A method for extracting jump layers from multiple types of sound velocity profiles

Abstract

The present application discloses a method for extracting multiple types of sound velocity profile jump layers, belonging to the field of marine survey. Aiming at the problems that the extraction range is too deep, the spatial distribution is too scattered, and the extraction cannot be accurate in the case of multiple sound velocity jump layers when the traditional optimal segmentation algorithm is used to extract the sound velocity jump layer, the present application simplifies the sound velocity profile into a four-layer structure on the basis of the traditional optimal segmentation method; the accurate identification of the double jump layer is realized according to the principle of the number of sound channel axes that meet the gradient threshold; the upper limit of the sound velocity jump layer is obtained by repeated iterative calculations with the gradient mutation point as the starting point; in the case of a single sound channel axis, the optimal segmentation algorithm is used to determine the lower limit of the jump layer; in the case of multiple sound channel axes, the optimal three-segmentation method is used to determine the upper and lower limits of the jump layer. The method of the present application can effectively and accurately identify the multi-jump layer structure, and can also realize the feature extraction of multiple types of sound velocity profile jump layers, which is helpful to improve the reliability of underwater communication and the concealment performance of submarines.

Description

Technical Field

[0001] The invention relates to a method for extracting multiple types of sound velocity profile jump layers, and belongs to the field of marine environment survey. Background Art

[0002] The traditional method of extracting the acoustic jump layer is usually to find a minimum value of the sound velocity gradient, and define the water layer that exceeds this gradient standard value as the acoustic jump layer. This method of calculating the acoustic jump layer is adopted by my country's "Ocean Survey Specifications" (State Administration of Technical Supervision, 2007). "Ocean Survey Specifications Part 7: Ocean Survey Data Exchange" stipulates that the threshold of the acoustic jump layer is 0.5 / s in shallow waters and 0.2 / s in deep waters, and the depth is less than 200m in shallow waters. However, this method only generally proposes a general algorithm for calculating the acoustic jump layer, and there are problems in extracting jump layers in complex acoustic environments in different sea areas.

[0003] Zhang Xu and others introduced the vertical gradient method and the optimal segmentation method to determine the upper and lower boundaries of the jump layer, and found that both methods are effective for typical profile data. However, for the case where the maximum vertical gradient is lower than the minimum standard, and the shallow sea and deep sea standards cannot match in the transition zone, these two methods cannot obtain satisfactory results. For example, the optimal segmentation method has the defect of selecting points that are too deep. Summary of the invention

[0004] The present application proposes a method for extracting jump layers of multiple types of sound velocity profiles, which can be used as an evaluation index to optimize any wave energy device or construct an array of wave energy devices based on the wave conditions in the target sea area.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is a method for extracting jump layers of multiple types of sound velocity profiles, comprising the following steps:

[0006] Step 1, calculating the ocean vertical sound velocity profile data based on the temperature, salinity and corresponding depth data;

[0007] Step 2, performing data interpolation and resampling on the ocean vertical sound velocity profile data obtained in step 1;

[0008] Step 3, using the difference method to calculate the sound velocity gradient;

[0009] Step 4, finding the vocal tract axis based on the sound velocity gradient and the gradient threshold;

[0010] Step 5, introduce the channel axis as a parameter for determining the jump layer structure, extract the acoustic jump layer, and obtain the upper and lower boundaries of the jump layer;

[0011] Step 6: Calculate the jump layer related attributes and record the jump layer related attribute information to construct the jump layer related data set.

[0012] Optimized, the above method for extracting multiple types of sound speed profile jump layers, in step 1, the ocean is densely divided into grid points, and the sound speed profile of each grid point is calculated to construct the global ocean sound speed field;

[0013] Substitute the temperature, salinity and corresponding depth data into the empirical formula of sound speed to obtain the vertical sound speed profile data of the ocean; the empirical formula of sound speed is:

[0014] c=1449.14+c t +c p +c s +c stp ;

[0015] c t =4.5721t-4.4532×10 -2 t 2 -2.6045×10 -4 t 3 +7.9851×10 -6 t;

[0016] c p =1.60272×10 -1 p+1.0268×10 -5 p 2 +3.5216×10 -9 p 3 -3.3603×10 -12 p 4 ;

[0017] c s =1.39799(s-35)+1.69202×10 -3 (s-35) 2

[0018]

[0019] Where c is the speed of sound; t is only related to temperature; c p is only related to static pressure; c s is only related to salinity; c stp is a term related to temperature, salinity and static pressure; t is temperature; s is salinity; p is static pressure.

