Underwater motion feature recognition method based on multi-source data

Abstract

The invention belongs to the field of underwater target navigation and positioning, and discloses an underwater motion feature recognition method based on multi-source data, which comprises the following specific steps of: 1) completing time-space registration of segmented observation data of a navigation body, and establishing a uniform position and speed parameter sample sequence; 2) defining a spline function, and establishing a function expression of underwater motion parameters of the navigation body; 3) identifying a parameter sequence function coefficient according to the unified sample sequence; and 4) reconstructing a parameter sequence, and identifying motion parameters of the feature points. According to the method, integrated and functional expression of multi-source observation data tracks is achieved, the continuous, smooth and robust fitting problem under the condition of high conformity of multi-source data is solved, the Runge phenomenon can be effectively avoided at the edge, the adaptability to the complex navigation process is good, the motion characteristic reduction degree is high, and the method is suitable for large-scale popularization and application. High-quality parameter sequences and feature point information can be provided for high-dynamic underwater navigation body testing, and then key data are provided for navigation body underwater motion performance analysis and evaluation.

Description

Technical Field

[0001] The present invention belongs to the field of underwater target navigation and positioning, and in particular relates to an underwater motion feature recognition method based on multi-source data, which is a method for recognizing motion parameters of underwater navigation feature points according to multi-source observation data. Background Art

[0002] During the underwater navigation test of the vehicle, it is necessary to obtain its trajectory parameters through external observation, and then analyze the motion characteristics of its time or space feature points. External measurement usually uses hydroacoustic positioning technology. When the navigation range is large or the track is more complex, different technical means are often used for segmented observation to obtain multi-source heterogeneous observation data. Due to differences in startup time, time accuracy, measurement performance, external environment, etc. between different sensors and different measurement equipment, the obtained observation data often have varying degrees of systematic errors, resulting in multi-source data being out of sync in time, mismatched in space, and misaligned in value, causing data fuzziness near the feature points, making it difficult to accurately determine their arrival time and corresponding motion parameters.

[0003] Although a series of methods and technologies for processing, associating, and fusing measurement data have been developed in the fields of transportation, aircraft measurement and control, etc., their application environment and data characteristics are quite different from the underwater navigation process, and the existing technology is difficult to be directly transplanted to the underwater field. The difficulties in solving this problem are: first, the temporal and spatial reference accuracy of multi-source heterogeneous data is not uniform, and the acquired data segments need to be registered in time and space; second, the function expression is inaccurate, and the trajectory generated by weighted fusion of data is prone to "jump points", and the continuity of its motion parameters is difficult to guarantee, while polynomial fitting has a significant Runge phenomenon, resulting in large fluctuations in the edge area; third, according to the unified navigation trajectory, feature point parameters can be extracted according to time or space criteria, but the accuracy of feature points is difficult to evaluate when there is data ambiguity. These problems restrict the effectiveness of underwater motion characteristic recognition. It is necessary to find a functional expression method based on the characteristics of underwater navigation data to integrate data of different types and different precisions to form a complete, accurate, and reliable motion parameter estimation result, thereby providing a basis for the determination and recognition of feature points. Summary of the invention

[0004] To solve the above problems, the present invention proposes a method for identifying motion parameters of underwater navigation feature points based on multi-source observation data. A set of spline functions and coefficients are used to describe the motion parameters. Through the discrete multi-source data identification coefficients, a unique and continuous navigation trajectory is fitted, and then the time or space criteria are used to realize the identification and extraction of feature points. This method has the characteristics of smooth, continuous and consistent expression of complex tracks, high data conformity and motion characteristic restoration, and has practical value for the analysis and evaluation of underwater motion performance of navigation bodies.

[0005] The technical solution of the present invention is:

[0006] A method for underwater motion feature recognition based on multi-source data, firstly, completes the spatiotemporal registration of the segmented observation data of the vehicle, and establishes a unified position and velocity parameter sample sequence; secondly, defines the spline function, and establishes the function expression of the underwater motion parameters of the vehicle; thirdly, identifies the parameter sequence function coefficients according to the unified sample sequence; finally, reconstructs the parameter sequence and identifies the motion parameters of the feature points.

