Dispersion compensation method and system for passive liquid crystal optical phased array wide spectrum imaging

Abstract

This invention discloses a dispersion compensation method and system for broadband imaging using a passive liquid crystal optical phased array. Dispersion parameters are obtained by calibrating the dispersion length and overall translation at different orders using a point source. The field of view is electrically deflected at different grating orders to acquire passive observation images. The dispersion process during diffraction is modeled, and an imaging matrix is ​​constructed based on a forward model. Compressed sensing reconstruction is solved using sparse priors as regularization terms to directly reconstruct a high-dimensional data cube of the entire field of view. After reconstruction, the entire field of view scene reconstruction result is directly projected downwards to obtain the reconstruction result of the scene during a scanning process. Field clipping is performed on the multi-order reconstruction results to obtain imaging results of different fields of view at different scanning angles. This invention significantly suppresses dispersion blurring along the deflection direction without changing the physical aperture and pixels of the passive liquid crystal optical phased array, improving the clarity and resolution of passive imaging. It is applicable to broadband passive scenes such as 380–760 nm visible / near-infrared.

Description

Technical Field

[0001] This invention relates to the field of imaging, and specifically to a dispersion compensation method and system for passive liquid crystal optical phased array broadband imaging. Background Technology

[0002] Numerous factors introduce chromatic aberration into traditional geometric optics imaging, such as axial and transverse chromatic aberration from conventional imaging lenses, and chromatic aberration caused by optical axis misalignment during optical path setup. While passive LC-OPA scanning and detection imaging is also unavoidable, these are all within the margin of error. However, the chromatic aberration introduced by the grating pattern loaded in the liquid crystal spatial light modulator is the most severe, causing significant blurring along the deflection direction and severely impacting image quality and resolution. Simple geometric optics chromatic aberration correction is insufficient to eliminate the "diffraction dispersion" introduced by the phase grating; therefore, a de-dispersion model and algorithm matching the OPA deflection mechanism are needed. Summary of the Invention

[0003] This invention addresses the problems of field-direction stretching and detail blurring, as well as significant reduction in edge field-direction resolution caused by diffraction dispersion in broadband imaging of passive liquid crystal optical phase arrays (LC-OPA). It proposes an integrated dispersion compensation method and system that includes "dispersion calibration, multi-order scanning, sparse reconstruction, and low-dimensional projection" to significantly improve imaging clarity and line-pair resolution without changing the physical aperture and pixel size of the LC-OPA.

[0004] The objective of this invention is achieved through the following technical solution: a dispersion compensation method for broadband imaging using a passive liquid crystal optical phased array, comprising the following steps:

[0005] (S1) Obtain the dispersion length and overall translation at different grating orders, and obtain the dispersion parameters based on the grating equation and image plane geometric calibration;

[0006] (S2) Drive the optical phased array at different grating orders to electrically deflect the target field of view and obtain multiple passive observation images with dispersion;

[0007] (S3) Based on the dispersion parameters, the coordinate displacement operator is obtained. A forward model of the dispersion process is constructed according to the spectral layer to be reconstructed. The forward model is combined with the passively observed image to perform optimization solution with sparse priors, and the dedispersed image or data cube is obtained.

[0008] (S4) Project the dedispersed data cube to a lower dimension to obtain a full-field achromatic image;

[0009] (S5) Perform independent cropping based on the dispersion parameters and output the de-dispersion imaging result of the target field of view.

[0010] Furthermore, the calibration in step S1 includes:

[0011] (a) For the grating order N, record the corresponding dispersion length. with overall translation ;

[0012] (b) By combining the grating equation, a scanning angle-dispersion displacement fitting is established, and the parameters between the angle and the distance are fitted using least squares to minimize the error of manual calibration;

[0013] (c) The dispersion parameters of different wavelength layers are: , For discrete spectral layers, This is expressed as resolution in the wavelength direction.

[0014] Furthermore, the forward model is as follows: ,in For field-of-view clipping operators, To be according to coordinate displacement operator, For the spectral layer to be reconstructed, This indicates different wavelength layers.

[0015] Furthermore, FISTA or Adam are used for automatic optimization solutions. This yields a dedispersed image or data cube, where... The variable to be optimized For the observed values, the regularization term For anisotropic / isotropic total variation TV or wavelet L1 regularization, For TV / wavelet operators, This represents the weight of the regularization term.

[0016] Furthermore, the displacement operator in the forward model A described in step S3 This is achieved using the grid_sample function in the pytorch package to support subpixel displacement correction for non-integer pixels.

