Polarization compass fitting orientation method based on sparse irregular polarization information

Abstract

The invention relates to the technical field of bionic polarization orientation, in particular to a polarization compass fitting orientation method based on sparse irregular polarization information, and solves the problem that the polarization compass fitting orientation performance is reduced under the limitation of the sparse irregular polarization information. According to the polarization compass fitting orientation method, based on the characteristic of anti-symmetric distribution of positive and negative polarization angles in an atmospheric polarization mode, a polarization angle image pixel region is segmented by using a self-adaptive unequal-interval ring domain, index values are calculated step by step by adopting a path step length to measure the symmetric distribution degree of positive and negative feature points in the segmented region, and a minimum error angle value is continuously optimized; global optimal search is continuously carried out on the ring domain number distribution scheme by means of an Archimedes optimization algorithm, the optimal fitness individual position is obtained as the optimal ring domain number, the position where the axis is located is determined as the meridian angle, the meridian is fitted according to the meridian angle, course information can be solved with high precision through reference transformation, and the polarization orientation precision is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of bionic polarization orientation, and specifically to a polarization compass fitting orientation method based on sparse and irregular polarization information. Background Art

[0002] When sunlight propagates, polarized light is generated through the scattering effect of atmospheric particles. The specific polarization state distribution formed by polarized light in the sky is the atmospheric polarization pattern. Bionic polarization orientation can achieve the orientation of a carrier by extracting the sun meridian information in the atmospheric polarization pattern. Compared with traditional satellite navigation and inertial navigation, bionic polarization orientation is not affected by electromagnetic field interference and does not produce cumulative errors, and has broad application prospects in the field of autonomous navigation. In the actual application process, complex atmospheric environments or local shielding situations are likely to interfere with the atmospheric polarization pattern, resulting in sparse and irregular polarization information, which in turn interferes with the calculation of the distribution of polarized angle pixel points and reduces the fitting orientation performance of the polarization compass, causing the polarization compass to be unable to accurately output the geographic heading angle. Based on this, it is necessary to invent a polarization compass fitting orientation method based on sparse and irregular polarization information to solve the problem of reduced fitting orientation performance of the polarization compass under the limitation of sparse and irregular polarization information. Summary of the Invention

[0003] In order to solve the problem of reduced fitting orientation performance of the polarization compass under the limitation of sparse and irregular polarization information, the present invention provides a polarization compass fitting orientation method based on sparse and irregular polarization information.

[0004] The present invention is implemented by the following technical solutions:

[0005] A polarization compass fitting orientation method based on sparse and irregular polarization information, which is implemented by the following steps:

[0006] Step S1: Input the atmospheric polarization angle map W(L) after value range expansion and resolution, set a threshold T to screen positive and negative feature points. Pixel points with a positive polarization angle are positive feature points, and pixel points with a negative polarization angle are negative feature points; mark the coordinate indexes of the screening results. The coordinate index of the positive feature point is z_s, and the coordinate index of the negative feature point is f_s;

[0007] Step S2: Define a circular pixel area with a radius of R, and determine whether the pixel I0 in this area meets the feature points screened in Step S1. Mark the coordinate indexes of the feature points in the circular pixel area as aop_x and aop_y respectively, and calculate the angles and distances j_a, j_d, f_a, f_d between the positive and negative feature points in the circular pixel area and the center of the circle;

[0008]

[0009] Among them, (num_x, num_y) is the coordinate of the central pixel of the circular area;

[0010] Step S3: Statistically sort the distance values zf_l between all feature points and the center of the circle within the circular pixel area, initialize the value of the number of annulus regions K, and statistically calculate the annulus interval value H i (K), discretize the distance variables j_d and f_d and obtain the allocated corresponding interval index; divide the positive and negative feature points into the annulus intervals:

[0011]

[0012] where num_z is the total length of the matrix zf_l, and Q a is the serial number index that returns the annulus interval to which each feature distance variable belongs;

[0013] Step S4: Discretize the angular variables j_a and f_a, determine the angular variables corresponding to the positive and negative feature points included within each annulus interval belonging to the annulus index, and calculate the number of positive and negative feature points within each annulus interval;

[0014] The calculation of the number of positive and negative feature points within each annulus interval is:

[0015]

[0016] where y 1i and y 2i are the angular variable matrices of the positive and negative feature points included within each annulus region, i is the corresponding serial number of each annulus, and w z m is the combination of the angular variable values arranged in sequence according to the annulus sequence;

