Optical imaging system based on optical ai encoding

Abstract

The present application relates to the field of optical imaging technology.The present application relates to an optical AI coding optical imaging system based on optics.The system comprises a mask plate adding unit, a blurred image unit, a phase parameter unit, an AI coding unit and a decoding recovery unit;the mask plate adding unit is used for obtaining a pupil specification, then a non-rotationally symmetric phase mask plate is added at the pupil according to the pupil specification, and the sensitivity of different parameter combinations to defocus aberration is simulated for the phase mask plate;by adding the non-rotationally symmetric phase mask plate at the pupil, combined with the non-focal point phase modulation plate to form a synergistic effect, the target can be imaged in a long dynamic range without zooming and focusing, meanwhile, the phase modulation depth and mask size of the phase mask plate are completely matched with the pupil specification, the non-focal point phase modulation plate adopts a non-rotationally symmetric phase distribution, and the two cooperate to strengthen the insensitivity of the system to the defocus of the image plane, so that the blurred modes of imaging at different defocus positions in a large focal depth range are consistent.

Description

An optical imaging system based on optical AI coding Technical Field

[0001] This invention relates to the field of optical imaging technology, and more specifically, to an optical imaging system based on optical AI coding. Background Technology

[0002] In military and special security fields such as high-intensity nighttime confrontations on land, air, and sea, optoelectronic imaging systems are core technology equipment for target detection, tracking, and identification. Their core function is to capture clear image information of targets in complex lighting and dynamic environments, providing data support for subsequent key tasks such as target locking and fire guidance.

[0003] Existing traditional optoelectronic imaging technologies are mostly used in relatively simple scenarios such as low-dynamic targets and short-range fixed monitoring during the day. However, they have significant shortcomings in high-complexity scenarios such as high-intensity combat at night on land, air, and sea. In such scenarios, targets are often highly maneuverable swarm targets, moving at high speeds and with complex trajectories. Traditional optoelectronic systems need to perform focusing and zooming operations through mechanical structures or optical elements during the imaging and tracking process to adapt to targets at different distances. This process introduces a tracking delay of 1 to 2 seconds, causing the system to fail to capture the movement trajectory of highly maneuverable targets in time, missing the best aiming and tracking opportunity, and seriously affecting the effective monitoring and rapid response to highly maneuverable swarm targets. To reduce this situation, an optical AI-coded optical imaging system is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide an optical imaging system based on optical AI coding to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, an optical imaging system based on optical AI coding is provided, including a mask addition unit, a blurred image unit, a phase parameter unit, an AI coding unit, and a decoding and restoration unit;

[0006] The mask addition unit is used to obtain the pupil specifications, and then add a non-rotationally symmetric phase mask at the pupil according to the pupil specifications. At the same time, it simulates the sensitivity of different parameter combinations to defocus aberration for the phase mask.

[0007] The blurred image unit is used to select the optimal parameter combination for the sensitivity of defocus aberration by combining PSF and OTF with different parameters, and then set the non-focus phase modulation plate combination based on the optimal parameter combination, and form a blurred image by cooperating with the phase mask and the non-focus phase modulation plate.

[0008] The phase parameter unit is used to set adjacent time ranges, form a new blurred image based on the adjacent time ranges and the blurred image unit, extract historical blurred images, combine historical blurred images with the new blurred image for interference data analysis, and adjust the phase parameters based on the interference data.

[0009] The AI ​​encoding unit is used to establish an optical AI that modulates the light phase, while collecting prior information of the incident light field, inputting the prior information of the incident light field into the optical AI to design constraints, and then using the optical AI to encode the blurred image.

[0010] The decoding and restoration unit is used to set the decoding algorithm based on optical AI, phase mask, and non-focus phase modulation board, and to use the decoding algorithm to uniformly decode and restore the encoded blurred image.

[0011] As a further improvement to this technical solution, the mask addition unit is connected to an optical system to obtain the pupil specifications. At the same time, a cubic odd-symmetric phase mask is selected. Based on the requirement of extended depth of field, the modulation principle of the non-rotational symmetric phase factor on the incident light wave is derived. Furthermore, the phase modulation depth and mask size of the phase mask must be completely matched with the pupil specifications.

