Binocular stereo vision 3D reconstruction method and system based on polarization state

Abstract

The present disclosure relates to a polarization-based binocular stereoscopic three-dimensional reconstruction method and system. The method comprises: acquiring polarization images at different polarization degrees from a polarization camera; calculating the angle of a normal vector based on the acquired polarization images and Fresnel theory, wherein the angle of the normal vector includes the zenith angle and azimuth angle of the object surface; eliminating the ambiguity of the azimuth angle by using the projection angle of the normal vector on the YOZ plane, and eliminating mismatches by using the angle constraint of the normal vector; and performing gradient integration on the angle of the normal vector after mismatches are eliminated to complete the three-dimensional reconstruction. The disclosed method utilizes the projection angle to assist in determining the quadrant of the normal line, thereby eliminating mismatches of corresponding points and improving measurement efficiency.

Description

Technical Field

[0001] The present disclosure relates to the field of computer vision technology, and in particular to a binocular stereo vision three-dimensional reconstruction method based on polarization state and a system thereof. Background Art

[0002] The application of non-contact 3D shape measurement technology in industrial inspection has gradually become a core technology with the rapid development of intelligent manufacturing in recent years. In particular, the requirements for high robustness and high efficiency have become key factors in improving production quality and ensuring product qualification rates. However, traditional binocular vision reconstruction methods often encounter difficulties when processing low-texture, highly reflective objects. In particular, when the target surface lacks obvious texture features, this often leads to data gaps or inaccurate depth information. To address this problem, polarization-based 3D reconstruction methods have emerged. Polarization technology utilizes the polarization properties of reflected light and, combined with Fresnel's theorem, derives polarization information differences when parallel incident light reflects off an object's surface. Polarization 3D reconstruction uses this information to infer the object's shape. This effectively suppresses interference from strong light reflections, improving the accuracy and robustness of 3D reconstruction. It is particularly suitable for reconstructing low-texture and highly reflective surfaces.

[0003] Maeda R et al. proposed a method for decomposing reflection components based on the rotation direction of the polarization plane, solving the problem of specular interreflection on metal objects and promoting the application of light polarization state in 3D measurement. Wolff LB et al. studied a polarization-based light reflection model, proposed using Fresnel reflection coefficients to analyze light reflection and transmission, and explored the application of polarization technology in computer vision. Ping Qianqian et al. proposed a 3D reconstruction method based on polarization binocular vision. By combining the world coordinates of a small number of feature points obtained by binocular stereo vision, the point cloud data obtained in the pixel coordinate system of polarization imaging is converted into absolute data in the world coordinate system, effectively solving the 3D reconstruction problem of smooth and highly reflective targets. Tian X et al. innovatively combined polarization imaging with binocular stereo vision, corrected for azimuth ambiguity, and introduced low-rank matrix decomposition constraints to improve the accuracy and quality of 3D reconstruction. However, the performance of this method is insufficient for highly reflective and low-texture objects, and the computational complexity is high, resulting in insufficient real-time performance. However, the ambiguity of the azimuth and zenith angle of the normal vector in polarization binocular vision reconstruction remains a major challenge. Summary of the Invention

[0004] To address the problem that traditional methods have multiple equivalent solutions for the definitions of azimuth and zenith angles, resulting in the same normal vector corresponding to different angle combinations, especially in complex binocular vision systems, this paper proposes a polarization-based binocular stereo 3D reconstruction method to solve this problem.

[0005] According to one aspect of the present disclosure, a method for 3D reconstruction of binocular stereo vision based on polarization state is provided, comprising:

[0006] S10, obtaining polarization images at different polarization degrees from a polarization camera;

[0007] S20. Calculating an angle of a normal vector based on the acquired polarization image and Fresnel theory, wherein the angle of the normal vector includes a zenith angle and an azimuth angle of the object surface;

[0008] S30, eliminating the ambiguity of the azimuth angle by using the projection angle of the normal vector on the YOZ plane, and eliminating mismatching by using the angle constraint of the normal vector;

[0009] S40, performing gradient integration on the angle of the normal vector after eliminating mismatching to complete the three-dimensional reconstruction.

