitch and heave in different wave directions are similar to Figure 8 The features in are similar.

[0134] By studying the 1 / 3 significant values ​​extracted based on the heave time history data under different working conditions, it can be found that the 1 / 3 significant values ​​of heave and pitch change with the wave direction. The 1 / 3 significant value of heave increases first and then decreases, reaching the maximum value near the 90° wave direction; the 1 / 3 significant value of pitch increases first and then decreases, reaching the minimum near 90°, and then continues to increase and then decrease, reaching the maximum value near 60° and 120°. According to the characteristics that the 1 / 3 significant values ​​of heave and pitch change with the wave direction, the significant values ​​extracted in different wave directions are quite different.

[0135] Therefore, the method of estimating wave height in this project based on the combination of features extracted from heave and pitch and the signal-to-noise ratio (SNR) needs to extract weight coefficients in different wave directions respectively, and then use the extracted weight coefficients to estimate the effective wave height in the wave direction.

[0136] Step 6-2, extracting the signal-to-noise ratio: starting from the starting time, according to the set time interval D, intercept the non-coherent radar image of the set time length L obtained by the simulation of any ship type A in step 5-2, and obtain a total of n non-coherent radar images arranged in chronological order, and record the effective wave height value b and the wave direction θ corresponding to the interception time; then, each non-coherent radar image is transformed into an image spectrum using Fourier transform, and the image spectrum is filtered and corrected to obtain the signal-to-noise ratio snr of the waves in the wave spectrum, and then obtain n groups of signal-to-noise ratio root mean squares corresponding to the n effective wave height values ​​b and n wave directions θ of ship type A arranged in chronological order.

[0137] In this embodiment, according to the actual situation of radar data acquisition, the simulated radar image sequence contains 32 images, the time resolution is 2.3 seconds, and the image sequence interval D = 4 minutes. In this way, under the same working condition, 1 hour corresponds to 15 groups of radar image sequences. The signal-to-noise ratio snr of each group of image sequences is obtained based on the spectrum analysis method of three-dimensional Fourier transform. At the same time, the ship swaying time history data is installed into 15 groups at intervals of 4 minutes, and the 1 / 3 significant value of the heave and 1 / 3 significant value of the pitch corresponding to each group of data is obtained. According to the significant wave height and wave period under 7 working conditions, n = 2100 groups of experimental data are generated, which are divided into two parts: training set and test set. The training set is used to determine the feature weight coefficient, and the test set is used to verify the accuracy of the multi-sensor information fusion inversion effective wave height model. In this embodiment, the training set experimental data is preferably m = 1050, which is used to determine the feature weight coefficient; the test set experimental data is preferably nm = 1050, which is used to verify the accuracy and effectiveness of the multi-sensor information fusion wave measurement model and method.

[0138] Step 7: Data normalization, including training set data normalization and test set data normalization.

[0139] A. Normalization of training set data

[0140] The RMS signal-to-noise ratio of the m groups of ship type A obtained in step 6 is m group l h 、m group l p The values ​​and the m wave directions θ are normalized to form m sets of dimensionless normalized data and

[0141] Above The normalization formula is preferably:

[0142]

[0143] in:

[0144]

[0145] In the formula, is the root mean square signal-to-noise ratio of m groups The column mean of .

[0146] is the root mean square signal-to-noise ratio of m groups The column variance of .

[0147] Above The normalization formula is preferably:

[0148]

[0149] In the formula, For m group l h The column mean of the values.

[0150] var(l hi ) is m group l h Column variance of the values.

[0151] Above The normalization formula is preferably:

[0152]

[0153] In the formula, For m group l p The column mean of the values.

[0154] var(l pi ) is m group l p Column variance of the values.

[0155] Above The normalization formula is preferably:

[0156]

[0157] In the formula, For m groups of θ i The column mean of the values.

[0158] var(θ i ) is m groups of θ i Column variance of the values.

[0159] B. Test set data normalization

[0160] In the test set, each group of experimental data uses the corresponding column mean and column variance in the training set when normalizing.

