All-sky polarization perception device based on curved surface biomimetic compound eye and heading solution method

Abstract

This invention provides an all-sky polarization sensing device and heading calculation method based on a curved surface biomimetic compound eye, relating to the fields of biomimetic navigation and optical sensing technology. The device includes: a curved substrate; several polarization imaging units fixedly arranged in a close-packed hexagonal array on the outer surface of the curved substrate; and a processing unit electrically connected to all polarization imaging units. The curved substrate includes first-level, second-level, and third-level sub-surfaces, each with a textured layer, forming a fractal moth-eye structure. The method employs an online self-calibration method based on field-of-view overlap constraints to correct the original image data acquired by each polarization imaging unit; maps all corrected image data to a virtual celestial coordinate system to generate an all-sky polarization angle distribution map; when some units are obstructed, the polarization information of the obstructed area is reconstructed using an interpolation algorithm, thereby calculating the absolute heading angle of the carrier. This invention achieves highly robust and high-precision all-weather autonomous navigation.

Description

Technical Field

[0001] This invention relates to the field of biomimetic navigation and optical sensing technology, and in particular to an all-sky polarization sensing device and heading calculation method based on a curved surface biomimetic compound eye. Background Technology

[0002] Navigation technology based on atmospheric polarization patterns is a passive, interference-resistant autonomous navigation method inspired by insects (such as ants and bees). This method detects the polarization angle distribution of scattered light in the sky and inverts the sun's azimuth to provide a heading reference for the vehicle.

[0003] Existing polarization navigation sensors are mainly divided into point sensors and imaging sensors. Among them, point sensors (such as Honeywell POL core point detection modules) have a small field of view and are easily affected by local cloud interference, resulting in insufficient accuracy and reliability. Imaging sensors are usually implemented using a single or a few upward-facing wide-angle lenses. Although they can acquire sky images with a larger field of view, they still have the following inherent defects: (1) Risk of single-point failure: Once a single lens is partially blocked (such as by leaves or the fuselage), the performance of the entire system will be severely reduced; (2) Contradiction between edge distortion and resolution: Wide-angle lenses have severe edge distortion and poor uniformity of resolution for the entire celestial sphere; (3) Complex calibration: Multi-lens systems require high-precision external parameter calibration, and the calibration parameters are easily lost due to vibration during use, resulting in high maintenance costs.

[0004] Given the limitations of existing technologies, insect compound eyes offer a natural, bio-inspired solution. In nature, insect compound eyes consist of thousands of ommatidia, possessing a near-panoramic field of vision and inherent redundancy, allowing them to effectively perceive visual information even when some ommatidia are damaged. Inspired by this, incorporating biomimetic compound eye structures into polarization navigation sensing holds promise for fundamentally solving the aforementioned problems. However, how to engineer this biological principle into a stable, reliable, and efficient data processing device remains a pressing technical challenge in this field. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, this invention proposes an all-sky polarization sensing device and heading calculation method based on the multi-unit and inherent redundancy of insect compound eyes. This aims to overcome the defects of existing polarization navigation sensors, such as limited field of view, weak anti-occlusion ability, cumbersome calibration and easy failure, and achieves highly robust and high-precision all-weather autonomous navigation.

[0006] On the one hand, the present invention proposes an all-sky polarization sensing device based on a curved bionic compound eye, the device comprising: a curved substrate, several polarization imaging units and a processing unit;

[0007] A polarization imaging unit is fixedly disposed at the center of the outer surface of the curved substrate. Based on the polarization imaging unit, several polarization imaging layers are divided along the meridian direction on the outer surface of the curved substrate at a preset interval. Several polarization imaging units are fixedly disposed on each polarization imaging layer to form a layered curved surface array. The number of polarization imaging units in each polarization imaging layer is the same, and the polarization imaging units between adjacent polarization imaging layers correspond one-to-one along the meridian direction of the curved substrate.

[0008] The processing unit is located inside or at the bottom of the curved substrate, and all polarization imaging units are independently electrically connected to the processing unit.

[0009] Furthermore, the outer surface of the curved substrate is a biomimetic curved surface; the biomimetic curved surface is a hemispherical surface, a hyperspherical surface, or a spherical surface.

[0010] Furthermore, the outer surface of the curved substrate adopts a fractal moth-eye structure design, and three levels of concentric sub-surfaces are evenly divided along the meridian direction on the outer surface of the curved substrate, namely: a first-level sub-surface, a second-level sub-surface, and a third-level sub-surface; the width of the three levels of concentric sub-surfaces is equal along the meridian direction;

[0011] Each of the three concentric sub-surfaces has a texture layer;

[0012] The texture layer on the first-level sub-surface is a low-curvature moth-eye texture, with a texture period of 400nm–600nm and a height of 200nm–400nm; the low curvature is defined as: the amount of curvature change per unit length along the texture surface does not exceed a preset threshold.

[0013] The texture layer on the secondary subsurface is a variable-period moth-eye texture; the texture period of the variable-period moth-eye texture is inversely proportional to the curvature of the location of the texture;

[0014] The texture layer on the tertiary subsurface is a slightly tilted moth-eye texture. The tilt angle of this texture matches the curvature of the tertiary subsurface, and the tilt angle is no greater than 5°.

[0015] Furthermore, the first-order subsurface, the second-order subsurface, and the third-order subsurface are each provided with a gradient refractive index coating, and the gradient refractive index coating and the fractal moth-eye structure form a dual anti-reflection mechanism;

[0016] For any subsurface among the first-order, second-order, and third-order subsurfaces, calculate the thickness of the gradient refractive index coating on the subsurface based on the ray incident angle distribution corresponding to the curvature of the subsurface.

[0017] The refractive index of the gradient refractive index coating decreases gradually from the surface layer to the inner layer; wherein the surface layer is the side of the gradient refractive index coating that is in contact with air; and the inner layer is the side of the gradient refractive index coating that is in contact with the curved substrate.

[0018] Furthermore, the polarization imaging unit is used to acquire raw image data;

[0019] The polarization imaging unit is implemented using a miniaturized linear polarization camera, which is integrated on a regular hexagonal modular packaging base.

[0020] The miniaturized linear polarization camera includes, in sequence along the optical path, a lens, a polarizer, and an image sensor; wherein the polarizer has a concave center, and the polarization directions of the polarizer are periodically arranged in an array at a preset angle.

[0021] For any polarization imaging unit, the optical axis of the polarization imaging unit is perpendicular to the tangent plane of the curved surface where the polarization imaging unit is located;

[0022] There is a field of view overlap area between any two adjacent polarization imaging units, and the angle range of the field of view overlap area is 5° to 20°.

[0023] Furthermore, the processing unit includes an FPGA and a multi-core processor; the processing unit is used to synchronously acquire the raw image data collected by each polarization imaging unit, and to solve the absolute heading angle of the carrier on which the all-sky polarization sensing device is located by running a preset heading calculation method.

[0024] On the other hand, this invention proposes a heading calculation method based on a curved surface bionic compound eye, which includes the following process:

[0025] For a carrier equipped with the all-sky polarization sensing device, the original image data collected by each polarization imaging unit in the all-sky polarization sensing device is acquired simultaneously.

