Voltage-adjustable liquid crystal box and multi-layer network alignment method and device thereof
Through voltage adjustable liquid crystal box and adaptive adjustment algorithm, the problem of low alignment accuracy of multi-layer optical diffraction neural network is solved, and efficient and automatic optical signal alignment is achieved, ensuring calculation accuracy and stability.
Patent Information
- Application Number
- CN202411314565.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-20
- Publication Date
- 2024-12-06
AI Technical Summary
The alignment methods of existing multi-layer optical diffraction neural networks rely on mechanical adjustment and manual operation, which are difficult to ensure accuracy and complex and inefficient operation, resulting in accumulated errors in optical signal transmission, affecting calculation accuracy and stability.
The voltage adjustable liquid crystal box is adopted to accurately control the arrangement of liquid crystal molecules through multi-layer voltage adjustable liquid crystal devices and signal generators. The spherical wave liquid crystal lens and camera are aligned, and the random forest algorithm and adaptive adjustment algorithm are combined to automatically adjust. Voltage value to achieve clear focus of the beam.
It significantly improves alignment accuracy and efficiency, reduces the influence of human factors, ensures the accuracy and stability of optical signal transmission, and simplifies the alignment process.
Smart Images
Figure CN119087713A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a voltage-adjustable liquid crystal box, and a multi-layer network alignment method and device thereof. Background Art
[0002] Optical Neural Networks (ONN) is an artificial neural network that uses light instead of traditional electronic signals for large-scale computing. It has large bandwidth, high interconnection, and parallel processing capabilities. In particular, the Diffractive Deep Neural Network (D2NN) architecture, the passive diffraction layer trained by deep learning can learn to implement various functions, and the calculation results are immediately presented in the process of wave propagation at the speed of light without the need for additional energy consumption. It has a wide range of applications in image classification, logical operations, optical encryption, quantum information processing, etc. .
[0003] However, the alignment problem of multi-layer optical diffraction neural networks has always been a major bottleneck in the development of this technology. Since it involves a multi-layer structure, precise alignment between each layer is crucial. Insufficient alignment accuracy will lead to the accumulation of errors in optical signal transmission, which in turn affects the computational accuracy and stability of the entire network. Existing alignment methods mainly rely on mechanical adjustment and manual operation. These methods usually require complex equipment and a lot of time, and are difficult to operate, easily affected by human factors, and it is difficult to ensure alignment accuracy. In addition, the alignment of multi-layer structures usually requires multiple repeated adjustments, which is cumbersome and inefficient. Summary of the invention
[0004] The main purpose of the present invention is to provide a voltage-adjustable liquid crystal box, and a multi-layer network alignment method and device thereof, so as to solve the technical problems that the existing multi-layer optical diffraction neural network alignment method mainly relies on mechanical adjustment and manual operation, which is difficult to ensure accuracy and the operation is complex and inefficient.
[0005] To achieve the above-mentioned purpose, the present invention provides a voltage-adjustable liquid crystal box, which includes: a multi-layer voltage-adjustable liquid crystal device, a plurality of signal generators, a plurality of spherical wave liquid crystal lenses, and a camera; the voltage-adjustable liquid crystal device changes the arrangement of internal liquid crystal molecules according to the applied voltage value, thereby changing the optical properties, wherein the optical properties include refractive index and diffraction efficiency; the plurality of signal generators are respectively connected to the multi-layer voltage-adjustable liquid crystal device, and are used to apply voltages to the multi-layer voltage-adjustable liquid crystal device respectively; the plurality of spherical wave liquid crystal lenses are respectively arranged between the multi-layer voltage-adjustable liquid crystal device, and the lens focal lengths of the plurality of spherical wave liquid crystal lenses are respectively the distances from the voltage-adjustable liquid crystal device corresponding to the spherical wave liquid crystal lens to the camera.
[0006] Optionally, the voltage-adjustable liquid crystal device is assembled frontally by bonding one SD1 substrate to another SD1 substrate using UV glue.
[0007] Optionally, the voltage-adjustable liquid crystal device is made by photo-aligning with a digital mask lithography machine and coating with E7 liquid crystal.
[0008] The present invention also provides a multi-layer network alignment method based on a voltage-adjustable liquid crystal box, wherein the voltage-adjustable liquid crystal box is any of the voltage-adjustable liquid crystal boxes described above, and the multi-layer network alignment method comprises: controlling a light source device corresponding to the voltage-adjustable liquid crystal box to generate a light beam; sequentially adjusting the voltage values applied to different layers of voltage-adjustable liquid crystal devices through multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface.
[0009] Optionally, the voltage values applied to different layers of voltage-adjustable liquid crystal devices are adjusted in sequence through multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface, including: performing the following operations on each voltage-adjustable liquid crystal device in sequence according to the distance of the voltage-adjustable liquid crystal device from the detection surface, from near to far: adjusting the voltage applied to the i-th layer of voltage-adjustable liquid crystal devices through the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, until the light beam passes through the different layers of voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface; applying a voltage higher than a preset threshold to the i-th layer of voltage-adjustable liquid crystal devices through the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, so that the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal devices are arranged vertically to prevent the diffraction of light from passing through.