[0020] Optimized, the above method for extracting multiple types of sound velocity profile jump layers, in step 2,

[0021] Use Akima interpolation method to interpolate the sound speed data points to obtain a set of continuous sound speed values;

[0022] Akima interpolation can produce smoother interpolation curves when dealing with irregular data point distribution;

[0023] The interpolated sound velocity data points are approximated by a cubic polynomial to obtain a smoother sound velocity profile curve;

[0024] The cubic polynomial approximation can be used to obtain a continuous curve representation by fitting the data points;

[0025] The Akima interpolation formula is

[0026] y=p0+p1(x-x3)+p2(x-x3) 2 +p3(x-x3) 3 ;

[0027] Use median filtering to filter out the subtle burrs produced by Akima interpolation;

[0028] Where y represents the sound velocity value of the interpolation point, and x represents the depth value of the interpolation point; suppose the known data point is

[0029] (x i ,y i )(i=1,2,3…n), find a smooth curve y=f(x) so that yi=f(xi);

[0030] p0, p1, p2, p3 are the required coefficients, which are determined by the slope t of the surrounding points. i The calculation formula is:

[0031]

[0032] t3 and t4 represent the derivatives of the third and fourth points respectively, where i = 3, 4; m i is the derivative at point i, replaced by the slope;

[0033] m i =(y i+1 -y i ) / (x i+1 -x i );

[0034] When m i =(y i+1 -y i ) / (x i+1 -x i ) is not true when the denominator is 0, then t i =(m i-1 +m i ) / 2.

[0035] Optimized, the above method for extracting multiple types of sound velocity profile jump layers, in step 3,

[0036] The calculation formula of the differential method is g = Δs / Δd;

[0037] Where g is the sound velocity gradient; Δs is the sound velocity difference between adjacent depth points; and Δd is the depth difference between adjacent points.

[0038] Optimized, the above method for extracting multiple types of sound velocity profile jump layers, in step 4,

[0039] Determine whether the maximum gradient of the entire sound velocity profile curve is less than or equal to 0.01 / s, and whether the extreme difference of the gradient is less than or equal to 0.02 / s;

[0040] If the maximum gradient of the sound velocity profile curve is less than or equal to 0.01 / s, and the extreme difference of the gradient is less than or equal to 0.02 / s, there is no sound channel axis that can produce a jump layer. Otherwise, it is further determined whether it is a negative jump layer.

[0041] After traversing each point on the sound velocity profile, the vocal tract axis is found.

[0042] Optimized, the above method for extracting jump layers of multiple types of sound velocity profiles, in step 5,

[0043] The channel axis is introduced as a parameter for judging the jump layer structure, and the number of channel axes corresponds to the number of jump layers.

[0044] The extraction of the sonic jump layer is to determine the upper and lower boundaries of the jump layer, and determine the water layer between the upper and lower boundaries as the sonic jump layer.

[0045] Optimized, the above method for extracting jump layers of multiple types of sound velocity profiles, in step 5,

[0046] The calculation method of the upper limit of the thermocline is the iteration of the sound velocity gradient mutation point and the optimal binary division method from the iteration starting point to the sound channel axis or the seabed.

[0047] The judgment of the lower boundary of the jump layer is divided into two cases:

[0048] In the case of a single channel axis, the end point of the optimal two-division process, that is, the channel axis, is taken as the lower boundary of the acoustic jump layer;

[0049] In the case of multi-channel axes, the end point of the optimal three-division is taken as the lower boundary of the sound jump layer.

[0050] The above method for extracting jump layers of multiple types of sound velocity profiles is optimized. After obtaining the upper and lower bounds of the jump layer, the characteristic quantities of the jump layer are calculated. The characteristic quantities of the jump layer are divided into jump layer depth dh, thickness th and intensity gy. The calculation of the characteristic quantities of the jump layer is

[0051] th=Zend -Z star t;

[0052]

[0053] In the formula, Z start Indicates the upper boundary depth of the acoustic jump layer, Z end represents the lower boundary depth of the acoustic jump layer, C start represents the speed of sound at the upper boundary of the sonic jump layer, C end Represents the sound velocity at the lower boundary of the sonic jump layer.