[0026] An online self-calibration method based on the field of view overlap region constraint is adopted to correct the original image data acquired by each polarization imaging unit, so as to obtain the corrected image data of all polarization imaging units.

[0027] The image data after correction of all polarization imaging units are mapped to a unified virtual celestial coordinate system to generate a full-sky polarization angle distribution map;

[0028] In the process of generating the full-sky polarization angle distribution map, when any polarization imaging unit is detected to be occluded, an interpolation algorithm based on spherical harmonic functions is used to reconstruct the corrected image data of the polarization imaging unit based on the corrected image data of the adjacent, unoccluded polarization imaging units.

[0029] Based on the full-sky polarization angle distribution map, calculate the absolute heading angle of the carrier.

[0030] Furthermore, the specific method for using the online self-calibration method based on the field-of-view overlap region constraint to correct the original image data acquired by each polarization imaging unit and obtain the corrected image data of all polarization imaging units is as follows:

[0031] For any pair of adjacent polarization imaging units A and B, feature point matching algorithm is used to match feature points in the overlapping area of ​​the field of view of polarization imaging units A and B to obtain several matching points.

[0032] For each matching point, obtain the Stokes parameters observed by polarization imaging unit A and the Stokes parameters observed by polarization imaging unit B;

[0033] Based on the Stokes parameters observed by polarization imaging unit A, calculate the polarization angle of the matching point relative to the theoretical polarization angle of polarization imaging unit A. ;

[0034] Based on the Stokes parameters observed by polarization imaging unit B, calculate the polarization angle of the matching point relative to the theoretical polarization angle of polarization imaging unit B. ;

[0035] Obtain the relative rotation error between polarization imaging units A and B, including: yaw error angle. Pitch error angle and roll error angle ;

[0036] Based on the relative rotation error and theoretical polarization angle and Calculate the observed polarization angle value and ;

[0037] According to the theoretical polarization angle and observed polarization angle value Calculate the polarization angle difference at the matching point. ;

[0038] To minimize the polarization angle difference of all matching points The sum of these values ​​is used as the optimization objective. The Levenberg-Marquardt nonlinear optimization algorithm is used to iteratively solve for the relative rotation error that minimizes the optimization objective, and this error is used as the relative installation error angle between polarization imaging units A and B.

[0039] Using the relative installation error angle, the original image data acquired by polarization imaging unit B is corrected to obtain the corrected image data of polarization imaging unit B.

[0040] Furthermore, the method for determining whether the polarization imaging unit is obstructed is as follows:

[0041] Real-time detection of the mean and variance of the raw image data acquired by each polarization imaging unit;

[0042] For any polarization imaging unit, when the mean value of the current frame raw image data acquired by the polarization imaging unit decreases by more than a preset first mean threshold compared to the mean value of the previous frame raw image data, and the variance of the current frame raw image data decreases by more than a preset first variance threshold compared to the variance of the previous frame raw image data, it is determined that the polarization imaging unit is occluded.

[0043] If any polarization imaging unit is continuous A polarization imaging unit is determined to be a stable unit when the raw image data acquired in a frame simultaneously meets the following conditions.

[0044] Condition 1: Continuous The mean fluctuation range of the original image data does not exceed the mean stability threshold;

[0045] Condition 2: Continuous The variance fluctuation range of the original image data of the frame does not exceed the variance stability threshold;

[0046] Condition 3: Continuous Polarization imaging units within the frame that are not identified as occluded;

[0047] For any polarization imaging unit, if the mean value of the original image data of the current frame acquired by the polarization imaging unit is... The mean of the original image data of the current frame acquired by the stabilizing unit The ratio between them does not exceed a preset second mean threshold, and the variance of the original image data of the current frame acquired by the polarization imaging unit is... The variance of the original image data of the current frame acquired by the stabilizing unit If the ratio between them does not exceed the preset second variance threshold, then the polarization imaging unit is determined to be suspected of being occluded;

[0048] If the polarization imaging unit is in continuous If all elements within a frame are identified as suspected occlusions or the amount of texture loss of the polarization imaging unit exceeds a preset standard, then the polarization imaging unit is determined to be occluded.

[0049] Furthermore, the specific details of calculating the absolute heading angle of the carrier based on the all-sky polarization angle distribution map are as follows:

[0050] Calculate the texture integrity coefficient of the full-sky polarization angle distribution map and compare the texture integrity coefficient with a preset switching threshold;

[0051] If the texture integrity coefficient is greater than the switching threshold, the solar meridian feature extraction method is used, and the absolute heading angle of the carrier is calculated by combining the current time and geographical location.

[0052] The specific method for calculating the absolute heading angle of the vehicle by employing the solar meridian feature extraction method and combining the current time and geographical location is as follows:

[0053] In the full-sky polarization angle distribution map, the polarization angle direction presents regular stripes with the solar meridian as the axis of symmetry;

[0054] The Hough transform detection is used to perform straight line detection on the full-sky polarization angle distribution map, extract the axis of symmetry that all fringes follow, and take the azimuth angle of this axis of symmetry as the direction of the solar meridian;

[0055] Obtain the current UTC time and the current geographical location of the carrier, and calculate the theoretical solar azimuth angle at the current moment according to the astronomical formula;

[0056] Calculate the difference between the theoretical solar azimuth angle and the solar meridian direction at the current moment, and use it as the absolute heading angle of the vehicle;

[0057] If the texture integrity coefficient is less than or equal to the switching threshold, the Rayleigh scattering model fitting method is used, and the absolute heading angle of the carrier is calculated by combining the current time and geographical location.

[0058] The specific method for calculating the absolute heading angle of the carrier using the Rayleigh scattering model fitting method, combined with the current time and geographical location, is as follows:

[0059] Obtain the current UTC time and the current geographical location of the carrier, and calculate the theoretical solar azimuth and altitude angles at the current moment according to astronomical formulas;

[0060] Using the heading angle of the carrier as the independent variable, a theoretical model of the polarization angle distribution across the entire sky is constructed based on Rayleigh scattering theory and the calculated theoretical solar azimuth and altitude angles.

[0061] The full-sky polarization angle distribution map is compared pixel by pixel with the theoretical model to construct a cost function; the cost function is used to quantify the degree of difference between the full-sky polarization angle distribution map and the theoretical model.

[0062] The cost function is optimized, and the heading angle of the carrier that minimizes the cost function value is taken as the absolute heading angle of the carrier.

[0063] The beneficial effects of adopting the above technical solution are as follows:

[0064] To address the limitations of existing polarization navigation sensors, such as limited field of view and poor resistance to local occlusion, this invention provides a biomimetic device and method with inherent redundancy and self-calibration capabilities, enabling highly robust all-sky polarization sensing and heading calculation. Its beneficial effects are analyzed in detail below:

[0065] 1. Biomimetic structure, no blind spots in the field of view: The device of this invention achieves truly seamless and uniform coverage of the sky in the upper hemisphere through a layered curved surface array design, solving the problems of edge distortion and blind spots in the field of view of a single lens.