[0012] As a further improvement to this technical solution, the mask template adding unit includes an installation module and a simulation analysis module;

[0013] The mounting module is used to mount a phase mask at the pupil position;

[0014] The simulation analysis module is used to build a simulation model using optical design software combined with an optical system and a phase mask, and then simulates the sensitivity of different parameter combinations to defocus aberration through the simulation model.

[0015] As a further improvement to this technical solution, the blurred image unit includes a parameter filtering module and a blurred image forming module;

[0016] The parameter filtering module is used to compare and filter different parameter combinations in the simulation model with PSF and OTF to obtain the optimal parameter combination with the smallest change in PSF shape and OTF value within a large focal depth range, and to determine that it is not sensitive to defocus aberration.

[0017] If the size and shape of the light spot change similarly at different defocus distances, it means that the shape change of the PSF is smaller and the imaging blur pattern is consistent.

[0018] If the OTF value fluctuates little within a large depth of focus range, it means that the degree of image detail retention is consistent at different defocus positions;

[0019] The blurred image forming module is used to set a non-focus phase adjustment plate based on a preferred parameter combination. By cooperating with the phase mask, the non-focus phase adjustment plate enhances the optical system's insensitivity to image plane defocusing, thereby forming a blurred image whose image quality is independent of the image plane position.

[0020] As a further improvement to this technical solution, when the blurred image unit is configured with a non-focus phase adjustment plate, the non-focus phase adjustment plate is installed at the exit pupil position of the optical system and has a non-rotationally symmetric phase distribution. By cooperating with the phase mask, the depth of focus is extended and the blurring is consistent at different defocus positions.

[0021] As a further improvement to this technical solution, the phase parameter unit includes an adjacent image extraction module and an interference data analysis module;

[0022] The adjacent image extraction module is used to set adjacent time ranges;

[0023] The interference data analysis module is used to monitor the blurred image forming module. When the blurred image forming module forms a new blurred image, it extracts historical blurred images based on the formation time of the new blurred image and the adjacent time range. Then, it combines the extracted historical blurred images with the new blurred image to perform interference data analysis and obtain interference data of the optical system scene.

[0024] As a further improvement to this technical solution, the phase parameter unit also includes a scene simulation module and a phase parameter adjustment module;

[0025] The scene simulation module is used to simulate the interference data corresponding to the interference scene using optical design software, and at the same time, it combines the parameters obtained by the simulation analysis module to perform phase parameter matching to form a mapping database of phase parameters corresponding to the interference data.

[0026] The phase parameter adjustment module is used to combine the interference data obtained by the interference data analysis module with the mapping database for adjustment and extraction, extract the phase parameters corresponding to the interference data in real time, and adjust the optical system according to the extracted phase parameters.

[0027] As a further improvement to this technical solution, the AI ​​encoding unit includes an AI creation module and an image encoding module;

[0028] The AI ​​establishment module is used to establish an optical AI that modulates the light phase. The optical AI includes a computational optics module and a deep learning module. The computational optics module calculates the phase change using optical formulas, and the deep learning module learns the optimal encoding rules.

[0029] The image encoding module is used to collect prior information of the incident light field and input the prior information of the incident light field into the optical AI for constraint design. At the same time, the optical AI is trained to generate a phase encoding scheme. Then, the optical AI is used to encode the blurred image to form an encoded blurred image.

[0030] As a further improvement to this technical solution, the decoding and restoration unit obtains the blur pattern of the blurred image based on phase modulation and optical AI encoding, then sets a unified decoding filter algorithm based on the blur pattern, and then integrates the decoding filter algorithm into the infrared computational imaging device to support instantaneous imaging and rapid recognition.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] 1. In this optical AI-coded optical imaging system, a non-rotationally symmetric phase mask is added at the pupil, which works synergistically with a focusless phase modulation plate. This enables the imaging of the target without zooming or focusing within a long dynamic range. At the same time, the phase modulation depth and mask size of the phase mask are perfectly matched with the pupil specifications, and the focusless phase modulation plate adopts a non-rotationally symmetric phase distribution. The combination of the two can enhance the system's insensitivity to image plane defocusing, making the blur pattern of imaging at different defocus positions within a large depth of focus range consistent.