[0010] Optionally, before applying a voltage higher than a preset threshold to the i-th layer of voltage-adjustable liquid crystal devices through a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices to vertically align the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal devices to prevent the diffraction of light from passing through, the multi-layer network alignment method further includes: acquiring historical data, wherein the historical data at least includes: image data of the focus position and focus clarity corresponding to the light beam passing through each layer of voltage-adjustable liquid crystal devices at different voltage values; constructing a network alignment initial model based on a random forest algorithm, wherein the construction content at least includes: setting initial parameters and data input format of the network alignment initial model; performing wavelet decomposition processing on the historical data, and performing soft threshold processing on the historical data after wavelet decomposition based on preset wavelet coefficients, and performing wavelet re-processing on the processing results. The method comprises the following steps: performing matrix data reorganization processing on the first initial data to obtain second initial data, wherein the second initial data is represented in matrix form, and each row represents a voltage value-image data pair; performing sparse coding conversion on each row of data in the second initial data, representing the data in the high-dimensional feature space of the second initial data as sparse features, and obtaining third initial data; calculating the importance of the sparse features in the third initial data based on the influence of the optical response of the image data, and screening and optimizing the sparse features in the third initial data according to the calculation results, and obtaining fourth initial data; training the network alignment initial model with the fourth initial data, and optimizing the parameters in the network alignment initial model with cross-validation and network search, and obtaining a final network alignment model.
[0011] Optionally, a voltage higher than a preset threshold is applied to the i-th layer of voltage-adjustable liquid crystal devices through a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, so that the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal devices are arranged vertically to prevent the diffraction of light from passing through, including: real-time image capture of the detection surface to obtain real-time image data of the focus; inputting the preset code of the i-th layer of voltage-adjustable liquid crystal devices and the real-time image data of the focus into the network alignment model to obtain feedback information, wherein the feedback information at least includes: the voltage deviation value of the i-th layer of voltage-adjustable liquid crystal devices; using an adaptive adjustment algorithm to iteratively adjust the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices for multiple times based on the feedback information until the light beam passes through the different layers of voltage-adjustable liquid crystal devices to produce a clear focus on the detection surface.
[0012] The present invention also provides a multi-layer network alignment device based on a voltage-adjustable liquid crystal box, comprising: a generating unit, used to control a light source device corresponding to the voltage-adjustable liquid crystal box to generate a light beam; an adjusting unit, used to sequentially adjust the voltage values applied to different layers of voltage-adjustable liquid crystal devices through multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface.
[0013] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.
[0014] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.
[0015] The present invention provides a voltage-adjustable liquid crystal box, and a multi-layer network alignment method and device thereof, wherein the voltage-adjustable liquid crystal box includes a voltage-adjustable liquid crystal device, a plurality of signal generators, a plurality of spherical wave liquid crystal lenses, and a camera. The voltage-adjustable liquid crystal device changes the arrangement of internal liquid crystal molecules by applying different voltages, thereby changing its optical properties, including refractive index and diffraction efficiency. The plurality of signal generators are respectively connected to the multi-layer voltage-adjustable liquid crystal devices, and are used to apply voltages to the liquid crystal devices respectively. The plurality of spherical wave liquid crystal lenses are arranged between different layers, and their focal lengths correspond to the distance from the liquid crystal device to the camera.
[0016] This technical solution uses voltage regulation to precisely control the arrangement of liquid crystal molecules, achieving precise alignment of multi-layer structures and avoiding the cumbersome and unstable mechanical adjustment and manual operation in traditional methods. The optical properties of the liquid crystal box can be adjusted in real time as needed, thereby ensuring the transmission accuracy and stability of the optical signal. Specifically, this method significantly simplifies the alignment process, improves efficiency and alignment accuracy, and reduces the impact of human factors. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 is a schematic structural diagram of a voltage-adjustable liquid crystal cell according to an embodiment of the present invention; Figure 2 It is a schematic diagram of the steps of a multi-layer network alignment method based on a voltage-adjustable liquid crystal cell in one embodiment of the present invention; Figure 3 Schematic diagram of XY plane alignment lens axis propagation based on a multi-layer network of voltage-adjustable liquid crystal cells in one embodiment of the present invention Figure 1 ; Figure 4 Schematic diagram of XY plane alignment lens axis propagation based on a multi-layer network of voltage-adjustable liquid crystal cells in one embodiment of the present invention Figure 2 ; Figure 5 It is a schematic diagram of lens axis propagation in the Z propagation direction of a multilayer network based on a voltage-adjustable liquid crystal cell in one embodiment of the present invention; Figure 6 It is a schematic diagram of light path propagation in a multi-layer network alignment experiment based on a voltage-adjustable liquid crystal cell in one embodiment of the present invention; Figure 7 It is a structural framework diagram of a multi-layer network alignment device based on a voltage-adjustable liquid crystal cell in one embodiment of the present invention; Figure 8 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.
