A method for establishing an accurate gun calibration positioning model

Abstract

The application discloses a kind of accurate artillery detection positioning model establishment methods, comprising the following steps: step S1: obtaining parameter;Step S2: according to the parameter obtained and using polynomial coefficient, the relationship of the horizontal offset x and vertical offset y of each detection point is characterized;Step S3: using polynomial coefficient is characterized to seabed interface;Step S4: according to the characterization information of seabed interface, obtain the feature for network training;Step S5: corresponding label is calculated by feature;Step S6: according to polynomial coefficient vector, direct wave travel and seabed reflection wave travel, it is trained network, obtains the neural network function of well-trained;Step S7: according to the neural network function of well-trained, reconstructs deep drag seismic shot detection positioning;Solve the problem of low efficiency of deep drag seismic shot detection positioning calculation.

Description

TECHNICAL FIELD

[0001] The application relates to a method for establishing a shot-receiver positioning model. BACKGROUND

[0002] Deep-towed multi-channel seismic exploration technology is a kind of multi-channel seismic exploration technology in which a seismic source and a receiving cable are deployed at a certain height near the seabed. Due to the deployment at a certain height near the seabed, the lateral and vertical resolution of offshore seismic exploration is greatly improved. Since the 1980s, with the urgent needs of offshore energy exploration, seabed geological disasters and the like, the deep-towed seismic technology has developed rapidly, and can provide good data for the exploration of seabed natural gas hydrate, landslide disaster evaluation, seabed tectonic activity and the like. With the development of the instrument industry and the progress of data processing technology, the deep-towed multi-channel seismic exploration technology will certainly play a greater role in future geophysical exploration.

[0003] Precise shot-receiver positioning plays a crucial role in marine seismic data processing, and greatly influences the resolution and signal-to-noise ratio of a seismic profile. However, conventional offshore multi-channel seismic exploration has acoustic water birds, positioning tail marks and the like to constrain the relative positions of the seismic source and the geophone, while the deep-towed seismic technology has a very high requirement for the precision of shot-receiver positioning. The current deep-towed seismic shot-receiver positioning method mostly uses a swarm intelligence algorithm for optimization, which needs to optimize multiple times in multiple shot gathers, and the time cost is significantly increased with the increase of the number of shots.

[0004] In view of the low calculation efficiency of the deep-towed seismic shot-receiver positioning, a method for establishing a precise shot-receiver positioning model is provided, which breaks through the limitation of the conventional calculation method of calculating the positioning of each shot gather.

[0005] Technical solutions of prior art

[0006] The existing deep-towed seismic shot-receiver positioning method is to convert the shot-receiver positioning problem into an optimization problem with the elevation angle of each channel as the optimization target, in a plane rectangular coordinate system, according to the fixed relationship between the offset distance of the seismic source and the geophone and the inter-channel distance, and the constraint condition when the direct wave and the seabed reflected wave arrive.

[0007] In order to describe the position of the geophone relative to the seismic source, a plane rectangular coordinate system is established with the seismic source point as the coordinate origin (x0,y0), it is assumed that the seismic source to geophone 1 is a straight line connection, the distance is the minimum offset distance L, the geophones are also connected in a straight line, the distance is the inter-channel distance D, and the number of seismic channels is j. The coordinates of the geophones are determined, and the coordinates (x1,y1) of the geophone 1 are (Lcosθ1,Lsinθ1), and the horizontal and vertical coordinates of the i-th geophone (i≥2) are: Figure 1

[0008] ​

[0009] where θ1 is the angle between the connecting line of the source and the receiver 1 and the horizontal direction, θ i is the angle between the connecting line of the i-th receiver and the i-1-th receiver and the horizontal direction, and the vector θ = [θ1, θ2,..., θ j-1 , θ j ] T is the vector of the inclination angle of each channel, and each θ corresponds to a streamer posture (shot-receiver positioning)

[0010] The arrival times of the direct wave and the sea bottom reflected wave reflect the relationship between the sea bottom interface and the shot-receiver positioning, and thus serve as the constraint condition of the problem. Since the depth of the sea bottom interface is known, different inclination angles correspond to different arrival times of the direct wave and the sea bottom reflected wave. The root mean square error (RMSE) of the arrival times of the direct wave and the sea bottom reflected wave is taken as the objective function of the optimization algorithm, and a more accurate shot-receiver positioning is calculated.