[0033] 2. In this optical AI-coded optical imaging system, the phase parameter unit presets a reasonable adjacent time range. When the blurred image unit forms a new blurred image, the historical blurred image is extracted and compared with the new image. The differences in features such as light spot shape and intensity distribution are analyzed, the interference data is quantified, and then the phase parameters of the optical system are adjusted in real time in combination with the interference scene simulation database built by the scene simulation module to counteract the influence of interference on imaging.

[0034] 3. In this optical AI-based optical imaging system, the AI ​​encoding unit collects prior information of the incident light field and designs constraints for the optical AI. Its computational optics module calculates phase changes based on physical optics formulas, and the deep learning module learns the optimal encoding rules. After encoding the blurred image, it ensures that the imaging quality of spatial objects within the same depth of field is consistent at each defocus position, achieving full-frame, focus-free, wide temperature adaptability, and aberration-free optical imaging. This multi-dimensional optimization design allows the system to maintain stable imaging effects in complex and ever-changing environments, adapting to more harsh application scenarios. Attached Figure Description

[0035] Figure 1 is a schematic diagram of the overall structure of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Designed for high-intensity nighttime combat scenarios involving land, air, and sea, this system utilizes computational optics technology to achieve instantaneous imaging and rapid multi-target identification of swarm-like targets in nighttime conditions. It eliminates the need for zooming and focusing within a long dynamic range, saving 1-2 seconds of tracking delay and enabling rapid targeting of highly maneuverable swarm targets.

[0038] Solve the problem of time delay caused by focusing and zooming during the imaging and tracking process in traditional optoelectronic systems;

[0039] Please refer to Figure 1. The purpose of this embodiment is to provide an optical imaging system based on optical AI coding, including a mask addition unit, a blurred image unit, a phase parameter unit, an AI coding unit, and a decoding and restoration unit.

[0040] The mask addition unit is used to obtain the pupil specifications, and then a non-rotationally symmetric phase mask is added at the pupil according to the pupil specifications. At the same time, the sensitivity of different parameter combinations to defocus aberration is simulated for the phase mask.

[0041] The mask addition unit connects to the optical system and obtains the pupil specifications (including key information such as size and shape) through the optical system. At the same time, a cubic odd-symmetric phase mask is selected. Based on the requirement of expanding depth of field, the modulation principle of the non-rotational symmetric phase factor on the incident light wave is derived. Furthermore, the phase modulation depth and mask size of the phase mask must be completely matched with the pupil specifications.

[0042] The mask template addition unit includes an installation module and a simulation analysis module;

[0043] The mounting module is used to mount the phase mask at the pupil position;

[0044] Define the basic parameters of the optical system, such as the focal length, aperture, and field of view. These parameters are important for building the simulation model. At the same time, determine the characteristics of the phase mask to be installed, including the phase modulation law and size. Then, install the phase mask at the pupil.

[0045] The simulation analysis module is used to build a simulation model using optical design software, combined with an optical system and a phase mask. Then, the simulation model is used to simulate the sensitivity of different parameter combinations to defocus aberration.

[0046] Open the selected optical design software, import the model file of the optical system into the software, set it as the basic optical system for simulation, and then add the corresponding phase mask model at the pupil position of the software according to the size and phase distribution characteristics of the phase mask, so that the software can simulate the optical system containing the phase mask. At the same time, determine the different parameter combinations (phase modulation depth and mask size) to be simulated, set a series of different values ​​for each parameter, and combine them to form multiple different parameter combinations.

[0047] Then, for each set of parameters, a simulation program is run in the optical design software. The software will simulate the propagation and imaging process of light after passing through an optical system with a phase mask, under the condition of defocus aberration, based on the principle of optical imaging. After the simulation is completed, the changes of these data under different parameter combinations are compared to analyze the sensitivity of different parameter combinations to defocus aberration and establish a basic parameter database.

[0048] The blurred image unit is used to select the optimal parameter combination for the sensitivity of defocus aberration by combining PSF and OTF with different parameters. Then, based on the optimal parameter combination, the non-focus phase modulation plate combination is set, and a blurred image is formed by the cooperation of the phase mask and the non-focus phase modulation plate.

[0049] The blurred image unit includes a parameter filtering module and a blurred image forming module;

[0050] The parameter filtering module is used to compare and filter different parameter combinations in the simulation model with PSF and OTF to obtain the optimal parameter combination with the smallest changes in PSF shape and OTF value within a large depth of focus range, and to determine the parameter combination that is not sensitive to defocus aberration.