[0007] The specific steps are as follows:

[0008] Step 1: Complete the spatiotemporal registration of the segmented observation data of the navigation body and establish a unified position and velocity parameter sample sequence.

[0009] For the N segmented observed navigation trajectories, the position and velocity parameter sequences of each segment are obtained and standardized to form a unified sample sequence, including:

[0010] (1.1) Interpolation is to equidistant time nodes;

[0011] (1.2) Through coordinate transformation, it is unified into the measurement coordinate system (the zero elevation point of the starting point is taken as the origin O, the OX axis points to the east, the OY axis points to the north, and the OZ axis is vertically upward).

[0012] (1.3) Combine the N segmented observation data segments into a unified sample sequence, which is in the form of:

[0013]

[0014] Among them, X traj , Y traj , Z traj is the combined position sample sequence, X i , Y i , Z i Corresponding to t i The sample value at the time, i = 1, 2, ..., N; is the combined velocity sample sequence, Corresponding to t i The sample value at the moment, i=1,2,…,N.

[0015] Step 2: Define the spline function and establish the function expression of the underwater motion parameters of the navigation body.

[0016] (2.1) Define the spline function.

[0017] In the observation interval, the time point Set as internal node, a total of M in Outside the observation interval, the time point Set as external node, a total of M out The number of external nodes distributed on the left and right sides of the observation interval is equal; the intervals between each time node are equal, and the node distance is Δt * , the total number of nodes is M = M in +M out .

[0018] At each node, a set of spline functions with independent variables as time t is set, where the spline function expression of node j is:

[0019]

[0020] Among them, B (j) (t) represents the spline function of node j, Represents the derivative of the spline function at node j; j = 1, 2, …, M; σ is the mean square error parameter.

[0021] (2.2) The underwater motion parameters are expressed by spline functions.

[0022] According to the above spline function, the position sequence x(t), y(t), z(t) and velocity sequence of underwater motion It can be expressed as:

[0023]

[0024] In the formula, and is the coefficient to be estimated.

[0025] Step 3: Identify parameter sequence function coefficients based on the unified sample sequence.

[0026] (3.1) Establish the matrix relationship between the parameter sequence function and the unified sample sequence.

[0027] The corresponding relationship between the parameter sequence function in the X direction, Y direction, and Z direction and the unified sample sequence is expressed by the following expression:

[0028] TC X =F X

[0029] TC Y =F Y

[0030] TC Z =F Z

[0031] in:

[0032]

[0033] (3.2) Identification parameter sequence function coefficients

[0034] The linear least squares regression method is used to obtain the estimated values ​​of the coefficients of the function expression in the three directions of X, Y, and Z:

[0035] C X =(T T T) -1 T T F X

[0036] C Y =(T T T) -1 T T F Y

[0037] C Z =(T T T) -1 T T F Z

[0038] Step 4: Reconstruct the parameter sequence and identify the motion parameters of feature points.

[0039] (4.1) Reconstruct the position and velocity parameters of the vehicle based on the parameter sequence function.

[0040] According to the function expression described in (2.2) and the coefficients obtained in (3.2), the functional relationship between position and velocity over time is established, and t i The estimated value of the location parameter at time (i=1,2,…,N) and the estimated values ​​of the velocity parameters

[0041]

[0042] In this way, a complete trajectory sequence based on the estimated values ​​is reconstructed.

[0043] (4.2) Identify the motion parameters of feature points.

