Depth measurement and edge extraction method based on single all-dielectric metasurface

Abstract

This invention discloses a depth measurement and edge extraction method based on a single all-dielectric metasurface, belonging to the fields of micro-nano optics and image information extraction applications. The phase distribution of the all-dielectric metasurface is obtained using the Fresnel zone method. Each Fresnel region corresponds to a spiral phase with a different topological charge number, increasing sequentially from the inside out, generating a double-helix point spread function. The metasurface is placed on the spectral plane of a 4f system, with a pair of linear polarizers and a quarter-wave plate placed before and after it respectively, achieving corresponding circular polarization selection and ensuring phase modulation based on the geometric phase modulation principle, thus obtaining the designed metasurface optical system for depth measurement and edge extraction. When incoherent light is incident, the main lobe rotation angle of the double-helix point spread function corresponding to the "twin image" is extracted using image cepstral analysis, thus deducing the object's depth information. When coherent light is incident, the incident light interacts with the central part of the metasurface. Through the convolution of the target object with the spiral phase, coherent destructive phase can be achieved for regions with uniform phase / amplitude, extracting the edge information of the target object. This invention enables both depth measurement and edge extraction.

Description

Technical Field

[0001] This invention relates to a depth measurement and edge extraction method based on a single all-dielectric metasurface, belonging to the field of micro-nano optics and image information extraction application technology. Background Technology

[0002] Depth measurement and edge extraction play crucial roles in machine vision, medical imaging, and intelligent recognition. Traditional cameras can only capture specific 2D intensity information projections from a 3D scene when the depth is unknown. However, rapidly developing 3D imaging technology can acquire depth features by capturing additional light information, with depth measurement being a key technology in this process. Meanwhile, edge information contains important geometric features for various image processing steps and can be obtained by filtering low spatial frequencies. How to simultaneously acquire the depth, edges, and other light information of a scene to accurately perceive the physical world is a significant challenge facing imaging systems.

[0003] Common depth measurement methods include stereo vision, time-of-flight, and point source / fringe projection, but these methods typically require active illumination or multiple viewpoints, increasing system complexity. Additionally, methods acquire depth information from a sequence of images with different defocus settings, but the measurement accuracy is fundamentally limited by the slow change of the point spread function with depth and the uncertainty of the defocus distance direction. Therefore, obtaining a point spread function capable of achieving more accurate depth characterization is crucial. The double-helix point spread function is an effective solution, generating a beam with two focal points that rotates continuously as the point source moves. Meanwhile, traditional edge extraction methods often utilize structured light for spatial differentiation. However, many spatial differentiation methods achieve one-dimensional differentiation, leading to anisotropic edge extraction and thus not being entirely suitable for imaging applications. In recent years, helical phase has provided a novel approach for edge extraction. Any radial line in the helical phase introduces a π-phase difference between the positive and negative spatial frequencies of the incident light, resulting in enhanced isotropic edge contrast of the observed object, thus achieving isotropic edge extraction. However, the spatial light modulator (SLM) or spiral phase plate that generates the spiral phase is relatively large, limiting the miniaturization and integration of optical systems. Furthermore, simultaneously acquiring depth and edge information of a scene typically requires designing more complex and bulkier optical components and active illumination conditions.

[0004] In recent years, optical metasurfaces have provided a novel approach for developing unconventional ultrathin devices with multiple functions, and can meet the needs of miniaturization and integration of optical systems, leading to rapid development in numerous practical applications. In the field of depth measurement, depth imaging methods using metalens arrays to construct light field cameras, and three-dimensional imaging methods using two spatially interleaved multiplexed off-axis focusing metalenses, have been proposed and experimentally verified. Regarding edge extraction, spatial differentiation methods using all-optical simulation computation can achieve optical edge extraction. Examples include optical spatial differentiators that use a photonic crystal plate as a Laplace operator to convert the image into its second derivative based on the surface plasmon polariton interference effect; or spatial differentiators that utilize overlapping electric and magnetic dipole resonances to locally and independently manipulate the amplitude and phase of orthogonally polarized emitted electromagnetic waves, combined with spatial Fourier transform. However, previous work has typically limited itself to performing a single function in either depth measurement or edge extraction. Detection systems combining multiple functions are often very bulky. There is an urgent need for a solution that integrates both depth measurement and edge extraction using a single metasurface. Summary of the Invention