[0054] Optimized, the above method for extracting multiple types of sound velocity profile jump layers, step 5 includes the following steps:

[0055] Step 5-1:

[0056] Iteration starts from the sound velocity gradient mutation point H, and the iteration end point is the minimum value among the sound channel axis, the seabed, and the depth value of 400 meters;

[0057] In the case of negative jump layers, the sound velocity gradient mutation point H is taken as the depth where the gradient is maximum; in other cases, the point where the sound velocity gradient changes from positive to negative or the gradient is minimum is considered to be the sound velocity gradient mutation point H;

[0058] Step 5-2:

[0059] Starting from the sound velocity gradient mutation point H, an iteration is performed every n meters;

[0060] Each iteration calculates the best binary segmentation result from the sound velocity gradient mutation point H to the current position, and records the segmentation result;

[0061] In the case of negative jump layer, the value of n in the sound velocity profile is 3, and the value of k is 4; in the case of the sound velocity profile without the formation of the sound channel axis, the value of n is 5, and the value of k is 3; in the case of the sound velocity profile with the formation of the sound channel axis, the value of n is 10, and the value of k is 5;

[0062] Step 5-3:

[0063] After the iteration is completed, if the difference between the segmentation results of two adjacent points is less than n / 2, it is considered to be a continuous value;

[0064] When the continuous value exceeds k, the average value of the continuous value is calculated as the upper bound of the jump layer;

[0065] Step 5-4:

[0066] For the sound velocity profile with the vocal tract axis, the optimal binary segmentation method is used for segmentation, and the starting point of the segmentation is set as the upper boundary of the jump layer, and the end point is set as the vocal tract axis to obtain the lower boundary of the jump layer;

[0067] Step 5-5:

[0068] Determine whether there are multiple channel axes; if there are multiple channel axes and the minimum gradient requirement is met, use the optimal three-segmentation method to obtain the jump layers between the channel axes;

[0069] The deep sea sound velocity profile structure is divided into mixing layer, main jump layer, acoustic channel layer and deep sea isothermal layer. This is used as the theoretical basis to simplify the jump layer division.

[0070] Optimized, the above method for extracting multiple types of sound velocity profile jump layers, the optimal two-segmentation algorithm is to convert the ordered sequence x k , at this time k = 2, divided into 2 groups;

[0071] The mean of the group from i to j can be recorded as Then there is

[0072]

[0073] The sum of squares of deviations of this group is defined as

[0074]

[0075] If the entire sequence is divided into 2 groups, the total variation under the 2-partition can be considered as the sum of the variations of the 2 groups.

[0076] V(k|α1,α2…,α k-1 )=V(1,α1)+V(α1+1,α2)+…+V(α k-1 +1, n)

[0077] Therefore, V n The smallest grouping is the optimal binary split;

[0078] The optimal three-partition algorithm is to transform the ordered sequence x k , at this time k = 3, divided into 3 groups;

[0079] The mean of the group from i to j can be recorded as Then there is

[0080]

[0081] The sum of squares of deviations for this group is defined as

[0082]

[0083] If the entire sequence is divided into 3 groups, the total variation under the 3-partition can be considered as the sum of the variations of the 3 groups.

[0084] V(k|α1,α2…,α k-1 )=V(1,α1)+V(α1+1,α2)+…+V(αk-1 +1, n)

[0085] Therefore, V n The smallest grouping is the optimal three-part split.

[0086] The beneficial effects of this application are:

[0087] The technical solution of the present application involves the identification of acoustic jump layers in the ocean acoustic environment and the extraction of their characteristic parameters. It mainly aims at the problems of too deep extraction range, too scattered spatial distribution and inability to accurately extract in the case of multiple acoustic jump layers when the traditional optimal segmentation algorithm is used to extract acoustic jump layers. A method for extracting acoustic jump layers that is adaptive to multiple sound velocity profiles and multiple jump layers is proposed - the ADMM method. Based on the traditional optimal segmentation method, this method uses the SSP gradient change and related environmental parameters to adaptively obtain the threshold, so as to obtain the number of sound channel axes that meet the gradient threshold requirements, and finally realize the identification of multiple jump layers. According to the sensitivity of the gradient change, the characteristic parameters of the acoustic jump layer are accurately extracted. The algorithm can effectively and accurately identify the multi-jump layer structure, and can also realize the feature extraction of various types of acoustic jump layers in the sound velocity profile.

[0088] In order to overcome the defects of the existing jump layer calculation method, such as inaccurate extraction of jump layer upper and lower boundaries, unreasonable threshold setting, and too deep extraction range, which leads to large jump layer extraction error, low efficiency, and poor applicability, this algorithm improves the traditional acoustic jump layer calculation method in the following ways:

[0089] The parameter of the number of channel axes is introduced to determine the number of acoustic jump layers. The gradient threshold is used to search for the channel axis and determine the number of channel axes. The number of jump layers is determined according to the principle that the number of channel axes determines the jump layer structure, thereby avoiding the problem that the multi-jump layer structure cannot be extracted.