[0044] Based on the position and velocity parameter sequence obtained by (4.1), according to the known conditions given by the test conditions and the position and velocity curve shapes, the time when the characteristic points appear is determined, and then the parameter estimation value at the corresponding time is obtained by interpolation, that is, for time t K , position eigenvalue The velocity characteristic value is

[0045] The beneficial effects of the present invention are as follows: the method is a functional data fusion method based on multi-source observation data, which makes full use of the symmetry, positive definiteness, second-order smoothness and other characteristics of spline functions, realizes the integration and functional expression of the tracks of multi-source observation data, solves the problems of continuous, smooth and robust fitting under the conditions of high conformity of multi-source data, and can effectively avoid the Runge phenomenon at the edges. It has good adaptability to complex navigation processes and high degree of motion characteristic restoration. It can provide high-quality parameter sequences and feature point information for high-dynamic underwater vehicle testing, and thus provide key data for the analysis and evaluation of the underwater motion performance of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is the basic process of the method proposed in the present invention.

[0047] Figure 2 (a) and (b) are respectively the position and velocity parameter sequence curves of the three observation segments described in the embodiment, wherein the horizontal axis is the data sequence number sequence and the vertical axis is the normalized parameter value.

[0048] FIG. 3( a ) and FIG. 3( b ) are respectively diagrams showing the coefficient variation of the parameter sequence function described in the embodiment, wherein the horizontal axis is the node sequence and the vertical axis is the coefficient value.

[0049] Figure 4 (a) and (b) are respectively the position and velocity parameter sequence curves calculated by the parameter sequence function as described in the embodiment, wherein the horizontal axis is the data sequence number sequence and the vertical axis is the normalized parameter value. The calculation results are compared with the original position and velocity data of the three observation segments.

[0050] Figure 5 (a) and (b) respectively show the position and speed sequence segment division based on the feature point recognition results described in the embodiment, where T K1 and T K2 is the characteristic point time, and the vertical dotted lines indicate the division intervals of the three segments; (c) shows the three-dimensional track diagram of the calculation results, where the characteristic point T K1 and T K2 The three segments are distinguished by curves of different colors. DETAILED DESCRIPTION

[0051] The specific implementation of the method proposed in the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0052] The main steps of the present invention are as follows Figure 1 The following is a set of underwater vehicle test results as an example. This test contains observation data from three intervals, and its goal is to obtain the characteristic time point T K1 and T K2 The speed parameter VK1 and V K2 , and obtain a unified segmented trajectory according to the characteristic time points to avoid data ambiguity in overlapping observation segments. The rationality and effectiveness of the proposed method are demonstrated by processing and calculating according to the above steps.

[0053] (1) During the sea test of the navigation body, different observation methods were used to measure its navigation process in sections, and the position and velocity parameter sequences of the three observation sections were obtained (the data were normalized), see Figure 2 (a) and (b).

[0054] (2) The three sets of data are interpolated into equidistant time nodes respectively, and unified into the measurement coordinate system through coordinate transformation (the elevation zero point of the starting point is taken as the origin O, the OX axis points to the east, the OY axis points to the north, and the OZ axis points vertically upward), and then combined into a unified sample sequence X traj , Y traj , Z traj , A total of 287 valid sample points are included.

[0055] (3) For the spline function described in (2.1), 8 outer nodes and 13 inner nodes are configured as initial parameters, with a total number of nodes of 21. Take σ = 2 to construct the parameter sequence function according to the expression described in (2.2), and combine the unified sample sequence to estimate the function coefficient according to the method described in (3.2), and then reconstruct the trajectory sequence according to the expression (4.1) (where i = 1, 2, ..., N); after 4 iterations, convergence was achieved, and the spline function configuration was finally adjusted to 16 outer nodes and 21 inner nodes as initial parameters, with a total number of nodes of 37, and the final function coefficients and parameter sequence were obtained. The function coefficients are shown in Figure 3 (a) and Figure 3 (b), and the parameter sequence is shown in Figure 4 (a) and (b).