[0005] The main objective of this invention is to provide a depth measurement and edge extraction method based on a single all-dielectric metasurface. This method utilizes the characteristic that the double-helix point spread function rotates around the optical axis along the propagation direction, and leverages the helical phase characteristics of the middle region of the phase profile corresponding to the single all-dielectric metasurface. This allows for depth measurement and edge extraction through a single all-dielectric metasurface.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] This invention discloses a depth measurement and edge detection method based on a single all-dielectric metasurface. Using a single metasurface optical system, under both incoherent and coherent light incident conditions, it leverages the double-helix point spread function generated by the metasurface and the helical phase characteristics of the metasurface's central region to achieve two different functions: depth measurement and edge detection. The phase distribution of the all-dielectric metasurface is obtained according to the Fresnel zone method, consisting of a series of Fresnel zones. Each Fresnel region corresponds to a helical phase with a different topological charge number, increasing sequentially from the inside out, generating a double-helix point spread function, and the central region of this phase distribution possesses helical phase characteristics. The all-dielectric metasurface is fabricated on a glass substrate through deposition, photolithography, lift-off, and etching. The metasurface is placed on the spectral plane of a 4f system, and by placing a pair of linear polarizers and a quarter-wave plate before and after it, respectively, the corresponding circularly polarized light is selected, ensuring phase modulation based on the geometric phase modulation principle, thus obtaining the designed metasurface optical system for depth measurement and edge detection. Experiments using incoherent light are conducted to calibrate the distance-angle relationship by recording the corresponding changes in the main lobe rotation angle of the double-helix point spread function generated by the metasurface optical system as the depth position of the point light source changes. When incoherent light is incident, it illuminates the object and is then imaged onto a CCD via a 4-f system and an all-dielectric metasurface encoding the double-helix point spread function on the spectral plane. This results in a pair of "twin images." By using image cepstral analysis, the main lobe rotation angle of the double-helix point spread function corresponding to the "twin images" can be extracted, allowing the derivation of the object's depth information. When coherent light is incident, the incident light interacts with the central part of the metasurface. Through convolution of the target object with the helical phase, coherent destructive phase / amplitude cancellation can be achieved in regions with uniform phase / amplitude, leaving only high-contrast regions (i.e., the object's edges), allowing the extraction of the target object's edge information. The metasurface optical system described in this invention enables both depth measurement and edge extraction.

[0008] The depth measurement and edge extraction method based on a single all-dielectric metasurface disclosed in this invention includes the following steps:

[0009] Step 1: Design the phase distribution of the all-dielectric metasurface based on the Fresnel zone method. This phase distribution consists of a series of Fresnel zones, each corresponding to a spiral phase with a different topological charge number, increasing sequentially from the inside to the outside. A double-helix point spread function is generated through this phase distribution, with the middle region exhibiting spiral phase characteristics.

[0010] The phase distribution of the generated double-helix point spread function is expressed as:

[0011]

[0012] in, Position in polar coordinates The phase at the point is given by R, where R is the maximum radius. The phase distribution of the generated double-helix point spread function consists of a series of Fresnel bands, where N is the number of Fresnel bands. Each Fresnel region corresponds to a spiral phase with a different topological charge number, increasing sequentially from the inside out, with each layer increasing by 2. The middle region of the phase profile, i.e., the innermost Fresnel band, exhibits spiral phase characteristics. Characterizing this phase distribution on the spectral surface of the 4-f system, the double-helix point spread function can be obtained near the image-side focal point.

[0013] The two-dimensional distribution of the double-helix point spread function on the focal plane consists of two discrete main lobes of equal size and brightness. The azimuth angle of the line connecting these two discrete main lobes rotates with defocusing, and the rotation angle increases with the amount of defocusing. The expression for the defocusing rotation rate of the main lobe is:

[0014]

[0015] In the formula, NA represents the numerical aperture of the 4-f system optical path. Within the depth measurement range, NA≈R / f. Δl represents the difference in topological charge of each layer. According to equation (1), Δl=2. According to equation (2), the rotation rate of the double-helix point spread function changes with the number of Fresnel bands N, while the depth measurement range changes in the opposite direction.

[0016] Step 2: Optimize the dielectric nanoantenna that forms the metasurface using rigorous coupled-wave analysis or finite-difference time-domain method. Then, use the in-plane orientation angle of the dielectric nanoantenna to encode the phase distribution of the metasurface through the geometric phase modulation principle.

