Optical diffraction neural network testing device and system

Abstract

The utility model provides an optical diffraction neural network testing device and system which can be applied to the technical field of optical testing, and the optical diffraction neural network testing device comprises a spatial light modulator which is used for generating a patterned light field with a variable shape; the spatial light modulator, the polarizing film, the first lens, the light splitting plain film, the first microscope objective, the to-be-detected sample, the second microscope objective, the second lens and the camera are sequentially arranged along the main optical axis direction; the third lens and the light source are vertically arranged below the light splitting plain film in sequence; the laser, the beam splitter prism, the half-wave plate and the reflector are sequentially arranged along the auxiliary optical axis direction, and the main optical axis direction is parallel to the auxiliary optical axis direction; and the reflecting mirror emits the laser in the auxiliary optical axis direction into the spatial light modulator. The variable-shape patterned light field is generated through the spatial light modulator, so that the limitation that the input light field of a traditional device is single is broken through, variable input is provided for testing of the optical diffraction neural network, and therefore a testing result can be conveniently obtained.

Description

Technical Field

[0001] The utility model relates to the technical field of optical testing, and particularly relates to an optical diffraction neural network testing device and system. Background Technique

[0002] Optical diffraction neural networks process information by utilizing the diffraction characteristics of light waves. Different from traditional electronic neural networks, optical diffraction neural networks achieve the propagation and processing of information by using a diffraction surface with a specific optical structure. Through the special design of the spatial phase modulation of the diffraction surface, optical diffraction neural networks can passively achieve the functions of a multi-layer perceptron and process complex inference and calculation tasks.

[0003] In the related optical diffraction neural network testing device, an optical mask with a fixed shape and characteristics is used as the module for modulating the input light field in the testing device. However, the optical mask with a fixed shape and characteristics causes the input light field to be unable to be flexibly changed, thus unable to meet the testing requirements with a large amount of complex and variable data. Summary of the Utility Model

[0004] In view of this, the utility model provides an optical diffraction neural network testing device. The above-mentioned optical diffraction neural network testing device is used to test a sample to be tested. The above-mentioned sample to be tested is an optical diffraction neural network computing device. The above-mentioned optical diffraction neural network testing device includes a laser, a beam splitter prism, a half-wave plate, a reflecting mirror, a spatial light modulator, a polarizer, a first lens, a first microscope objective, a second microscope objective, a second lens, a camera, a beam splitter flat plate, a third lens and a light source. The above-mentioned spatial light modulator is used to generate a patterned light field with a variable shape. The above-mentioned spatial light modulator, the above-mentioned polarizer, the above-mentioned first lens, the above-mentioned beam splitter flat plate, the above-mentioned first microscope objective, the above-mentioned sample to be tested, the above-mentioned second microscope objective, the above-mentioned second lens and the above-mentioned camera are sequentially arranged along the main optical axis direction. The above-mentioned third lens and the above-mentioned light source are sequentially and vertically arranged under the above-mentioned beam splitter flat plate. The above-mentioned laser, the above-mentioned beam splitter prism, the above-mentioned half-wave plate and the above-mentioned reflecting mirror are sequentially arranged along the auxiliary optical axis direction. The above-mentioned main optical axis direction and the above-mentioned auxiliary optical axis direction are parallel. The above-mentioned reflecting mirror incident the laser in the above-mentioned auxiliary optical axis direction into the above-mentioned spatial light modulator.

[0005] Further, the laser light emitted by the above-mentioned laser is incident into the above-mentioned beam splitter prism, and linearly polarized light is obtained after being processed by the above-mentioned beam splitter prism and is incident into the above-mentioned half-wave plate.

[0006] Further, the above-mentioned half-wave plate performs polarization processing on the linearly polarized light to obtain a target linearly polarized light that meets the polarization direction requirements of the above-mentioned spatial light modulator and is incident into the above-mentioned spatial light modulator.

[0007] Further, the above-mentioned spatial light modulator modifies the phase information of the target linearly polarized light by performing phase modulation on the target linearly polarized light, and the light is incident on the above-mentioned polarizer.