[0066] 2. Built-in redundancy and extremely strong anti-occlusion capability: The device of this invention adopts a multi-unit design based on a biomimetic compound eye, providing natural data redundancy. Specifically, when a polarization imaging unit is occluded, the data is reconstructed using a spherical harmonic function interpolation algorithm; and based on the integrity coefficient of the sky polarization texture, the solar meridian extraction or Rayleigh scattering model fitting strategy is adaptively switched to calculate a highly robust absolute heading angle.

[0067] 3. Intelligent online self-calibration, maintenance-free: This invention utilizes the physical constraints of the overlapping field of view for online calibration to automatically compensate for installation errors, enabling the device to have self-calibration capabilities. This significantly reduces manufacturing and installation accuracy requirements and maintenance costs during use, and improves long-term reliability.

[0068] 4. Hardware and software synergy, performance optimization: This invention protects not only the hardware device structure, but also the dedicated data processing method that is deeply coupled with it and gives full play to its structural advantages, forming a complete technical solution closed loop with a clear scope of protection. Attached Figure Description

[0069] Figure 1 This is a structural diagram of the all-sky polarization sensing device based on curved surface bionic compound eye in this embodiment;

[0070] Figure 2 This is a top view of the structure of the all-sky polarization sensing device described in this embodiment;

[0071] Figure 3 This is a schematic diagram of the overlapping field of view of two adjacent polarization imaging units in this embodiment;

[0072] Figure 4 This is a flowchart of the heading calculation method based on curved surface bionic compound eye in this embodiment;

[0073] Figure 5 This is a schematic diagram of the heading calculation method based on curved surface bionic compound eye in this embodiment;

[0074] Figure 6 This is a schematic diagram of the online self-calibration in this embodiment;

[0075] Figure 7 This is a schematic diagram of the anti-occlusion data reconstruction effect in this embodiment; where (a) is the full-space polarization angle distribution image acquired by the polarization imaging unit under normal conditions; (b) is the polarization angle distribution image acquired by the polarization imaging unit that is determined to be occluded; and (c) is the polarization angle distribution image reconstructed by the occluded polarization imaging unit after spherical harmonic function interpolation.

[0076] In the figure: 101-curved substrate, 102-polarization imaging unit, 301-field of view overlap area. Detailed Implementation

[0077] To facilitate understanding of this application, specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following embodiments are illustrative of the invention but are not intended to limit its scope. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0078] Example 1:

[0079] This embodiment presents an all-sky polarization sensing device based on a curved surface biomimetic compound eye, such as... Figure 1 and Figure 2 As shown, the device includes: a curved substrate 101, a plurality of polarization imaging units 102 and a processing unit.

[0080] A polarization imaging unit 102 is fixedly disposed at the center of the outer surface of the curved substrate 101. Based on the polarization imaging unit 102, several polarization imaging rings are divided along the meridian direction on the outer surface of the curved substrate 101 at preset intervals. Several polarization imaging units 102 are fixedly disposed on each polarization imaging ring to form a ring-shaped curved surface array. The number of polarization imaging units 102 in each polarization imaging ring is the same, and the polarization imaging units 102 between adjacent polarization imaging rings correspond one-to-one along the meridian direction of the curved substrate 101.

[0081] It should be noted that the meridian direction refers to the direction that extends from the center of the curved surface along the edge of the curved surface and is perpendicular to the concentric circles (parallels of latitude).

[0082] The preset interval is the zenith angle difference between adjacent polarization imaging layers, with a value ranging from 15° to 30°. The zenith angle refers to the central angle of the sphere corresponding to the arc length between the center of the outer surface of the curved substrate 101 and any point on that outer surface.

[0083] In this embodiment, as Figure 1As shown, the curved substrate 101 is a hemispherical surface with a radius of R = 50 mm. On the outer surface of the curved substrate 101, 37 polarization imaging units 102 are fixed. Besides the polarization imaging unit fixed at the center of the outer surface of the curved substrate 101, three polarization imaging rings are arranged at predetermined intervals on the outer surface of the curved substrate 101. Each polarization imaging ring has 12 polarization imaging units uniformly fixed, and the polarization imaging units between adjacent polarization imaging rings correspond one-to-one along the meridian direction of the curved substrate 101, forming a ring-shaped curved surface array. Specifically, taking the polarization imaging unit at the center as the reference point (the line connecting the reference point and the center of the sphere is perpendicular to the bottom surface of the curved substrate 101), the arc length distances between the center of the polarization imaging unit and the reference point on the same meridian direction on the outer surface of the curved substrate 101 are as follows: 1st polarization imaging layer: 17.5 mm (corresponding to a zenith angle of approximately 20°); 2nd polarization imaging layer: 34.9 mm (corresponding to a zenith angle of approximately 40°); 3rd polarization imaging layer: 52.4 mm (corresponding to a zenith angle of approximately 60°). This arrangement allows for the arrangement of a larger number of units per unit area on a given curved surface, thereby achieving full spectral coverage of the outer surface of the curved substrate 101, ensuring high resolution and sufficient redundancy.

[0084] The processing unit is located inside or at the bottom of the curved substrate 101, and all polarization imaging units 102 are independently electrically connected to the processing unit.

[0085] It should be noted that in this embodiment, the processing unit is disposed inside the curved substrate 101 and electrically connected to each polarization imaging unit 102. Those skilled in the art should understand that when the processing unit is located inside the curved substrate 101, the processing unit... Figure 1 Although it is not visible, its structure and connection relationship are well-known technologies in the field and do not affect the understanding of this technical solution.

[0086] The outer surface of the curved substrate 101 is a biomimetic curved surface; the biomimetic curved surface is a hemispherical surface, a super-hemispherical surface, or a spherical surface.

[0087] In this embodiment, the outer surface of the curved substrate 101 is preferably a hemispherical biomimetic curved surface to most effectively cover the upper hemisphere sky. Furthermore, the hemispherical curved substrate used in this embodiment is made of lightweight alloy or engineering plastic, and the concentric curved surface array on its outer surface is biomimetic to the ommatidia arrangement of an insect's compound eye.

[0088] The outer surface of the curved substrate 101 adopts a fractal moth-eye structure design, and three levels of concentric sub-surfaces are evenly divided along the meridian direction on the outer surface of the curved substrate 101, namely: first-level sub-surface, second-level sub-surface and third-level sub-surface; the width of the three levels of concentric sub-surfaces is equal along the meridian direction.

[0089] In this embodiment, the fractal moth-eye structure refers to a cross-scale design concept inspired by the nanoscale anti-reflective structure on the surface of a moth's eye, applied to the macroscopic surface division of the curved substrate 101. Specifically, starting from the center of the outer surface of the curved substrate 101, first-level sub-surfaces, second-level sub-surfaces, and third-level sub-surfaces are uniformly divided sequentially along the meridian direction of the outer surface of the curved substrate 101. In this embodiment, the zenith angle range of the first-level sub-surface is... The zenith angle range of the second-order subsurface is: The zenith angle range of the third-order subsurface is Furthermore, the widths of the three concentric sub-surfaces are equal along the meridian direction.

