CN-122021237-A - Visible light broadband edge detection system and diffraction optical element design method

CN122021237ACN 122021237 ACN122021237 ACN 122021237ACN-122021237-A

Abstract

The invention discloses a visible light broadband edge detection system and a diffraction optical element design method, and relates to the technical field of optical neural networks. The system comprises an information input module, an all-optical edge detection module and an information acquisition module, wherein the information input module comprises a light source and a mask plate and is used for converting object information into intensity information of light beams so as to obtain an input light field, the all-optical edge detection module comprises a 4f system and a diffraction optical element and is used for outputting the light field after phase modulation, the information acquisition module acquires the intensity information of the output light field through a photoelectric detector, and the output light field is subjected to phase modulation through the 4f system and the diffraction optical element so as to realize broadband edge detection. The design of the diffraction optical element realizes the micro-mapping from the diffraction element structural parameters to the output light field by establishing a micro end-to-end light field transmission simulation model, and optimizes the diffraction element structural parameters based on a deep learning technology. The end-to-end reverse design method provided by the invention can realize full-optical broadband edge detection with high diffraction efficiency without physical priori knowledge.

Inventors

CHANG WEIJIE
WANG HEWEN
XU SHENGYAO
HUANG FENG

Assignees

福州大学

Dates

Publication Date: 20260512
Application Date: 20251210

Claims (10)

1. A visible light broadband edge detection system, comprising: the information input module comprises a light source and a mask plate and is used for converting object information into intensity information of light beams so as to obtain an input light field; The full-optical edge detection module comprises a 4f system and a diffraction optical element DOE, wherein an input light field is subjected to Fourier transform through a first lens in the 4f system, and the light field subjected to Fourier transform is subjected to phase modulation through a second lens in the 4f system after being subjected to Fourier transform through the diffraction optical element DOE to obtain an output light field subjected to phase modulation; the information acquisition module is used for acquiring the output light field after the phase modulation through the photoelectric detector to obtain the intensity information of the output light field.
2. The system of claim 1, wherein the end-to-end inverse learning of the phase modulation parameters of the diffractive optical element DOE comprises: designing a diffraction optical system, and establishing a phase modulation model of the DOE of the diffraction optical element on the light field by utilizing a light field forward propagation model and a Rayleigh-Soxhlet equation according to the light source type, the detector, the lens and the material of the DOE of the diffraction optical element; giving an input image set according to task requirements and constructing a joint loss function according to the characteristic loss and diffraction efficiency ratio between an output light field and a target light field; the input image obtains a target image through an electrical convolution operator, and obtains an output image through a DOE (data of eye) on a phase modulation model of the light field; And (3) utilizing a deep learning and error back propagation algorithm according to the joint loss function, adjusting the system structure in the training process, and optimizing the phase modulation matrix of the DOE of the diffraction optical element.
3. The system of claim 2, wherein constructing the joint loss function comprises constructing a first loss function based on a mean square error loss between the predicted light field and the target light field, constructing a second loss function based on a difference in diffraction efficiency ratio between 1 and the predicted light field and the target light field, and constructing the joint loss function based on the first loss function and the second loss function.
4. The system of claim 1, wherein the information input module comprises a laser, a pinhole filter, a collimating lens and a mask plate, wherein the laser emits laser light, the laser light is expanded by the filter, and the laser light passes through the collimating lens and is subjected to amplitude modulation by the mask plate to obtain an input light field.
5. A method of designing a diffractive optical element, comprising: Designing a diffraction optical system, and establishing a phase modulation model of the diffraction optical element DOE on a light field by utilizing a light field forward propagation model and a Rayleigh-Soxhlet equation according to the light source type, the detector, the lens and the material of the diffraction optical element DOE; The loss function construction process comprises the steps of constructing a joint loss function according to the characteristic loss and diffraction efficiency ratio between an output light field and a target light field; The parameter learning process comprises the steps of obtaining a target image from an input image set through an electrical convolution operator, obtaining an output image through a phase modulation model of a DOE of a diffraction optical element, adjusting a system structure in a training process according to a joint loss function by utilizing a deep learning and error back propagation algorithm, and optimizing the phase modulation matrix of the DOE of the diffraction optical element.
6. The method of claim 5, wherein constructing a first loss function based on the mean square error loss between the predicted light field and the target light field, constructing a second loss function based on the difference between 1 and the diffraction efficiency ratio between the predicted light field and the target light field, constructing a joint loss function based on the first loss function and the second loss function; The formula of the first loss function is: where n is the number of pixels of the image, Representing the output light field intensity of the ith sample point, I i representing the target light field intensity of the ith sample point; The formula of the second loss function is: Wherein, the Representing the output light field intensity for the ith sample point, I imax represents the maximum value of the target light field intensity for the ith sample point.
7. The method of claim 5, wherein the step height matrix of the DOE is converted from the learned phase modulation matrix to achieve phase modulation, and the conversion formula is: Wherein, the For the difference between the refractive indices of DOE material and air, lambda is the wavelength of light incident into the all-optical edge detection module, Is a phase matrix.
8. The method of claim 7, wherein the diffractive optical element DOE is fabricated according to a step height matrix of DOEs.
9. The method of claim 5, wherein the light source is a coherent light source and the material of the diffractive optical element DOE is SK1300 material.
10. The method of claim 5, wherein the diffractive optical element DOE is an eighth order quantized DOE device.