[0008] Further, the above-mentioned polarizer performs polarization filtering on the light field to obtain a target patterned light field, and successively undergoes scaling processing by the above-mentioned first lens and the above-mentioned first microscope objective lens, and is projected onto the input plane of the above-mentioned sample to be measured.

[0009] Further, the illumination light emitted by the above-mentioned light source is incident on the above-mentioned beam splitter flat plate through the above-mentioned third lens, and is irradiated onto the above-mentioned sample to be measured through the above-mentioned beam splitter flat plate.

[0010] Further, the above-mentioned second microscope objective lens receives the output light field of the above-mentioned sample to be measured, scales the above-mentioned output light field to obtain a scaled target output light field, and the above-mentioned target output light field is captured by the photosensitive surface of the above-mentioned camera through the above-mentioned second lens.

[0011] Further, the polarization direction requirement includes that the included angle between the polarization direction of the linearly polarized light modulated by the above-mentioned half-wave plate and the extraordinary axis of the above-mentioned spatial light modulator is 45 degrees.

[0012] Further, the above-mentioned spatial light modulator is used to set the phase value at the corresponding position of the pixel to a preset phase value when the pixel gray value of the target generated light field is a preset value.

[0013] According to another aspect of the present invention, there is also provided an optical diffraction neural network testing system, including: the optical diffraction neural network testing device described in any one of the above.

[0014] The present invention has the following beneficial effects:

[0015] (1) According to the embodiments of the present invention, a variable-shaped patterned light field is generated by a spatial light modulator, thereby breaking through the limitation of the single input light field of traditional devices, providing variable inputs for the testing of optical diffraction neural networks, and thus obtaining test results more conveniently;

[0016] (2) Through the second microscope objective lens, the second lens and the camera, the spatial distribution of the output light field of the optical diffraction neural network can be obtained in real time, providing high-resolution data support for testing, so as to facilitate performance analysis and optimization;

[0017] (3) Illumination light is provided by a light source and a shared objective lens to achieve precise observation and positioning of the structure of the optical diffraction neural network, ensuring the accuracy of test operations. Description of the Drawings

[0018] Through the following description of the embodiments of the present invention with reference to the drawings, the above and other objects, features and advantages of the present invention will become clearer. In the drawings:

[0019] Figure 1 Shows an optical path schematic diagram according to an embodiment of the present utility model.

[0020] 101 - Laser; 102 - Beam splitting prism; 103 - Half-wave plate; 104 - Reflecting mirror; 105 - Spatial light modulator; 106 - Polarizer; 107 - First lens; 108 - Beam splitting flat plate; 109 - 40x microscope objective; 110 - Sample to be measured; 111 - 60x microscope objective; 112 - Second lens; 113 - CCD camera; 114 - Third lens; 115 - LED light source. Detailed implementation manners

[0021] In order to make the objectives, technical solutions and advantages of the present utility model more clear and understandable, the following further describes the present utility model in detail with reference to specific embodiments and the accompanying drawings.

[0022] The terms used herein are only for describing specific embodiments and are not intended to limit the present utility model. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations and / or components, but do not exclude the presence or addition of one or more other features, steps, operations or components.

[0023] All terms used herein, including technical and scientific terms, have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0024] In the case of using expressions such as "at least one of A, B, and C, etc.", generally, it should be interpreted according to the meaning commonly understood by those skilled in the art. For example, "a system having at least one of A, B, and C" should include, but not be limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C. In the case of using expressions such as "at least one of A, B, or C, etc.", generally, it should be interpreted according to the meaning commonly understood by those skilled in the art. For example, "a system having at least one of A, B, or C" should include, but not be limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C.

[0025] It should also be noted that the directional terms mentioned in the embodiments, such as "up", "down", "front", "back", "left", "right", etc., are only for reference to the drawings and do not limit the protection scope of the present invention. Throughout the drawings, the same elements are denoted by the same or similar reference numerals. When it may cause confusion in the understanding of the present invention, the conventional structures or configurations will be omitted.