[0090] Each of the three concentric sub-surfaces has a texture layer; wherein, the texture layer on the first-level sub-surface is a low-curvature moth-eye texture, the texture period of which is 400nm–600nm and the height is 200nm–400nm; the low curvature is defined as: the amount of curvature change per unit length along the texture surface does not exceed a preset threshold.

[0091] In this embodiment, a low-curvature moth-eye texture is processed on the primary sub-surface, with a texture period of 500 nm and a height of 300 nm. Specifically, this embodiment measures the curvature change per unit length along the texture surface. The condition is considered low curvature. For a hemispherical or super-hemispherical curved substrate 101, since its curvature is equal everywhere and the amount of curvature change is zero, the first-level sub-surfaces on the curved substrate 101 automatically meet the low curvature requirement; however, for a spherical curved substrate 101, since its curvature changes with position, it is necessary to ensure that the curvature change of the first-level sub-surfaces on the curved substrate 101 meets the low curvature requirement in order to ensure the processing feasibility and optical performance of the moth-eye texture.

[0092] The texture layer on the secondary subsurface is a variable-period moth-eye texture; the texture period of the variable-period moth-eye texture is inversely proportional to the curvature of the location of the texture.

[0093] Specifically, the texture period of the first-level subsurface is used as the reference period. In this embodiment The average curvature of the region with the smallest curvature change in the first-order subsurface is used as the reference curvature. For a spherical curved substrate 101, the region of minimum curvature variation in its first-order sub-surface is located near the zenith (the top of the surface) of the substrate 101. For a hemispherical or hyperspherical curved substrate 101, since its curvature is constant everywhere, the reference curvature of the first-order sub-surface is... It is equal to the curvature of the curved substrate 101.

[0094] The texture period of the variable-period moth-eye texture on the second-level subsurface is scaled proportionally with the curvature relative to the texture period of the first-level subsurface, and is expressed as follows:

[0095] ;

[0096] in, It refers to the local curvature on the second-order subsurface, that is, the curvature at a certain position on the second-order subsurface; The local texture period on the second-order subsurface, i.e., the period of curvature. The corresponding texture period. In this embodiment, each texture on the secondary subsurface is considered as a bump, and its texture period is inversely proportional to the curvature of its location. For example, the top period is 400nm, and the side period is 600nm, corresponding to different local curvatures. For every 1% increase in curvature, the period decreases. This conforms to an inverse proportional relationship.

[0097] The texture layer on the tertiary subsurface is a slightly tilted moth-eye texture. The tilt angle of this texture matches the curvature of the tertiary subsurface, and the tilt angle is no greater than 5°.

[0098] In this embodiment, the texture on the tertiary subsurface is distributed with a slight tilt, the tilt angle of which matches the curvature of the tertiary subsurface, and the tilt angle is ≤5°, thus forming a two-layer structure of macroscopic variable curvature fractal + microscopic fractal texture. In this embodiment, the micro-tilted moth-eye period can follow the range matching the secondary subsurface, avoiding parameter fragmentation.

[0099] The first-level subsurface, second-level subsurface, and third-level subsurface are each provided with a gradient refractive index coating, and the gradient refractive index coating and the fractal moth-eye structure form a dual anti-reflection mechanism.

[0100] For any subsurface among the first-order, second-order, and third-order subsurfaces, the thickness of the gradient refractive index coating on the subsurface is calculated based on the ray incident angle distribution corresponding to the curvature of the subsurface.

[0101] The refractive index of the gradient refractive index coating decreases gradually from the surface layer to the inner layer; wherein the surface layer is the side of the gradient refractive index coating that is in contact with air; and the inner layer is the side of the gradient refractive index coating that is in contact with the curved substrate 101.

[0102] In this embodiment, based on the aforementioned texture layer, synergistic optimization is achieved by superimposing gradient refractive index coatings. Specifically, an optical simulation software is used to obtain the light incident angle distribution corresponding to the curvature of each sub-surface, and then the optimal coating thickness is calculated. This embodiment employs ray tracing and gradient descent optimization algorithms, which solve for the optimal thickness by traversing the incident angle distribution and minimizing reflectivity. Rigorous Coupled-Wave Analysis (RCWA) and a genetic algorithm are used to accurately calculate the microstructure reflection spectrum. The genetic algorithm iteratively optimizes the coating thickness, and the Finite-Difference Time-Domain method (FDTD) and Particle Swarm Optimization (PSO) algorithms are used to simulate the optical response across the entire wave, and the PSO algorithm quickly converges to the optimal solution. In this embodiment, the gradient refractive index coating thickness of the first-order subsurface is 50 nm, with a surface refractive index of 1.25 and an inner refractive index of 1.50, a difference of 0.25; the gradient refractive index coating thickness of the second-order subsurface is 60 nm-70 nm, with a surface refractive index of 1.22 and an inner refractive index of 1.52, a difference of 0.30; and the gradient refractive index coating thickness of the tertiary subsurface is 80 nm, with a surface refractive index of 1.20 and an inner refractive index of 1.55, a difference of 0.35. Thus, the refractive index of the gradient refractive index coating on each subsurface decreases gradually from the surface to the inner layer, with a difference of 0.2-0.3, forming a dual anti-reflection mechanism with the micro / nano texture.

[0103] The polarization imaging unit 102 is used to acquire raw image data.

[0104] The polarization imaging unit 102 is implemented using a miniaturized linear polarization camera, which is integrated on a regular hexagonal modular packaging base.

[0105] The miniaturized linear polarization camera includes, in sequence along the optical path, a lens, a polarizer, and an image sensor; wherein the polarizer is concave at the center, and the polarization directions of the polarizer are periodically arranged in an array at a preset angle.

[0106] In this embodiment, each polarization imaging unit is a package, including: a miniaturized linear polarization camera and a hexagonal modular packaging base. The miniaturized linear polarization camera is integrated on the base, and the maximum outer circle diameter of the package is less than 8 mm. Inside the miniaturized linear polarization camera, there are, in sequence: a lens, an optional macro lens, a polymer linear polarizer, and a CMOS image sensor, to facilitate rapid calculation of the Stokes parameters and polarization angle of each pixel. The polarizer has a concave center. The polarization directions of the polarizers are arranged periodically in the array in the order of 0°, 45°, 90°, and 135°, that is, every 4 adjacent units constitute a complete polarization measurement group.

[0107] It should be noted that the polarization imaging units 102 are arranged periodically in a 0° layer and a 45° layer, i.e., every 45° is a layer. Within the same layer, the lens centers of all polarization imaging units 102 form a regular hexagon. However, due to the tiny gaps between the polarization imaging units 102, when sunlight shines on the curved substrate 101, some of the sunlight that passes through these tiny gaps and reaches the outer surface of the curved substrate 101 is prevented from being reflected back onto the lenses of the polarization imaging units, thus avoiding interference with the judgment.

[0108] For any polarization imaging unit 102, the optical axis of the polarization imaging unit 102 is perpendicular to the tangent plane of the curved surface where the polarization imaging unit 102 is located.