[0051] If the size and shape of the light spot change similarly at different defocus distances, it means that the shape change of the PSF is smaller and the imaging blur pattern is consistent.

[0052] If the OTF value fluctuates little within a large depth of focus range, it indicates that the image detail retention is consistent across different out-of-focus positions. The specific steps are as follows:

[0053] Organize all parameter combinations to be screened in the simulation model, clarify the optical system configuration corresponding to each set of parameters, and extract the point spread function (PSF) and optical transfer function (OTF) data of each set of parameters at different defocus distances from the simulation results to ensure that the data covers the preset large focal depth range. Then, for each set of parameters, observe the spot size and shape of the PSF at different defocus distances, evaluate the degree of change of the spot at different defocus positions, and screen out parameter combinations with similar spot size and shape changes and high consistency in morphology.

[0054] Next, for each parameter combination, the OTF value variation over a large depth of field was analyzed. The peak value, mean value, and standard deviation of OTF at different defocus positions were statistically analyzed. Parameter combinations with small OTF value fluctuations and consistent detail retention at each defocus position were selected. Then, combined with PSF morphology analysis and OTF value stability analysis results, a comprehensive evaluation of the parameter combinations was conducted (prioritizing parameter combinations that simultaneously satisfy small PSF morphology changes and small OTF value fluctuations over a large depth of field, which were determined as the preferred parameter combinations insensitive to defocus aberrations). The formula is as follows:

[0055]

[0056] Where S is the PSF morphological similarity index (value ranges from 0 to 1, with 1 representing identical shapes), N is the number of out-of-focus positions, and PSF i Let PSF be the point spread function at the i-th out-of-focus position. p The average PSF value for all out-of-focus positions, max(PSF) i PSF p () represents the maximum value;

[0057]

[0058] Where V is the OTF numerical variation coefficient, σ OTF μ represents the standard deviation of OTF values ​​over a wide depth of focus range. OTF This represents the average OTF value over a wide focal depth range.

[0059] The blurred image forming module is used to set the non-focus phase adjustment plate based on the optimal parameter combination. By cooperating with the phase mask, the non-focus phase adjustment plate enhances the optical system's insensitivity to image plane defocusing, thereby forming a blurred image whose image quality is independent of the image plane position.

[0060] The non-rotationally symmetric phase distribution of the focal-free phase modulation board is designed to complement the phase modulation of the phase mask. This complementary design cancels out the aberration effects caused by different defocus positions and enhances the system's insensitivity to image plane defocus. The designed focal-free phase modulation board is then installed at the exit pupil of the optical system, and the orientation of the modulation board is calibrated using optical adjustment tools to align it with the optical center of the phase mask, ensuring the phase modulation effect when the two work together.

[0061] Then, when the optical system is started to acquire images, the phase mask and the non-focal phase modulation plate work together to form a blurred image that is independent of image quality and image plane position within a large depth of focus.

[0062] When the blurred image unit is set with a focusless phase adjustment plate combination, the focusless phase adjustment plate is installed at the exit pupil position of the optical system. It also has a non-rotationally symmetric phase distribution. By cooperating with the phase mask, the depth of focus is extended and the blur is consistent at different defocus positions.

[0063] The phase parameter unit is used to set the adjacent time range. Based on the adjacent time range, a new blurred image is formed by combining it with the blurred image unit to extract the historical blurred image. The historical blurred image is then combined with the new blurred image to analyze the interference data, and the phase parameter is adjusted based on the interference data.

[0064] The phase parameter unit includes a neighboring image extraction module and an interference data analysis module;

[0065] The adjacent image extraction module is used to set the adjacent time range;

[0066] Based on the imaging frame rate of the optical system and the dynamic changes of the scene, a reasonable adjacent time range (such as 50ms-500ms) is preset to avoid interference and data distortion caused by excessively long time intervals;

[0067] The interference data analysis module is used to monitor the blurred image formation module. When the blurred image formation module forms a new blurred image, it extracts historical blurred images based on the formation time of the new blurred image and the adjacent time range. Then, it combines the extracted historical blurred images with the new blurred image to perform interference data analysis and obtain interference data of the optical system scene.