[0017] Using dielectric nanoantennas with high transmittance due to anti-rotation circular polarization, the in-plane azimuth angles of the dielectric nanoantennas at various positions on the metasurface are determined according to the spatial phase distribution of the target light field. Based on the chiral selectivity of the geometric phase, when left- or right-rotation circularly polarized incident light is incident on a dielectric nanoantenna with an azimuth angle of θ, it can form a mutually conjugate phase modulation of ±2θ for right- or left-rotation circularly polarized outgoing light. Here, "+" or "-" is determined by the specific polarization state combination of the incident and outgoing light. Based on the above principle, the phase distribution encoding of the metasurface is achieved by arranging the in-plane azimuth angles of the dielectric nanoantennas. The polarization state combinations include left- or right-rotation combinations and right- or left-rotation combinations.

[0018] The dielectric nanoantenna is an amorphous silicon nanorod antenna.

[0019] The shape and size of the amorphous silicon nanorod antenna are determined using either Rigorous Coupled-Wave Analysis (RCWA) or Finite-Difference Time-Domain (FDTD). During the optimization of the amorphous silicon nanorod antenna structure, the height, period, and operating wavelength of the amorphous silicon nanorod antenna are fixed. Then, under these conditions, the length and width of the amorphous silicon nanorod antenna are scanned to select structural dimensions that exhibit high anti-rotational circular polarization transmittance and low in-rotational circular polarization transmittance at the operating wavelength. The phase distribution encoding of the metasurface is achieved using the in-plane orientation angle of the dielectric nanoantenna through geometric phase modulation.

[0020] Step 3: The all-dielectric metasurface, composed of dielectric nanoantennas, is fabricated on a glass substrate using deposition, photolithography, lift-off, and etching methods. The all-dielectric metasurface is placed on the spectral plane of a 4-f system, and a pair of linear polarizers and a quarter-wave plate are placed before and after it, respectively, to achieve the selection of corresponding circularly polarized light. This ensures phase modulation based on the geometric phase modulation principle, resulting in the designed metasurface optical system for depth measurement and edge extraction.

[0021] Step 4: Under incoherent light incidence, using the metasurface optical system constructed in Step 3, record the corresponding changes in the main lobe rotation angle of the double-helix point spread function generated by the metasurface optical system when the depth position of the point source changes, thereby achieving the calibration of the distance-angle relationship.

[0022] Step 5: Under incoherent light incidence, the object is imaged as a pair of "twin images" through the metasurface optical system constructed in Step 3. By using image cepstral analysis, the main lobe rotation angle of the double helix point spread function corresponding to the "twin images" is extracted, and the depth information of the object is deduced.

[0023] Incoherent incident light illuminates an object, passes through a metasurface optical system, and is ultimately imaged onto a CCD, resulting in an image composed of a pair of "twin images." The distance between the two images corresponds to the distance between the main lobes of the point spread function, while the rotation angle depends on the depth of the sample. Therefore, by using image cepstral analysis to extract the main lobe rotation angle of the double-helix point spread function corresponding to the image, the depth information of the object can be deduced.

[0024] The image cepstral analysis method includes the following steps:

[0025] 1) Incoherent incident light irradiates the object, and then passes through the 4-f system and the all-dielectric metasurface encoding the double-helix point diffusion function on the metasurface optical system constructed in step 3. The image is then imaged on the CCD to obtain image I composed of a pair of "twin images".

[0026] 2) The twin image I is processed using a two-dimensional Hann window to obtain image I', thereby improving the reliability of peak recognition. The Hann window used is shown in formula (3):

[0027]

[0028] Where m and n are the spatial coordinates of the entire image on the two axes, S w It refers to the size of the window.

[0029] 3) Perform cepstrum calculation on image I' to obtain its cepstrum C. The cepstrum is the inverse Fourier transform of the power spectrum in logarithmic coordinates and is often used as a key tool in signal processing, such as filtering, denoising, and reconstruction. The cepstrum operation of an image is represented as:

[0030] C = C{I′} : = F -1 {log(|F{I′}| 2 (4)

[0031] Where F represents the Fourier transform.