[0109] In this embodiment, the optical axis of each polarization imaging unit 102 is mounted perpendicular to the tangential plane of the substrate at the mounting point of the unit. This mounting method allows the optical axes of all units to naturally diverge radially outward from the center of the sphere, thereby seamlessly and uniformly covering the entire target sky region (such as the upper hemisphere).

[0110] There is a field of view overlap area between any two adjacent polarization imaging units, and the angle range of the field of view overlap area is 5° to 20°.

[0111] In this embodiment, as Figure 3 As shown, due to the divergence of the optical axes, there exists an angle between adjacent polarization imaging units A and B. The field of view overlap region 301. In this embodiment, by designing the radius R of the curved substrate 101 and the field of view angle of the polarization imaging unit 102, the field of view overlap region 301 is achieved. The angle is controlled at approximately 10°, thus providing sufficient data for online self-calibration.

[0112] In this embodiment, all polarization imaging units 102 are connected to the processing unit located inside or at the bottom of the curved substrate 101 via flexible printed circuits.

[0113] The processing unit includes an FPGA and a multi-core processor. The processing unit is used to synchronously acquire the raw image data collected by each polarization imaging unit, and to solve the absolute heading angle of the carrier on which the all-sky polarization sensing device is located by running a preset heading calculation method.

[0114] In this embodiment, the processing unit includes an FPGA and a multi-core processor, which are used to synchronously control the exposure of all units and run subsequent solution algorithms.

[0115] Example 2:

[0116] This embodiment presents a heading calculation method based on a curved bionic compound eye, implemented using the all-sky polarization sensing device based on a curved bionic compound eye described in Embodiment 1. The core innovation of this method lies in utilizing the inherent structural characteristics of the bionic array to achieve online self-calibration and anti-occlusion reconstruction. Specifically, online self-calibration is performed using sky polarization information from the overlapping fields of view of adjacent units to compensate for installation and response errors between polarization imaging units; an all-sky polarization angle distribution map is constructed; when some units are occluded, the polarization information of the occluded area is reconstructed using redundant unit data through an interpolation algorithm; finally, the absolute heading of the carrier is calculated based on the Rayleigh scattering model.

[0117] like Figure 4 and Figure 5 As shown, the heading calculation method includes the following process:

[0118] For a carrier equipped with the all-sky polarization sensing device, the raw image data collected by each polarization imaging unit in the all-sky polarization sensing device is acquired simultaneously.

[0119] In this embodiment, the processing unit issues a global trigger signal and simultaneously acquires one frame of image from all 37 polarization imaging units as raw image data.

[0120] An online self-calibration method based on the field of view overlap region constraint is adopted to correct the original image data acquired by each polarization imaging unit, thereby obtaining the corrected image data of all polarization imaging units.

[0121] In this embodiment, the core of the online self-calibration method lies in: Figure 6 As shown, for each pair of adjacent polarization imaging units with overlapping fields of view, there is a field of view overlap center. The polarization observation information of the same sky point within the overlapping area is extracted, and then the relative installation angle error and polarization response inconsistency between units are calculated and compensated by minimizing the observation difference.

[0122] like Figure 6 As shown, the specific method for using an online self-calibration method based on field-of-view overlap constraints to correct the original image data acquired by each polarization imaging unit and obtain the corrected image data of all polarization imaging units is as follows:

[0123] For any pair of adjacent polarization imaging units A and B, feature point matching algorithms are used to match feature points in the overlapping area of ​​the fields of view of polarization imaging units A and B, resulting in several matching points.

[0124] In this embodiment, algorithms such as SIFT can be used for feature point matching.

[0125] For each matching point, obtain the Stokes parameters observed by polarization imaging unit A. Stokes parameters observed by polarization imaging unit B ;in This represents the total intensity parameter observed by polarization imaging unit A; This represents the linear polarization intensity difference parameter observed by polarization imaging unit A; This represents the 45-degree linear polarization intensity difference parameter observed by polarization imaging unit A; This represents the total intensity parameter observed by polarization imaging unit B; This represents the linear polarization intensity difference parameter observed by polarization imaging unit B; This represents the 45-degree linear polarization intensity difference parameter observed by polarization imaging unit B.

[0126] Based on the Stokes parameters observed by polarization imaging unit A Calculate the theoretical polarization angle of the matching point relative to polarization imaging unit A. .

[0127] Stokes parameters observed by polarization imaging unit B Calculate the polarization angle of the matching point relative to the theoretical polarization angle of polarization imaging unit B. .

[0128] For each matching point, based on the Stokes parameters observed by polarization imaging units A and B respectively. and Calculate the theoretical polarization angle of this feature point. and ,and and The consistency constraint equations that must be satisfied. The method for calculating the theoretical polarization angle is as follows:

[0129] ;

[0130] in, Indicates the theoretical polarization angle; This represents the intensity difference parameter of linear polarization at 45 degrees. This represents the linear polarization intensity difference parameter.

[0131] The consistency constraint equation is expressed as:

[0132] ;

[0133] Obtain the relative rotation error between polarization imaging units A and B, including: yaw error angle. Pitch error angle and roll error angle .

[0134] Based on the relative rotation error and theoretical polarization angle and Calculate the observed polarization angle and .

[0135] Under ideal, error-free conditions Should equal to However, due to the actual existence of relative rotation errors (three Euler angles)... , , An error model is constructed to describe the relationship between the observed polarization angle and the theoretical polarization angle. This model consists of a rotation matrix. A decision is expressed as:

[0136] ;

[0137] in, Represents the observed polarization angle; rotation matrix Represented as:

[0138] ;

[0139] When the error is very small (in this embodiment, when the yaw error angle is small) Pitch error angle and roll error angle When the absolute values ​​of all values ​​are less than 5° (assuming the error is very small), the rotation matrix... It can be approximated as:

[0140] ;

[0141] According to the theoretical polarization angle and observed polarization angle value Calculate the polarization angle difference at the matching point. .

[0142] To minimize the polarization angle difference of all matching points The sum of these values ​​is used as the optimization objective. The Levenberg-Marquardt nonlinear optimization algorithm is employed to iteratively solve for the relative rotation error that minimizes the optimization objective, which is then used as the relative installation error angle between polarization imaging units A and B.

[0143] In this embodiment, an optimization problem is constructed using the constraint equations of all matching points, and the problem is solved using the constraints between adjacent units. , , To optimize the variables, the solution is performed with the objective of minimizing the sum of all polarization angle differences, yielding the solution that minimizes the objective function. , , The relative installation error angle between these two units is used to correct the observation data of polarization imaging unit B in real time. This process is performed in parallel for all adjacent unit pairs. Calibration parameters are obtained and used for data correction.

[0144] In this embodiment, the relative installation error angle is solved using the least squares method, and is expressed as:

[0145] ;

[0146] ;

[0147] in, The coefficient matrix consists of the theoretical polarization angles of the matching points and the partial derivatives of the error model. The residual vector is composed of the difference between the observed polarization angle and the theoretical polarization angle at the matching point; and and The required data all come from matching points in the overlapping area of ​​the fields of view of adjacent polarization imaging units (extracted by algorithms such as SIFT).