[0026] Figure 1 Fig. shows the optical path schematic diagram according to an embodiment of the present invention.

[0027] As Figure 1 shown, the present invention at least includes a laser 101, a beam splitting prism 102, a half-wave plate 103, a mirror 104, a spatial light modulator 105, a polarizer 106, a first lens 107, a beam splitting flat plate 108, a 40x microscope objective 109, a 60x microscope objective 111, a second lens 112, a CCD camera 113, a third lens 114, and an LED light source 115.

[0028] According to an embodiment of the present invention, the sample to be measured 110 is a sample for optical diffraction neural network calculation. The optical diffraction neural network is an optical-based computing system that utilizes the diffraction and interference characteristics of light to achieve the computing function of the neural network. The sample to be measured 110 can be, for example, a diffractive optical element or material, and its characteristics need to be evaluated through optical tests.

[0029] According to an embodiment of the present invention, the optical diffraction neural network test device can be arranged on an optical platform. The optical platform is an experimental platform for supporting and fixing components and is widely used in optical experiments. The main optical axis is selected as Figure 1 the direction of light transmission in the spatial light modulator 105 to the 40x microscope objective 109, and the auxiliary optical axis is selected as the direction of light transmission in the laser 101 to the mirror 104. The mirror 104 is used to guide the laser emitted by the laser 101 into the spatial light modulator 105. The mirror 104 and the spatial light modulator 105 can be placed at any angle as long as it ensures that the laser can be guided into the spatial light modulator 105.

[0030] According to an embodiment of the present utility model, the laser 101, the beam splitting prism 102, the half-wave plate 103, and the mirror 104 can be summarized as a light source module, which is used to provide stable linearly polarized light for the optical diffraction neural network testing device; the spatial light modulator 105, the polarizer 106, the first lens 107, and the 40x microscope objective 109 can be summarized as a beam shaping module, which can modulate the input linearly polarized light into an arbitrarily patterned light field and project it onto the input plane of the sample to be tested 110; the 60x microscope objective 111, the second lens 112, and the CCD camera 113 can be summarized as an information acquisition module, which is used to collect the light field information of the output plane of the sample to be tested 110; the beam splitting flat plate 108, the third lens 114, and the LED light source 115 can be summarized as an imaging module, which is used to assist in observing the position of the sample to be tested 110, that is, the optical diffraction neural network structure, to achieve precise positioning.

[0031] According to an embodiment of the present utility model, the laser 101 can be a semiconductor laser, the beam splitting prism 102 can be a polarization beam splitting prism, and the spatial light modulator 105 can be a liquid crystal spatial light modulator.

[0032] The principle of the present utility model will be described below. By analyzing the self - characteristics of optical elements and basic optical principle knowledge, the laser 101 is a device capable of emitting laser light. Its principle is based on stimulated emission, and through a specific physical mechanism, the coherent amplification and output of light are realized. The laser 101 outputs a test laser beam (which can be a collimated laser beam). After being adjusted by the beam - splitting prism 102 and the half - wave plate 103, it is converted into linearly polarized light with a specific polarization direction. The beam - splitting prism 102 is an optical element that can decompose incident light into two beams of light with perpendicular polarization directions. It is usually composed of two high - precision right - angled prisms. The inclined surface of one prism is coated with multiple layers of polarization - splitting films and then glued to the other prism to form a cube structure. When light is incident on the polarization - splitting film at a specific angle (usually the Brewster angle), the polarized light parallel to the incident plane is almost completely transmitted, while the polarized light perpendicular to the incident plane is reflected. This characteristic enables the beam - splitting prism 102 to decompose unpolarized light or partially polarized light into two beams of linearly polarized light, and the polarization directions of the two beams of light are perpendicular to each other. The half - wave plate 103 is an optical element, usually made of a birefringent crystal, and its thickness is precisely controlled so that the phase difference between the ordinary light and the extraordinary light passing through the wave plate is π. Through the combined adjustment of the beam - splitting prism 102 and the half - wave plate 103, the test laser beam is converted into linearly polarized light with a specific polarization direction that meets the requirements of the spatial light modulator 105. For the embodiments of the present utility model, the angle between the polarization direction of the linearly polarized light modulated by the half - wave plate 103 and the extraordinary axis of the spatial light modulator 105 can be 45 degrees. The present utility model does not limit this, and it can be specifically set according to the requirements of the spatial light modulator 105.