[0018] Description of reference numerals: 101, SD1 substrate; 102, curing glue 1; 103, SD1 substrate; 104, signal generator; 105, signal generator; 106, signal generator; 107, signal generator; 108, signal generator; 109-134, liquid crystal lens; 135, camera; 136-140, voltage-adjustable liquid crystal box; 401. Laser; 402. Objective lens; 403. Lens; 404. Spatial light modulator; 405. Lens; 406. Lens; 407. Circular polarizer; 408. First voltage adjustable liquid crystal device; 409. Second voltage adjustable liquid crystal device; 410. Third voltage adjustable liquid crystal device; 411. Fourth voltage adjustable liquid crystal device; 412. Fifth voltage adjustable liquid crystal device; 413. Camera.
[0019] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0021] Reference Figure 1 , an embodiment of the present invention provides a voltage-adjustable liquid crystal box, the voltage-adjustable liquid crystal box comprising: a voltage-adjustable liquid crystal device, a plurality of signal generators, a plurality of spherical wave liquid crystal lenses, and a camera; The voltage-adjustable liquid crystal device changes the arrangement of internal liquid crystal molecules according to the applied voltage value, thereby changing the optical properties, wherein the optical properties include refractive index and diffraction efficiency; The plurality of signal generators are respectively connected to the multi-layer voltage-adjustable liquid crystal device, and are used to apply voltages to the multi-layer voltage-adjustable liquid crystal device respectively; The multiple spherical wave liquid crystal lenses are respectively arranged between the multiple layers of voltage-adjustable liquid crystal devices, and the lens focal lengths of the multiple spherical wave liquid crystal lenses are respectively the distances from the voltage-adjustable liquid crystal devices corresponding to the spherical wave liquid crystal lenses to the camera.
[0022] The present invention provides a voltage-adjustable liquid crystal box, which includes a voltage-adjustable liquid crystal device, a plurality of signal generators, a plurality of spherical wave liquid crystal lenses and a camera. The voltage-adjustable liquid crystal device changes the arrangement of liquid crystal molecules inside according to the applied voltage value, thereby changing its optical properties, including refractive index and diffraction efficiency. The plurality of signal generators are respectively connected to the multi-layer voltage-adjustable liquid crystal device, and are used to apply voltages to the liquid crystal devices respectively. The plurality of spherical wave liquid crystal lenses are arranged between different layers, and their focal lengths correspond to the distance from the liquid crystal device to the camera.
[0023] This technical solution uses voltage regulation to precisely control the arrangement of liquid crystal molecules, achieving precise alignment of multi-layer structures and avoiding the cumbersome and unstable mechanical adjustment and manual operation in traditional methods. The optical properties of the liquid crystal device can be adjusted in real time as needed, thereby ensuring the transmission accuracy and stability of the optical signal. Specifically, this method significantly simplifies the alignment process, improves efficiency and alignment accuracy, and reduces the impact of human factors.
[0024] In one example, the voltage-adjustable liquid crystal device is assembled frontally by bonding one SD1 substrate to another SD1 substrate using UV glue.
[0025] In this example, the assembly process of the voltage-adjustable LC device involves two SD1 substrates. First, place one SD1 substrate on the workbench, and then attach the voltage-adjustable LC device to this SD1 substrate by using UV glue (ultraviolet light curing glue). The UV glue will cure quickly under the irradiation of ultraviolet light, thus firmly fixing the LC device on the SD1 substrate. Next, place another SD1 substrate on top of the voltage-adjustable LC device, align it with the SD1 substrate and LC device below, and ensure that all layers are tightly attached together.
[0026] Frontal assembly means that during the assembly process, all components and layers are installed and aligned from the same side. This method can simplify the assembly process, reduce the possibility of errors, and ensure that each layer can be accurately aligned. Through the use of UV glue, fast and reliable bonding can be achieved, making the entire device structure more stable and precise. This method not only improves assembly efficiency, but also ensures the precise alignment of liquid crystal devices in multi-layer structures, thereby ensuring the accuracy and stability of optical signal transmission.
[0027] In one example, the voltage-adjustable liquid crystal device is made by photo-aligning with a digital mask lithography machine and coating with E7 liquid crystal.
[0028] In this example, the manufacturing process of the voltage-adjustable liquid crystal device uses digital mask lithography technology and E7 liquid crystal. First, a digital mask lithography machine is used to perform photo-controlled orientation on the substrate. Specifically, the lithography machine uses a digitally controlled mask to accurately irradiate ultraviolet light to specific areas on the substrate. The illumination of these areas changes the chemical properties of the substrate surface, thereby controlling the alignment direction of the liquid crystal molecules. The photo-controlled orientation step is very important because it determines the arrangement of the liquid crystal molecules under the action of the electric field, which affects the optical performance of the liquid crystal device.
[0029] After completing the photo-controlled orientation, the next step is to fill the E7 liquid crystal. E7 is a commonly used liquid crystal material with good electro-optical properties and stability. The E7 liquid crystal material is filled into the substrate that has been photo-controlled oriented so that the liquid crystal molecules are arranged in a pre-set direction. The filling process needs to be carried out in a dust-free environment to avoid the influence of impurities on the performance of the liquid crystal device. After the filling is completed, the liquid crystal device is usually sealed and cured to ensure that the liquid crystal material does not leak during use and maintains its predetermined arrangement state.