[0148] Using the relative installation error angle, the original image data acquired by polarization imaging unit B is corrected to obtain the corrected image data of polarization imaging unit B.

[0149] In this embodiment, due to the designed field-of-view overlap between adjacent imaging units, the same sky region is observed within this overlap. Theoretically, the observations of the polarization state of this region by each unit should be consistent. Utilizing this consistency as a natural constraint, relative angular deviations and differences in photoelectric response caused by installation errors, vibration, and temperature drift between units can be estimated and compensated in real time and online. This completely eliminates the dependence on high-cost offline calibration equipment and complex calibration fields, achieving "plug and play" and long-term stability.

[0150] The image data after correction of all polarization imaging units are mapped to a unified virtual celestial coordinate system to generate a full-sky polarization angle distribution map.

[0151] In the process of generating the full-sky polarization angle distribution map, when any polarization imaging unit is detected to be occluded, an interpolation algorithm based on spherical harmonic functions is used to reconstruct the corrected image data of the polarization imaging unit based on the corrected image data of the adjacent, unoccluded polarization imaging units.

[0152] In this embodiment, the image data corrected by all polarization imaging units are unified into a virtual celestial coordinate system through coordinate transformation, and then stitched together to generate a complete, high-resolution all-sky polarization angle map. During this process, if some polarization imaging units are detected to be completely or partially occluded (judged by analyzing abnormal image brightness and contrast), the polarization information of the occluded area is intelligently reconstructed using data from the surrounding unoccluded units through a spherical interpolation algorithm (such as spherical harmonic function fitting). This greatly improves the survivability of this method in complex real-world environments (such as urban canyons and tree-lined roads).

[0153] The virtual celestial coordinate system is as follows: with the observation center of the carrier as the origin, the direction opposite to the direction of the carrier's gravity is defined as the positive z-axis, the geographic north direction is defined as the positive x-axis, and the positive y-axis is determined according to the right-hand rule, thus forming a local horizontal coordinate system fixed to the carrier.

[0154] The method for determining whether the polarization imaging unit is obstructed is as follows:

[0155] The mean and variance of the raw image data acquired by each polarization imaging unit are detected in real time.

[0156] For any polarization imaging unit, when the mean value of the current frame raw image data acquired by the polarization imaging unit decreases by more than a preset first mean threshold compared to the mean value of the previous frame raw image data, and the variance of the current frame raw image data decreases by more than a preset first variance threshold compared to the variance of the previous frame raw image data, the polarization imaging unit is determined to be occluded.

[0157] In this embodiment, if any polarization imaging unit experiences a sudden and significant drop in mean and disappearance of texture, that polarization imaging unit is determined to be occluded. Since the variance is large when the texture is clear and drops sharply when the texture disappears, the relationship between variance and texture is used to quantify texture intensity / complexity. Therefore, in this embodiment, if the mean decreases... (In this embodiment, the first mean threshold is set to) And the variance decreased. (In this embodiment, the first variance threshold is set to) If the polarization imaging unit is blocked, then it is determined that the current polarization imaging unit is blocked.

[0158] If any polarization imaging unit is continuous A polarization imaging unit is determined to be a stable unit when the raw image data acquired in a frame simultaneously meets the following conditions.

[0159] Condition 1: Continuous The mean fluctuation range of the original image data does not exceed the mean stability threshold.

[0160] Condition 2: Continuous The variance fluctuation range of the original image data does not exceed the variance stability threshold.

[0161] Condition 3: Continuous Polarization imaging units within the frame that are not identified as occluded.

[0162] In this embodiment, the quantization judgment condition of the stabilization unit is: the mean fluctuation of 10 consecutive frames of images. Variance fluctuation Furthermore, no polarization imaging units exhibited a significant drop in mean and loss of texture (i.e., were not identified as being occluded).

[0163] For any polarization imaging unit, if the mean value of the original image data of the current frame acquired by the polarization imaging unit is... The mean of the original image data of the current frame acquired by the stabilizing unit The ratio between them does not exceed a preset second mean threshold. In this embodiment, the second mean threshold is set to... Furthermore, the variance of the original image data of the current frame acquired by the polarization imaging unit The variance of the original image data of the current frame acquired by the stabilizing unit The ratio between them does not exceed a preset second variance threshold. In this embodiment, the second variance threshold is set to... If so, the polarization imaging unit is determined to be suspected of being obstructed.

[0164] If the polarization imaging unit is in continuous If all elements within a frame are identified as suspected occlusions or the amount of texture loss of the polarization imaging unit exceeds a preset standard, then the polarization imaging unit is determined to be occluded.

[0165] In this embodiment, a single-frame determination is first performed on the arbitrary polarization imaging unit, that is:

[0166] ;

[0167] If any polarization imaging unit meets the above-mentioned suspected occlusion criteria for three consecutive frames, or if the amount of texture loss exceeds a preset standard (in this embodiment, the preset standard is set to 8.3% of the total texture amount in the stabilizing unit), then the polarization imaging unit is determined to be occluded. Furthermore, occlusion can be further confirmed through angle verification, for example, by using the difference between the heading angle calculated based on the polarization imaging unit and the heading angle calculated by the stabilizing unit. In this case, it can also be considered that the polarization imaging unit is blocked.

[0168] In this embodiment, for the celestial coordinates corresponding to the occluded region, spherical harmonic function interpolation is used to reconstruct the coordinates using data from surrounding unoccluded cells. Specifically, the polarization angle values ​​of known points on the celestial sphere are used as samples to fit a low-order spherical harmonic function, and then this function is used to calculate the polarization angle values ​​of the occluded region, resulting in the reconstructed distribution map, as shown below. Figure 7 As shown.

[0169] The spherical harmonic function is expressed as:

[0170] ;

[0171] in, This represents the reconstructed image data; Zenith angle, ; It is the azimuth angle. ; Let be the order of the azimuth angle, satisfying , used to indicate the frequency of change in the azimuth direction; The order of the spherical harmonic function is taken as 2 in this embodiment, that is... The higher the order, the higher the fitting accuracy; Let be real spherical harmonic basis functions, such as ; Let be a spherical harmonic function, and we have:

[0172] ;

[0173] in, These are coefficient vectors, each corresponding to the weights under different orders of spherical basis functions; The zero-order coefficient is used to control the overall mean of the polarization angle distribution; All are first-order coefficients, used to control the linear gradient change of the polarization angle along different directions, reflecting the main directional trend of the solar meridian; All are second-order coefficients used to control the quadrupole moment characteristics of the polarization angle distribution, corresponding to the key symmetric features of the polarization angle distribution in the Rayleigh scattering mode; is the basis function matrix.

[0174] Based on the full-sky polarization angle distribution map, calculate the absolute heading angle of the carrier.

[0175] The specific steps for calculating the absolute heading angle of the carrier based on the full-sky polarization angle distribution map are as follows:

[0176] Calculate the texture integrity coefficient of the full-sky polarization angle distribution map and compare the texture integrity coefficient with a preset switching threshold.