[0033] According to an embodiment of the present utility model, the phase modulation of linearly polarized light can be performed by the spatial light modulator 105. The spatial light modulator 105 is an optical device capable of dynamically controlling the spatial distribution of light waves. It can change parameters such as the amplitude, phase, and polarization state of the light field under the control of an electrical drive signal or other signals, thereby loading information into the light wave. For a liquid - crystal spatial light modulator, its core principle is based on the birefringence or electro - optic effect of liquid - crystal molecules. When a voltage is applied, the arrangement of liquid - crystal molecules changes, thus changing the propagation characteristics of light. In the present utility model, we use the spatial light modulator 105 to perform phase modulation on linearly polarized light and perform polarization filtering with the polarizer 106 to obtain a patterned light field. The role of the polarizer 106 is to filter out the unwanted polarization directions and only retain the polarized light in a specific direction.

[0034] According to an embodiment of the present utility model, after obtaining the patterned light field, the first lens 107 and the 40x microscope objective lens 109 are used to reduce the patterned light field so as to focus on the input plane of the sample to be measured 110, thereby obtaining a high-precision patterned light field that meets the input requirements of the optical diffraction neural network. The sum of the focal lengths of the first lens 107 and the 40x microscope objective lens 109 is 4f. The present utility model does not limit the magnification of the microscope objective lens, and can be specifically set according to the actual situation of the sample to be measured 110.

[0035] According to an embodiment of the present utility model, the spatial light modulator 105 is used to set the phase value at the corresponding position to a preset phase value when the pixel gray value of the target generated light field is a preset value.

[0036] According to an embodiment of the present utility model, after determining the target light field to be generated and clarifying its spatial distribution and optical characteristic requirements, the gray value can be set. When the pixel gray value is set to 128, the phase value at the corresponding position is set to π, and when the gray value is 0, the phase value at the corresponding position is set to 0.

[0037] According to an embodiment of the present utility model, the LED light source 115 provides illumination light, which is introduced into the beam shaping optical path through the third lens 114 and then through the beam splitting flat plate 108, and is irradiated onto the sample to be measured 110 through the 40x microscope objective lens 109, so that the structural positioning and observation of the sample to be measured 110 can be realized.

[0038] According to an embodiment of the present utility model, the 60x microscope objective lens 111 receives the output light field of the sample to be measured 110, scales the output light field to obtain a scaled target output light field, and the target output light field is captured by the photosensitive surface of the CCD camera 113 via the second lens 112.

[0039] According to an embodiment of the present utility model, the output light field of the sample to be measured 110 is collected by the 60x microscope objective lens 111 and the second lens 112, and is captured in real time by the CCD camera 113, and can be sent to a computer for further analysis. Among them, the CCD camera 113 is a Charge-Coupled Device.

[0040] According to an embodiment of the present utility model, the optical diffraction neural network test device of the present utility model can be operated through the following operation steps:

[0041] Step S10, turn on the laser 101, load the calculated phase diagram onto the spatial light modulator 105, and observe the generated patterned light field through the CCD camera 113;

[0042] Step S20: Place the sample to be tested 110 into the device, turn on the LED light source 115, and observe the position of the structure of the sample to be tested 110 through the CCD camera 113.

[0043] Step S30: Adjust the position of the sample to be tested 110 to make the patterned light field exactly coincide with the designed input plane of the sample to be tested 110.

[0044] Step S40: Adjust the position of the 60x microscope objective lens 111 so that the object focal plane of the 60x microscope objective lens 111 coincides with the output plane of the sample to be tested 110.