[0056] (4) The arrival of the vehicle at a specific location is used as the feature point criterion, and the arrival time is determined by the position sequence. The time of the feature point is T K1 and T K2 , and then the velocity parameters V of these two feature points are obtained by interpolating the sequence K1 and V K2 ,See Figure 5 (a) and (b). The whole sequence is divided into segments using the characteristic points as critical conditions, and a clear result of the trajectories of the three segments is obtained, see Figure 5 Middle (c).

[0057] like Figures 2 to 5As shown, compared with the track crossing, bifurcation and overlapping segment data ambiguity caused by the original data, the method proposed in the present invention accurately and effectively unifies the navigation trajectory constrained by multi-source data, and clearly identifies the characteristic point parameters, providing high-value information for the performance evaluation of the navigation body.

Claims

1. A method for underwater motion feature recognition based on multi-source data, characterized in that: The specific steps are as follows: Step 1: Complete the spatiotemporal registration of the segmented observation data of the navigation body and establish a unified position and velocity parameter sample sequence; For the N segmented observed navigation trajectories, the position and velocity parameter sequences of each segment are obtained and standardized to form a unified sample sequence, including: (1.1) Interpolation is to equidistant time nodes; (1.2) Through coordinate conversion, the coordinate system is unified into the measurement coordinate system; the zero elevation point of the starting point is taken as the origin O, the OX axis points to the east, the OY axis points to the north, and the OZ axis is vertically upward; (1.3) Combine the N segmented observation data segments into a unified sample sequence, which is in the form of: Among them, X traj , Y traj , Z traj is the combined position sample sequence, X i , Y i , Z i Corresponding to t i The sample value at the time, i = 1, 2, ..., N; is the combined velocity sample sequence, Corresponding to t i The sample value at the time, i = 1, 2, ..., N; Step 2: Define the spline function and establish the function expression of the underwater motion parameters of the navigation body; (2.1) Define the spline function; In the observation interval, the time point Set as internal node, a total of M in Outside the observation interval, the time point Set as external node, a total of M out The number of external nodes distributed on the left and right sides of the observation interval is equal; the intervals between each time node are equal, and the node distance is Δt * , the total number of nodes is M = M in +M out ; At each node, a set of spline functions with independent variables as time t is set, where the spline function expression of node j is: Among them, B (j) (t) represents the spline function of node j, represents the derivative of the spline function at node j; j = 1, 2, ..., M; σ is the mean square error parameter; (2.2) The underwater motion parameters are expressed by spline functions; According to the above spline function, the position sequence x(t), y(t), z(t) and velocity sequence of underwater motion It is expressed as: In the formula, and is the coefficient to be estimated, j = 1, 2, …, M; Step 3: Identify the parameter sequence function coefficients based on the unified sample sequence; (3.1) Establish the matrix relationship between the parameter sequence function and the unified sample sequence; The corresponding relationship between the parameter sequence function in the X direction, Y direction, and Z direction and the unified sample sequence is expressed by the following expression: TC X =F X TC Y =F Y TC Z =F Z in: (3.2) Identification parameter sequence function coefficients The linear least squares regression method is used to obtain the estimated values ​​of the coefficients of the function expression in the three directions of X, Y, and Z: C X =(T T T) -1 T T F X C Y =(T T T) -1 T T F Y C Z =(T T T) -1 T T F Z Step 4: Reconstruct the parameter sequence and identify the motion parameters of the feature points; (4.1) Reconstruct the position and velocity parameters of the navigation body based on the parameter sequence function; According to the function expression described in (2.2) and the coefficients obtained in (3.2), the functional relationship between position and velocity over time is established, and t i The estimated value of the location parameter at time and the estimated values ​​of the velocity parameters In this way, a complete trajectory sequence based on the estimated values ​​is reconstructed; (4.2) Identify the motion parameters of feature points; Based on the position and velocity parameter sequence obtained by (4.1), according to the known conditions given by the test conditions and the position and velocity curve shapes, the time when the characteristic points appear is determined, and then the parameter estimation value at the corresponding time is obtained by interpolation, that is, for time t K , position eigenvalue The velocity characteristic value is

Images