[0068] By comparing and analyzing the differences between historical blurred images and newly blurred images, the focus is on identifying changes in features such as light spot morphology, intensity distribution, and edge sharpness. Through feature difference extraction, interference data caused by factors such as scene jitter, temperature fluctuations, and light source changes are identified, forming an interference information dataset. The formula is as follows:

[0069]

[0070] Where D is the quantization value of the interference data (the larger the value, the stronger the interference), m and o are the pixel size of the image, and I x (j, k) represents the pixel grayscale value in the j-th row and k-th column of the new blurred image. l (j, k) represents the pixel grayscale value in the j-th row and k-th column of the historical blurred image;

[0071] The phase parameter unit also includes a scene simulation module and a phase parameter adjustment module;

[0072] The scene simulation module is used to simulate the interference data corresponding to the interference scene using optical design software. At the same time, it combines the parameters obtained by the simulation analysis module to perform phase parameter matching and form a mapping database of phase parameters corresponding to the interference data.

[0073] Using optical design software, a simulation scenario containing various typical interference factors is built. The simulation program is run to generate optical system imaging data under different interference levels. The corresponding interference type, intensity and corresponding imaging deviation data are recorded to form an interference dataset. Then, the optical system parameters obtained by the simulation analysis module are called and correlated with the interference data obtained by the simulation. Through multiple sets of comparative experiments, the optimal matching phase parameters corresponding to different interference data are determined, and a one-to-one correspondence between the two data and phase parameters is established.

[0074] The phase parameter adjustment module is used to combine the interference data obtained by the interference data analysis module with the mapping database for adjustment and extraction, extract the phase parameters corresponding to the interference data in real time, and adjust the optical system according to the extracted phase parameters.

[0075] The extracted phase parameters are converted into executable control commands and sent to the phase adjustment components of the optical system (such as piezoelectric controllers, variable phase modulators, etc.) to adjust the phase state of the system in real time and counteract the influence of current interference on imaging.

[0076] The AI ​​encoding unit is used to establish an optical AI that modulates the phase of light, while collecting prior information of the incident light field and inputting the prior information of the incident light field into the optical AI to design constraints. Then, the optical AI is used to encode the blurred image.

[0077] The AI ​​encoding unit includes an AI creation module and an image encoding module;

[0078] The AI ​​building module is used to build an optical AI that modulates the phase of light. The optical AI includes a computational optics module and a deep learning module. The computational optics module calculates the phase change using optical formulas, and the deep learning module learns the optimal encoding rules.

[0079] The image encoding module is used to collect prior information of the incident light field and input the prior information of the incident light field into the optical AI for constraint design. At the same time, the optical AI is trained to generate a phase encoding scheme. Then, the optical AI is used to encode the blurred image to form an encoded blurred image (ensuring that the imaging quality of spatial objects within the same depth of field is consistent at each defocus position, laying the foundation for subsequent unified decoding).

[0080] By encoding, a full-frame, focus-free, wide-temperature-adaptive, and aberration-free optical imaging system is achieved. For spatial objects with the same depth of field, the system produces images of consistent quality at all out-of-focus positions. This series of images can be restored using the same filter to obtain a clearer image within the focus-free range. The formula is as follows:

[0081] Φ AI (x,y)=Φ calc (x,y)+Φ learn (x,y);

[0082] Where, Φ AI (x, y) represents the total phase modulation of the optical AI output, Φ calc (x, y) represents the phase modulation output of the optical module, calculated based on physical optics formulas, Φ learn (x, y) represents the optimized phase modulation output by the deep learning module, which is obtained through training.

[0083] L=α1·||Φ pred -Φ ture ||2+α2·R(Φ pred );

[0084] Where L is the loss function value, α1 and α2 are weight coefficients, and Φ pred Φ is the phase modulation amount predicted by the deep learning module. ture R(Φ) represents the phase modulation amount of the actual tag. pred ) represents the regularization term, and ||·|| represents the L2 norm;

[0085] I enc (x,y)=I blur (x,y)·exp[v·Φ AI [(x,y)];

[0086] Among them, I enc (x, y) is the encoded blurred image, I blur (x, y) represents the amplitude distribution of the original blurred image, v is the imaginary unit, and exp is the exponential function (used to convert the phase modulation into a complex form of the modulation factor).

