Target-enhanced three-dimensional reconstruction method based on multi-wavelength polarized line structured light

Abstract

The application provides a target enhancement three-dimensional reconstruction method based on a multi-wavelength polarized line structured light. The method introduces a polarization imaging technology, which is independent of light source intensity and depends on an incident angle and surface properties, and solves a measurement ambiguity problem caused by insufficient contrast of multi-band light source intensity. In addition, in order to alleviate the problem of inconsistent response of different wavelengths under wideband illumination, a linear polarization degree is used as an observation reference. Then, by decoupling the incident angle and material response and using the inherent correlation between the wavelength and the polarization, the estimation of the root mean square absolute roughness of the object is realized. Finally, the estimated absolute roughness is taken as a keyword and integrated into the three-dimensional reconstruction process to enhance the target extraction effect.

Description

Technical Field

[0001] This invention belongs to the field of optical measurement and three-dimensional imaging technology, specifically relating to a target enhancement three-dimensional reconstruction method based on multi-wavelength polarized structured light. Background Technology

[0002] Linear structure 3D reconstruction is an active optical imaging technique that typically relies on highly coherent laser sources rather than traditional illumination. However, the monochromaticity of lasers limits the acquisition of complete surface features, while using multi-band sources can lead to measurement ambiguity due to insufficient intensity contrast. Summary of the Invention

[0003] To address the challenges of obtaining complete surface features from laser sources and the measurement ambiguity caused by insufficient intensity contrast of multi-band light sources, this invention proposes a three-dimensional reconstruction framework for linear structures based on multi-band light sources. It introduces polarization imaging technology, which relies solely on the incident angle and surface properties, independent of light source intensity. Furthermore, to alleviate the inconsistency in response across different wavelengths under broadband illumination, linear polarization is used as the observation benchmark. By decoupling the incident angle from the material response and utilizing the intrinsic correlation between wavelength and polarization, the root mean square absolute roughness of the object is estimated. Finally, the estimated absolute roughness features are integrated into the three-dimensional reconstruction process to enhance target extraction.

[0004] Therefore, this invention proposes a target enhancement 3D reconstruction method based on multi-wavelength polarized structured light, comprising the following steps: S1: Project a multi-band light source linear polarization stripe pattern onto the surface of the object under test, wherein the surface of the object under test is made of a dielectric material; S2: A linearly polarized stripe pattern group is acquired by camera imaging, and the degree of linear polarization is calculated when the incident angle between the linearly polarized light source and the camera imaging angle is close to zero. The incident direction of the linearly polarized light source is normal incidence. S3: Obtain the absolute roughness of the surface of the object to be tested based on the degree of linear polarization; S4: Using the absolute roughness obtained in S3 as the observation value, perform interest domain selection and 3D reconstruction on the object to be measured.

[0005] The present invention has the following beneficial effects: 1. This invention introduces surface roughness, which is independent of light source intensity and depends only on the incident angle and the surface characteristics of the object, as the core observation value to improve measurement robustness and contrast.

[0006] 2. This invention combines the correlation analysis of wavelength and polarization characteristics to correct systematic errors caused by inconsistent behavior of multiple wavelengths. By leveraging polarization characteristic modulation, it fully utilizes the advantages of multi-band light sources, effectively overcoming the limitations of traditional laser light sources in changing the appearance of target surfaces. Based on surface roughness as a threshold, it more efficiently enhances targets in complex environments while maintaining measurement accuracy comparable to traditional methods. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the technical process of the present invention; Figure 2 This is a schematic diagram comparing the linearly polarized stripe patterns under the method in this embodiment of the invention with those under the laser light source projection method; Figure 3 This is a diagram illustrating random rough surfaces of the target surface under different roughnesses in an embodiment of the present invention. Figure 4 These are various feature curves of the relative roughness of the target surface under test in the embodiments of the present invention; Figure 5 This is a comparison diagram showing the depth maps reconstructed by the method in this embodiment of the invention and the line laser scanning method. Figure 6 This is a diagram illustrating the measurement depth curves of the same location of the reconstructed target using the measurement method and the line laser measurement method provided in this embodiment of the invention. Figure 7 This is a schematic diagram showing a comparison of depth maps between the filtering method provided in this embodiment of the invention and the polarization coding filtering method; Figure 8 This is a diagram showing the proportion of point cloud reconstruction of the target under different screening methods in an embodiment of the present invention. Detailed Implementation

[0008] This invention proposes a target enhancement 3D reconstruction method based on multi-wavelength polarized structured light, such as... Figure 1 As shown, it includes the following steps: S1: Project a multi-band light source linear polarization stripe pattern onto the surface of the object under test, the surface of which is made of dielectric material.