[0017] Step S5: Initialize the measurement parameter min_j, calculate the mode most_z and most_f in the angular variable matrices j_a and f_a, set the traversal path length θ, divide the effective and interference regions of the sectors within the annulus by the π / 4 interval for the reference axis, and statistically calculate the distribution of the number of positive and negative feature points in the adjacent effective and interference regions;

[0018] Statistical process of traversal distribution within the divided region:

[0019]

[0020] where β j is the screening value of the positive feature angular variable, δ j is the screening value of the negative feature angular variable, t is the sampling step, and z_x and f_x are the numbers of positive and negative feature points in the adjacent sector regions respectively;

[0021] Step S6: Calculate the weight u corresponding to the annulus i, calculate the test index value γ by cyclic iteration according to the sampling step and the annular domain sequence, count the superposition result and compare it with the measurement parameter min_j. If γ < min_j, the measurement parameter is replaced with the test index value and the iteration test continues within the annular domain. If γ > min_j, this γ value is discarded and the target value within the range less than the measurement parameter is continued to be measured; after the iteration within the annular domain sequence ends, continue the annular domain iteration in the π / 4 sector area divided according to the path length θ until the minimum test index value γ is obtained after all path length processes end;

[0022] The calculation formulas for the weight sum and the test index value corresponding to the annular domain are as follows:

[0023]

[0024] Among them, z_sum is the total number of feature points, and z_y i , f_y i are the numbers of positive and negative feature points in the i-th annular domain of adjacent sector areas respectively, and I_sum is the total number of feature points in the interference area;

[0025] Step S7: Introduce the Archimedes optimization algorithm. Use the number of annular domains K as the evaluation initial population, and select the free individual K best as the best global optimization solution. Repeat steps S3 - S7 through N iterations. When the γ scheme error obtained by the fitness function F is the smallest, obtain the position x of the optimal fitness individual best as the best number of annular domains, and determine the position of the axis as the meridian angle;

[0026] The best parameter update scheme of the Archimedes optimization is as follows:

[0027]

[0028] Among them, represents the position vector of the j-th individual at the N-th iteration. c1, c2, and c3 are constants, rand is a random number in the interval (0, 1), and x rand represents the position vector of the j-th random individual at the N-th iteration. d is the density reduction factor, TF is the transfer factor, and F determines the update direction of the optimal individual;

[0029] Step S8: Relying on the most accurate meridian angle α1 obtained by the best K best , obtain the slope best_k, fit the meridian. After the reference transformation, the absolute heading angle can be obtained from the solar azimuth angle in the navigation system

[0030] The orientation calculation process is as follows:

[0031] ​

[0032] The solar azimuth It is calculated from the local solar altitude angle, azimuth angle, latitude of the observation point, solar declination angle, and solar hour angle.

[0033] The invention is based on the characteristic of antisymmetric distribution of positive and negative polarization angles in an atmospheric polarization mode, uses adaptive non-equidistant ring domains to segment polarization angle map pixel areas, adopts path step length to gradually calculate index values to measure the degree of symmetric distribution of positive and negative feature points in the segmented area, continuously optimizes the minimum error angle value, and uses Archimedean optimization algorithm to continuously perform global optimal search on the number of ring domain allocation schemes, obtains the individual position with the best fitness as the optimal number of ring domains, determines the position of the axis as the meridian angle, and fits the meridian based on this, and can solve the heading information with high precision after reference transformation; the invention realizes compensation for sparse and irregular polarization information in this way, effectively solves the problem that sparse and irregular polarization information reduces the fitting orientation performance of a polarization compass, resulting in the inability of the polarization compass to accurately output the geographic heading angle, and improves the environmental adaptability of the bionic polarization compass. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the model of the fitting orientation method of the present invention.