[0011] Disadvantages of the prior art

[0012] First, as shown in FIG. 1, the shot-receiver system model assumes that the channels are straight lines, while the actual situation should be an arc curve, which is determined by the inherent nature of the receiving cable and will cause cumulative positioning errors as the number of channels increases. Figure 1 Secondly, due to the movement and retraction of the deep-towed source and the receiver array, the shot-receiver positioning is different at each time of exciting the seismic wave, and thus the shot-receiver positioning needs to be calculated each time, which greatly increases the calculation time cost.

[0013] Summary of the application

[0014] To solve the problems in the prior art, the application provides a method for establishing an accurate shot-receiver positioning model, which solves the problem of low calculation efficiency of deep-towed seismic shot-receiver positioning.

[0015] The specific technical solutions are as follows:

[0016] A method for establishing an accurate shot-receiver positioning model, comprising the following steps:

[0017] Step S1: obtaining parameters;

[0018] Step S2: characterizing the relationship between the horizontal offset x and the vertical offset y of each receiver point according to the obtained parameters and using polynomial coefficients;

[0019] Step S3: characterizing the sea bottom interface using polynomial coefficients;

[0020] Step S4: obtaining features for network training according to the characterization information of the sea bottom interface; ​

[0021] Step S5: calculating the corresponding label by feature;

[0022] Step S6: training the network according to the polynomial coefficient vector, direct wave travel and sea bottom reflected wave travel, to obtain a trained neural network function;

[0023] Step S7: reconstructing the deep-towed seismic shot location according to the trained neural network function.

[0024] Preferably, the parameters include the number of source excitations N, the depth measurement H of the source excitation time from the sea level D and the height measurement H of the source from the sea bottom interface, the depth observation H of the sea bottom interface from the sea level at the source excitation time D +H, the direct wave and sea bottom reflected wave travel time observation T recorded by each geophone D and T F , and the near-bottom water sound speed v.

[0025] Preferably, step S2 includes the following sub-steps:

[0026] Sub-step S21: obtaining a polynomial matrix P(x), which is expressed as:

[0027] P(x) = R T X

[0028] wherein: R = [r n r n-1 ...r2 r1] T is the polynomial coefficient vector, X = [x n x n-1 ...x 2 x] T is the power vector of the variable;

[0029] Sub-step S22: the derivative of the polynomial is P'(x), and the coordinates of the geophone 1 can be expressed as (x1, P(x1)), wherein x1 satisfies:

[0030]

[0031] The coordinates of the i-th geophone (i ≥ 2) are (x i , P(x i )), wherein each x i satisfies:

[0032]

[0033] Thus, the horizontal offset x and the vertical offset y of each geophone are represented by the coefficient vector R.

[0034] Preferably, step S3 is specifically:

[0035] The depth of the sea floor interface is fitted by a polynomial F(x) of degree k, which can be expressed as

[0036] F(x) = x k x k + s k-1 x k-1 +... + s2x 2 + s1x + s0

[0037] The matrix representation of the polynomial F(x) is:

[0038] F(x) = S T X

[0039] wherein S = [s k s k-1 ... s1 s0] T is a polynomial coefficient vector, X = [x k x k-1 ... x1] T is a power vector of variables, the number of polynomial coefficients of the depth of the sea floor interface is k+1, and the coefficient vector S is used to represent the depth of the sea floor interface.

[0040] Preferably, step S4 comprises the following sub-steps:

[0041] Sub-step S41: setting a reasonable interval of the height of the seismic source from the bottom;

[0042] Sub-step S42: setting a reasonable interval of the vertical offset of the seismic source from the farthest offset distance;

[0043] Sub-step S43: setting a reasonable interval of the slope angle of the sea floor;

[0044] Sub-step S44: obtaining a plurality of sets of polynomial coefficients R and coefficients S that meet the conditions by means of target optimization.

[0045] Preferably, step S5 comprises the following sub-steps:

[0046] Sub-step S51: calculating the direct wave travel time T D = [d1 d2...d j-1 d j ] T according to the offset x and the offset y represented by the plurality of sets of polynomial coefficients R;

[0047] Sub-step S52: calculating the sea floor reflected wave travel time T F = [f1 f2...f j-1 fj ] T .

[0048] Preferably, step S6 is specifically:

[0049] characterized by: I = [R, S], and the label is: O = [T D ,T F ], and the mapping relationship is:

[0050] O = f θ (I)

[0051] wherein f θ is a trained neural network function.