[0087] The decoding and restoration unit is used to set the decoding algorithm based on optical AI, phase mask, and focusless phase modulation board, and uses the decoding algorithm to uniformly decode and restore the encoded blurred image.

[0088] The decoding and restoration unit obtains the blur pattern of the blurred image based on phase modulation and optical AI encoding, then sets a unified decoding filter algorithm based on the blur pattern, and then integrates the decoding filter algorithm into the infrared computational imaging device to finally output a clear image, supporting the application requirements of instant imaging and rapid recognition.

[0089] Based on the extracted fundamental parameters, a unified decoding model is designed. This model integrates the encoding logic of optical AI, the inverse modulation function of the phase mask, and the compensation algorithm of the non-focus phase modulation plate to form a complete decoding link. This ensures that the phase changes introduced during the encoding process can be offset. Then, based on the characteristics of the blurred image after encoding, a targeted decoding filter is designed. The filter needs to adapt to the unified blur pattern at different defocus positions. The image blur caused by phase modulation is eliminated through inverse operation to restore the original scene information. The formula is as follows:

[0090] I rec (x,y)=F -1 {F{I enc (x,y)}·H dec (f x ,f y )};

[0091] Among them, I rec (x, y) represents the decoded and restored clear image, and F{·} is the Fourier transform, which converts the spatial domain image to the frequency domain. -1 {·} represents the inverse Fourier transform, converting the frequency domain result back to the spatial domain. H dec (f x f y ) is the frequency response function of the decoding filter, which integrates the inverse characteristics of optical AI, phase mask and focusless phase modulation board.

[0092] H dec (f x ,f y )=H phy (f x ,f y )·H scene (f x ,f y )·H noise (f x ,f y );

[0093] Among them, H phy (f x f y ) is the physical reverse compensation term, which cancels out the physical modulation in the coding stage, H scene (f x f yThis is a scene adaptation optimization item, adapting to different out-of-focus and light field scenes to improve detail restoration. noise (f x f y () is a noise robustness enhancement term that suppresses high-frequency noise amplified during the decoding process.

[0094] By deeply integrating optical design and AI technology, the entire chain of optimization from optical modulation to intelligent processing has been achieved, which not only ensures the improvement of the physical performance of the optical system, but also solves the complex imaging problems that are difficult to handle by traditional optics through AI algorithms.

[0095] The encoding part is implemented by the receiving optical lens, and the back end of the image sensor is connected to a circuit board with an FPGA. The decoding part is implemented by the decoding algorithm on the circuit board. It adopts the characteristics of butterfly interlacing and pipeline arrangement, which is highly complex. The algorithm architecture has high computing power and low bandwidth requirements, which meets the low latency, high real-time performance and low power consumption requirements of computational optics technology in the era of high image resolution. It greatly reduces the technical threshold for the application of core computational optics algorithms in military applications.

[0096] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An optical imaging system based on optical AI coding, characterized in that: The system includes a mask addition unit, a blurred image unit, a phase parameter unit, an AI encoding unit, and a decoding and restoration unit. The mask addition unit acquires the pupil specifications and then adds a non-rotationally symmetric phase mask at the pupil location based on these specifications. Simultaneously, it simulates the sensitivity of different parameter combinations to defocus aberrations using the phase mask. The blurred image unit uses PSF and OTF to select optimal parameter combinations based on the sensitivity of different parameters to defocus aberrations. Then, based on the optimal parameter combinations, it sets a focusless phase modulation plate combination and forms a blurred image through the cooperation of the phase mask and the focusless phase modulation plate. The phase parameter unit sets the adjacent time range and, based on adjacent time... The system extracts historical blurred images by combining the formation time of the new blurred image with the blurred image unit within the range, and analyzes the interference data by combining the historical blurred image with the new blurred image, and adjusts the phase parameters according to the interference data; the AI ​​encoding unit is used to establish an optical AI that modulates the light phase, and at the same time collects the prior information of the incident light field, and inputs the prior information of the incident light field into the optical AI to design the constraint conditions, and then uses the optical AI to encode the blurred image formed by the blurred image unit; the decoding and restoration unit is used to set the decoding algorithm according to the optical AI, the phase mask, and the non-focal phase modulation plate, and uses the decoding algorithm to uniformly decode and restore the encoded blurred image.

