KR-20260062702-A - BEAM PROFILE DERIVATION STRUCTURE DESIGN SYSTEM, DESIGN METHOD, LIGHT SOURCE SYSTEM DESIGNED THEREBY, AND APPLICATION INCLUDING THE SAME

KR20260062702AKR 20260062702 AKR20260062702 AKR 20260062702AKR-20260062702-A

Abstract

A beam profile derivation structure design system, a design method, a light source system designed thereby, and an application including the same are disclosed. The design system is a beam profile derivation structure design system performed by a computing device, wherein the computing device may perform: (a) a step of setting a target beam profile; (b) a step of deriving a design phase filter through a deep learning model; (c) a step of deriving an expected beam profile using the design phase filter through a forward model; (d) a step of deriving a loss function from the target beam profile and the expected beam profile; and (e) a step of updating the deep learning model from the loss function.

Inventors

주철민
조진우
김영훈
조담빈
김인경

Assignees

연세대학교 산학협력단

Dates

Publication Date: 20260507
Application Date: 20241029

Claims (12)

As a beam profile derivation structure design system performed by a computing unit, The above computing device is, (a) Step of setting the target beam profile; (b) A step of deriving a design phase filter through a deep learning model; (c) A step of deriving an expected beam profile using the design phase filter through a forward model; (d) a step of deriving a loss function from the target beam profile and the expected beam profile; and (e) a step of updating the deep learning model from the above loss function; and Repeating steps (b) to (e) above one or more times, Beam profile derivation structure design system.
In paragraph 1, The above deep learning model comprises one or more models selected from the group including a convolutional neural network (CNN) composed of multiple convolutional layers, a fully connected network, U-net, ResNet, VGGNet, InceptionNet, DenseNet, MobileNet, Transformer, BERT, GPT, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), GAN (Generative Adversarial Network), Autoencoder, Variational Autoencoder (VAE), Deep Belief Network (DBN), Deep Q-Network (DQN), Capsule Network, Siamese Network, Attention Network, and NASNet. Beam profile derivation structure design system.
In paragraph 1, The above forward model receives the phase pattern of a design phase filter as input, simulates the propagation characteristics of a beam based on a physical optical illumination system, calculates a beam profile formed through the design phase filter, and outputs the calculated beam profile as an expected beam profile. Beam profile derivation structure design system.
In paragraph 1, In the step of deriving the design phase filter through the above deep learning model, The above-mentioned design phase filter is derived by a linear combination of one or more phase maps, and The above deep learning model derives the weights of each topology map used in the above linear combination, Beam profile derivation structure design system.
In paragraph 4, Among the above one or more phase maps, at least the nth phase map is a phase map expressed as the nth power of the distance from the center, Beam profile derivation structure design system.
In paragraph 1, The above computing device comprises one or more devices selected from the group including computers, servers, cloud computing platforms, and GPUs. Beam profile derivation structure design system.
As a beam profile derivation structure design method performed by a computing device, (a) Step of setting the target beam profile; (b) A step of deriving a design phase filter through a deep learning model; (c) A step of deriving an expected beam profile using the design phase filter through a forward model; (d) a step of deriving a loss function from the target beam profile and the expected beam profile; and (e) a step of updating the deep learning model from the loss function; comprising, Repeating steps (b) to (e) above one or more times, Beam profile derivation structure design method.
In Paragraph 7, The above deep learning model comprises one or more models selected from the group including a convolutional neural network (CNN) composed of multiple convolutional layers, a fully connected network, U-net, ResNet, VGGNet, InceptionNet, DenseNet, MobileNet, Transformer, BERT, GPT, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), GAN (Generative Adversarial Network), Autoencoder, Variational Autoencoder (VAE), Deep Belief Network (DBN), Deep Q-Network (DQN), Capsule Network, Siamese Network, Attention Network, and NASNet. Beam profile derivation structure design method.
In Paragraph 7, The above forward model receives the phase pattern of a design phase filter as input, simulates the propagation characteristics of a beam based on a physical optical illumination system, calculates a beam profile formed through the design phase filter, and outputs the calculated beam profile as an expected beam profile. Beam profile derivation structure design method.
In Paragraph 7, In the step of deriving the design phase filter through the above deep learning model, The above-mentioned design phase filter is derived by a linear combination of one or more phase maps, and The above deep learning model derives the weights of each topology map used in the above linear combination, Beam profile derivation structure design method.
In Paragraph 10, Among the above one or more phase maps, at least the nth phase map is a phase map expressed as the nth power of the distance from the center, Beam profile derivation structure design method.
In Paragraph 7, The above computing device comprises one or more devices selected from the group including computers, servers, cloud computing platforms, and GPUs. Beam profile derivation structure design method.