[0030] This manufacturing method can precisely control the optical properties of voltage-tunable liquid crystal devices, including refractive index and diffraction efficiency, to achieve high-precision control of optical signal transmission. Compared with traditional methods, digital mask lithography technology provides higher resolution and accuracy, making the performance of liquid crystal devices more superior. This technical solution has significant advantages in the alignment problem of multi-layer optical diffraction neural networks, helping to improve calculation accuracy and stability.
[0031] Reference Figure 2 The embodiment of the present invention provides a multi-layer network alignment method based on a voltage-adjustable liquid crystal box. The voltage-adjustable liquid crystal box is the voltage-adjustable liquid crystal box provided in the above embodiment. Specifically, the multi-layer network alignment method includes: S1, controlling the light source device corresponding to the voltage-adjustable liquid crystal box to generate a light beam; S2, adjusting the voltage values applied to different layers of voltage-adjustable liquid crystal devices in sequence through multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and a clear focus is generated on the detection surface.
[0032] In an embodiment of the present invention, a multi-layer network alignment method based on a voltage-adjustable liquid crystal cell is provided. The voltage-adjustable liquid crystal cell is the liquid crystal cell described above. The specific steps are as follows: First, the light source device corresponding to the voltage-adjustable liquid crystal box is controlled to generate a light beam. The purpose of this step is to generate a beam of light as a test signal for aligning the multi-layer liquid crystal device. The light source device may include a laser or other type of beam generating device, and its beam needs to have sufficient intensity and stability to ensure accurate detection in subsequent steps. Next, the voltage applied to different layers of the voltage-adjustable liquid crystal device is adjusted in turn through multiple signal generators. Each voltage-adjustable liquid crystal device corresponds to a signal generator, and these signal generators can independently control the voltage value of each layer of liquid crystal device. The purpose of adjusting the voltage is to change the arrangement of the liquid crystal molecules, thereby adjusting the optical properties of each layer of the liquid crystal device (such as refractive index and diffraction efficiency) to ensure that the light beam can accurately pass through all liquid crystal layers.
[0033] The adjustment process is performed layer by layer, gradually changing the voltage of each layer until the light beam passes through all the different layers of voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface. The detection surface may be a photodetector or camera that is used to detect and image the light spot after the light beam passes through the liquid crystal device. Producing a clear focus means that the path of the light signal is accurately aligned after the light beam passes through all the liquid crystal layers, and the light spot has no obvious scattering or blurring.
[0034] Through the above steps, the multi-layer network can be accurately aligned to ensure that the liquid crystal devices of each layer can correctly adjust the path of the light beam, thereby improving the accuracy and stability of the optical signal transmission of the entire system. This method is more efficient and reliable than traditional mechanical adjustment and manual operation, reduces the impact of human errors, and improves the automation and accuracy of the alignment process.
[0035] In one example, the voltage values applied to different layers of voltage-adjustable liquid crystal devices are adjusted in sequence by multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface, including: performing the following operations on each voltage-adjustable liquid crystal device in sequence according to the distance of the voltage-adjustable liquid crystal device from the detection surface, from near to far: adjusting the voltage applied to the i-th layer of voltage-adjustable liquid crystal devices by the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, until the light beam passes through the different layers of voltage-adjustable liquid crystal devices and produces a clear focus on the detection surface; applying a voltage higher than a preset threshold to the i-th layer of voltage-adjustable liquid crystal devices by the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, so that the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal devices are arranged vertically to prevent the diffraction of light from passing through.
[0036] In this example, the multi-layer alignment method of the voltage-adjustable liquid crystal cell adjusts the voltage layer by layer through multiple signal generators until the light beam produces a clear focus on the detection surface. The specific steps include operating each voltage-adjustable liquid crystal device from near to far from the detection surface.
[0037] First, the signal generator of the voltage-adjustable liquid crystal device of the i-th layer is controlled to adjust the applied voltage until the light beam passes through all layers and produces a clear focus on the detection surface. This means that when the light beam passes through the adjusted voltage-adjustable liquid crystal device of the i-th layer, its path is accurately aligned, and finally a clear light spot is formed on the detection surface. Then, a voltage higher than the preset threshold is applied through the signal generator of the voltage-adjustable liquid crystal device of the i-th layer, so that the liquid crystal molecules in the liquid crystal device of the i-th layer are arranged vertically, thereby preventing the diffraction of light from passing through this layer. This step ensures that the liquid crystal device of the i-th layer no longer affects the adjustment of subsequent layers.
[0038] Next, repeat the above steps for the next layer of voltage-adjustable liquid crystal devices, adjusting the voltage of each layer in turn until all layers are aligned and the light beam produces a clear focus on the detection surface.
[0039] The following takes a five-layer voltage-adjustable liquid crystal device as an example to explain the entire alignment process in detail: First, start from the first layer, which is closest to the detection surface. Through the signal generator of the voltage-adjustable liquid crystal device of the first layer, adjust the applied voltage until the light beam passes through the first layer and forms a clear focus on the detection surface. Then, apply a voltage higher than the preset threshold to the liquid crystal device of the first layer to align its liquid crystal molecules vertically, preventing the diffraction of light from passing through the first layer.