[0090] The adaptive threshold is set by using the change of sound velocity gradient to realize the extraction of jump layers in different sound field environments. A typical sound velocity profile data set is constructed to verify the applicability of this algorithm under different sound velocity profile types in different regions. The results show that this algorithm is applicable to deep sea, shallow sea and transitional sound velocity profiles.

[0091] The upper boundary of the jump layer is regarded as the starting point of the seasonal jump layer and the permanent jump layer, which simplifies the jump layer structure and improves the accuracy of sound speed jump layer detection, thus avoiding the problem of extracting the jump layer too deep by traditional methods.

[0092] The average value of multiple optimal segmentation results is used to replace the single optimal segmentation result. Based on the improved calculation of the single optimal segmentation result, it is further proposed to calculate the multiple optimal segmentation results. Instead of simply using the single optimal segmentation algorithm, multiple optimal segmentations are used by changing the segmentation end point to obtain a continuous result sequence, calculate its average value, and determine the upper boundary of the jump layer. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is the flow chart of ADMM algorithm in this application;

[0094] Figure 2 Comparison of the extraction results of binaural axonocline using the traditional optimal segmentation method (left) and the ADMM method (right);

[0095] Figure 3 Comparison of the extraction results under different sound velocity profiles using the traditional optimal segmentation method (left) and the ADMM method (right); DETAILED DESCRIPTION

[0096] The technical features of the present invention are further described below in conjunction with specific embodiments.

[0097] The present invention provides a method for extracting multiple types of sound velocity profile jump layers, comprising the following steps:

[0098] Step 1

[0099] Based on the temperature, salinity and corresponding depth data, the ocean vertical sound velocity profile data is calculated.

[0100] GLORYS12V1 data is a global ocean reanalysis data product provided by EU CMEMS. This data set contains multi-year temperature and salinity data of global oceans, with a spatial resolution of (1 / 12)°×1 / 12)°, a temporal resolution of 1d, and a vertical resolution of 50 layers from 0.5 meters on the sea surface to 5,500 meters on the bottom. GLORYS12V1 can be used to create boundaries or initial conditions for various ocean models and perform ocean numerical simulations. It has the characteristics of neat arrangement and ease of use.

[0101] The sound velocity profile is a sound velocity-depth curve that varies with depth. It reflects the change of sound velocity at a certain point with depth. The global ocean is divided into dense grid points, and the sound velocity profile of each grid point is calculated. In this way, the global ocean sound velocity field can be constructed. The sound velocity field can be constructed through temperature, salinity and depth data. At present, most empirical formulas are used. Wilson (Wilson) n ) The empirical formula for the speed of sound is considered to be more stable in shallow and deep seas. The formula is as follows

[0102] c=1449.14+c t +c p +c s +c stp ;

[0103] c t =4.5721t-4.4532×10 -2 t 2 -2.6045×10 -4 t 3 +7.9851×10-6 t;

[0104] c p =1.60272×10 -1 p+1.0268×10 -5 p 2 +3.5216×10 -9 p 3 -3.3603×10 -12 p 4 ;

[0105] c s =1.39799(s-35)+1.69202×10 -3 (s-35) 2

[0106]

[0107] Where c is the speed of sound; t is only related to temperature; c p is only related to static pressure; c s is only related to salinity; c stp are items related to temperature, salinity and static pressure, all of which are in units of m / s; t is temperature, in ℃; s is salinity, in ‰; and p is static pressure, in Pa.

[0108] By substituting the temperature, salinity and corresponding depth data of GLORYS12V1 into the empirical formula of sound speed, the vertical sound speed profile data of the ocean can be obtained, thereby constructing the global ocean sound speed field.

[0109] Step 2

[0110] Perform data interpolation and resampling on the ocean vertical sound velocity profile data.

[0111] The sound velocity profile constructed using temperature, salinity and depth data has a large number of noise points, so the sound velocity profile data needs to be preprocessed. First, Akima interpolation is performed on the sound velocity profile, and the data points are approximated using a cubic polynomial to smooth the sound velocity profile.

[0112] y=p0+p1(x-X3)+p2(x-X3) 2 +p3(x-X3) 3 (19)

[0113] Secondly, the median filter is used to filter out the tiny burrs generated after Akima interpolation, which can eliminate the noise caused by interpolation and avoid the problem of the extracted acoustic jump layer having too scattered spatial distribution.

[0114] The Akima interpolation method interpolates between two measured points and requires the use of data from three measured points before and after. Therefore, the Akima interpolation method requires at least six measured points, and two additional points are required for interpolation at the endpoints.