[0010] Preferably, calculating the angle of the normal vector based on Fresnel theory includes: calculating the Stokes parameters according to the acquired polarization image, and further obtaining the polarization degree and polarization azimuth, which are expressed as:

[0011] ,

[0012] ,

[0013] ,

[0014] ,

[0015] ,

[0016] Where, 、 、 and The polarization images represent the linear polarization degrees at 0°, 45°, 90°, and 135°, respectively. is the degree of polarization, is the polarization azimuth.

[0017] Preferably, calculating the angle of the normal vector based on Fresnel theory further includes: obtaining the azimuth angle of the object surface according to the relationship between the polarization azimuth angle and the azimuth angle, which is expressed as:

[0018] ,

[0019] Where, is the azimuth.

[0020] Preferably, calculating the angle of the normal vector based on Fresnel theory further includes: solving a relationship between the degree of polarization and the zenith angle through numerical iteration to obtain the zenith angle of the object surface, wherein the relationship between the degree of polarization and the zenith angle is expressed as:

[0021] ,

[0022] Where, is the zenith angle, is the complex refractive index, =n(1 + ik), i is the imaginary unit, and k is the attenuation exponent.

[0023] Preferably, the ambiguity of the azimuth angle is eliminated by the projection angle of the normal vector in the YOZ plane, including: introducing the projection angle of the normal vector in the YOZ plane according to the relationship between the polarization azimuth angle and the azimuth angle, simulating rotation through the binocular vision system, judging the sign of the projection angle, and determining the quadrant in which the azimuth angle is located.

[0024] Preferably, the angle of the normal vector after eliminating the mismatch is subjected to gradient integration to complete the three-dimensional reconstruction, including: converting the zenith angle and the azimuth angle into a three-dimensional normal vector, expressed as:

[0025] ,

[0026] Where, is the normal vector of any point on the surface.

[0027] Preferably, performing gradient integration on the angle of the normal vector after eliminating mismatching to complete the three-dimensional reconstruction further includes:

[0028] Convert the three-dimensional normal vector into a discrete approximate gradient through local linear approximation;

[0029] The discrete approximate gradient is fitted to the surface gradient field by minimizing the cost function, and the surface shape of the object is reconstructed by numerical integration.

[0030] According to one aspect of the present disclosure, a binocular stereoscopic vision 3D reconstruction system based on polarization state is provided, comprising:

[0031] A polarization image acquisition module is used to acquire polarization images of different polarization degrees from a polarization camera;

[0032] A Fresnel theory calculation module, which calculates the angle of the normal vector based on the acquired polarization image and Fresnel theory, wherein the angle of the normal vector includes the zenith angle and azimuth angle of the object surface;

[0033] The mismatch elimination module eliminates the ambiguity of the azimuth angle by the projection angle of the normal vector on the YOZ plane and eliminates mismatches by using the angle constraint of the normal vector;

[0034] The 3D reconstruction module performs gradient integration on the angle of the normal vector after eliminating mismatching to complete the 3D reconstruction.

[0035] According to one aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to: execute the above-mentioned polarization-state-based binocular stereo vision three-dimensional reconstruction method.

[0036] According to one aspect of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the above-mentioned binocular stereo vision three-dimensional reconstruction method based on polarization state is implemented.

[0037] Compared with the prior art, the beneficial effects of the present disclosure are:

[0038] 1) This paper proposes a new polarization-based stereo vision disambiguation (PBAE) method to solve the ambiguity problem in normal integrals in high-precision surface reconstruction.

[0039] 2) This paper uses the projection angle to assist in determining the normal quadrant, eliminating point mismatch, improving measurement efficiency, and effectively solving the point cloud hole problem.

[0040] 3) This disclosure provides a simple and effective solution to correct normal integration and alleviate missing data in point clouds with high robustness and accuracy.

[0041] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

[0042] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings herein are incorporated into and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and, together with the specification, are used to explain the technical solutions of the present disclosure.