[0040] Next, the same operation is performed on the second layer of liquid crystal devices. The voltage of the second layer is adjusted so that the light beam passes through the second layer and forms a clear focus on the detection surface. Subsequently, a voltage higher than the preset threshold is applied to the second layer, so that its liquid crystal molecules are aligned vertically, preventing the diffraction of light from passing through the second layer.
[0041] Repeat the same steps to adjust the third layer of liquid crystal devices until the light beam passes through the third layer and forms a clear focus on the detection surface. Then, apply a voltage higher than the preset threshold to the third layer to align its liquid crystal molecules vertically, preventing the diffraction of light from passing through the third layer.
[0042] The 4th layer of liquid crystal devices is adjusted to ensure that the light beam passes through the 4th layer and forms a clear focus on the detection surface. Then, a voltage higher than the preset threshold is applied to the 4th layer to align its liquid crystal molecules vertically, preventing the diffraction of light through the 4th layer.
[0043] Finally, the fifth layer of liquid crystal devices is adjusted so that the light beam passes through the fifth layer and forms a sharp focus on the detection surface. Since this is the last layer, there is no need to prevent the diffraction of light.
[0044] By adjusting the voltage layer by layer, the optical properties of each layer of voltage-adjustable liquid crystal devices are precisely controlled, thereby achieving high-precision alignment of the multi-layer network and ensuring the transmission accuracy and stability of the optical signal in the multi-layer structure. In other words, if Figure 3-6 As shown, in an embodiment of the present invention, a multi-layer network alignment method based on a voltage-adjustable liquid crystal box is provided, and the voltage is adjusted layer by layer by multiple signal generators until the light beam produces a clear focus on the detection surface. First, the light source device corresponding to the voltage-adjustable liquid crystal device is controlled to generate a light beam as a test signal. Next, the voltage values applied to different layers of voltage-adjustable liquid crystal devices are adjusted in turn by multiple signal generators. The voltage adjustment process of each layer of voltage-adjustable liquid crystal devices is performed layer by layer until the light beam passes through all liquid crystal layers and produces a clear focus on the detection surface.
[0045] In the specific alignment process, the voltage is adjusted layer by layer through the signal generator, starting from the first layer closest to the detection surface. The applied voltage is adjusted through the signal generator of the voltage-adjustable liquid crystal device of the first layer until the light beam passes through the first layer and forms a clear focus on the detection surface. Then, a voltage higher than the preset threshold is applied to the first layer to align its liquid crystal molecules vertically and prevent the diffraction of light from passing through. Next, the same operation is performed on the second layer of liquid crystal devices, and the voltage of the second layer is adjusted so that the light beam passes through the second layer and forms a clear focus on the detection surface. Subsequently, a voltage higher than the preset threshold is applied to the second layer to align its liquid crystal molecules vertically and prevent the diffraction of light from passing through.
[0046] This alignment method has significant advantages. By using the external voltage adjustability of the voltage-adjustable liquid crystal device, the effective refractive index of the liquid crystal molecules can be optimized by precisely controlling the applied voltage, thereby optimizing the diffraction efficiency according to the operating wavelength. This control can suppress non-diffracted light and improve the signal-to-noise ratio. In addition, when a sufficiently high voltage is applied, all liquid crystal molecules in the voltage-adjustable liquid crystal device are almost vertically aligned, exhibiting a closed state, preventing the diffraction of light from passing through. This switchable property can further improve the alignment accuracy of the light propagation path in the Z direction and XY direction by introducing a set of additional lenses.
[0047] The specific alignment steps include: in liquid crystal spherical wave lenses with different focal lengths, the spherical wave liquid crystal lenses are recorded in layers 1-5 respectively, and the focal length of the lens is the distance from each layer to the camera. Multiple layers of liquid crystal devices are combined together, and after the phase of each layer is carried by light, the alignment strategy of all diffraction layers includes positioning the last layer at a set distance in front of the camera. Through collimated input light, a signal generator voltage signal is applied to the voltage-adjustable liquid crystal device, and the reference layer is aligned when the lens focuses the beam to the smallest and brightest point on the camera. The previous layer is placed where the focal length of the lens is met, and the voltage-adjustable liquid crystal device is turned off by applying an external voltage, and the lens is controlled to align the plane in the XY axis, so that the lens in the voltage-adjustable liquid crystal device focuses the collimated beam on the camera.
[0048] For example, for the alignment process of a five-layer voltage-adjustable liquid crystal device, first adjust the first layer so that the light beam passes through and forms a clear focus on the detection surface, and then align its liquid crystal molecules vertically. Next, adjust the second, third, fourth, and fifth layers layer by layer to ensure that each layer of light beam forms a clear focus on the detection surface, and align the liquid crystal molecules vertically after each layer is aligned to prevent the diffraction of light. Ultimately, through this layer-by-layer adjustment method, high-precision alignment of multi-layer networks is achieved to ensure the transmission accuracy and stability of optical signals.