[0115] Assume that the known data point is (x i ,y i )(i=1,2,3,…n), we need to find a smooth curve y that satisfies y i =f(x i ). At the same time, it is agreed that between any two adjacent data points, a cubic polynomial is used to approximate. Therefore, the derivative of each data point is the key to solving the interpolation. Assume that the interpolation point is

[0116] Function y = p0 + p1 (x-x3) + p2 (x-x3) 2 +p3(x-x3) 3 superior:

[0117]

[0118] Where: t3 and t4 represent the derivatives of the third and fourth points respectively, where i = 3, 4. i is the derivative at point i, calculated by the following formula: i =(y i+1 -y i ) / (x i+1 -x i ).

[0119] When m i =(y i+1 -y i ) / (x i+1 -x i ) is not true when the denominator is 0, then t i =(m i-1 +m i ) / 2.

[0120] Step 3

[0121] Using the difference method, the sound velocity gradient is calculated;

[0122] The sound velocity gradient reflects the rate of change of sound velocity with depth. The gradient is calculated using the traditional differential method, and the calculation formula is as follows: g = Δs / Δd. Among them, g represents the sound velocity gradient, Δs represents the sound velocity difference between adjacent depth points, and Δd represents the depth difference between adjacent points.

[0123] Step 4

[0124] The vocal tract axis is found based on the sound velocity gradient and the gradient threshold.

[0125] a) Preliminary screening of vocal channel axis

[0126] Determine whether the maximum gradient of the entire sound velocity profile is less than or equal to 0.01 / s, and determine whether the extreme difference of the gradient is less than or equal to 0.02 / s. If the conditions are met, it means that the sound velocity gradient is small and there is no sound channel axis that can produce jump layers.

[0127] Negative jump layer judgment: Negative jump layer is a special jump layer structure. At a certain depth underwater, the sound speed is increased. Contrary to the positive jump layer, it is also distributed in a step-like manner. This standard is used to judge whether there is a negative jump layer.

[0128] Traverse each point, determine the minimum point of the sound velocity profile as the vocal tract axis, and preliminarily screen it as the vocal tract axis.

[0129] After completing the above steps, we get the original set of vocal tract axes, among which there may be some fine structures that are not real vocal tract axes and need to be further screened.

[0130] b) Expand the initial screening results

[0131] The interval of x meters is extended above and below each channel axis. According to the different sea depths, when the sea depth is less than 50 meters, x is taken as 10 meters; when the sea depth is greater than 50 meters, x is taken as 100 meters.

[0132] Find the point with the minimum sound speed within the extended 2x interval. If the point found is not the minimum sound speed, discard it.

[0133] It is determined whether the gradients between the channel axes are all less than or equal to 0.005 / s. If not, it is considered that although there are multiple channel axes, there is no jump layer here.

[0134] Step 5

[0135] The acoustic channel axis is introduced as a parameter for judging the jump layer structure and the acoustic jump layer is extracted.

[0136] This algorithm introduces the channel axis as an important parameter for determining the jump layer structure. The traditional jump layer extraction algorithm does not take the channel axis into consideration, which results in incomplete jump layer structure extraction. In this algorithm, the number of channel axes is considered to correspond to the number of jump layers.

[0137] The extraction of the sonic jump layer is to determine the upper and lower boundaries of the jump layer on the basis of the above processing. The water layer between the upper and lower boundaries is the sonic jump layer. The calculation method of the upper boundary of the jump layer adopts the iteration of the sound velocity gradient mutation point and the optimal two-division method from the iteration starting point to the channel axis or the seabed; the judgment of the lower boundary of the jump layer is divided into two cases: in the case of a single channel axis, the end point of the optimal two-division process, that is, the channel axis, is used as the lower boundary of the sonic jump layer; in the case of multiple channel axes, the end point of the optimal three-division is used as the lower boundary of the sonic jump layer.

[0138] The present invention extracts acoustic jump layers for three different sound velocity profiles, including a sound velocity profile structure with a negative jump layer, a sound velocity profile structure with a sound channel axis, and a sound velocity profile structure without a sound channel axis. For the above three sound velocity profile structures, the following steps are performed:

[0139] Step 5-1, first, start the iteration from the sound velocity gradient mutation point H, and the iteration end point can be selected as the channel axis, the seabed or the depth value of 400 meters (the minimum value among the three is taken to prevent the iteration from being too deep). In the case of a negative jump layer, the sound velocity gradient mutation point H is the depth where the gradient is the maximum; in other cases, the point where the sound velocity gradient changes from positive to negative or the gradient is the minimum is considered to be the sound velocity gradient mutation point H.