[0032] 4) The cepstral results are cropped based on the distance between the main lobes and the width of the main lobe of the double-helix point spread function. The two peak positions marked by the dashed annular region are then obtained in the cepstral C. Gaussian filtering is then applied to the dashed annular region in the cepstral C to reduce the influence of noise on the peak positions. Finally, the rotation angles corresponding to the positions of the two peaks are calculated.

[0033] 5) By matching the extracted peak rotation angle with the calibrated distance-angle relationship, the axial depth information of the object can be measured.

[0034] Step Six: Under coherent light incidence, the object passes through the metasurface optical system constructed in Step Three. Any two points symmetrical about the origin in the central region of the metasurface have the same amplitude, but their phases differ by π. Through convolution of the object with the spiral phase, regions with uniform phase / amplitude will coherently cancel each other out due to destructive interference, leaving only high-contrast regions, thus achieving edge extraction.

[0035] Beneficial effects:

[0036] 1. The depth measurement and edge detection method based on a single all-dielectric metasurface disclosed in this invention utilizes the same metasurface optical system under both incoherent and coherent light incident conditions. It leverages the double-helix point spread function generated by the metasurface and the helical phase characteristics of the intermediate region of the metasurface to achieve two different functions: depth measurement and edge detection. This invention is more integrated and compact than traditional optical detection systems.

[0037] 2. The depth measurement and edge extraction method based on a single all-dielectric metasurface disclosed in this invention utilizes the coherence of incident light as a new multiplexing dimension to achieve different functions under different incident conditions, thereby realizing the multiplexing of functions of a single system.

[0038] 3. The depth measurement and edge extraction method based on a single all-dielectric metasurface disclosed in this invention changes the depth measurement range by adjusting the number N of Fresnel wavebands.

[0039] 4. The depth measurement and edge extraction method based on a single all-dielectric metasurface disclosed in this invention allows any two points symmetrical about the origin in the central region of the metasurface to have the same amplitude but a phase difference of π. By convolving the object with the spiral phase, regions with uniform phase / amplitude will coherently cancel each other out due to destructive interference, leaving only high-contrast regions, thus achieving edge extraction.

[0040] 5. The depth measurement and edge extraction method based on a single all-dielectric metasurface disclosed in this invention can be applied to fields such as optical inspection, machine vision, pattern recognition, autonomous vehicles, virtual reality, and augmented reality. Based on the depth measurement and edge extraction of a single all-dielectric metasurface, it improves the effects of optical inspection, machine vision, pattern recognition, autonomous vehicles, virtual reality, and augmented reality. Attached Figure Description

[0041] Figure 1 This is a flowchart of the depth measurement and edge extraction method based on a single all-dielectric metasurface disclosed in this invention;

[0042] Figure 2 This is a schematic diagram of the experimental apparatus used in Example 1 for calibrating the distance-angle relationship, measuring the depth of the target object, and extracting the edge of the target object;

[0043] Figure 3 (a) is an experimental result diagram of the double-helix point diffusion function generated by point light sources at different depths through a metasurface optical system in Example 1. Figure 3 (b) is a graph showing the relationship between the rotation angle of the calibrated double helix point spread function and the distance to the target object;

[0044] Figure 4 The flowchart of the image cepstral analysis algorithm used in Example 1 to extract the main lobe rotation angle of the double helix point spread function to obtain the depth information of the target object is shown.

[0045] Figure 5 The figure shows the experimental results of depth measurement and edge extraction of the all-dielectric metasurface sample processed in Example 1. Figure 5 (a) is the imaging result corresponding to the distance Δz between the object and the object-side focal point when performing depth measurement under incoherent illumination conditions; Figure 5 (b) is the position of the cepstral peak when the object is at a distance Δz from the object-side focal point; Figure 5 (c) represents the main lobe rotation angle of the double helix point spread function extracted from images at different depths, and the double helix phase rotation angle generated by point light sources at different locations; Figure 5 (d) shows the experimental results of edge extraction under coherent light illumination.

[0046] Among them, 1-coherent incident light, 2-incoherent incident light, 3-semi-transparent mirror, 4-target object, 5-first focusing lens, 6-first linear polarizer, 7-metasurface, 8-second linear polarizer, 9-second focusing lens, 10-first quarter-wave plate, 11-second quarter-wave plate, 12-CCD camera. Detailed Implementation

[0047] To better illustrate the purpose and advantages of the present invention, the invention will be further described below in conjunction with the accompanying drawings and examples.