[0049] In one example, before a voltage higher than a preset threshold is applied to the i-th layer of voltage-adjustable liquid crystal devices through a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices to vertically align the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal devices to prevent the diffraction of light from passing through, the multi-layer network alignment method further includes: acquiring historical data, wherein the historical data at least includes: image data of the focus position and focus clarity corresponding to the light beam passing through each layer of voltage-adjustable liquid crystal devices at different voltage values; constructing a network alignment initial model based on a random forest algorithm, wherein the construction content at least includes: setting initial parameters and data input format of the network alignment initial model; performing wavelet decomposition processing on the historical data, and performing soft threshold processing on the historical data after wavelet decomposition based on preset wavelet coefficients, and performing wavelet decomposition on the processing results. Wave reconstruction processing is performed to obtain the first initial data after denoising; the first initial data is subjected to matrix data reorganization processing to obtain the second initial data, wherein the second initial data is represented in matrix form, and each row represents a voltage value-image data pair; each row of data in the second initial data is subjected to sparse coding conversion, and the data in the high-dimensional feature space of the second initial data is represented as sparse features to obtain the third initial data; based on the influence of the optical response of the image data, the importance of the sparse features in the third initial data is calculated, and the sparse features in the third initial data are screened and optimized according to the calculation results to obtain the fourth initial data; the network alignment initial model is trained using the fourth initial data, and the parameters in the network alignment initial model are optimized using cross-validation and network search to obtain the final network alignment model.
[0050] In this example, the multi-layer network alignment method builds and optimizes the network alignment model by introducing historical data and random forest algorithm, thereby achieving more accurate beam alignment. First, historical data needs to be obtained. These data include the optical response data of each layer of voltage-adjustable liquid crystal devices at different voltage values, that is, the image of the focus position and focus clarity on the detection surface when the light beam passes through the liquid crystal box. These data record the optical characteristics of each layer of liquid crystal devices under different voltage conditions as the basis for model training.
[0051] Next, the random forest algorithm is used to build the initial network alignment model. The construction process includes setting the initial parameters of the algorithm, such as the number of trees and the maximum depth, and defining the data input format. In this way, the initial model can preliminarily reflect the optical behavior of the voltage-adjustable liquid crystal device, providing a basis for subsequent optimization. Then, the historical data is processed by wavelet decomposition. In this step, the data is soft-thresholded by the preset wavelet coefficients to remove noise. The processed data is then reconstructed by wavelet to obtain the first initial data after denoising. The process of wavelet decomposition and reconstruction can effectively filter out noise and retain useful signal features, thereby improving the quality of the data.
[0052] The first initial data is then subjected to matrix data reorganization processing to obtain the second initial data. The second initial data is represented in matrix form, with each row representing a voltage value and a corresponding image data pair. This step structures the data to make it more suitable for subsequent sparse coding processing. In order to improve the efficiency and effect of data processing, the second initial data is subjected to sparse coding conversion, and the data in the high-dimensional feature space is represented as sparse features to obtain the third initial data. Sparse coding helps reduce data dimensions, highlight key features, and make model training more efficient.
[0053] Based on the influence of the optical response of the image data, the sparse features in the third initial data are calculated and processed to determine the importance of each sparse feature. According to the calculation results, the sparse features are screened and optimized to obtain the fourth initial data. This step further improves the representativeness and quality of the data through feature selection and optimization. Finally, the initial network alignment model is trained using the fourth initial data, and the model parameters are optimized through cross-validation and network search to obtain the final network alignment model. Cross-validation and network search ensure that the model parameters are optimal, so that the model has stronger generalization ability and prediction accuracy.
[0054] In actual operation, when adjusting the voltage of the i-th layer, the signal generator will refer to the data in the network alignment model, and quickly find the optimal voltage value by comparing the current voltage and image data, so that the light beam can accurately pass through the i-th layer of liquid crystal devices and form a clear focus on the detection surface. This process greatly simplifies the alignment steps, reduces the adjustment time, and improves the overall alignment effect. In this way, high-precision alignment of multi-layer voltage-adjustable liquid crystal devices can be achieved, ensuring the transmission accuracy and stability of optical signals.
[0055] In one example, a voltage higher than a preset threshold is applied to the i-th layer of voltage-adjustable liquid crystal devices through a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, so that the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal devices are arranged vertically to prevent the diffraction of light from passing through, including: real-time image capture of the detection surface to obtain real-time image data of the focus; the preset code of the i-th layer of voltage-adjustable liquid crystal devices and the real-time image data of the focus are input into the network alignment model to obtain feedback information, wherein the feedback information at least includes: the voltage deviation value of the i-th layer of voltage-adjustable liquid crystal devices; an adaptive adjustment algorithm is used to iteratively adjust the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices based on the feedback information for multiple times until the light beam passes through the different layers of voltage-adjustable liquid crystal devices to produce a clear focus on the detection surface.