[0052] Preferably, step S7 includes the following sub-steps:

[0053] Sub-step S71: according to step S3, the polynomial of H D +H at N excitation moments is represented and the polynomial coefficients S0 are extracted;

[0054] Sub-step S72: combined with T D and T F obtained in step S1, the corresponding coefficient R is calculated by inputting the network, and further combined with step S2, step S5 obtains the shot location.

[0055] The beneficial effects of the method for establishing the accurate shot location model are as follows:

[0056] 1. The present application uses polynomial coefficients to represent the relative positions of the seismic source and the geophone in deep-towed seismic exploration, and based on the parameterized expression of the polynomial, it can effectively solve the problems of large cumulative error and low calculation efficiency in existing methods.

[0057] 2. The present application improves the calculation efficiency: the traditional method needs to perform shot location one by one at each excitation of seismic wave, while the present application can directly derive the accurate shot location after obtaining the arrival time of direct wave and sea bottom reflected wave by using neural network to train a large number of polynomial coefficients, which greatly reduces the calculation time.

[0058] 3. The present application reduces the cumulative error: the existing technology assumes that the connecting line between each geophone is a straight line, ignoring the influence of the curvature of the streamer posture. The present application uses a polynomial to fit the spatial position of the geophone array, which is more in line with the actual situation and reduces the cumulative positioning error.

[0059] 4. The model of the present application has strong expandability: by introducing polynomial coefficients to represent shot location, the method of the present application has good expandability. Under different conditions, only the order of the polynomial or the range of the training data needs to be adjusted, so that different source excitation depths, geophone distributions and sea bottom interface shapes can be adapted.

[0060] 5. The application is not strongly dependent on the sound speed of the near-sea bottom water: the prior art requires high accuracy of the sound speed of the water, and when there is a slight difference between the set sound speed and the actual sound speed, it still affects the position of the overall multiple detection points. In the present application, since the sound speed of the water is not directly reflected in the output and input of the neural network, it has good fault tolerance for the sound speed of the water. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 A schematic diagram of the conventional shot location positioning method of the application.

[0062] Figure 2 A schematic diagram of the shot location positioning method of the application.

[0063] Figure 3 A technical roadmap of the scheme of the application.

[0064] Figure 4 A schematic diagram of the propagation of the sea bottom reflected wave in a single sample of the application.

[0065] Figure 5 A relative position diagram of the shot-detection point of multiple samples of the application.

[0066] Figure 6 A positioning error distribution diagram of the network prediction of the application. DETAILED DESCRIPTION

[0067] The embodiments of the application will be described in detail below with reference to the accompanying drawings.

[0068] The specific embodiments of the application will be described below to facilitate understanding of the application by those skilled in the art, but it should be clear that the application is not limited to the scope of the specific embodiments, and for those skilled in the art, as long as various changes are within the spirit and scope of the application defined and determined by the appended claims, these changes are obvious, and all applications utilizing the concept of the application are within the scope of protection.

[0069] A method for establishing an accurate shot location positioning model, comprising the following steps:

[0070] Step S1: obtaining parameters;

[0071] Step S2: according to the obtained parameters and using polynomial coefficients, the relationship between the horizontal offset x and the vertical offset y of each detection point is characterized;

[0072] Step S3: using polynomial coefficients to characterize the sea bottom interface;

[0073] Step S4: according to the characterization information of the sea bottom interface, obtaining features for network training;

[0074] Step S5: calculating the corresponding label by feature;

[0075] Step S6: training the network according to the polynomial coefficient vector, direct wave travel and sea bottom reflected wave travel, to obtain a trained neural network function;

[0076] Step S7: reconstructing the deep-towed seismic shot location according to the trained neural network function.

[0077] The parameters of the embodiment include the number of source excitations N, the depth measurement H of the source excitation time from the sea level D and the height measurement H of the source from the sea bottom interface, the depth observation H of the sea bottom interface from the sea level at the source excitation time D +H, the direct wave and sea bottom reflected wave travel time observation T recorded by each geophone D and T F , and the near-bottom water sound speed v.