[0017] Furthermore, the multiple passively observed images containing dispersion include at least five different grating orders N, and the set of orders is adaptively selected based on the signal-to-noise ratio, imaging speed, and the condition number of the forward matrix A.

[0018] On the other hand, the present invention also provides a passive optical phased array dispersion compensation imaging system, comprising:

[0019] (a) LC-OPA scanning and acquisition module, used to generate electrically controlled deflection and acquire multiple passive observation images containing dispersion at different orders N.

[0020] (b) Dispersion calibration module, used to determine the dispersion length and overall translation of different orders and obtain dispersion parameters based on grating equation and image plane geometry calibration;

[0021] (c) Forward model and reconstruction module, used to obtain coordinate displacement operator based on dispersion parameters, construct forward model of dispersion process according to the spectral layer to be reconstructed, and perform optimization solution with sparse prior in combination with passive observation image to obtain dedispersion image or data cube.

[0022] (d) Control and storage module, used to store calibration parameters and timing and data streams for controlling the LC-OPA / detector;

[0023] (e) Output module, used to independently crop according to dispersion parameters and output dedispersed image or spectral data cube.

[0024] Furthermore, the forward model and reconstruction module incorporates TV / wavelet regularization and supports GPU acceleration and adaptive parameter selection for sparse regularization.

[0025] Furthermore, the dispersion calibration module includes a least-squares calculation unit for order-displacement fitting and an image plane parameter solving unit based on the calibration plate / point target.

[0026] The beneficial effects of this invention are:

[0027] (1) While keeping the OPA hardware unchanged, the scanning direction dispersion blur is significantly suppressed.

[0028] (2) Although the present invention is described in terms of scanning in one-dimensional direction of OPA, it is also applicable to scanning in two-dimensional OPA. It is only necessary to decompose the two-dimensional scanning direction into two one-dimensional scanning and dispersion methods.

[0029] (3) Combining subpixel displacement and sparse prior, it can be compatible with different wavelengths (visible light and long-wave infrared) and different deflection angles. Therefore, the dispersion compensation system proposed in this invention is applicable to any spectral band, any scene, and any target deflection angle.

[0030] (4) The present invention effectively solves the problem that the dispersion effect of passive optical phased arrays causes large color difference in the imaging system and reduces the detection capability. It can significantly improve the final imaging effect, so that optical field control devices such as OPA can freely scan imaging targets in different fields of view under a wide spectrum and in a fully solid-state hardware system. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of a process for calibrating dispersion parameters provided in an embodiment of the present invention.

[0033] Figure 2 This is a schematic diagram of the calibration results of dispersion parameters provided in an embodiment of the present invention.

[0034] Figure 3 This is a schematic diagram of a dispersion forward model for passive optical phased array broadband imaging provided in an embodiment of the present invention.

[0035] Figure 4 This is a schematic diagram of a system optical path provided in an embodiment of the present invention.

[0036] Figure 5 This is a comparison diagram of dispersion compensation effects provided in an embodiment of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] like Figure 1 The diagram illustrates the process of calibrating dispersion parameters according to this invention. First, the dispersion parameters required for the reconstruction algorithm are obtained. Using a point source as the imaging target, the dispersion length and overall translation are recorded for each grating order N. Since the specific dispersion coordinate displacement of light in each wavelength band needs to be known, the distance between the center point of the first-order diffraction order and the center point of the zero-order diffraction order is used as the overall translation, and the total number of dispersion pixels visible to the human eye from blue to red light is used as the dispersion length. A scanning angle-dispersion displacement fitting is established based on the grating equation to obtain the displacement of each wavelength layer. As a hyperparameter of the reconstruction algorithm, This is a discrete spectral layer. Specifically, it is determined by the grating equation. We can obtain, ), where m represents the diffraction order. Therefore, light of different wavelengths will be located at different positions when it hits the detector. The parameter between angle and distance is represented by K. The dispersion length can be expressed as... This value is represented as a long colored line on the image; the overall translation can be expressed as... That is, the distance between the center of the colored long line and the center of the 0th order.

[0039] After labeling multiple sets of data, record the dispersion length and overall translation at different orders, denoted as vectors respectively. and The fitting coefficients K1 and K2 can be obtained using the least squares method:

[0040]

[0041] This is a classic convex optimization problem. The derivation is omitted here; the closed-form solution is given directly, and the fitting coefficients are derived:

[0042]

[0043] Therefore, the final first-order dispersion length is The first-order total displacement is .