[0056] In this example, the signal generator of the i-th layer of voltage-adjustable liquid crystal devices is adjusted to ensure that the light beam can pass through different layers of voltage-adjustable liquid crystal devices and produce a clear focus on the detection surface. The specific steps include: first, capturing the image of the detection surface in real time to obtain real-time image data of the focus. This step continuously monitors the light spot on the detection surface through a camera or other optical sensor, and records the image after the light beam passes through the liquid crystal device.
[0057] Next, the preset code of the i-th layer of voltage-adjustable liquid crystal devices and the focus image data captured in real time are input into the network alignment model. The preset code refers to the initial voltage setting and other related parameters related to the i-th layer of voltage-adjustable liquid crystal devices. The network alignment model uses these input data to generate feedback information. The feedback information includes at least the voltage deviation value of the i-th layer of voltage-adjustable liquid crystal devices, which indicates the difference between the current voltage and the ideal voltage. Then, an adaptive adjustment algorithm is used to iteratively adjust the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices based on the feedback information. The adaptive adjustment algorithm continuously adjusts the voltage applied by the signal generator according to the voltage deviation value, gradually approaching the optimal voltage value. This process is iterative, that is, through multiple adjustments and feedback, the voltage deviation is gradually reduced until the light beam forms a clear focus on the detection surface.
[0058] The technical effect of this method is that, through real-time image capture and feedback adjustment, precise control of the i-th layer of voltage-adjustable liquid crystal devices is achieved. The network alignment model provides intelligent voltage adjustment guidance, and the adaptive adjustment algorithm ensures rapid convergence to the optimal voltage value. This method significantly improves the efficiency and accuracy of multi-layer network alignment and reduces the complexity and error of manual adjustment. Through iterative optimization, each layer of voltage-adjustable liquid crystal devices can be precisely aligned to ensure the transmission accuracy and stability of optical signals in the multi-layer structure.
[0059] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.
[0060] Process flow for fabricating voltage-tunable LC cell for D2NN. First, 0.5% SD1 was spin-coated at 3000 rpm for 30 s on a clean indium tin oxide (ITO) glass substrate and then baked on a hot plate at 100 °C for 10 min to completely remove the N, N-dimethylformamide solvent. Subsequently, a 2 μm spacer mixed with UV glue was coated along the edge of the substrate and then assembled with another SD1 substrate front side. The UV glue was cured using a 365 nm UV LED to form the wall and maintain the LC cell gap. The assembled LC cell was irradiated by a digital micromirror device (DMD)-based pattern alignment exposure system. By rotating the polarizer in front of the LC cell, patterns with different polarization azimuth angles were sequentially printed on the LC cell to build an LC device with the desired azimuth angle distribution. LC E7 (HCCH, China) was filled into the LC cell on a hot plate at 80 °C. As the temperature decreased, the LC molecules were arranged according to the SD1 micropattern, serving as a voltage-tunable LC cell for D2NN.
[0061] The working principle of light passing through the entire multi-layer network is as follows Each optical neuron can be considered a new source of light as it passes through the diffractive layers. The amplitude and phase of these sources are affected by the input wave and the complex transmission coefficients at that specific point. Optical neural networks enable these sources to pass through diffraction and interact with the neurons in the next layer. Each layer of the network adjusts the signal to accomplish a specific task. The free space diffraction of each neuron between the diffractive layers can be described by the following formula: , in l Represents the network l layer, l Indicates that it is located at l layer( , , ) No. l neurons, is the wavelength of the incident light, , .. Since the amplitude and phase of this wave are determined by the complex-valued input wave and the transmission coefficient of the optical neuron ( The product of is determined, so it is located in l layer( , , )’s neuron output function ( ) can be expressed as: , Defined as 1 Tier i The input wave of a neuron, k For all neurons in the above diffraction layer. Since the transmission coefficient of a neuron consists of an amplitude term and a phase term , so. In this work, it is assumed The amplitude is a constant because the LC component manipulates only the The phase term, plus the phase delay encountered by the secondary wave and through layer diffraction completely connects all artificial optical neurons.
[0062] Reference Figure 7 The embodiment of the present invention provides a multi-layer network alignment device based on a voltage-adjustable liquid crystal cell, comprising: A generating unit 1, used for controlling a light source device corresponding to the voltage-adjustable liquid crystal cell to generate a light beam; The adjustment unit 2 is used to sequentially adjust the voltage values applied to different layers of voltage-adjustable liquid crystal devices through multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and a clear focus is generated on the detection surface.
[0063] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.
[0064] Reference Figure 8 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.
[0065] Those skilled in the art will understand that Figure 8 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.
[0066] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.
[0067] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.
[0068] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.
[0069] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A voltage adjustable liquid crystal cell, It is characterized in that The voltage-adjustable liquid crystal box comprises: a multi-layer voltage-adjustable liquid crystal device, a plurality of signal generators, a plurality of spherical wave liquid crystal lenses, and a camera; The voltage-adjustable liquid crystal device changes the arrangement of internal liquid crystal molecules according to the applied voltage value, thereby changing the optical properties, wherein the optical properties include refractive index and diffraction efficiency; The plurality of signal generators are respectively connected to the multi-layer voltage-adjustable liquid crystal device, and are used to apply voltages to the multi-layer voltage-adjustable liquid crystal device respectively; The multiple spherical wave liquid crystal lenses are respectively arranged between the multiple layers of voltage-adjustable liquid crystal devices, and the lens focal lengths of the multiple spherical wave liquid crystal lenses are respectively the distances from the voltage-adjustable liquid crystal devices corresponding to the spherical wave liquid crystal lenses to the camera.