[0044] Figure 1 The flowchart of the binocular stereo vision 3D reconstruction method based on polarization state is shown;

[0045] Figure 2 A schematic diagram of obtaining polarization images under different polarization degrees of a polarization camera in an example of the present disclosure is shown;

[0046] Figure 3 A schematic diagram of assisting binocular disambiguation by projecting an angle in an example of the present disclosure is shown;

[0047] Figure 4A schematic diagram of performing gradient integration on the normal vector information after eliminating mismatching in an example of the present disclosure is shown;

[0048] Figure 5 The structure block diagram of the binocular stereo vision 3D reconstruction system based on polarization state in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0049] Various exemplary embodiments, features, and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0050] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0051] The term "and / or" herein simply describes an association relationship between associated objects, indicating that three relationships can exist. For example, "A and / or B" can represent the existence of three situations: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" herein refers to any combination of at least two of any one or more of a plurality of items. For example, "at least one of A, B, and C" can represent any one or more elements selected from the set consisting of A, B, and C.

[0052] In addition, numerous specific details are provided in the following detailed description to better illustrate the present disclosure. Those skilled in the art will appreciate that the present disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main points of the present disclosure.

[0053] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] Based on the above ideas, the present invention proposes a binocular stereo vision 3D reconstruction method based on polarization state. Figure 1 A flow chart of a method for binocular stereoscopic vision 3D reconstruction based on polarization state is shown. The method comprises:

[0056] S10, obtaining polarization images at different polarization degrees from a polarization camera;

[0057] S20. Calculating an angle of a normal vector based on the acquired polarization image and Fresnel theory, wherein the angle of the normal vector includes a zenith angle and an azimuth angle of the object surface;

[0058] S30, eliminating the ambiguity of the azimuth angle by using the projection angle of the normal vector on the YOZ plane, and eliminating mismatching by using the angle constraint of the normal vector;

[0059] S40, performing gradient integration on the angle of the normal vector after eliminating mismatching to complete the three-dimensional reconstruction.

[0060] The present disclosure provides a method for 3D reconstruction of binocular stereo vision based on polarization states, which specifically includes the following steps:

[0061] S10, obtaining polarization images at different polarization degrees from a polarization camera.

[0062] In this embodiment, the light waves are transverse waves, with their vibration direction perpendicular to the propagation direction. In nature, the vibration direction of natural light is uniformly distributed in all directions, making it unpolarized. For any plane light wave, its light vector can be decomposed into two mutually orthogonal components: the s component, which vibrates perpendicular to the plane of incidence, and the p component, which vibrates parallel to the plane of incidence. When the s and p components in a beam of light are unevenly distributed, polarization of the light occurs.

[0063] According to the Fresnel equation, the reflectivity R will change with the change of the light vector vibration direction. For vertically polarized light (s-polarization), the reflectivity and the angle of incidence The calculation formula is:

[0064] ,

[0065] Where, is the refractive index of the incident medium, is the refractive index of the transmission medium, is the angle of incident light, is the angle of refracted light.

[0066] For parallel polarized light (p-polarization), the reflectivity The calculation formula is:

[0067] ,

[0068] According to Malus's law, the polarization state of light can be inferred by detecting the light intensity received by a detector placed after a linear polarizer, which is expressed as:

[0069] ,

[0070] Where, is the angle between the transmission axis of the polarizer and the starting position (reference position), and They represent the maximum and minimum light intensities observed at a certain point during the rotation of the polarizer, is the polarization azimuth, which can be defined as The maximum brightness is observed when ,when The minimum brightness is observed when .

[0071] S20. Calculate the angle of the normal vector based on the acquired polarization image and Fresnel theory, wherein the angle of the normal vector includes the zenith angle and the azimuth angle of the object surface.

[0072] In this embodiment, the polarization information of the light reflected from the surface of the object can be obtained by polarization imaging technology. The polarization information of the reflected light includes the normal vector information of the surface of the object. The schematic diagram of obtaining polarization images under different polarization degrees of the polarization camera is as follows: Figure 2 Specifically, by analyzing the polarization state of the reflected light and calculating the Stokes parameters based on the acquired polarization image, the degree of polarization and polarization azimuth are further obtained, which are expressed as:

[0073] ,

[0074] ,

[0075] ,

[0076] ,

[0077] ,

[0078] Where, 、 、 and The polarization images represent the linear polarization degrees at 0°, 45°, 90°, and 135°, respectively. is the degree of polarization, is the polarization azimuth.

[0079] In the mirror reflection light, the incident angle is equal to the zenith angle. According to the relationship between polarization degree and zenith angle, the relationship is solved by numerical iteration to obtain the zenith angle of the object surface and the linear polarization degree of the mirror reflection of the metal target surface. The geometric expression of is:

[0080] ,

[0081] Where, is the zenith angle, is the complex refractive index, =n(1 + ik), i is the imaginary unit, and k is the attenuation exponent.