[0140] Step 5-2, starting from the sound velocity gradient mutation point H, iterate every n meters. Each iteration calculates the best binary segmentation result from the sound velocity gradient mutation point H to the current position, and records the segmentation result. In the case of a negative jump layer, the n value of the sound velocity profile is 3, and the k value is 4; in the sound velocity profile without forming the sound channel axis, the n value is 5, and the k value is 3; in the sound velocity profile with forming the sound channel axis, the n value is 10, and the k value is 5.

[0141] Step 5-3, after the iteration is completed, if the difference between the segmentation results of two adjacent points is less than nn / 2, it is considered to be a continuous value. When the continuous value exceeds k, the average value of the continuous value is calculated as the upper bound of the jump layer.

[0142] Step 5-4, for the sound velocity profile with the vocal tract axis, the optimal binary segmentation method is used for segmentation, and the starting point of the segmentation is set as the upper boundary of the jump layer, and the end point is set as the vocal tract axis to obtain the lower boundary of the jump layer.

[0143] The optimal binary split algorithm is to transform the ordered sequence x k , at this time k = 2, divided into 2 groups. The mean of the group from i to j can be recorded as Then there is

[0144]

[0145] The sum of squares of deviations for this group is defined as

[0146]

[0147] If the entire sequence is divided into 2 groups, the total variation under this 2-partition can be considered as the sum of the variations of the 2 groups.

[0148] V(k|α1,α2…,α k-1 )=V(1,α1)+V(α1+1,α2)+…+V(α k-1 +1, n)

[0149] Therefore, V n The smallest grouping is the optimal binary split.

[0150] Step 5-5, determine whether there are multiple acoustic channel axes. If there are multiple acoustic channel axes and the minimum gradient requirement is met, the optimal three-segmentation method is used to obtain the jump layers between the acoustic channel axes. In view of the inaccurate extraction of jump layers by the original optimal segmentation method, the present invention adopts the deep sea sound velocity profile structure division theory proposed by Zhang Xu et al., and divides the sound velocity profile into four parts: mixing layer, main jump layer, acoustic channel layer and deep sea isotherm. Based on this theory, the jump layer division is simplified. The optimal three-segmentation algorithm is to convert the ordered sequence x into a k , at this time k = 3, divided into 3 groups. The mean of the group from i to j can be recorded as Then there is

[0151]

[0152] The sum of squares of deviations for this group is defined as

[0153]

[0154] If the entire sequence is divided into 3 groups, the total variation under this 3-partition can be considered as the sum of the variations of the 3 groups.

[0155] V(k|α1,α2…,α k-1 )=V(1,α1)+V(α1+1,α2)+…+V(α k-1 +1, n)

[0156] Therefore, V n The smallest grouping is the optimal three-part split.

[0157] Step 6

[0158] After obtaining the upper and lower bounds of the jump layer, the characteristic quantities of the jump layer can be obtained. The characteristic quantities are divided into jump layer depth dh, thickness th and intensity gy. The jump layer related attributes are calculated as follows:

[0159] th=Z end -Z start

[0160]

[0161] In the formula, Z start Indicates the upper boundary depth of the acoustic jump layer, Z end represents the lower boundary depth of the acoustic jump layer, C start represents the speed of sound at the upper boundary of the sonic jump layer, C end Represents the sound velocity at the lower boundary of the sonic jump layer.

[0162] Record the attribute information related to the jump layer to construct the jump layer related data set.

[0163] The technical solution of the present application can completely extract the multi-layer structure. The technical solution of the present application improves the traditional optimal segmentation method and innovatively introduces the key parameter of the number of channel axes to a priori judge the overall number of sound velocity jump layers, thereby solving the problem that the traditional method cannot identify all jump layers in the case of multiple channel axes. After experimental verification, the improved ADMM algorithm can completely and accurately identify all jump layers when multiple channel axes exist.

[0164] Figure 2 The recognition results of the jump layer by the traditional optimal segmentation method and the improved ADMM algorithm in the case of binaural axis are shown. Figure 2 In the left figure, the traditional optimal segmentation method can only identify the jump layer corresponding to the upper channel axis. Figure 2 In the right figure, the improved ADMM algorithm can extract two complete jump-layer structures. The comparison between the two clearly shows the superiority of the ADMM algorithm for multi-jump-layer identification.