[0045] Step S50: Change the phase map loaded on the spatial light modulator 105 to generate different patterned light fields, capture the light field information output by the sample to be tested 110 in real time through the CCD camera 113, and convert it into a digital signal for analysis.

[0046] According to an embodiment of the present invention, the optical diffraction neural network test system may include the optical diffraction neural network test device corresponding to any embodiment. Through the modulation and demodulation of the light field and the multi-level cascaded diffraction structure, the system can achieve complex transformations of the input light field, thereby completing information encoding and decoding to obtain the test result. Exemplarily, the system can be used to observe the results output by the optical diffraction neural network test for identifying handwritten numbers or letters, and further analyze them via devices such as servers. By designing a specific optical network structure, handwritten numbers or letters can be quickly identified and classified with high precision.

[0047] The embodiments of the present invention have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present invention is defined by the appended claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and these substitutions and modifications should fall within the scope of the present invention.

Claims

1. An optical diffraction neural network testing device, characterized in that: The optical diffraction neural network testing device is used to test a sample to be tested, and the sample to be tested is an optical diffraction neural network computing device. The optical diffraction neural network testing device includes a laser, a beam splitter prism, a half-wave plate, a reflector, a spatial light modulator, a polarizer, a first lens, a first microscope objective lens, a second microscope objective lens, a second lens, a camera, a beam splitter plate, a third lens and a light source. The spatial light modulator is used to generate a patterned light field with a variable shape; The spatial light modulator, the polarizer, the first lens, the beam splitter, the first microscope objective, the sample to be measured, the second microscope objective, the second lens and the camera are sequentially arranged along the main optical axis direction; The third lens and the light source are arranged vertically below the beam splitter plate in sequence; The laser, the beam splitter, the half-wave plate and the reflector are sequentially arranged along the auxiliary optical axis direction, and the main optical axis direction and the auxiliary optical axis direction are parallel; The reflecting mirror allows the laser light in the auxiliary optical axis direction to be incident on the spatial light modulator.

2. The optical diffraction neural network testing device according to claim 1, characterized in that: The laser light emitted by the laser is incident on the beam splitter prism, processed by the beam splitter prism to obtain linearly polarized light and then incident on the half wave plate.

3. The optical diffraction neural network testing device according to claim 1, characterized in that: The half-wave plate performs polarization processing on the linearly polarized light to obtain target linearly polarized light that meets the polarization direction requirement of the spatial light modulator and then injects the target linearly polarized light into the spatial light modulator.

4. The optical diffraction neural network testing device according to claim 1, characterized in that: The spatial light modulator performs phase modulation on the target linear polarized light to modify the phase information of the target linear polarized light, and then the target linear polarized light is incident on the polarizer.

5. The optical diffraction neural network testing device according to claim 1, characterized in that: The polarizer performs polarization filtering on the patterned light field to obtain a target patterned light field, and the target patterned light field is sequentially zoomed through the first lens and the first microscope objective lens, and projected onto the input plane of the sample to be tested.

6. The optical diffraction neural network testing device according to claim 1, characterized in that: The illumination light emitted by the light source is incident on the spectroscopic plate via the third lens, and is then irradiated onto the sample to be tested via the spectroscopic plate.

7. The optical diffraction neural network testing device according to claim 1, characterized in that: The second microscope objective lens receives the output light field of the sample to be tested, scales the output light field to obtain a scaled target output light field, and the target output light field is captured by the photosensitive surface of the camera via the second lens.

8. The optical diffraction neural network testing device according to claim 3, characterized in that: The polarization direction requirement includes that the angle between the polarization direction of the linearly polarized light modulated by the half-wave plate and the extraordinary axis of the spatial light modulator is 45 degrees.

9. The optical diffraction neural network testing device according to claim 1, characterized in that: The spatial light modulator is used to set the phase value of the position corresponding to the pixel to the preset phase value when the pixel gray value of the target generated light field is the preset value.

10. An optical diffraction neural network testing system, characterized in that: include: A device as claimed in any one of claims 1 to 9.

Images