[0082] The polarization phase angle is the angle between the vibration direction of the linear polarization component in partially polarized light and the starting axis. In the mirror-reflected light, the reflectivity of the vertical component is higher than that of the horizontal component. The vibration direction of the linear polarization component in the generated partially polarized light is perpendicular to the reflection plane, that is, the azimuth angle and the polarization phase angle differ by 90°.

[0083] Azimuth and polarization azimuth Directly related, according to the relationship between the polarization azimuth and the azimuth, the azimuth of the object surface is obtained, which is expressed as:

[0084] ,

[0085] Where, is the azimuth.

[0086] The disambiguation of the azimuth angle is to add constraints, from two angles with a difference of π The final solution is selected from the two most reliable ones. The normal vector in the three-dimensional coordinate system has directionality. The angle between the projection of the normal vector on the YOZ plane and the Z axis is defined as the projection angle , expressed as:

[0087] ,

[0088] Where, is the projection angle.

[0089] The angle between the normal vector and the Z axis Different quadrants may have different symbols. Taking the XOZ plane as the quadrant, the normal vector is defined to point to the right. Is positive, the normal vector points to the left Is negative. Therefore, the polarization azimuth of the normal vector can be combined and projection angle The positive or negative value of the normal vector is used to determine the quadrant in which the normal vector is located. The XYZ coordinate system divides the space into 8 quadrants. The normal vector in this embodiment only exists in the four quadrants of the positive half axis of the Z axis. The four quadrants where the positive axis of the Z axis is located are divided into quadrants 1, 2, 3, and 4 in a clockwise direction.

[0090] Polarization phase angle .when , , the normal vector falls in the 2nd or 4th quadrant, ;when , , the normal vector falls in the 1st or 3rd quadrant, .

[0091] As long as the polarization phase angle is determined Size and The positive or negative value of the normal vector can determine the quadrant in which the normal vector is located, and we can also use the above and Specifically, when , the normal falls in the 2nd or 4th quadrant, If it is positive, it falls in the second quadrant. ; If it is negative, it falls in the second quadrant. .

[0092] S30 , eliminating the ambiguity of the azimuth angle by using the projection angle of the normal vector on the YOZ plane, and eliminating mismatching by using the angle constraint of the normal vector.

[0093] Using rotational motion to judge The positive and negative angles of the normal vector and the difficulty in controlling the rotation of the target object or the rotation of the camera around the target object are relatively small. This embodiment uses a binocular vision system to replace the rotational motion. Specifically, the ambiguity of the azimuth angle is eliminated by the projection angle of the normal vector on the YOZ plane, including: introducing the projection angle of the normal vector on the YOZ plane based on the relationship between the polarization azimuth and the azimuth angle, simulating the rotation through the binocular vision system, determining the sign of the projection angle, and determining the quadrant in which the azimuth angle is located.

[0094] Assume that the target point before rotation is measured , the angle of rotation , then the target point is measured after rotation .like ,but .like ,but As long as we find the positive or negative value of the target point after rotation, we can determine the positive or negative value of the point before rotation.

[0095] In this embodiment, the schematic diagram of using projection angle to assist binocular disambiguation is as follows: Figure 3 As shown in the figure, first, multi-angle polarization images are collected synchronously to calculate the Stokes parameters and polarization degree information. Then, the zenith angle and azimuth angle are calculated based on the Fresnel reflection law. The three-dimensional reconstruction is further completed by using binocular disambiguation constrained by the projection angle. The core idea of ​​the binocular disambiguation algorithm can be summarized as follows:

[0096] First, by establishing the angle based on the normal vector Then, by introducing a series of geometric and physical constraints, we filter out the wrong point pairs that do not meet the matching conditions, and finally obtain the correct correspondence and angle between the same target point in two images from different perspectives. According to the above theory, we can further determine symbols for pairing.

[0097] Assume that the normal vector angle of all reference points in the image before rotation is lie in range, among which It is the maximum acceptable angle range defined by the camera rotation angle and field of view limit. In the rotated image, the angle of the target point Changes to , that is, by performing a translation transformation on the rotation angle, the potential matching points in the rotated image can be searched.