[0035] Figure 2 It is the processing result of the present invention based on the condition of sparse irregular polarization information. DETAILED DESCRIPTION

[0036] A polarization compass fitting orientation method based on sparse irregular polarization information is implemented by the following steps:

[0037] Step S1: Input the atmospheric polarization angle map W(L) after the range expansion solution, set the threshold T to filter positive and negative feature points, the pixel points with positive polarization angle are positive feature points, and the pixel points with negative polarization angle are negative feature points; mark the coordinate index of the screening result, the coordinate index of the positive feature point is z_s, and the coordinate index of the negative feature point is f_s;

[0038] Step S2: Define a circular pixel area with a radius of R, determine whether the pixel I0 in this area meets the feature point screening step S1, mark the feature point coordinate indexes in the circular pixel area as aop_x and aop_y respectively, and calculate the angles and distances j_a, j_d, f_a, f_d between the positive and negative feature points in the circular pixel area and the center of the circle;

[0039]

[0040] Among them, (num_x, num_y) is the pixel coordinate of the center of the circular area;

[0041] Step S3: Count and sort the distance values zf_l between all feature points and the center of the circle within the circular pixel region, initialize the value of the number of annular regions K, and count the annular region spacing value H i (K), discretize the distance variables j_d and f_d and obtain the allocated corresponding spacing interval indices; divide the positive and negative feature points into the annular regions:

[0042]

[0043] Among them, num_z is the total length of the matrix zf_l, Q a is the serial number index that returns the annular region intervals to which each feature distance variable belongs;

[0044] Step S4: Discretize the angle variables j_a and f_a, determine the angle variables corresponding to the positive and negative feature points within each annular region spacing belonging to the annular region index, and calculate the number of positive and negative feature points within each annular region spacing;

[0045] The calculation of the number of positive and negative feature points within each annular region spacing is:

[0046]

[0047] Among them, y 1i and y 2i are the angle variable matrices of the positive and negative feature points contained in each annular region, i is the corresponding serial number of each annular region, w z m is the combination of the angle variable values arranged in sequence according to the annular region sequence;

[0048] Step S5: Initialize the measurement parameter min_j, calculate the mode most_z and most_f in the angle variable matrices j_a and f_a, set the traversal path length θ, divide the effective and interference regions of the sectors within the annular region at intervals of π / 4 along the reference axis, and count the distribution of the number of positive and negative feature points in the adjacent effective and interference regions;

[0049] The statistical process of traversing and distributing within the divided regions:

[0050]

[0051] Among them, β j is the positive feature angle variable screening value, δ j is the negative feature angle variable screening value, t is the sampling step, and z_x and f_x are the numbers of positive and negative feature points in the adjacent sector regions respectively;

[0052] Step S6: Calculate the weight u corresponding to the annular region i, calculate the test index value γ by cyclic iteration according to the sampling step size and the annular domain sequence, count the superposition result and compare it with the measurement parameter min_j. If γ < min_j, the measurement parameter is replaced with the test index value and the iteration test continues within the annular domain. If γ > min_j, this γ value is discarded and the target value within the range less than the measurement parameter is continued to be measured; after the iteration within the annular domain sequence ends, continue the annular domain iteration in the π / 4 sector area divided according to the path length θ until the minimum test index value γ is obtained after all path length processes end;

[0053] The calculation formulas for the weight sum and the test index value corresponding to the annular domain are as follows:

[0054]

[0055] Among them, z_sum is the total number of feature points, and z_y i , f_y i are the numbers of positive and negative feature points in the i-th annular domain of adjacent sector areas respectively, and I_sum is the total number of feature points in the interference area;

[0056] Step S7: Introduce the Archimedes optimization algorithm. Use the number of annular domains K as the evaluation initial population, and select the free individual K best as the best global optimization solution. Repeat steps S3 - S7 through N iterations. When the γ scheme error obtained by the fitness function F is the smallest, obtain the position x of the optimal fitness individual best as the best number of annular domains, and determine the position of the axis as the meridian angle;

[0057] The best parameter update scheme of the Archimedes optimization is as follows:

[0058]

[0059] Among them, represents the position vector of the j-th individual in the N-th iteration. c1, c2, and c3 are constants, rand is a random number in the interval (0, 1), and x rand represents the position vector of the j-th random individual in the N-th iteration. d is the density reduction factor, TF is the transfer factor, and F determines the update direction of the optimal individual;

[0060] Step S8: Relying on the most accurate meridian angle α1 obtained by the best K best obtained, obtain the slope best_k, fit the meridian, and after the reference transformation, the absolute heading angle can be obtained from the solar azimuth angle in the navigation system

[0061] The orientation calculation process is as follows:

[0062] ​

[0063] Among them, the solar azimuth angle is calculated from the local solar altitude angle, azimuth angle, the latitude of the observation point, the solar declination angle, and the solar hour angle.