[0048] Example 1

[0049] Based on the property that the double-helix point spread function rotates around the optical axis along the propagation direction, and the helical phase characteristics of the corresponding phase profile in the middle region, depth measurement and edge extraction are achieved through a single all-dielectric metasurface.

[0050] like Figure 1 As shown, this embodiment includes the following steps:

[0051] Step 1: Generate a double-helix point spread function based on the Fresnel zone method to obtain a phase distribution that can generate a double-helix point spread function and has helical phase characteristics in the middle region.

[0052] The phase distribution can be directly expressed as:

[0053]

[0054] in, Position in polar coordinates The phase at the given location is defined by R, where R is the maximum radius. Clearly, this phase consists of a series of Fresnel zones, with N representing the number of Fresnel zones. Each Fresnel zone corresponds to a spiral phase with a different topological charge number, increasing sequentially from the inside out, with each layer increasing by 2. The middle region of the phase profile, i.e., the innermost Fresnel zone, exhibits spiral phase characteristics. The variation in the value of N in the formula affects the shape and rotation rate of the double-helix point spread function (DFD) of the 4-f system. A larger N value leads to excessively large leaf spacing and a low rotation rate in the DFD, resulting in poor imaging performance and limited depth measurement accuracy. Conversely, a smaller N value results in excessively small leaf spacing and a high rotation rate, making it difficult to distinguish and also limiting the depth measurement range. Therefore, this embodiment selects N=4 for phase generation, ensuring that the double-helix point spread function generated by the metasurface has good distinguishability and high measurement accuracy, thereby achieving excellent depth measurement functionality.

[0055] Step 2: Design the dielectric nanoantenna that forms the metasurface hologram using rigorous coupled-wave analysis or the finite-difference time-domain method. Encode the phase distribution of the metasurface hologram using the in-plane orientation angle of the dielectric nanoantenna through geometric phase modulation.

[0056] Through a series of simulations based on rigorous coupled-wave analysis and finite-difference time-domain methods, amorphous silicon nanorod antennas were selected as the basic building blocks of all-dielectric metasurface holograms, and the phase distribution of the metasurface holograms was encoded using the geometric phase modulation principle.

[0057] During the structural design process, the height of the amorphous silicon nanorod antenna was fixed at 600 nm, the period in both the x and y directions was fixed at 300 nm, and the operating wavelength was set to 633 nm. The length and width of the amorphous silicon nanorod antenna were then scanned under these conditions. Due to the use of geometric phase modulation, structural dimensions with high anti-rotational circular polarization transmittance and low in-rotational circular polarization transmittance at the operating wavelength were selected. Considering the balance between fabrication accuracy and structural performance, this embodiment selected an amorphous silicon nanorod antenna with a length of 134 nm and a width of 70 nm.

[0058] Step 3: The all-dielectric metasurface composed of dielectric nanoantennas is fabricated on a glass substrate by deposition, photolithography, stripping and etching methods, and placed on the spectral plane of the 4-f system to form the designed metasurface optical system for depth measurement and edge extraction.

[0059] Its specific implementation method includes the following steps:

[0060] 1) A 600 nm thick amorphous silicon film was prepared by plasma-enhanced chemical vapor deposition (PECVD). Subsequently, a polymethyl methacrylate (PMMA) resist layer was spin-coated onto the amorphous silicon film, and the film was baked on a hot plate to remove the solvent.

[0061] 2) The desired structure was fabricated using standard electron beam lithography, then the sample was developed in MIBK:IPA solution and washed with IPA. After that, a chromium layer was coated by electron beam evaporation.

[0062] 3) The stripping process is completed in hot acetone;

[0063] 4) The desired structure is converted from chromium to silicon using inductively coupled plasma reactive ion etching (ICP-RIE), and finally the residual chromium mask is removed using a standard wet etching process.

[0064] According to the design, the all-dielectric metasurface sample fabricated in the embodiment has a size of 480 μm × 480 μm and contains 1600 × 1600 pixels. Each amorphous silicon nanorod antenna has a length of 134 nm and a width of 70 nm.

[0065] Step 4: Conduct an experiment using incoherent light to obtain the specific change in the main lobe rotation angle of the double helix point spread function generated by the metasurface optical system when the depth position of the point light source changes, and calibrate the distance-angle relationship.