2. The voltage-adjustable liquid crystal cell according to claim 1, It is characterized in that The voltage adjustable liquid crystal device is assembled by bonding one SD1 substrate to another SD1 substrate through UV glue.
3. The voltage-adjustable liquid crystal cell according to claim 1, It is characterized in that The voltage-adjustable liquid crystal device is made by light-controlled orientation using a digital mask lithography machine and coating E7 liquid crystal.
4. A multilayer network alignment method based on voltage-adjustable liquid crystal cells, It is characterized in that The voltage-adjustable liquid crystal cell is the voltage-adjustable liquid crystal cell according to any one of claims 1 to 3, and the multilayer network alignment method comprises: Controlling a light source device corresponding to the voltage-adjustable liquid crystal box to generate a light beam; The voltage values applied to different layers of voltage-adjustable liquid crystal devices are adjusted in sequence by multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and a clear focus is generated on the detection surface.
5. The multi-layer network alignment method according to claim 4, It is characterized in that The voltage values applied to different layers of voltage-adjustable liquid crystal devices are adjusted in sequence by multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and a clear focus is generated on the detection surface, including: According to the distance between the voltage-adjustable liquid crystal device and the detection surface, from near to far, the following operations are performed on each voltage-adjustable liquid crystal device in sequence: By means of a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices, the voltage applied to the i-th layer of voltage-adjustable liquid crystal devices is adjusted until the light beam passes through the voltage-adjustable liquid crystal devices of different layers and a clear focus is generated on the detection surface; A voltage higher than a preset threshold is applied to the i-th voltage-adjustable liquid crystal device through a signal generator corresponding to the i-th voltage-adjustable liquid crystal device, so that the liquid crystal molecules in the i-th voltage-adjustable liquid crystal device are arranged vertically to prevent the diffraction of light from passing through.
6. The multi-layer network alignment method according to claim 4, It is characterized in that Before applying a voltage higher than a preset threshold to the i-th layer of voltage-adjustable liquid crystal device through a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal device to vertically align the liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal device to prevent the diffraction of light from passing through, the multi-layer network alignment method further includes: Acquiring historical data, wherein the historical data at least includes: image data of focus positions and focus clarity corresponding to light beams passing through each layer of voltage-adjustable liquid crystal devices at different voltage values; Constructing a network alignment initial model based on a random forest algorithm, wherein the construction content at least includes: setting initial parameters and data input format of the network alignment initial model; Performing wavelet decomposition processing on the historical data, performing soft threshold processing on the historical data after wavelet decomposition based on preset wavelet coefficients, and performing wavelet reconstruction processing on the processing result to obtain first initial data after denoising; Performing matrix data reorganization processing on the first initial data to obtain second initial data, wherein the second initial data is represented in a matrix form, and each row represents a voltage value-image data pair; Performing sparse coding conversion on each row of data in the second initial data, representing the data in the high-dimensional feature space of the second initial data as sparse features, and obtaining third initial data; Based on the optical response influence of the image data, the importance of the sparse features in the third initial data is calculated, and the sparse features in the third initial data are screened and optimized according to the calculation results to obtain fourth initial data; The fourth initial data is used to train the network alignment initial model, and cross-validation and network search are used to optimize the parameters in the network alignment initial model to obtain a final network alignment model.
7. The multi-layer network alignment method according to claim 6, It is characterized in that Applying a voltage higher than a preset threshold to the i-th layer of voltage-adjustable liquid crystal device through a signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal device, so that liquid crystal molecules in the i-th layer of voltage-adjustable liquid crystal device are vertically aligned to prevent diffraction of light, including: Capturing the image of the detection surface in real time to obtain focal point real-time image data; Inputting the preset code of the i-th layer of voltage-adjustable liquid crystal device and the focus real-time image data into the network alignment model to obtain feedback information, wherein the feedback information at least includes: the voltage deviation value of the i-th layer of voltage-adjustable liquid crystal device; Adopting an adaptive adjustment algorithm, the signal generator corresponding to the i-th layer of voltage-adjustable liquid crystal devices is adjusted multiple times iteratively based on the feedback information until the light beam passes through the voltage-adjustable liquid crystal devices of different layers and produces a clear focus on the detection surface.
8. A multi-layer network alignment device based on a voltage-adjustable liquid crystal cell, It is characterized in that include: A generating unit, used for controlling the light source device corresponding to the voltage-adjustable liquid crystal box to generate a light beam; The adjustment unit is used to sequentially adjust the voltage values applied to different layers of voltage-adjustable liquid crystal devices through multiple signal generators until the light beam passes through multiple layers of the voltage-adjustable liquid crystal devices and a clear focus is generated on the detection surface.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 4 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 4 to 7 are implemented.