[0044] The purpose of fitting is to minimize the error inherent in manually calibrated data, and fitting helps to make it conform to the theory as closely as possible. Note that all quantities in the formula are vectors, so the numerator is a matrix product. The final fitting coefficient K minimizes the error of manual calibration and can directly derive the dispersion length of the second-order dispersion. Furthermore, since the number of grating orders N that can be calibrated during the calibration process is limited, without fitting, OPA imaging would be restricted to discretely selecting only a few pre-calibrated angles, which limits the effectiveness in practical use. This calibration process only needs to be performed once the system is determined, and thereafter the system can be used for imaging tasks in any scenario.

[0045] Figure 2 The results are the fitting results from the experimental process of this invention. It can be seen that the overall displacement fitting is good, while the dispersion fitting is slightly worse. This is because the true dispersion length is unknown, and the light intensity at the edge of the spectrum is weak, making it almost impossible to see after dispersion. This effect is more pronounced when scanning at large angles, so only a portion of the dispersion length can be calibrated.

[0046] Finally passed The dispersion parameters are obtained; these are hyperparameters that must be used in the forward model A. This represents the resolution in the wavelength direction. The 1 in the formula represents the dispersion caused by the first-order diffraction of the grating. If the second-order dispersion is significant in the experiment, it is only necessary to calibrate the second-order dispersion once by following this process.

[0047] A fixed optical system is used to acquire dispersive observations through multi-order scanning. The OPA is driven at different grating orders N to electrically deflect the target field of view, acquiring multiple frames of passively observed images, which are then used as input for subsequent inversion. This setup can be equivalent to a one-dimensional projection sequence along the main scanning direction within a two-dimensional OPA scene, providing the necessary "multi-measurement" for compressed sensing reconstruction.

[0048] By establishing a physical model for broadband passive imaging with dispersion, the observed values ​​can be... Write it in the following form

[0049]

[0050] in The variable to be optimized is a three-dimensional variable, including the spatial axis (X, Y) and the wavelength axis (…). ), This refers to noise during the actual imaging process. For field-of-view clipping operators, To be according to coordinate displacement operator, For the spectral layer to be reconstructed, This represents different wavelength layers. With the forward model A, inversion reconstruction based on sparse priors can be performed to solve for... For optimization issues, automatic optimization can be performed using the FISTA framework or PyTorch's Adam optimizer. Represents the regularization term, Represents the regularization coefficient. Displacement operator. It supports dispersion shift parameters for non-integer pixels. This model requires a relatively large number of samples for robust reconstruction, depending on the resolution of the wavelength direction of the reconstructed data cube.

[0051] like Figure 3 The diagram shown illustrates the forward model, including the dispersion parameters. Then a definite forward model A(x) = Displacement operator This is implemented using the `grid_sample` function in PyTorch, enabling the handling of sub-pixel dispersion parameters. (Crop operator) This is because the reconstructed image is a full-field image from all scanning angles, so only an image cropped from the center to the same size as the measurement data can be used to calculate the MSE loss.

[0052] After reconstructing the high-dimensional data cube, it is projected onto the low dimension to obtain the achromatic full-field reconstructed image. If you want to obtain the corresponding images from different viewpoints, you can perform field cropping on the achromatic full-field reconstructed image to output the target field image.

[0053] To practically test the imaging effect of the method of the present invention, the present invention constructed as follows: Figure 4 The optical path shown illustrates a comparative experiment on the dispersion correction effect. A white LED surface light source was used as the light source in the liquid crystal phased array dispersion compensation imaging experimental system. The object was illuminated and its angle was deflected by a spatial light modulator before observation on a CMOS sensor. The effect of dispersion compensation before and after a certain field of view was compared using a resolution panel.

[0054] The specific experimental procedure is as follows: The illumination source is a white LED surface light source in the visible light band. The light source illuminates the target, and the light carrying the target information enters the system and illuminates the FSLM-4K62-P02 liquid crystal spatial light modulator. By changing the grating phase diagram on the loaded liquid crystal spatial light modulator, the field of view is deflected.

[0055] Dispersion patterns with chromatic aberration were acquired at different fields of view and reconstructed using a dispersion compensation algorithm. According to the Rayleigh criterion, if the distance between two Airy disks is such that the intensity maxima of one spectral line coincides with the intensity minima of the other, then the two Airy disks are just distinguishable. In this case, the peak-to-valley ratio at the centers of the two Airy disks is approximately 0.81. The resolution corresponding to the line pairs in the image was read out and used as the resolution of the instantaneous field of view at the maximum deflection angle before and after dispersion compensation. and Calculate the instantaneous field-of-view dispersion compensation resolution boost factor for the maximum deflection angle. .