[0066] The schematic diagram of the experimental setup used in this embodiment to calibrate the distance-angle relationship of the double-helix point spread function is shown below. Figure 2 As shown in the figure. The metasurface fabricated in step three is placed in the frequency domain of the 4-f system. To avoid additional limitations imposed on the optical path system by the lens NA, a large-diameter lens with a focal length of 50mm was selected to construct the 4-f system. A pair of linear polarizers and a quarter-wave plate were placed before and after the metasurface, respectively, to select the corresponding circularly polarized light and ensure phase modulation based on the geometric phase modulation principle. When calibrating the double-helix point spread function, an incoherent point light source was used as the incident light. The experimental results of the double-helix point spread function generated by the metasurface optical system through point light sources at different depths are shown in the figure. Figure 3 As shown in (a), in the experiment, the position of the point light source was adjusted using a rack and pinion translation stage, while the image of the double helix point spread function at the object-side focal point was recorded using a CCD. As the point light source moved back and forth from the object-side focal point, both main lobes of the double helix point spread function rotated around the image center, and the rotation angle increased with the defocusing amount of the point light source. During the rotation, the intensity, size, and distance between the main lobes remained essentially constant. The relationship between the rotation angle of the double helix point spread function and the distance to the target object, calibrated based on the experimental results, is shown in the figure below. Figure 3 As shown in (b).

[0067] Step 5: Under incoherent light incidence, the target object is imaged using a metasurface optical system. The main lobe rotation angle of the double-helix point spread function of the captured image is extracted by image cepstral analysis, thereby obtaining the depth information of the object.

[0068] A schematic diagram of the experimental setup used in this embodiment to measure the depth of the target object is shown below. Figure 2 As shown, incoherent incident light illuminates the object, then passes through a 4-f system and an all-dielectric metasurface encoding the double-helix point spread function on its spectral plane, ultimately forming an image on a CCD. This results in an image composed of a pair of "twin images," the distance between which corresponds to the distance between the main lobes of the point spread function, while the rotation angle depends on the depth of the sample. Therefore, the rotation angle of the main lobe of the double-helix point spread function corresponding to the image can be extracted using image cepstral analysis, thereby deducing the object's depth information. The flowchart of the image cepstral analysis algorithm used in this embodiment to extract the main lobe rotation angle of the double-helix point spread function to obtain the target object's depth information is shown below. Figure 4 As shown, its specific implementation method includes the following steps:

[0069] 1) Incoherent incident light irradiates the object, then passes through the 4-f system and the all-dielectric metasurface on its spectral surface that encodes the double-helix point spread function, and is finally imaged on the CCD to obtain image I composed of a pair of "twin images";

[0070] 2) The twin image I is processed using a two-dimensional Hann window to obtain image I', thereby improving the reliability of peak recognition. The Hann window used can be written as:

[0071]

[0072] Where m and n are the spatial coordinates of the entire image on the two axes, S w It refers to the size of the window.

[0073] 3) Perform cepstrum calculation on image I' to obtain its cepstrum C. The cepstrum is the inverse Fourier transform of the power spectrum in logarithmic coordinates and is often used as a key tool in signal processing, such as filtering, denoising, and reconstruction. The cepstrum operation of the image can be expressed as:

[0074] C = C{I′} : = F -1 {log(|F{I′}| 2 (4)

[0075] Where F represents the Fourier transform.

[0076] 4) The cepstral results are cropped based on the distance between the main lobes and the width of the main lobe of the double-helix point spread function. The two peak positions marked by the dashed annular region can be obtained in the cepstral C. Then, a Gaussian filter is applied to reduce the influence of noise on the peak positions, and the rotation angles corresponding to the positions of the two peaks are calculated.

[0077] 5) Finally, the extracted peak rotation angle is matched with the calibrated distance-angle relationship to measure the axial depth information of the object.

[0078] Figure 5 The figure shows the experimental results of depth measurement and edge extraction of the all-dielectric metasurface sample processed in Example 1. Figure 5 (a) is the imaging result corresponding to the distance Δz between the object and the object-side focal point when performing depth measurement under incoherent illumination conditions; Figure 5 (b) is the position of the cepstral peak when the object is at a distance Δz from the object-side focal point; Figure 5 (c) shows the main lobe rotation angle of the double helix point spread function extracted from images at different depths, and the double helix phase rotation angle generated by point light sources at different locations. It can be seen that the two are in good agreement.