[0078] Step S2 of the embodiment includes the following sub-steps:

[0079] Sub-step S21: obtaining a polynomial matrix P(x), which is expressed as:

[0080] P(x) = R T X

[0081] wherein: R = [r n r n-1 ...r2 r1] T is the polynomial coefficient vector, X = [x n x n-1 ...x 2 x] T is the power vector of the variable;

[0082] Sub-step S22: the derivative of the polynomial is P'(x), and the coordinates of the geophone 1 can be expressed as (x1, P(x1)), wherein x1 satisfies:

[0083]

[0084] The coordinates of the i-th geophone (i ≥ 2) are (x i , P(x i )), wherein each x i satisfies:

[0085]

[0086] Thus, the horizontal offset x and the vertical offset y of each geophone are represented by the coefficient vector R.

[0087] Step S3 of the embodiment is specifically:

[0088] The sea bottom interface depth is fitted by a k-th order polynomial F(x), which can be expressed as

[0089] F(x) = s k x k +s k-1 x k-1 +...+s2x 2 +s1x+s0

[0090] The matrix representation of the polynomial F(x) is:

[0091] F(x) = S T X

[0092] wherein s = [s k s k-1 ... s1 s0] T is a polynomial coefficient vector, X = [x k x k-1 ... x1] T is a power vector of variables, the number of polynomial coefficients of the sea bottom interface depth is k+1, and the coefficient vector S is used to represent the sea bottom interface depth.

[0093] Step S4 of the embodiment includes the following sub-steps:

[0094] Sub-step S41: setting a reasonable interval of the height of the seismic source from the bottom;

[0095] Sub-step S42: setting a reasonable interval of the vertical offset of the seismic source from the farthest offset distance;

[0096] Sub-step S43: setting a reasonable interval of the sea bed slope angle;

[0097] Sub-step S44: obtaining a plurality of sets of polynomial coefficients R and coefficients S meeting the conditions by means of target optimization.

[0098] Step S5 of the embodiment includes the following sub-steps:

[0099] Sub-step S51: calculating the direct wave travel time T D =[d1 d2...d j-1 d j ] T according to the offset x and offset y represented by the plurality of sets of polynomial coefficients R;

[0100] Sub-step S52: calculating the sea bottom reflected wave travel time T F =[f1 f2...f j-1 f ] according to the plurality of sets of polynomial coefficients R and coefficients S, in combination with the near-sea-bottom sound speed v, by using Fermat's principle.f j ] T .

[0101] Step S6 of the embodiment is specifically:

[0102] The feature is: I = [R, S], the label is: O = [T D ,T F ], and the mapping relationship is:

[0103] O = f θ (I)

[0104] Wherein, f θ is a trained neural network function.

[0105] Step S7 of the embodiment includes the following sub-steps:

[0106] Sub-step S71: According to step S3, the polynomial of H D +H at N excitation moments is represented, and the polynomial coefficients S0 are extracted;

[0107] Sub-step S72: In combination with T D and T F obtained in step S1, the corresponding coefficient R is calculated by inputting the network, and further in combination with step S2 and step S5, the shot location is obtained.

[0108] When the embodiment is implemented,

[0109] (1) Obtain the number of source excitations N; obtain the depth measurement value H D of the source excitation moment from the sea level and the height measurement value H of the source from the sea bottom interface; obtain the depth observation value H D +H of the sea bottom interface from the sea level at the source excitation moment; obtain the direct wave and sea bottom reflection wave travel time observation values T D and T F recorded by each receiver; obtain the near-bottom water sound speed v;

[0110] (2) Use the polynomial coefficients to represent the relationship between the horizontal offset x and the vertical offset y of each receiver

[0111] As shown in FIG. Figure 2 , the source and multiple receivers can be represented in a polynomial P(x) of n times, and since the polynomial passes through the coordinate origin, it can be represented as:

[0112] P(x) = r n x n +r n-1 x n-1 +...+r2x 2 +r1x (2)

[0113] The matrix representation of the polynomial P(x) is:

[0114] P(x) = R T X (3)

[0115] wherein R = [r n r n-1 ... r2 r1] T is the polynomial coefficient vector, X = [x n x n-1 ... x 2 x] T is the power vector of the variable. It can be seen that the number of polynomial coefficients of the model is n.

[0116] The derivative of the polynomial is P'(x). According to the arc length formula, the coordinates of the detector 1 can be represented as (x1, P(x1)), wherein x1 satisfies:

[0117]

[0118] By analogy, the coordinates of the i-th detector (i ≥ 2) are (x i , P(x i )), wherein each x i satisfies:

[0119]

[0120] Thus, the horizontal offset x and the vertical offset y of each detector are represented using the coefficient vector R.