[0064] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A polarization compass fitting and orientation method based on sparse irregular polarization information, characterized in that: The method is implemented by the following steps: Step S1: Input the atmospheric polarization angle map W(L) after range expansion solution, set a threshold T to screen positive and negative feature points. Pixel points with positive polarization angles are positive feature points, and pixel points with negative polarization angles are negative feature points; mark the coordinate indices of the screening results. The coordinate index of the positive feature point is z_s, and the coordinate index of the negative feature point is f_s; Step S2: Define a circular pixel area with a radius of R, and determine whether the pixels I0 in this area meet the feature points screened in Step S1. Mark the coordinate indices of the feature points in the circular pixel area as aop_x and aop_y respectively, and calculate the angles and distances j_a, j_d, f_a, and f_d between the positive and negative feature points and the center of the circle in the circular pixel area; Among them, (num_x, num_y) is the coordinate of the central pixel of the circular area; Step S3: Statistically sort the distance values zf_l between all feature points and the center of the circle within the circular pixel area, initialize the value of the number of annular regions K, and statistically calculate the annular region spacing value H i (K), discretize the distance variables j_d and f_d and obtain the indexes for allocating the corresponding spacing intervals; divide the positive and negative feature points into the annular region intervals: where num_z is the total length of the number of matrices zf_l, and Q a is the serial number index that returns the attribution ring domain interval of each feature distance variable; Step S4: Discretize the angular variables j_a and f_a, and determine the angular variables corresponding to positive and negative feature points within each annular domain spacing. The index of the belonging annular domain, and calculate the number of positive and negative feature points within each annular domain spacing; The number of positive and negative feature points in each ring domain spacing is calculated as: Among them, y 1i and y 2i are the angular variable matrices of the positive and negative characteristic points contained in each annular region, i is the corresponding serial number of each annular region, and w z m is the arrangement and combination of the angular variable values in sequence according to the annular region sequence; Step S5: Initialize the measurement parameter min_j, calculate the modes most_z and most_f in the angle variable j_a and f_a matrices, set the traversal path length θ, divide the effective and interference areas of the fan-shaped area in the ring domain at intervals of π / 4 along the reference axis, and count the distribution of the number of positive and negative feature points in adjacent effective and interference areas; Statistical process of traversing and distributing within the divided area: Among them, β j is the positive characteristic angle variable screening value, δ j is the negative characteristic angle variable screening value, t is the sampling step, and z_x and f_x are the numbers of positive and negative characteristic points in adjacent fan-shaped regions respectively; Step S6: Calculate the weight u corresponding to the annular region i , and cyclically iterate to calculate the test index value γ according to the sampling step size and the annular region sequence. Statistically superimpose the results and compare them with the measurement parameter min_j. If γ < min_j, the measurement parameter is replaced with the test index value and the iteration test continues within the annular region. If γ > min_j, this γ value is discarded and the target value within the range less than the measurement parameter continues to be measured. After the iteration within the annular region sequence ends, continue to perform annular region iteration in the π / 4 sector region divided according to the path length θ until the minimum test index value γ is obtained after all path length processes end; The calculation formulas for the sum of weights and test index values corresponding to the ring domain weights are: where z_sum is the total number of feature points, and z_y i and f_y i are the numbers of positive and negative feature points in the i-th annular region of adjacent fan-shaped regions respectively, and I_sum is the total number of feature points in the interference region; Step S7: Introduce the Archimedes optimization algorithm, use the number of annular domains K as the evaluation of the initial population, and select the free individual K best As the best global optimization solution, repeat steps S3 - S7 through N iterations. When the error of the γ solution obtained by the fitness function F is the smallest, obtain the position x of the optimal fitness individual best As the best number of annular domains, determine the position of the axis as the meridian angle; The update scheme for the best parameters of Archimedes optimization is: where x j N represents the position vector of the j-th individual at the N-th iteration, c1, c2, and c3 are constants, rand is a random number in the interval (0, 1), and x rand represents the position vector of the j-th random individual at the N-th iteration, d is the density reduction factor, TF is the transfer factor, and F determines the update direction of the optimal individual; Step S8: Rely on the optimal K best The most accurate meridian angle obtained is α1, the slope best_k is obtained, the meridian is fitted, and after the reference transformation, from the solar azimuth angle φ in the navigation system s the absolute heading angle can be calculated The orientation calculation process is as follows: Among them, the solar azimuth angle is calculated from the local solar altitude angle, azimuth angle, the latitude of the observation point, the solar declination angle, and the solar hour angle.

Images