[0079] Step 6: Under coherent light incident, the target object is imaged using a metasurface optical system. Due to the destructive interference, the regions of the target object with uniform phase / amplitude are coherently canceled out, leaving only high-contrast regions, which enables edge extraction.

[0080] A schematic diagram of the experimental setup used in this embodiment to extract the edge information of the target object is shown below. Figure 2 As shown in the figure, in the experiment, the incoherent incident light was replaced with coherent incident light, while other experimental setups remained unchanged. The experimental results are as follows. Figure 5 As shown in (d), when the incident coherent light interacts with the central portion of the metasurface, the target object convolves with the spiral phase. Regions with uniform phase / amplitude coherently cancel each other out due to destructive interference, leaving only high-contrast regions (i.e., the object's edges). Ultimately, the edge information of the target object can be obtained.

[0081] In summary, this embodiment provides a depth measurement and edge extraction method based on a single all-dielectric metasurface. This method places the metasurface on the spectral plane of a 4f system to form an optical system. It utilizes the coherence of incident light as a new multiplexing dimension. When incoherent light is incident, the main lobe rotation angle of the double-helix point spread function can be extracted using image cepstral analysis to obtain the object's depth information. When coherent light is incident, destructive interference can be used to achieve edge extraction. The metasurface optical system described in this invention has a compact structure, providing a framework and platform for constructing more integrated and compact optical detection systems. It holds great potential in fields such as optical detection, machine vision, pattern recognition, autonomous vehicles, virtual reality, and augmented reality.

[0082] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for depth measurement and edge extraction based on a single all-dielectric metasurface, characterized in that: Includes the following steps, Step 1: Design the phase distribution of the all-dielectric metasurface based on the Fresnel zone method. This phase distribution consists of a series of Fresnel zones, each Fresnel zone corresponding to a spiral phase with a different topological charge number, increasing sequentially from the inside to the outside. A double-helix point spread function is generated through the phase distribution, and the middle region has spiral phase characteristics. Step 2: Optimize the dielectric nanoantenna that makes up the metasurface using the rigorous coupled-wave analysis method or the finite-difference time-domain method. Then, use the in-plane orientation angle of the dielectric nanoantenna to encode the phase distribution of the metasurface through the geometric phase modulation principle. Step 3: The all-dielectric metasurface composed of dielectric nanoantennas is fabricated on a glass substrate by deposition, photolithography, lift-off and etching methods; the all-dielectric metasurface is placed on the spectral plane of the 4-f system, and by placing a pair of linear polarizers and a quarter-wave plate in front of it respectively, the selection of corresponding circularly polarized light is achieved, ensuring phase modulation based on the geometric phase modulation principle, and obtaining the designed metasurface optical system for depth measurement and edge extraction. Step 4: Under incoherent light incidence, using the metasurface optical system constructed in Step 3, record the corresponding change in the main lobe rotation angle of the double helix point spread function generated by the metasurface optical system when the depth position of the point source changes, thereby achieving the calibration of the distance-angle relationship. Step 5: Under incoherent light incident, the object is imaged as a pair of "twin images" through the metasurface optical system constructed in Step 3. The main lobe rotation angle of the double helix point spread function corresponding to the "twin images" is extracted by using image cepstral analysis, and the depth information of the object is deduced. Step 6: Under coherent light incident, the object passes through the metasurface optical system constructed in Step 3. Any two points in the metasurface central region that are symmetrical about the origin have the same amplitude, but the phase difference is π. Through the convolution of the object with the spiral phase, the regions with uniform phase / amplitude will coherently cancel each other out due to destructive interference, leaving only the high-contrast region, thus achieving edge extraction.

2. The depth measurement and edge extraction method based on a single all-dielectric metasurface as described in claim 1, characterized in that: In step one, The phase distribution of the generated double-helix point spread function is expressed as: in, Position in polar coordinates The phase at the point is R, where R is the maximum radius; the phase distribution of the generated double-helix point spread function consists of a series of Fresnel bands, where N is the number of Fresnel bands. Each Fresnel region corresponds to a spiral phase with a different topological charge number, increasing sequentially from the inside out, with each layer increasing by 2. The middle region of the phase profile, i.e., the innermost Fresnel band, has spiral phase characteristics; characterizing the phase distribution on the spectral surface of the 4-f system, the double-helix point spread function can be obtained near the image-side focal point; The two-dimensional distribution of the double-helix point spread function on the focal plane consists of two discrete main lobes of equal size and brightness. The azimuth angle of the line connecting the two discrete main lobes rotates with defocusing, and the rotation angle increases with the amount of defocusing. The expression for the defocusing rotation rate of the main lobe is: In the formula, NA represents the numerical aperture of the optical path of the 4-f system. Within the depth measurement range, NA≈R / f. Δl represents the difference in topological charge of each layer. According to equation (1), Δl=2. According to equation (2), the rotation rate of the double helix point spread function will change with the number of Fresnel bands N, while the depth measurement range will change in the opposite direction.