[0121] (3) Characterizing the sea bottom interface using the polynomial coefficients

[0122] Suppose the depth of the sea bottom interface is fitted using a polynomial F(x) of degree k, which can be represented as:

[0123] F(x) = s k x k + s k-1 x k-1 +... + s2x 2 + s1x + s0 (6)

[0124] The matrix representation of the polynomial F(x) is:

[0125] F(x) = S T X (7)

[0126] wherein S = [s k s k-1 ... s1 s0] T is the polynomial coefficient vector, X = [x kx k-1 ... x 1] T is the power vector of variable. Thus, the number of polynomial coefficients of the seafloor interface depth is k+1.

[0127] Thus, the seafloor interface depth is characterized by using the coefficient vector S.

[0128] (4) Obtain features for network training

[0129] Set a reasonable interval for the height of the source from the bottom; set a reasonable interval for the vertical offset of the source and the farthest offset; set a reasonable interval for the seabed slope angle; obtain multiple sets of polynomial coefficients R and coefficients S that meet the conditions through target optimization,

[0130] (5) Calculate corresponding labels through features

[0131] According to the offset x and offset y represented by multiple sets of polynomial coefficients R, calculate the direct wave travel time T D =[d1 d2... d j-1 d j ] T ; according to multiple sets of polynomial coefficients R and coefficients S, combined with the near-seafloor sound speed v, calculate the seafloor reflected wave travel time T F =[f1 f2... f j-1 f j ] T .

[0132] (6) Train the network

[0133] The feature is I=[R,S], the label is O=[T D ,T F ], and the mapping relationship is:

[0134] O=f θ (I) (8)

[0135] Where f θ is the trained neural network function.

[0136] (7) Deep-towed seismic shot location reconstruction

[0137] According to step S3, the polynomial of H D +H of N shooting times is represented, and the polynomial coefficient S0 is extracted, combined with T D and T F obtained in step S1, the corresponding coefficient R is calculated by inputting the network, and further combined with step S2 and step S5, the shot location is obtained.

[0138] In this example, the deep-towed seismic system is equipped with altimeter and depth meter, and the height of the towed body from the bottom and the depth of the sink are recorded continuously during operation. The system receives a 48-channel cable with a channel spacing of 3.125 m and a minimum offset of 12.5 m.

[0139] Figure 3 As the technical roadmap of the present solution, as can be known from the drawings, the deep-towed seismic shot location based on neural network consists of two parts, i.e. network training and network prediction.

[0140] Before network training, first, a simulation data set is constructed (for example, fitting the source and receiver points with a fourth-order polynomial, and fitting the sea bottom interface height observation value with a third-order polynomial), and a single sample of the data set contains Figure 4 ):

[0141] (1) 4 polynomial coefficients for fitting the source and receiver points;

[0142] (2) 4 polynomial coefficients for fitting the sea bottom interface height;

[0143] (3) 48 direct wave travel times corresponding to the sample;

[0144] (4) 48 sea bottom reflected wave travel times corresponding to the sample.

[0145] In combination with the relative positions of the source and receiver points and the depth variation range of the sea bottom interface during actual operation, the range of the polynomial coefficients is further constrained, so as to construct a more suitable data set, for example, the height difference between the source and the farthest offset receiver point should not be too large, and the pitch angle of a single receiver point should not be too large.

[0146] Figure 5 is a group of single sample source-receiver point relative positions in which the height difference between the source and the farthest offset receiver point is limited within ±50 m, and the pitch angle of a single receiver point is limited within ±30°. Similarly, the depth variation of the sea bottom interface should not be too large, and similar constraint means are adopted.

[0147] 100,000 samples are generated in the above manner, and the 48 direct wave travel times corresponding to the polynomial coefficients of the fitted source and receiver points and the polynomial coefficients of the fitted sea bottom interface height in a single sample are calculated, and the 48 sea bottom reflected wave travel times are calculated by Fermat's principle.

[0148] 80% of the data set is used as the training set, 10% as the test set, and 10% as the validation set, a single-layer feedforward neural network with 20 hidden layer neurons is input, the mean square error is used as the loss function, and 100 training rounds are set. After training, the performance of the neural network model on the validation set reaches a satisfactory accuracy, and the mean square error (MSE) is 3.03 x 10 -7 .