[0009] Broadband composite white light is used as a multi-band light source. It adopts a DC regulated fiber-coupled light source and then collimates the light into parallel light through a convex lens. Then, the stray light generated by ambient lighting and lens aberration is filtered out by an aperture. After that, the filtered light is focused by a second convex lens onto a combined unit consisting of a cylindrical lens and a polarizer, and finally a linearly polarized stripe pattern is projected onto the surface of the object under test. The surface of the object under test is made of dielectric material.

[0010] like Figure 2As shown in (b), the area within the dashed box is an enlarged image of the stripe details. The stripes obtained using the method of this invention retain the true features of the text surface, while Figure 2 (a) shows that the stripes using a laser light source cause text distortion and disappearance.

[0011] S2: The linearly polarized stripe pattern group is obtained by camera imaging, and the degree of linear polarization is calculated when the incident angle between the linearly polarized light source and the camera imaging angle is close to zero. The incident direction of the linearly polarized light source is normal incidence.

[0012] The linearly polarized fringe pattern set is obtained by camera imaging based on a three-color model. The linearly polarized fringe pattern set is an orthogonal image pair. The degree of linear polarization is obtained by substituting the obtained orthogonal image pairs into polarization difference imaging. (1) in, Indicates the degree of linear polarization. and These represent the light intensities of orthogonal image pairs that are parallel to and perpendicular to the reference direction, respectively.

[0013] Without loss of generality, the degree of linear polarization at the air-medium interface, i.e., when the incident angle of the linearly polarized light source and the camera imaging angle are close to zero, is: (2) in, and The elements in the Mueller matrix describe the effect of polarization elements on the polarization state. This demonstrates the target object's ability to transmit, scatter, and reflect the intensity of incident light, with the linearly polarized light source incident in the positive direction. This demonstrates the target object's ability to depolarize and delay the phase of incident light. and It can be represented as: (3) In the formula The scattering coefficient is... The first character in the subscript indicates the polarization direction of the incident light, and the second character indicates the polarization direction of the outgoing light, corresponding to the s-component (vibration perpendicular to the incident plane) and the p-component (vibration parallel to the incident plane), respectively. Both the incident and outgoing light contain both s and p polarization components. , , , Indicates four types Scattering coefficient.

[0014] S3: Obtain the absolute roughness of the surface of the object to be measured based on the degree of linear polarization.

[0015] According to classical micro-element scattering theory, for a surface whose surface height follows a Gaussian random distribution, its specular reflection component attenuates with increasing relative roughness (σ / λ), where λ represents wavelength and σ represents absolute roughness. When the relative roughness is between 0.05 and 0.3, scattering is enhanced, specular and diffuse reflection coexist, and the degree of linear polarization varies with the relative roughness. The Beckmann model shows that this attenuation follows... In the form of . When the angle between the incident and imaging points is close to zero, the equation simplifies to Therefore, the theoretical dependence of linear polarization degree in the mirror direction on relative roughness can be expressed as follows: That is, with Relatedly, the physical basis for establishing the empirical formula of the relationship between linear polarization degree and roughness of various materials under incident light is established. The amplitude a represents the maximum linear polarization degree under the zero roughness limit (σ→0), which depends on the Fresnel reflectivity of the surface medium. By introducing the depolarization coefficient b and the contribution of diffuse reflection polarization degree c, the fitting relationship model is obtained as follows (4): (4) Calculations obtained from different materials and Substitute into formula (2) to obtain the corresponding linear polarization degree. Finally, based on the fitting relationship model, inversely calculate the model parameters to obtain the fitting results, thereby obtaining the values ​​of amplitude a, depolarization coefficient b and diffuse reflection polarization degree contribution c.

[0016] Simulation of the Mueller matrix at different wavelengths using the finite-difference time-domain method and The influence of the model was investigated to obtain a Gaussian random rough surface. The model parameters were solved by rigorous coupled-wave analysis, and the electromagnetic response of the rough surface was calculated using the finite-difference time-domain method.

[0017] The height fluctuations of dielectric material surfaces with different roughnesses all follow a Gaussian distribution law. For this type of surface with specific distribution characteristics, the height value of any point on its random rough surface is further obtained by using the frequency domain filtering method, generating random rough surfaces with different roughnesses. Then, the scattering coefficients obtained in the time domain method of the finite difference time domain method are substituted into formula (3) to obtain the height value. and Therefore, the calculation constants for different dielectric materials can be determined, and the calculation results are shown in Table 1 below: Table 1: Roughness Calculation Constants for Different Dielectric Materials

[0018] Among them, R 2 R represents the coefficient of determination of the fitted model, RMSE is the root mean square error, and RMSE < 0.1 indicates a positive R-squared value. 2>0.9 reflects the data trend within a certain precision range.