3. The depth measurement and edge extraction method based on a single all-dielectric metasurface as described in claim 2, characterized in that: In step two, Using a dielectric nanoantenna with high transmittance through anti-rotation circular polarization, the in-plane azimuth angles of the dielectric nanoantennas at various locations on the metasurface are determined based on the spatial phase distribution of the target light field. Based on the chiral selectivity of phase modulation characteristics of geometric phase, when left- or right-hand circularly polarized incident light is incident on a dielectric nanoantenna with an azimuth angle of θ, it can form a mutually conjugate phase modulation of ±2θ for right- or left-hand circularly polarized outgoing light. The "+" or "-" is determined by the specific polarization state combination of the incident and outgoing light. According to the above principle, the phase distribution encoding of the metasurface is realized by arranging the in-plane azimuth angle of the dielectric nanoantenna. The polarization state combination includes left-hand / right-hand combination and right-hand / left-hand combination. The dielectric nanoantenna is an amorphous silicon nanorod antenna; The shape and size of the amorphous silicon nanorod antenna are determined by the rigorous coupled-wave analysis (RCWA) method or the finite-difference time-domain (FDTD) method. During the optimization of the amorphous silicon nanorod antenna structure, the height, period, and operating wavelength of the amorphous silicon nanorod antenna are fixed. Then, under these conditions, the length and width of the amorphous silicon nanorod antenna are scanned to select a structural size with high anti-rotational circular polarization transmittance and low co-rotational circular polarization transmittance at the operating wavelength. The phase distribution encoding of the metasurface is achieved by utilizing the in-plane orientation angle of the dielectric nanoantenna through the geometric phase modulation principle.

4. The depth measurement and edge extraction method based on a single all-dielectric metasurface as described in claim 3, characterized in that: In step five, Incoherent incident light illuminates an object and is then imaged onto a CCD through a metasurface optical system, resulting in an image composed of a pair of "twin images." The distance between the two images corresponds to the distance between the main lobes of the point spread function, while the rotation angle depends on the depth of the sample. Therefore, by using image cepstral analysis to extract the rotation angle of the main lobe of the double-helix point spread function corresponding to the image, the depth information of the object can be deduced.

5. The depth measurement and edge extraction method based on a single all-dielectric metasurface as described in claim 4, characterized in that: The image cepstral analysis method includes the following steps: 1) Incoherent incident light irradiates the object, and then passes through the 4-f system and the all-dielectric metasurface encoding the double-helix point diffusion function on the metasurface optical system constructed in step 3. The image is then imaged on the CCD to obtain image I composed of a pair of "twin images". 2) The "twin" image I is processed using a two-dimensional Hann window to obtain image I', thereby improving the reliability of peak recognition; the Hann window used is shown in formula (3): Where m and n are the spatial coordinates of the entire image on the two axes, S w It refers to the size of the window; 3) Perform cepstrum calculation on image I' to obtain its cepstrum C; the cepstrum is the inverse Fourier transform of the power spectrum in logarithmic coordinates, and is often used as a key tool in signal processing, including filtering, denoising, and reconstruction; the cepstrum operation of an image is expressed as: C=C{I′}:=F -1 {log(|F{I′}| 2 )} (4) Where F represents the Fourier transform; 4) The cepstral results are cropped based on the distance between the main lobes and the width of the main lobe of the double helix point spread function. The positions of the two peaks marked by the dashed ring region are obtained in the cepstral C. Then, Gaussian filtering is applied to the dashed ring region obtained in the cepstral C to reduce the influence of noise on the peak positions. The rotation angles corresponding to the positions of the two peaks are then calculated. 5) By matching the extracted peak rotation angle with the calibrated distance-angle relationship, the axial depth information of the object can be measured.

Images