[0149] Using the trained network, the receiver positioning is predicted when the sea bottom interface, direct wave and sea bottom reflected wave travel. The positioning error is represented by the difference between the predicted value and the theoretical value of the receiver positioning, 100 new samples are generated for prediction, and the results show that the average error of each channel positioning is less than 0.162m, and the error distribution is as shown in Figure 6 (a). When there is an error in picking up the direct wave or the sea bottom reflected wave travel (assuming that there is a Gaussian distribution of picking error with an expected value of 2ms and a variance of 1), the overall positioning error is less than 0.3m, and the error distribution is as shown in Figure 6 (b).

[0150] The technical key points and points to be protected of the present application are the method of describing the shot-receiver positioning using polynomials and the shot-receiver positioning calculation method taking the polynomial coefficients as the input and output of the neural network.

Claims

1. A method for establishing a precise gun calibration positioning model, characterized in that, The method comprises the following steps: Step S1: obtaining parameters; Step S2: representing the relationship between the horizontal offset x and the vertical offset y of each receiver according to the obtained parameters and using polynomial coefficients; Step S3: representing the sea floor interface using polynomial coefficients; Step S4: obtaining features for network training according to the representation information of the sea floor interface; Step S5: calculating corresponding labels through the features; Step S6: training the network according to the polynomial coefficient vector, the direct wave travel and the sea floor reflected wave travel, to obtain a trained neural network function; Step S7: reconstructing the deep-towed seismic shot-receiver positioning according to the trained neural network function; The parameters include the number of shots N, the depth of the shot from the sea level H D and the height of the shot from the sea bottom interface H, the depth of the sea bottom interface from the sea level H D + H, the travel time observation value T of the direct wave and the sea bottom reflected wave recorded by each geophone D and T F , the near-bottom sound speed v; The step S2 comprises the following sub-steps: Sub-step S21: obtaining a polynomial matrix P(x), which is expressed as: P(x) = R T X Where R = [r n r n-1 … r2 r1] T It is the coefficient vector of the polynomial, X = [x n x n-1 ... x 2 x] T It is a vector of powers of variables; Sub-step S22: the derivative of the polynomial is P'(x), and the coordinates of the receiver 1 can be expressed as (x1, P(x1)) according to the arc length formula, wherein x1 satisfies: The coordinates of the ith detector (i≥2) are (x i , P(x i )), where each x i satisfies: The coefficient vector R is used to represent the horizontal offset x and the vertical offset y of each receiver; The step S3 is specifically: The depth of the sea floor interface is fitted using a polynomial F(x) of a certain order, and the polynomial can be expressed as F(x) = s k x k +s k-1 x k-1 +...+s2x 2 +s1x+s0 The matrix representation of the polynomial F(x) is: F(x) = S T X where: S = [s k s k-1 ...s1s0] T is a polynomial coefficient vector, X = [x k x k-1 ...x1] T is a power vector of variables, the number of polynomial coefficients of the seafloor interface depth is k+1, and the seafloor interface depth is represented by using the coefficient vector S; The step S4 comprises the following sub-steps: Sub-step S41: setting a reasonable interval for the height of the source from the bottom; Sub-step S42: setting a reasonable interval for the vertical offset of the source and the farthest offset distance; Sub-step S43: setting a reasonable interval for the sea floor slope angle; Sub-step S44: obtaining a plurality of sets of polynomial coefficients R and coefficients S that meet the conditions through target optimization; The step S5 comprises the following sub-steps: Sub-step S51: Calculate the direct wave travel time T according to the offset x and the offset y represented by the multiple sets of polynomial coefficients R D = [d1 d2...d j-1 d j ] T ; Sub-step S52: According to the multiple sets of polynomial coefficients R and coefficients S, in combination with the near-bottom water acoustic speed v, the bottom reflection wave travel time T is calculated by using Fermat's principle F = [f1f2...f j-1 f j ] T ; The step S6 is specifically: The feature is: I = [R, S], the label is: O = [T D ,T F ], the mapping relationship is: O = f θ (I) where f θ is a trained neural network function; The step S7 comprises the following sub-steps: Sub-step S71 : from the polynomial of H D + H and extracting the polynomial coefficients S0; Sub-step S72: combine T D and T F , input network to calculate the corresponding coefficient R, further combined with step S2, step S5 to get the shot positioning.

Images