[0177] If the texture integrity coefficient is greater than the switching threshold, the solar meridian feature extraction method is used, and the absolute heading angle of the carrier is calculated by combining the current time and geographical location.

[0178] If the texture integrity coefficient is less than or equal to the switching threshold, the Rayleigh scattering model fitting method is used, and the absolute heading angle of the carrier is calculated by combining the current time and geographical location.

[0179] In this embodiment, the solar meridian feature extraction method is suitable for clear / partially cloudy scenes, relying on the symmetry of the polarization angle distribution; excessively thick clouds will disrupt the stripe symmetry. The Rayleigh scattering model fitting method, on the other hand, is suitable for cloudy / low-light scenes. By fitting the polarization angle distribution across the entire sky, it is more robust to local occlusion / noise, but requires more computation. Therefore, a switching threshold is set; in this embodiment, it is set to 80%. When the texture integrity exceeds 80%, the solar meridian feature extraction method is used; otherwise, the Rayleigh scattering model fitting method is switched to.

[0180] The specific method for calculating the absolute heading angle of the vehicle by employing the solar meridian feature extraction method and combining the current time and geographical location is as follows:

[0181] In the full-sky polarization angle distribution map, the polarization angle direction presents regular stripes with the solar meridian as the axis of symmetry.

[0182] In this embodiment, for the full-sky polarization angle distribution map, a curve whose polarization angle direction changes continuously and is symmetrically distributed is identified. The axis of symmetry of this curve corresponds to the direction of the solar meridian.

[0183] The Hough transform was used to detect straight lines in the full-sky polarization angle distribution map, and the axis of symmetry that all fringes follow was extracted. The azimuth angle of this axis of symmetry was then taken as the direction of the solar meridian. .

[0184] Obtain the current UTC time and the current geographical location of the carrier, and calculate the theoretical solar azimuth angle at the current moment using astronomical formulas. .

[0185] Calculate the theoretical solar azimuth at the current moment. In the direction of the sun meridian The difference is used as the absolute heading angle of the carrier. .

[0186] In this embodiment, in the reconstructed full-sky polarization angle distribution map, the polarization angle direction presents regular fringes with the solar meridian as the axis of symmetry. The Hough transform is used to detect the axis of symmetry of these fringes, and their azimuth angle is the direction of the solar meridian. By combining a high-precision clock and initial position (or GNSS assistance), the theoretical solar azimuth angle at the current moment is calculated using astronomical formulas. Then, the absolute heading angle of the carrier can be calculated. , is represented as:

[0187] ;

[0188] The specific method for calculating the absolute heading angle of the carrier using the Rayleigh scattering model fitting method, combined with the current time and geographical location, is as follows:

[0189] Obtain the current UTC time and the current geographical location of the carrier, and calculate the theoretical solar azimuth and altitude angles at the current moment based on astronomical formulas.

[0190] Using the heading angle of the carrier as the independent variable, a theoretical model of the polarization angle distribution across the entire sky is constructed based on Rayleigh scattering theory and the calculated theoretical solar azimuth and altitude angles.

[0191] The full-sky polarization angle distribution map is compared pixel by pixel with the theoretical model to construct a cost function; the cost function is used to quantify the degree of difference between the full-sky polarization angle distribution map and the theoretical model.

[0192] The cost function is optimized, and the heading angle of the carrier that minimizes the cost function value is taken as the absolute heading angle of the carrier.

[0193] In this embodiment, based on the generated full-sky polarization angle map, a more robust algorithm (such as Hough transform to extract the meridian or global fitting of the Rayleigh scattering model) can be used to invert the sun's azimuth, and then the carrier's heading can be calculated by combining accurate spatiotemporal information. Since the input is complete and reliable sky polarization information, its calculation accuracy and anti-interference ability are far superior to traditional methods based on local sky information.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the present invention.

Claims

1. A full-sky polarization perception device based on curved biomimetic compound eye, characterized in that, The device includes: a curved substrate, several polarization imaging units, and a processing unit; A polarization imaging unit is fixedly disposed at the center of the outer surface of the curved substrate. Based on the polarization imaging unit, several polarization imaging layers are divided along the meridian direction on the outer surface of the curved substrate at a preset interval. Several polarization imaging units are fixedly disposed on each polarization imaging layer to form a layered curved surface array. The number of polarization imaging units in each polarization imaging layer is the same, and the polarization imaging units between adjacent polarization imaging layers correspond one-to-one along the meridian direction of the curved substrate. The processing unit is located inside or at the bottom of the curved substrate, and all polarization imaging units are independently electrically connected to the processing unit. 2.The full-view polarization perception device based on curved biomimetic compound eye according to claim 1, wherein, The outer surface of the curved substrate is a biomimetic curved surface; the biomimetic curved surface is a hemispherical, a hyperspherical, or a spherical surface. 3.The full-view polarization perception device based on curved biomimetic compound eye according to claim 1, wherein, The outer surface of the curved substrate adopts a fractal moth-eye structure design, and three levels of concentric sub-surfaces are evenly divided along the meridian direction on the outer surface of the curved substrate, namely: first-level sub-surface, second-level sub-surface and third-level sub-surface; the width of the three levels of concentric sub-surfaces is equal along the meridian direction; Each of the three concentric sub-surfaces has a texture layer; The texture layer on the first-level sub-surface is a low-curvature moth-eye texture, with a texture period of 400nm–600nm and a height of 200nm–400nm; the low curvature is defined as: the amount of curvature change per unit length along the texture surface does not exceed a preset threshold. The texture layer on the secondary subsurface is a variable-period moth-eye texture; the texture period of the variable-period moth-eye texture is inversely proportional to the curvature of the location of the texture; The texture layer on the tertiary subsurface is a slightly tilted moth-eye texture. The tilt angle of this texture matches the curvature of the tertiary subsurface, and the tilt angle is no greater than 5°.

4. The full-view polarization perception device based on curved biomimetic compound eye according to claim 3, characterized in that, The first-level subsurface, the second-level subsurface, and the third-level subsurface are each provided with a gradient refractive index coating, and the gradient refractive index coating and the fractal moth-eye structure form a dual anti-reflection mechanism; For any subsurface among the first-order, second-order, and third-order subsurfaces, calculate the thickness of the gradient refractive index coating on the subsurface based on the ray incident angle distribution corresponding to the curvature of the subsurface. The refractive index of the gradient refractive index coating decreases gradually from the surface layer to the inner layer; wherein the surface layer is the side of the gradient refractive index coating that is in contact with air; and the inner layer is the side of the gradient refractive index coating that is in contact with the curved substrate.

5. The full-sky polarization perception device based on curved biomimetic compound eye according to claim 1, wherein, The polarization imaging unit is used to acquire raw image data; The polarization imaging unit is implemented using a miniaturized linear polarization camera, which is integrated on a regular hexagonal modular packaging base. The miniaturized linear polarization camera includes, in sequence along the optical path, a lens, a polarizer, and an image sensor; wherein the polarizer has a concave center, and the polarization directions of the polarizer are periodically arranged in an array at a preset angle. For any polarization imaging unit, the optical axis of the polarization imaging unit is perpendicular to the tangent plane of the curved surface where the polarization imaging unit is located; There is a field of view overlap area between any two adjacent polarization imaging units, and the angle range of the field of view overlap area is 5° to 20°.