[0165] The technical solution of the present application is applicable to complex sound field environments under various types of sound velocity profiles. Compared with the original optimal segmentation algorithm, the improved ADMM algorithm searches for the best segmentation points as the upper and lower bounds of the sound velocity jump layer according to the principle of minimizing the sum of squared deviations of the gradients within the sound velocity jump layer, and sets adaptive thresholds through multiple iterative calculations to achieve jump layer recognition. Therefore, compared with the traditional method of setting a fixed threshold to extract the sound velocity jump layer, ADMM is more applicable and has a higher accuracy rate in jump layer recognition.

[0166] Figure 3 The results of the original optimal segmentation method and the improved ADMM method for the recognition of jump layers of six typical sound speed profiles in the North Atlantic are shown. It can be seen that the jump layer distribution identified by the improved ADMM algorithm is more compact, and the positions of the extracted upper and lower boundaries of the jump layer are always located at the inflection point of the sound speed profile curve or the position where the sound speed changes sharply in the local area, so the extraction results are more reasonable.

[0167] The accuracy of layer jump identification has been significantly improved. A large number of test points were randomly selected to compare the layer jump feature parameters extracted by the original algorithm and the improved ADMM algorithm. The test results show that the accuracy of the improved algorithm has been greatly improved.

[0168] Table 1 Comparison of jump layer characteristic parameters extracted by the original algorithm and the improved algorithm

[0169]

[0170]

[0171] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Any changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention should fall within the protection scope of the present invention.

Claims

1. A method for extracting multiple types of sound velocity profile jump layers, characterized in that: The following steps are involved: Step 1, calculating the ocean vertical sound velocity profile data based on the temperature, salinity and corresponding depth data; Step 2, performing data interpolation and resampling on the ocean vertical sound velocity profile data obtained in step 1; Step 3, using the difference method to calculate the sound velocity gradient; Step 4, finding the vocal tract axis based on the sound velocity gradient and the gradient threshold; Step 5, introduce the channel axis as a parameter for determining the jump layer structure, extract the acoustic jump layer, and obtain the upper and lower boundaries of the jump layer; Step 6, calculate the jump layer related attributes and record the jump layer related attribute information to construct the jump layer related data set; In step 1, the ocean is divided into dense grid points, and the sound speed profile of each grid point is calculated to construct the global ocean sound speed field; Substitute the temperature, salinity and corresponding depth data into the empirical formula of sound speed to obtain the vertical sound speed profile data of the ocean; the empirical formula of sound speed is: c=1449.14+c t +c p +c s +c stp ; c t =4.5721t-4.4532×10 -2 t 2 -2.6045×10 -4 t 3 +7.9851×10 -6 t; c p =1.60272×10 -1 p+1.0268×10 -5 p 2 +3.5216×10 -9 p 3 -3.3603×10 -12 p 4 ; c s =1.39799(s-35)+1.69202×10 -3 (s-35) 2 Where c is the speed of sound; t is only related to temperature; c p is only related to static pressure; c s is only related to salinity; c stp is a term related to temperature, salinity and static pressure; t is temperature; s is salinity; p is static pressure; In step 2, the Akima interpolation method is used to interpolate the sound speed data points to obtain a set of continuous sound speed values; Akima interpolation produces smoother interpolation curves when dealing with irregular data point distributions; The interpolated sound velocity data points are approximated by a cubic polynomial to obtain a smoother sound velocity profile curve; The cubic polynomial approximation obtains a continuous curve representation by fitting the data points; The Akima interpolation formula is y=p0+p1(x-x3)+p2(x-x3) 2 +p3(x-x3) 3 ; Use median filtering to filter out the subtle burrs produced by Akima interpolation; Where y represents the sound velocity value of the interpolation point, and x represents the depth value of the interpolation point; let the known data point be (x i ,y i )(i=1,2,3…n), find a smooth curve y=f(x) that satisfies y i =f(x i ); p0, p1, p2, p3 are the required coefficients, which are determined by the slope t of the surrounding points. i The calculation formula is: t3 and t4 represent the derivatives of the third and fourth points respectively, where i = 3, 4; m i is the derivative at point i, replaced by the slope; m i =(y i+1 -y i ) / (x i+1 -x i ); When m i =(y i+1 -y i ) / (x i+1 -x i ) is not true when the denominator is 0, then t i =(m i-1 +m i ) / 2; In step 3, the differential method calculation formula is g = Δs / Δd; Where g is the sound velocity gradient; Δs is the sound velocity difference between adjacent depth points; Δd is the depth difference between adjacent points; In step 4, it is determined whether the maximum gradient of the entire sound velocity profile curve is less than or equal to 0.01 / s, and whether the extreme difference of the gradient is less than or equal to 0.02 / s; If the maximum gradient of the sound velocity profile curve is less than or equal to 0.01 / s, and the extreme difference of the gradient is less than or equal to 0.02 / s, there is no sound channel axis that can produce a jump layer. Otherwise, it is further determined whether it is a negative jump layer. After traversing each point on the sound velocity profile, the sound channel axis is found; In step 5, the channel axis is introduced as a parameter for determining the jump layer structure, and the number of channel axes corresponds to the number of jump layers; The extraction of the acoustic jump layer is to determine the upper and lower boundaries of the jump layer, and determine the water layer between the upper and lower boundaries as the acoustic jump layer; In step 5, the upper limit of the thermocline is calculated by iterating the sound velocity gradient mutation point and the optimal binary division method from the iteration starting point to the sound channel axis or the seabed; The judgment of the lower boundary of the jump layer is divided into two cases: In the case of a single channel axis, the end point of the optimal two-division process, that is, the channel axis, is taken as the lower boundary of the acoustic jump layer; In the case of multiple sound channel axes, the end point of the optimal three-division is taken as the lower boundary of the acoustic jump layer; The optimal binary split algorithm is to transform the ordered sequence x k , at this time k = 2, divided into 2 groups; The mean of the group from i to j can be recorded as Then there is The sum of squares of deviations of this group is defined as If the entire sequence is divided into 2 groups, the total variation under the 2-partition is considered to be the sum of the variations of the 2 groups. V(k|α1,α2…,α k-1 )=V(1,α1)+V(α1+1,α2)+…+V(α k-1 +1,n); Therefore, V n The smallest grouping is the optimal binary split; The optimal three-partition algorithm is to transform the ordered sequence x k , at this time k = 3, divided into 3 groups; The mean of the group from i to j can be recorded as Then there is The sum of squares of deviations of this group is defined as If the entire sequence is divided into 3 groups, the total variation under the 3-partition is considered to be the sum of the variations of the 3 groups. V(k|α1,α2…,α k -1)=V(1,α1)+V(α1+1,α2)+…+V(α k-1 +1,n) Therefore, V n The smallest grouping is the optimal three-part split.