Description

Beam profile derivation structure design system, design method, light source system designed thereby, and application including the same The present invention relates to a beam profile derivation structure design system, a design method, a light source system designed thereby, and an application including the same. An optical system refers to a device or technology that achieves various purposes by utilizing the properties, propagation, and interactions of light. Such systems are composed of various optical elements, including lenses, mirrors, prisms, filters, and optical fibers, and are used in multiple fields such as imaging, lighting, communications, and sensing. Optical systems are designed to manipulate and control light to obtain desired results and play a crucial role in diverse sectors including scientific research, medicine, industry, and communications. A beam profile refers to the manner and shape in which light emitted by a light source is distributed in space, including beam intensity distribution and energy concentration. It describes how a beam generated by a laser or other optical device spreads with a specific shape and intensity, and is important in various optical applications. An accurate beam profile is essential for high-resolution imaging, precision machining, and efficient communication systems, and can directly determine system performance. Therefore, accurately designing and implementing a beam profile is critical. In conventional technology, methods for forming beam profiles primarily involved controlling the beam's path and intensity using physical lenses or filters. For example, it was common practice to use spherical or cylindrical lenses to focus or disperse the beam into a specific shape. Additionally, methods were employed to obtain a desired beam profile by controlling the beam's phase using diffraction gratings or phase filters. However, these methods required complex mathematical calculations and iterative experiments, and had limitations in deriving the optimal beam profile under various conditions. Deep learning is a machine learning technique based on artificial neural networks, referring to a technology that learns data and recognizes complex patterns through multi-layered neural networks. Deep learning is particularly strong in large-scale data processing and pattern recognition, and is utilized in various fields such as image and speech recognition and natural language processing. Deep learning models learn from large amounts of data and, based on this, acquire the ability to predict or classify new data. This characteristic is highly useful for solving complex problems such as beam profile design. Designing beam profiles using deep learning allows for the derivation of optimal beam profiles faster and more accurately than traditional physical methods. Fig. 1. Framework for Deep Learning-Based Optical Phase Structure Design 1 Fig. 2. Framework for Deep Learning-Based Optical Phase Structure Design 2 Fig. 3. Framework 3 for Deep Learning-Based Optical Phase Structure Design Fig. 4. Example of application of a physical optical illumination system and optical phase structure Fig. 5. Beam depth expansion result according to optical phase structure pattern and optical transfer model. Fig. 6. Result of multi-focus formation according to optical phase structure pattern and optical transmission model Fig. 7. Result according to optical phase structure pattern Fig. 8. Framework for Deep Learning-Based Optical Phase Structure Design Fig. 9. Beam depth expansion result according to optical phase structure pattern and optical transfer model Fig. 10. Binary optical phase structure design framework for extracting fluorescent molecule positions distributed in three-dimensional space. Fig. 11. Binary optical phase structure pattern and accurate function result according to depth Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. Since the present invention is susceptible to various modifications and may take various forms, specific embodiments are illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to the specific disclosed forms, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the present invention. Similar reference numerals have been used for similar components in the description of each drawing. In the attached drawings, the dimensions of the structures are shown enlarged compared to the actual dimensions for the clarity of the present invention. The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as “comprising” or “having” are intended to specify th