2. The optical imaging system based on optical AI coding according to claim 1, characterized in that: The mask addition unit is connected to an optical system to obtain the pupil specifications. At the same time, a cubic odd-symmetric phase mask is selected. Based on the requirement of extended depth of field, the modulation principle of the non-rotational symmetric phase factor on the incident light wave is derived. Furthermore, the phase modulation depth and mask size of the phase mask must be completely matched with the pupil specifications.

3. The optical imaging system based on optical AI coding according to claim 1, characterized in that: The mask addition unit includes an installation module and a simulation analysis module; the installation module is used to install a phase mask at the pupil position; the simulation analysis module is used to build a simulation model using optical design software in combination with the optical system and the phase mask, and then simulate the sensitivity of different parameter combinations to defocus aberration through the simulation model.

4. The optical imaging system based on optical AI coding according to claim 1, characterized in that: The blurred image unit includes a parameter filtering module and a blurred image forming module. The parameter filtering module is used to compare and filter different parameter combinations in the simulation model with PSF and OTF to obtain the optimal parameter combination with the smallest changes in PSF shape and OTF value within a large depth of focus range, and to determine that it is not sensitive to defocus aberration. If the size and shape of the light spot change similarly at different defocus distances, it means that the PSF shape change is smaller and the imaging blur mode is consistent. Small fluctuations in OTF values ​​within a large depth of focus indicate consistent retention of imaging details at different defocus positions. The blurred image forming module is used to set a non-focus phase adjustment plate based on a preferred parameter combination. By cooperating with the phase mask, the non-focus phase adjustment plate enhances the optical system's insensitivity to image plane defocus, thereby forming a blurred image whose image quality is independent of image plane position.

5. The optical imaging system based on optical AI coding according to claim 1, characterized in that: When the blurred image unit is configured with a non-focus phase modulation plate, the non-focus phase modulation plate is installed at the exit pupil position of the optical system. It also has a non-rotationally symmetric phase distribution. By cooperating with the phase mask, the depth of focus is extended and the blur pattern is consistent at different defocus positions.

6. The optical imaging system based on optical AI coding according to claim 1, characterized in that: The phase parameter unit includes an adjacent image extraction module and an interference data analysis module. The adjacent image extraction module is used to set adjacent time ranges. The interference data analysis module is used to monitor the blurred image forming module. When the blurred image forming module forms a new blurred image, it extracts historical blurred images based on the formation time of the new blurred image and the adjacent time range. Then, it combines the extracted historical blurred images with the new blurred image to perform interference data analysis and obtain interference data of the optical system scene.

7. The optical imaging system based on optical AI coding according to claim 1, characterized in that: The phase parameter unit further includes a scene simulation module and a phase parameter adjustment module. The scene simulation module is used to simulate interference data corresponding to the interference scene using optical design software, and at the same time, it performs phase parameter matching by combining the parameters obtained by the simulation analysis module to form a mapping database of phase parameters corresponding to the interference data. The phase parameter adjustment module is used to adjust and extract the interference data obtained by the interference data analysis module in combination with the mapping database, extract the phase parameters corresponding to the interference data in real time, and adjust the optical system according to the extracted phase parameters.

8. The optical AI-coded optical imaging system according to claim 1, characterized in that: The AI ​​encoding unit includes an AI establishment module and an image encoding module. The AI ​​establishment module is used to establish an optical AI that modulates the light phase. The optical AI includes a computational optics module and a deep learning module. The computational optics module calculates the phase change using optical formulas, and the deep learning module learns the optimal encoding rules. The image encoding module is used to collect prior information of the incident light field and input the prior information of the incident light field into the optical AI for constraint design. At the same time, it trains the optical AI to generate a phase encoding scheme. Then, the optical AI is used to encode the blurred image formed by the blurred image unit to form an encoded blurred image.

9. The optical AI-coded optical imaging system according to claim 1, characterized in that: The decoding and restoration unit obtains the blur pattern of the blurred image based on phase modulation and optical AI encoding, then sets a unified decoding filter algorithm based on the blur pattern, and then integrates the decoding filter algorithm into the infrared computational imaging device to support instantaneous imaging and rapid recognition.

Images