[0019] like Figure 3 As shown, it is the surface of the object to be tested selected. In this material, a linearly polarized point light source with a wavelength of 0.4 μm is incident orthogonally onto a random rough surface with a fixed correlation length of 0.1 μm and different roughness variations, where the roughness variation is a relative roughness change from 0.05 to 0.3.

[0020] Figure 3 Random rough surfaces under varying relative roughness and Characteristic curve diagram as follows Figure 4 As shown in (a), it can be seen that as the relative roughness of the surface of the object to be measured increases, its and Consequently, the scattering intensity decreases.

[0021] For example Figure 4 As shown in (b), this is to adopt The characteristic curve of the linear polarization degree of the test object and its relative roughness shows that the linear polarization degree decreases as the relative roughness increases.

[0022] At the same time, such as Figure 4 As shown in (c), it is a characteristic curve showing the relationship between the linear polarization degree and the relative roughness of the object under test using three dielectric materials. The linear polarization degree of all three materials decreases as the relative roughness increases.

[0023] Therefore, by transforming formula (4), the absolute roughness is obtained as: (5) Therefore, the absolute roughness is obtained by using DOLP and wavelength. Since a camera based on the RGB three-color model is used, the three center wavelengths of red, green and blue are obtained at the same pixel. Then, the DOLP corresponding to different wavelengths is calculated by formula (1). At the same time, the absolute roughness is calculated by substituting the wavelength and taking the average value, so as to obtain the absolute roughness of the surface of the object to be measured. (6) in, The wavelength values ​​represent different wavelengths, and n is the number of wavelengths involved in the calculation. By using formula (5) to introduce the analysis of different wavelengths, the problem of different linear polarization degrees of the same material at different wavelengths is transformed into solving for absolute roughness.

[0024] S4: Using the absolute roughness obtained in S3 as the observation value, perform interest domain selection and 3D reconstruction on the object to be measured.

[0025] The absolute roughness of the object under test is used as a keyword for filtering regions of interest, and a three-dimensional model of the multi-band light source line structure is performed.

[0026] Linear structured light, with intensity as the observed value, and multi-band polarized structured light are compared using the gray-scale centroid method at the same center point. Figure 5 The image shows depth maps reconstructed by the method of this invention and the classical line laser scanning method. Figure 5 (a) is the reconstruction target provided in this embodiment, and the dashed lines in the figure represent... Figure 6 The target location shown by the depth curve indicates that the left side represents a sculpture with higher roughness, while the right side represents a triangular prism with lower roughness. Both objects have a surface medium of... ; Figure 5 (b) is a depth map reconstructed using the line laser scanning method, where some discrete points represent reconstruction noise from the measurement results. Figure 5 (c) is the depth map reconstructed by the method of the present invention. It can be seen that the method of the present invention can also perform high-precision reconstruction of the target surface, just like the line laser scanning method.

[0027] at the same time, Figure 6 The image shows the depth curves of the measurement method of this invention and the conventional line laser measurement method at the same location of the reconstructed target. Figure 5 (a) As shown by the dashed lines on the triangular prism, although the central part of the prism surface should be flat, there is always an error tilt due to relative position relative to the camera. Therefore, a linear regression method is used to generate a local reference line. The coefficient of determination R1 of the depth curve reconstructed by the method of this invention is... 2 The coefficient of determination R² for depth curves reconstructed by conventional line laser measurement is 0.992. 2 The value is 0.984, and the root mean square errors of the two are 1.152 and 1.275, respectively. Figure 6 The linear regression results show that the depth curve measured by the method of the present invention is comparable to the data measured by traditional line laser.

[0028] Then, a thorough quantitative analysis is performed on the discrete 3D point cloud, and... Figure 5 (a) The triangular prism on the right is considered the region of interest. Figure 7 The image shows the region-of-interest depth maps retrieved by the roughness screening method provided by this invention and the traditional polarization coding screening method. Some discrete points in the image are noise point clouds caused by screening errors. Figure 7 (a) shows the results of the traditional polarization coding screening method. For conventional polarized structured light, under the same threshold width and the condition of completely recovering the triangular prism, Figure 5 (a) The higher roughness of the sculpture on the left cannot be well suppressed; while Figure 7(b) shows the roughness screening method of the present invention. The figure shows that the method of the present invention helps to identify targets from complex environments and maintains the integrity of the targets well.