2. The method for extracting multiple types of sound velocity profile jump layers according to claim 1, characterized in that: After obtaining the upper and lower bounds of the jump layer, the characteristic quantities of the jump layer are calculated. The characteristic quantities of the jump layer are divided into jump layer depth dh, thickness th and intensity gy. The calculation of the characteristic quantities of the jump layer is th=Z end -WITH start ; In the formula, Z start Indicates the upper boundary depth of the acoustic jump layer, Z end represents the lower boundary depth of the acoustic jump layer, C start represents the speed of sound at the upper boundary of the sonic jump layer, C end Represents the sound velocity at the lower boundary of the sonic jump layer.

3. The method for extracting multiple types of sound velocity profile jump layers according to claim 1, characterized in that: Step 5 includes the following steps: Step 5-1: Iteration starts from the sound velocity gradient mutation point H, and the iteration end point is the minimum value among the sound channel axis, the seabed, and the depth value of 400 meters; In the case of negative jump layers, the sound velocity gradient mutation point H is taken as the depth where the gradient is maximum; in other cases, the point where the sound velocity gradient changes from positive to negative or the gradient is minimum is considered to be the sound velocity gradient mutation point H; Step 5-2: Starting from the sound velocity gradient mutation point H, an iteration is performed every n meters; Each iteration calculates the best binary segmentation result from the sound velocity gradient mutation point H to the current position, and records the segmentation result; In the case of negative jump layer, the value of n in the sound velocity profile is 3, and the value of k is 4; in the case of the sound velocity profile without the formation of the sound channel axis, the value of n is 5, and the value of k is 3; in the case of the sound velocity profile with the formation of the sound channel axis, the value of n is 10, and the value of k is 5; Step 5-3: After the iteration is completed, if the difference between the segmentation results of two adjacent points is less than n / 2, it is considered to be a continuous value; When the continuous value exceeds k, the average value of the continuous value is calculated as the upper bound of the jump layer; Step 5-4: For the sound velocity profile with the vocal tract axis, the optimal binary segmentation method is used for segmentation, and the starting point of the segmentation is set as the upper boundary of the jump layer, and the end point is set as the vocal tract axis to obtain the lower boundary of the jump layer; Step 5-5: Determine whether there are multiple channel axes; if there are multiple channel axes and the minimum gradient requirement is met, use the optimal three-segmentation method to obtain the jump layers between the channel axes; The deep sea sound velocity profile structure is divided into the mixing layer, the main jump layer, the acoustic channel layer and the deep sea isotherm layer.

Images