[0098] S40, performing gradient integration on the angle of the normal vector after eliminating mismatching to complete the three-dimensional reconstruction.

[0099] In this embodiment, the normal vector information after eliminating mismatching is subjected to gradient integration for 3D reconstruction. Figure 4 As shown in the figure, the 3D coordinates of the normal are used to obtain a local point cloud image, which is then used to obtain a 3D reconstruction image. The angle of the normal vector after eliminating mismatches is then gradient-integrated to complete the 3D reconstruction. This includes: converting the 3D normal vector into a discrete approximate gradient through local linear approximation; fitting the discrete approximate gradient to the surface gradient field by minimizing the cost function; and reconstructing the surface shape of the object through numerical integration.

[0100] Assume the target surface is a Cartesian surface , is the normal vector of any point on the surface, Zenith angle and azimuth Expression, converting the zenith angle and azimuth angle into a three-dimensional normal vector, is expressed as:

[0101] ,

[0102] Where, is the normal vector of any point on the surface.

[0103] Specifically, the discrete approximate gradient (p, q) can be expressed as the polarization normal vector To calculate, where

[0104] ,

[0105] ,

[0106] The direction of the normal vector is the same as the direction of the surface gradient, and the magnitude of the normal vector is proportional to the rate of surface change. The change of the normal vector can be expressed as the partial derivative of the gradient:

[0107] ,

[0108] ,

[0109] shape It can be calculated by minimizing the following cost function, expressed as:

[0110] ,

[0111] Getting the surface normal vector N and surface gradient ( Z x , Z y ) and the cost function of the discretized approximate gradient (p, q), the gradient field is integrated using the following integration method.

[0112] ,

[0113] ,

[0114] ,

[0115] when( u , v )≠(0,0), solving the above equation yields:

[0116] .

[0117] The disclosed embodiments propose a polarization-based binocular stereo 3D reconstruction method that combines polarization information with binocular vision to more effectively achieve high-precision 3D reconstruction results in complex, low-texture, and highly reflective environments. This method offers a new approach to resolving azimuth ambiguity and provides technical support for the promotion of polarization binocular vision in practical industrial applications. The disclosed method significantly improves point cloud quality, reduces normal errors, and effectively resolves the problem of point cloud data holes.

[0118] Example 2

[0119] As another aspect of the embodiment of the present disclosure, a binocular stereoscopic vision 3D reconstruction system 100 based on polarization state is also provided. Figure 5 Shown, including:

[0120] Polarization image acquisition module 1, acquires polarization images under different polarization degrees of the polarization camera;

[0121] A Fresnel theory calculation module 2 calculates the angle of the normal vector based on the acquired polarization image and the Fresnel theory, wherein the angle of the normal vector includes the zenith angle and the azimuth angle of the object surface;

[0122] Mismatch elimination module 3, eliminating the ambiguity of the azimuth angle by the projection angle of the normal vector on the YOZ plane, and eliminating mismatches by using the angle constraint of the normal vector;

[0123] The 3D reconstruction module 4 performs gradient integration on the angle of the normal vector after eliminating mismatching to complete the 3D reconstruction.

[0124] In the absence of any contradiction, the above modules in the system of the embodiment of the present disclosure can implement any implementation of the above method.

[0125] Based on the description of the above embodiments, it can be seen that the embodiments of the present disclosure can achieve the following technical effects:

[0126] 1) This paper proposes a new polarization-based stereo vision disambiguation (PBAE) method to solve the ambiguity problem in normal integrals in high-precision surface reconstruction.

[0127] 2) This paper uses the projection angle to assist in determining the normal quadrant, eliminating point mismatch, improving measurement efficiency, and effectively solving the point cloud hole problem.

[0128] 3) This disclosure provides a simple and effective solution to correct normal integration and alleviate missing data in point clouds with high robustness and accuracy.

[0129] The present disclosure also provides an electronic device comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to implement the aforementioned polarization-based binocular stereoscopic 3D reconstruction method. The electronic device can be provided as a terminal, server, or other device.