6. The full-sky polarization perception device based on curved biomimetic compound eye according to claim 1, wherein, The processing unit includes an FPGA and a multi-core processor. The processing unit is used to synchronously acquire the raw image data collected by each polarization imaging unit, and to solve the absolute heading angle of the carrier on which the all-sky polarization sensing device is located by running a preset heading calculation method.

7. The heading solution method based on the curved-surface biomimetic compound eye, which is implemented by using the all-sky polarization perception device based on the curved-surface biomimetic compound eye according to any one of claims 1-6, characterized in that, This method includes the following steps: For a carrier equipped with the all-sky polarization sensing device, the original image data collected by each polarization imaging unit in the all-sky polarization sensing device is acquired simultaneously. An online self-calibration method based on the field of view overlap region constraint is adopted to correct the original image data acquired by each polarization imaging unit, so as to obtain the corrected image data of all polarization imaging units. The image data after correction of all polarization imaging units are mapped to a unified virtual celestial coordinate system to generate a full-sky polarization angle distribution map; In the process of generating the full-sky polarization angle distribution map, when any polarization imaging unit is detected to be occluded, an interpolation algorithm based on spherical harmonic functions is used to reconstruct the corrected image data of the polarization imaging unit based on the corrected image data of the adjacent, unoccluded polarization imaging units. Based on the full-sky polarization angle distribution map, calculate the absolute heading angle of the carrier.

8. The heading solution method based on the curved-surface biomimetic compound eye according to claim 7, characterized in that, The specific method for using an online self-calibration method based on field-of-view overlap constraints to correct the original image data acquired by each polarization imaging unit and obtain the corrected image data of all polarization imaging units is as follows: For any pair of adjacent polarization imaging units A and B, feature point matching algorithm is used to match feature points in the overlapping area of ​​the field of view of polarization imaging units A and B to obtain several matching points. For each matching point, obtain the Stokes parameters observed by polarization imaging unit A and the Stokes parameters observed by polarization imaging unit B; According to the Stokes parameters observed by the polarization imaging unit A, the matching point is calculated compared to the theoretical polarization angle of the polarization imaging unit A ; According to the Stokes parameters observed by the polarization imaging unit B, the matching point is calculated compared to the theoretical polarization angle of the polarization imaging unit B ; Obtaining the relative rotation error of the polarization imaging units A and B includes a yaw error angle , a pitch error angle , and a roll error angle ; Based on the relative rotation error and theoretical polarization angle and Calculate the observed polarization angle value and ; According to the theoretical polarization angle and the observed polarization angle value , the polarization angle difference of the matching point is calculated; to minimize the sum of the polarization angle differences of all matching points The Levenberg-Marquardt nonlinear optimization algorithm is used to iteratively solve the relative rotation error that minimizes the optimization target, and the relative installation error angle of the polarization imaging units A and B. Using the relative installation error angle, the original image data acquired by polarization imaging unit B is corrected to obtain the corrected image data of polarization imaging unit B.

9. The heading solution method based on the curved-surface biomimetic compound eye according to claim 8, characterized in that, The method for determining whether the polarization imaging unit is obstructed is as follows: Real-time detection of the mean and variance of the raw image data acquired by each polarization imaging unit; For any polarization imaging unit, when the mean value of the current frame raw image data acquired by the polarization imaging unit decreases by more than a preset first mean threshold compared to the mean value of the previous frame raw image data, and the variance of the current frame raw image data decreases by more than a preset first variance threshold compared to the variance of the previous frame raw image data, it is determined that the polarization imaging unit is occluded. If any of the polarization imaging units is continuous The raw image data of frame acquisition simultaneously satisfies the following conditions, and the polarization imaging unit is determined as a stable unit; Condition 1: continuous a mean fluctuation range of the frame raw image data does not exceed a mean stability threshold; Condition 2: consecutive a variance fluctuation range of the frame raw image data does not exceed a variance stability threshold; Condition 3: consecutive Intra-polarization imaging units not determined to be occluded; For any polarization imaging unit, if the ratio between the mean of the current frame raw image data acquired by the polarization imaging unit and the mean of the current frame raw image data acquired by the stable unit does not exceed a preset second mean threshold, and the ratio between the variance of the current frame raw image data acquired by the polarization imaging unit and the variance of the current frame raw image data acquired by the stable unit does not exceed a preset second variance threshold, it is determined that the polarization imaging unit is suspected to be blocked. If the polarization imaging unit is in continuous If all elements within a frame are identified as suspected occlusions or the amount of texture loss of the polarization imaging unit exceeds a preset standard, then the polarization imaging unit is determined to be occluded.

10. The heading calculation method based on curved surface bionic compound eye according to claim 9, characterized in that, The specific steps for calculating the absolute heading angle of the carrier based on the full-sky polarization angle distribution map are as follows: Calculate the texture integrity coefficient of the full-sky polarization angle distribution map and compare the texture integrity coefficient with a preset switching threshold; If the texture integrity coefficient is greater than the switching threshold, the solar meridian feature extraction method is used, and the absolute heading angle of the carrier is calculated by combining the current time and geographical location. The specific method for calculating the absolute heading angle of the vehicle by employing the solar meridian feature extraction method and combining the current time and geographical location is as follows: In the full-sky polarization angle distribution map, the polarization angle direction presents regular stripes with the solar meridian as the axis of symmetry; The Hough transform detection is used to perform straight line detection on the full-sky polarization angle distribution map, extract the axis of symmetry that all fringes follow, and take the azimuth angle of this axis of symmetry as the direction of the solar meridian; Obtain the current UTC time and the current geographical location of the carrier, and calculate the theoretical solar azimuth angle at the current moment according to the astronomical formula; Calculate the difference between the theoretical solar azimuth angle and the solar meridian direction at the current moment, and use it as the absolute heading angle of the vehicle; If the texture integrity coefficient is less than or equal to the switching threshold, the Rayleigh scattering model fitting method is used, and the absolute heading angle of the carrier is calculated by combining the current time and geographical location. The specific method for calculating the absolute heading angle of the carrier using the Rayleigh scattering model fitting method, combined with the current time and geographical location, is as follows: Obtain the current UTC time and the current geographical location of the carrier, and calculate the theoretical solar azimuth and altitude angles at the current moment according to astronomical formulas; Using the heading angle of the carrier as the independent variable, a theoretical model of the polarization angle distribution across the entire sky is constructed based on Rayleigh scattering theory and the calculated theoretical solar azimuth and altitude angles. The full-sky polarization angle distribution map is compared pixel by pixel with the theoretical model to construct a cost function; the cost function is used to quantify the degree of difference between the full-sky polarization angle distribution map and the theoretical model. The cost function is optimized, and the heading angle of the carrier that minimizes the cost function value is taken as the absolute heading angle of the carrier.

Images