[0161] Step 8: Training weight coefficient vector

[0162] Step 8-1, training: training using the training set

[0163] The m groups of normalized data after normalization in step 7 and Substitute the observation matrix A constructed in step 1 into the observation matrix A, and then combine it with m groups of normalized data and The m corresponding effective wave height values ​​b are substituted into the effective wave height vector constructed in step 1 for training, wherein the constructed feature vector is fitted using the least squares method to determine the weight coefficients of each value in the feature vector, thereby obtaining the weight coefficient vector x. That is, according to the MP inverse solution equation Ax = b, x0 = A + b is its unique minimum norm least squares solution, that is, the weight coefficient vector x is obtained.

[0164] Next, the trained weight coefficient vector x is substituted into the wave measurement model in step 1 to obtain an inversion wave measurement model.

[0165] Step 8-2, Verification: Use the normalized test data set to verify the inversion wave measurement model; during verification, the correlation coefficient r and root mean square error of the effective wave height b are used for evaluation.

[0166] The present invention takes the ship swaying time history data and radar images of 0° wave direction simulation as examples to verify the effectiveness of the combined feature method in estimating the significant wave height, and uses the correlation coefficient r and root mean square error (RMSE) to evaluate the inversion method and model performance.

[0167] Fig.10 is the effective wave height extracted based on the signal-to-noise ratio (SNR) feature. The red line with a cross is the true value, and the blue dotted line with a triangle is the effective wave height estimated using only the SNR feature. It can be seen that the effective wave height estimated using only the SNR can increase with the increase of working conditions, but it deviates greatly from the true value.

[0168] Fig.11 The scatter plot of the effective wave height estimated based on the signal-to-noise ratio characteristics and the true value is as follows: the correlation coefficient is 0.83 and the root mean square error RMSE is 1.08m.

[0169] The significant wave height extracted based on the significant value extracted from the sway time history data and the signal-to-noise ratio combined feature is as follows: Fig.12 As shown in the figure. The red realization with a cross is the true value, and the blue dotted line with a triangle is the effective wave height estimated using the combined features. It can be seen from the figure that the effective wave height estimated by the meaningful value and signal-to-noise ratio extracted from the swaying time history data can increase with the increase of working conditions, and the changes in the estimated value and the true value are basically consistent.

[0170] Fig.13 The scatter plot of the effective wave height estimated based on the significant value extracted from the swaying time history data and the signal-to-noise ratio combined characteristics and the true value is shown in the figure. The root mean square error RMSE is 0.22m and the correlation coefficient is 0.99. It can be seen from the figure that the deviation between the effective wave height estimated by the combined method and the true value is relatively small under different sea conditions.

[0171] By comparing the effective wave height estimated by the simulated ship swaying time history data and the features extracted from the radar image under 0° wave downward, it can be found that the effective wave height extracted based on the significant value extracted from the swaying time history data and the signal-to-noise ratio combined features has a correlation coefficient of 0.16 greater than that extracted only using the signal-to-noise ratio features, and the root mean square error RMSE is 0.86m smaller. The experimental results show that the model and method for combining the ship swaying information and the non-coherent X-band radar image to invert the effective wave height not only has high wave height accuracy and small error, but also has a lower cost for marine environmental monitoring.

[0172] Based on the simulated ship swaying time history data and radar images, the experimental results under different wave conditions are shown in Table 1. The experimental results show that the method of fusing radar images and ship motion features proposed in the present invention can improve the accuracy of effective wave height estimation under different wave conditions.

[0173] Table 1 Performance of feature fusion methods at different wavelet levels

[0174]

[0175] Step 9, inversion, includes the following steps:

[0176] Step 9-1, obtain the actual signal-to-noise ratio of ship B: obtain the non-coherent radar image of the actual ship B at time K, and refer to the method of step 6-2 to extract the root mean square signal-to-noise ratio of the actual ship B at time K

[0177] Step 9-2, obtain the motion response time history data of the actual ship B: use the sway sensor installed on the actual ship B to obtain the wave direction θ at any time K k , 1 / 3 of the meaningful value l hk and pitch 1 / 3 meaningful value l pk .

[0178] Step 9-3: Normalize the actual data: Use the method in step 7 to normalize the data in step 9-1. l in step 9-2 hk , l pk ,θ k , are normalized to form the normalized and

[0179] Above l hk , l pk ,θ k The normalization of each set of experimental data in the test set uses the corresponding column mean and column variance in the training set, specifically:

[0180] A. The actual signal-to-noise ratio of ship B at time K When normalizing, use the column mean of the root mean square of the signal-to-noise ratio of the m groups in step 7 and column variance

[0181] B. The actual ship B at time K hk When normalizing, use the m group l in step 7 h The column mean of and column variance var(l hi ).