[0130] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implements the aforementioned polarization-based binocular stereoscopic 3D reconstruction method. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0131] Those skilled in the art will understand that in the above-mentioned polarization-based binocular stereo vision three-dimensional reconstruction method and system of the specific implementation method, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0132] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the prescribed logical function. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the prescribed function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0133] While various embodiments of the present disclosure have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technical improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A three-dimensional reconstruction method for binocular stereo vision based on polarization state, characterized in that: The steps include: S10, obtaining polarization images at different polarization degrees from a polarization camera; S20. Calculate the angle of the normal vector based on the acquired polarization image and Fresnel theory, wherein the angle of the normal vector includes the zenith angle and azimuth angle of the object surface; and obtain the azimuth angle of the object surface based on the relationship between the polarization azimuth angle and the azimuth angle, which is expressed as: , where is the azimuth, is the polarization azimuth; Calculating the angle of the normal vector based on Fresnel theory also includes: solving the relationship between the degree of polarization and the zenith angle through numerical iteration to obtain the zenith angle of the object surface, wherein the relationship between the degree of polarization and the zenith angle is expressed as: ， Where, is the zenith angle, is the complex refractive index, =n(1 + ik), i is the imaginary unit, k is the decay exponent; S30, eliminating the ambiguity of the azimuth angle by the projection angle of the normal vector on the YOZ plane, and eliminating mismatching by using the angle constraint of the normal vector; introducing the projection angle of the normal vector on the YOZ plane based on the relationship between the polarization azimuth and the azimuth angle, simulating rotation through the binocular vision system, determining the sign of the projection angle, and determining the quadrant in which the azimuth angle is located; S40, performing gradient integration on the angle of the normal vector after eliminating mismatching to complete the three-dimensional reconstruction.

2. The method according to claim 1, characterized in that The angle of the normal vector is calculated based on Fresnel theory, including: calculating the Stokes parameters based on the acquired polarization image, and further obtaining the polarization degree and polarization azimuth, which are expressed as: ， ， ， ， ， Where, 、 、 and The polarization images represent the linear polarization degrees at 0°, 45°, 90°, and 135°, respectively. is the degree of polarization.

3. The method according to claim 1, characterized in that The gradient integration of the normal vector angle after eliminating mismatching is performed to complete the 3D reconstruction, including: converting the zenith angle and azimuth angle into a 3D normal vector, expressed as: ， Where, is the normal vector of any point on the surface.

4. The method according to claim 3, characterized in that The 3D reconstruction is completed by performing gradient integration on the angle of the normal vector after eliminating the mismatch, and also includes: Convert the three-dimensional normal vector into a discrete approximate gradient through local linear approximation; The discrete approximate gradient is fitted to the surface gradient field by minimizing the cost function, and the surface shape of the object is reconstructed by numerical integration.

5. A binocular stereo vision 3D reconstruction system based on polarization state, characterized in that: include: A polarization image acquisition module is used to acquire polarization images of different polarization degrees from a polarization camera; The Fresnel theory calculation module calculates the angle of the normal vector based on the acquired polarization image and Fresnel theory, where the angle of the normal vector includes the zenith angle and azimuth angle of the object surface. Based on the relationship between the polarization azimuth angle and the azimuth angle, the azimuth angle of the object surface is obtained, which is expressed as: , where is the azimuth, is the polarization azimuth; Calculating the angle of the normal vector based on Fresnel theory also includes: solving the relationship between the degree of polarization and the zenith angle through numerical iteration to obtain the zenith angle of the object surface, wherein the relationship between the degree of polarization and the zenith angle is expressed as: ， Where, is the zenith angle, is the complex refractive index, =n(1 + ik), i is the imaginary unit, k is the decay exponent; The mismatch elimination module eliminates the ambiguity of the azimuth angle by using the projection angle of the normal vector on the YOZ plane and eliminates mismatches by using the angle constraint of the normal vector. Based on the relationship between the polarization azimuth and the azimuth angle, the projection angle of the normal vector on the YOZ plane is introduced. The binocular vision system simulates rotation, determines the sign of the projection angle, and determines the quadrant of the azimuth angle. The 3D reconstruction module performs gradient integration on the angle of the normal vector after eliminating mismatching to complete the 3D reconstruction.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the polarization-state-based binocular stereo vision three-dimensional reconstruction method according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the polarization-state-based binocular stereo vision three-dimensional reconstruction method according to any one of claims 1 to 4 is implemented.

Images