[0056] Once the target scanning field of view angle is determined, the reconstruction process described in the invention is executed, wherein TV regularization is selected as the regularization term, Adam is used as the optimizer, and a high-dimensional data cube with a wavelength resolution of 50 layers is obtained through inversion reconstruction. Then, the images of each field of view are obtained through the projection and cropping processes described above.

[0057] The comparison results before and after dispersion compensation are as follows: Figure 5 As shown, the three sets of images represent grating orders of N=20, 17, and 14, respectively. A smaller order indicates a larger deflection angle. The images show that as the field-of-view deflection angle increases, the original achromatic image becomes increasingly blurry. This is because the dispersion effect intensifies as N decreases. After applying the dispersion compensation method of this invention, the reconstructed achromatic image shows that the blurring caused by dispersion has been largely eliminated, and the images in each field of view are clearer, significantly improving the resolution. This directly proves the feasibility of the method of this invention.

[0058] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A dispersion compensation method for broadband imaging using a passive liquid crystal optical phased array, characterized in that, The method includes the following steps: (S1) Obtain the dispersion length and overall translation under different grating orders, and obtain the dispersion parameters based on the grating equation and image plane geometry calibration; the overall translation is the distance between the center point of the first-order diffraction order and the center point of the zero-order diffraction order, the dispersion length is the number of dispersion pixels visible to the human eye from blue light to red light, and the dispersion parameters are the displacement of each wavelength layer obtained by combining the grating equation to establish a scanning angle-dispersion displacement fitting. (S2) Drive the optical phased array at different grating orders to electrically deflect the target field of view and obtain multiple passive observation images with dispersion; (S3) Based on the dispersion parameters, the coordinate displacement operator is obtained. A forward model of the dispersion process is constructed according to the spectral layer to be reconstructed. The forward model is combined with the passively observed image to perform optimization solution with sparse priors, and the dedispersed image or data cube is obtained. (S4) Project the dedispersed data cube to a lower dimension to obtain a full-field achromatic image; (S5) Perform independent cropping based on the dispersion parameters and output the de-dispersion imaging result of the target field of view.

2. The method according to claim 1, characterized in that, The calibration in step S1 includes: (a) For the grating order N, record the corresponding dispersion length. with overall translation ; (b) By combining the grating equation, a scanning angle-dispersion displacement fitting is established, and the parameters between the angle and the distance are fitted using least squares to minimize the error of manual calibration; (c) The dispersion parameters of different wavelength layers are: , For discrete spectral layers, This is expressed as resolution in the wavelength direction.

3. The method according to claim 2, characterized in that, The forward model is: ,in For field-of-view clipping operators, To be according to Coordinate displacement operator for performing displacement. For the spectral layer to be reconstructed, Representing different wavelength layers; the coordinate displacement operator Implemented using the `grid_sample` function in the PyTorch package to support sub-pixel displacement correction for non-integer pixels; the field-of-view clipping operator Used to crop an image from the center of the full-field image to the same size as the measurement data.

4. The method according to claim 3, characterized in that, Use FISTA or Adam for automatic optimization solution This yields a dedispersed image or data cube, where... The variable to be optimized For the observed values, the regularization term For anisotropic / isotropic total variation TV or wavelet L1 regularization, For TV / wavelet operators, This represents the weight of the regularization term.

5. The method according to claim 1, characterized in that, The multiple passively observed images containing dispersion include at least five different grating orders N, and the set of orders is adaptively selected based on the signal-to-noise ratio, imaging speed, and the condition number of the forward matrix A.

6. A passive optical phased array dispersion compensation imaging system, characterized by: (a) LC-OPA scanning and acquisition module, used to generate electrically controlled deflection and acquire multiple passive observation images with dispersion at different orders N; (b) Dispersion calibration module, used to determine the dispersion length and overall translation of different orders and obtain dispersion parameters based on grating equation and image plane geometry calibration; (c) Forward model and reconstruction module, used to obtain coordinate displacement operator based on dispersion parameters, construct forward model of dispersion process according to the spectral layer to be reconstructed, and perform optimization solution with sparse prior in combination with passive observation image to obtain dedispersion image or data cube. (d) Control and storage module, used to store calibration parameters and timing and data streams for controlling the LC-OPA / detector; (e) Output module, used to independently crop according to dispersion parameters and output dedispersed image or spectral data cube.

7. The system according to claim 6, characterized in that, The forward model and reconstruction module incorporates TV / wavelet regularization and supports GPU acceleration and adaptive parameter selection for sparse regularization.

8. The system according to claim 6, characterized in that, The dispersion calibration module includes a least-squares calculation unit for order-displacement fitting and an image plane parameter solving unit based on the calibration plate / point target.

Images