[0182] C. The actual ship B at time K pk When normalizing, use the m group l in step 7 p The column mean and column variance var(l pi ).

[0183] D. The actual θ of ship B at time K k When normalizing, use the m groups of θ in step 7 i The column mean of and column variance var(θ i ).

[0184] Step 9-4, Inversion: and Substitute into the observation matrix A of the inversion wave measurement model in step 8 to obtain the actual effective wave height b of ship B at time K k .

[0185] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A technology for inverting effective wave height based on the fusion of radar and ship sway information, Features: The steps include: Step 1: Construct a wave measurement model, specifically: Ax=b (1) Among them, A is the observation matrix, x is the weight coefficient vector, and b is the effective wave height vector. The specific expressions are: In the formula, and 1 are the i-th group of observation row vectors in the observation matrix A; where 1≤i≤m; m is the total number of groups of observation row vectors; and They are l hi , l pi and θ i The normalized value of snr i , l hi , l pi and θ i are the signal-to-noise ratio, heave 1 / 3 significant value, pitch 1 / 3 significant value and wave direction of the i-th group of observation row vectors in the observation matrix A respectively; x 1 、x 2 、x 3 、x 4 and x 5 are the five column vectors in each set of observation vectors and a weight coefficient of 1; b 1 、b 2 、b i and b m are the significant wave height values ​​corresponding to the first group of observation row vectors, the second group of observation row vectors, the i-th group of observation row vectors and the m-th group of observation row vectors in the observation matrix A respectively; Step 2: Determine the ship type to be simulated: Determine the ship type to be simulated by investigating the historical navigable ship types in the ocean to be monitored; Step 3, determine the direction of the waves to be simulated: according to the ship type determined in step 2, and by investigating the historical data of the ocean waves to be monitored, determine the direction θ of all waves to be simulated for each type of ship; Step 4, determine the working condition to be simulated: according to the ship type determined in step 2, the wave direction θ determined in step 3, and by investigating the historical data of the ocean waves to be monitored, determine the number of working conditions to be simulated for each type of ship under each wave direction, as well as the effective wave height value b and wave period to be simulated for each working condition; Step 5: Working condition simulation, specifically including the following steps: Step 5-1, time history data simulation based on ship motion response: for each type of ship in each working condition of each wave, a time history data simulation based on ship motion response is performed for a set time L, and a swaying sensor is used to monitor the swaying data of the ship in real time; Step 5-2, radar image simulation: using a non-coherent radar installed in the ocean to be monitored, a non-coherent radar image simulation of a set duration L is performed for each type of ship in each working condition of each wave downward; Step 6: Data extraction, specifically including the following steps: Step 6-1, Time history data extraction based on ship motion response: Starting from the starting time, according to the set time interval D, the swaying data obtained by the simulation of any ship type A in step 5-1 is subjected to the heave 1 / 3 significant value l h and pitch 1 / 3 meaningful value l p The extraction of the effective wave height b and the wave direction θ corresponding to the acquisition time is recorded, so as to obtain n groups l corresponding to the n effective wave height values ​​b and n wave directions θ of ship type A arranged in time sequence h and l p Value; where n>m; Step 6-2, extracting the signal-to-noise ratio: starting from the starting time, according to the set time interval D, intercept the non-coherent radar image of the set time length L obtained by the simulation of any ship type A in step 5-2, and obtain a total of n non-coherent radar images arranged in chronological order, and record the effective wave height value b and the wave direction θ corresponding to the interception time; then, each non-coherent radar image is transformed into an image spectrum using Fourier transform, and the image spectrum is filtered and corrected to obtain the signal-to-noise ratio snr of the waves in the wave spectrum, and then obtain n groups of signal-to-noise ratio root mean squares corresponding to the n effective wave height values ​​b and n wave directions θ of ship type A arranged in chronological order. Step 7: Data normalization: The root mean square of the signal-to-noise ratio of the m groups of ship type A obtained in step 6 is m group l h 、m group l p The values ​​and the m wave directions θ are normalized to form m sets of dimensionless normalized data and Step 8: Train the weight coefficient vector: normalize the m groups of data after step 7 and Substitute the observation matrix A constructed in step 1 into the observation matrix A, and then combine it with the m groups of normalized data and Substitute the m corresponding effective wave height values ​​b into the effective wave height vector constructed in step 1 for training, thereby obtaining a weight coefficient vector x; substitute the trained weight coefficient vector x into the wave measurement model in step 1, thereby obtaining an inversion wave measurement model; Step 9, inversion, includes the following steps: Step 9-1, obtain the actual signal-to-noise ratio of ship B: obtain the non-coherent radar image of the actual ship B at time K, and refer to the method of step 6-2 to extract the root mean square signal-to-noise ratio of the actual ship B at time K Step 9-2, obtain the motion response time history data of the actual ship B: use the sway sensor installed on the actual ship B to obtain the wave direction θ at any time K k , 1 / 3 of the meaningful value l hk and pitch 1 / 3 meaningful value l pk ; Step 9-3: Normalize the actual data: Use the method in step 7 to normalize the data in step 9-1. l in step 9-2 hk , l pk ,θ k , are normalized to form the normalized and Step 9-4, Inversion: and Substitute into the observation matrix A of the inversion wave measurement model in step 8 to obtain the actual effective wave height b of ship B at time K k .

2. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 3, the values ​​of the wave directions θ of all waves are 0°, 20°, 40°, 60°, 80°, 100°, 120°, 140°, 160° and 180° respectively.

3. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 4, each type of ship needs to simulate 7 working conditions at each wave level, and the 7 working conditions need to be selected from sea conditions of level 3 to 6 according to the type of ship; among them, the effective wave height of waves corresponding to level 3 sea conditions is 0.5-1.25m; the effective wave height of waves corresponding to level 4 sea conditions is 1.25-2.5m; the effective wave height of waves corresponding to level 5 sea conditions is 2.5-4.0m; the effective wave height of waves corresponding to level 6 sea conditions is 4.0-6.0m.

4. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 6-1, the swing 1 / 3 has a significant value l h and pitch 1 / 3 meaningful value l p The calculation formula is: Where Q is the number of large waveforms in the first 1 / 3 of the heave time history data; h q is the height of the corresponding single waveform; R is the number of large waveforms in the first 1 / 3 of the pitch history data; h r is the corresponding single waveform height.

5. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 7, The normalized formula is: in: In the formula, is the root mean square signal-to-noise ratio of m groups The column mean of ; is the root mean square signal-to-noise ratio of m groups Column variance of In step 9-3, the actual signal-to-noise ratio of ship B at time K is When normalizing, use the column mean of the root mean square of the signal-to-noise ratio of the m groups in step 7 and column variance 6. The technology for inverting significant wave height based on the fusion of radar and ship sway information according to claim 1, Features: In step 7, The normalized formula is: In the formula, For m group l h The column mean of the values; var(l hi ) is m group l h Column variance of values; In step 9-3, the actual ship B is at time K. hk When normalizing, use the m group l in step 7 h The column mean and column variance var(l hi ).

7. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 7, The normalized formula is: In the formula, For m group l p The column mean of the values; var(l pi ) is m group l p Column variance of values; In step 9-3, the actual ship B is at time K. pk When normalizing, use the m group l in step 7 p The column mean and column variance var(l pi ).

8. The technology for inverting significant wave height based on the fusion of radar and ship sway information according to claim 1, Features: In step 7, The normalized formula is: In the formula, For m groups of θ i The column mean of the values; var(θ i ) is m groups of θ i Column variance of values; In step 9-3, the actual θ of ship B at time K k When normalizing, use the m groups of θ in step 7 i The column mean and column variance var(θ i ).

9. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 7, the RMS signal-to-noise ratio of the remaining nm groups is l h , l p The values ​​and wave direction θ are normalized; the normalized nm group l h , l p and θ, together with the remaining nm groups of significant wave height values ​​b, form the validation data set; when in step 8, m groups of normalized data and After the wave measurement model training is completed, the inversion wave measurement model is verified using the validation data set. During the verification, the correlation coefficient r and the root mean square error of the effective wave height b are used for evaluation.

10. The technology for inverting significant wave height based on fusion of radar and ship sway information according to claim 1, Features: In step 4, the wave period ranges from 6.5 to 10.5 s.

Images