[0029] like Figure 8 As shown, during stripe scanning, no background is set, only the background is scanned. Figure 5 (a) The triangular prism is scanned. When the number of effective point clouds and the total number of pixel point clouds are equal, and the integrity of the triangular prism itself is close to 1, the screening performance is evaluated by calculating the coverage ratio of the triangular prism in the overall reconstructed scene. Figure 8 This shows the percentage of point cloud reconstructions of the target object under different selection methods. Figure 8 (b) Shows the results of the absolute roughness screening method of the present invention, with absolute roughness between 0.05 and 0.1. Figure 8 (a) shows the results of the polarization coding screening method, with linear polarization degrees between 0.9 and 0.95; Figure 8 (b) It can be seen that, for the absolute roughness screening method proposed in this invention, the proportion of the target point cloud in the entire reconstructed scene is higher than that of the target point cloud. Figure 8 (a) Results of the polarization coding screening method.

[0030] Therefore, the method of the present invention can screen surfaces with different roughnesses for high-precision reconstruction while preserving the surface appearance.

Claims

1. A target enhancement 3D reconstruction method based on multi-wavelength polarized structured light, characterized in that, Includes the following steps: S1: Project a multi-band light source linear polarization stripe pattern onto the surface of the object under test, wherein the surface of the object under test is made of a dielectric material; S2: A linearly polarized stripe pattern group is acquired by camera imaging, and the degree of linear polarization is calculated when the incident angle between the linearly polarized light source and the camera imaging angle is close to zero. The incident direction of the linearly polarized light source is normal incidence. S3: Obtain the absolute roughness of the surface of the object to be tested based on the degree of linear polarization; S4: Using the absolute roughness obtained in S3 as the observation value, perform interest domain selection and 3D reconstruction on the object to be measured.

2. The target enhancement 3D reconstruction method based on multi-wavelength polarized structured light as described in claim 1, characterized in that, The specific implementation steps of step S1 are as follows: broadband composite white light is used as the multi-band light source. The broadband composite white light adopts a DC regulated fiber-coupled light source. Then, the light beam is collimated into parallel light through a convex lens. Subsequently, stray light generated by ambient lighting and lens aberration is filtered out by an aperture. After that, the filtered light beam is focused by a second convex lens onto a combined unit composed of a cylindrical lens and the polarizer, and finally the linearly polarized stripe pattern is projected onto the surface of the object to be tested.

3. The target enhancement 3D reconstruction method based on multi-wavelength polarized structured light as described in claim 2, characterized in that, The specific implementation steps of step S2 are as follows: the linearly polarized stripe pattern group is an orthogonal image pair. The orthogonal image pair is substituted into polarization difference imaging to obtain the degree of linear polarization, as shown in the following formula (1): （1） in, Indicates the degree of linear polarization. and The light intensities of the orthogonal image pairs, respectively, are parallel and perpendicular to the reference direction; The degree of linear polarization when the incident angle between the linearly polarized light source and the camera imaging angle is close to zero is shown in the following formula (2): （2） in, and For the elements in the Mueller matrix, and Represented as: （3） In the formula The scattering coefficient is... The first character in the subscript indicates the polarization direction of the incident light, and the second character indicates the polarization direction of the outgoing light, corresponding to the s-component (vibration perpendicular to the incident plane) and the p-component (vibration parallel to the incident plane), respectively. Both the incident and outgoing light contain both s and p polarization components. , , , Indicates four types Scattering coefficient.

4. The target enhancement 3D reconstruction method based on multi-wavelength polarized structured light as described in claim 3, characterized in that, The specific implementation steps of step S3 are as follows: The theoretical dependence of the linear polarization degree on the relative roughness in the mirror direction is expressed as follows: Where λ represents wavelength, σ represents absolute roughness, and σ / λ represents relative roughness, the relative roughness is between 0.05 and 0.3, and the degree of linear polarization varies with the relative roughness; then, depending on the Fresnel reflectivity of the surface medium, the depolarization coefficient b and the contribution of diffuse reflection polarization c are introduced to obtain the fitting relationship model, as shown in the following formula (4): （4） Where the amplitude 'a' represents the maximum degree of linear polarization under the zero roughness limit; Calculated based on the scattering coefficient of the medium material and Substitute into formula (2) to obtain the corresponding linear polarization degree. Finally, according to the fitted relational model, inversely calculate the model parameters to obtain the values ​​of the amplitude a, the depolarization coefficient b, and the diffuse reflection polarization degree contribution c. Transforming formula (4), we obtain the absolute roughness as follows: （5） Then, the linear polarization degree corresponding to different wavelengths is calculated using formula (1), and the absolute roughness is calculated by substituting the wavelength and taking the average value. Finally, the absolute roughness of the surface of the object to be measured is obtained: （6） in, This represents the wavelength values ​​of the different wavelengths, where n is the number of wavelengths involved in the calculation.

Images