Description

Visible light broadband edge detection system and diffraction optical element design method Technical Field The invention relates to the technical field of optical neural networks, in particular to a visible light broadband edge detection system and a diffraction optical element design method. Background With the rapid development of artificial intelligence technology, image processing plays an increasingly important role in various fields such as medical imaging, automatic driving, industrial detection and the like. However, conventional electronic computers typically employ von neumann architecture, inevitably suffering from "memory wall" and "power wall" bottlenecks. Furthermore, the development of moore's law has slowed significantly and the size of silicon-based chip transistors is approaching its physical limits. These challenges make it increasingly difficult for existing electronic computing to meet the exponentially growing computing demands of the artificial intelligence era. Optical computing has the inherent advantages of high speed, low power consumption, large bandwidth, and massive parallel processing, as compared to electronic computing, thereby providing a new computing paradigm for addressing these limitations. Optical image edge detection has become an emerging target recognition and detection technique. In recent years, with the development of micro-nano processing technology and material science, a super surface becomes a compact and multifunctional all-optical edge detection platform due to the fact that the amplitude, the phase and the polarization of an optical field can be manipulated. Generally, there are two methods for plenoptic image edge detection, namely, the green function method and the spatial fourier phase shift method. By using the green function method, the super-surface structure can be directly designed in the space domain so as to meet specific optical transfer functions required by all-optical space differentiation and edge detection. However, the conventional method, although realizing miniaturization and integration of the system, inevitably has limitations such as specific incidence angle, polarization dependence, narrow working bandwidth, complex device design and the like, and is difficult to realize complex and changeable practical application scenes. As a classical alternative, the optical 4f system utilizes the fourier transform capability of the conventional lens, combines spatial frequency domain filtering, realizes angle-insensitive polarization-independent broadband edge detection, and has stronger robustness and stability. However, most edge detection architectures based on 4f systems employ amplitude-type filtering elements, resulting in low light energy utilization. In recent years, phase modulation devices such as a super surface and a diffractive optical element DOE have been integrated into an edge detection system, and the purposes of miniaturization and integration are achieved. However, conventional phase-type schemes typically rely on forward designs based on a priori knowledge and can only operate at a single wavelength, failing to achieve wideband edge detection, limiting their potential application in real-world scenarios. Therefore, how to effectively implement broadband edge detection with polarization independence, insensitivity to incidence angle, and high diffraction efficiency while eliminating the reliance on physical prior knowledge remains to be explored. Disclosure of Invention The technical problem to be solved by the invention is to provide a visible light broadband edge detection system and a diffraction optical element design method, wherein a diffraction optical element DOE is introduced into a 4f system, a phase modulation matrix is learned through end-to-end reverse design, and polarization independence, insensitivity of an incident angle, high diffraction efficiency and broadband edge detection are realized under the condition of no physical priori knowledge. In a first aspect, the present invention provides a visible light broadband edge detection system, comprising: the information input module comprises a light source and a mask plate and is used for converting object information into intensity information of light beams so as to obtain an input light field; The full-optical edge detection module comprises a 4f system and a diffraction optical element DOE, wherein an input light field is subjected to Fourier transform through a first lens in the 4f system, and the light field subjected to Fourier transform is subjected to phase modulation through a second lens in the 4f system after being subjected to Fourier transform through the diffraction optical element DOE to obtain an output light field subjected to phase modulation; the information acquisition module is used for acquiring the output light field after the phase modulation through the photoelectric detector to obtain the intensity information of the output light fie