CN114839789A - Diffraction focusing method and device based on binarized spatial modulation - Google Patents

Abstract

The invention discloses a diffraction focusing method and device based on binarized spatial modulation. The method comprises: randomly generating several binarized arrays and respectively controlling the binarized spatial modulator to spatially modulate the incident wave wave surface, and the diffracted wave intensity Obtain several diffracted wave intensities corresponding to the several binarized arrays in the collection area; build an evaluation model, and perform the evaluation model according to the several diffracted wave intensities, the selected position to be focused, and several binarized arrays. training; constructing a strategy model, and using a well-trained evaluation model to train the strategy model; using the well-trained strategy model to obtain the optimal array, binarizing the optimal array and setting the modulation of each binarized spatial modulator state to achieve focusing at the position to be focused. The invention has the advantages of simple implementation, strong scalability of the model construction method, short calculation time, and less influence by noise.

Description

Diffraction focusing method and device based on binarized spatial modulation

technical field

The invention relates to the technical field of computational imaging, in particular to a diffraction focusing method and device based on binarization spatial modulation.

Background technique

The wavefront propagating in space is divided into multiple independent units by means of spatial modulation, and amplitude or phase modulation is applied to each unit to control the diffraction propagation of the subsequent wavefront, thereby achieving focusing at a specified position. Compared with continuous phase or continuous amplitude modulation, binary modulation is easier to implement and the modulation speed is faster. For example, it is difficult to perform continuous phase modulation for extreme ultraviolet light or X-rays, but binary modulation can be achieved by controlling the "transmission" and "non-transmission" of different positions in space. A typical example is a zone plate. In the visible light band, the binary spatial modulator based on digital micromirror device can reach the modulation frequency of tens of thousands of hertz, which is much higher than the current liquid crystal-based continuous phase or continuous amplitude spatial light modulator.

When the wavefront has an unknown spatial distribution of amplitude and phase, or undergoes a process of non-free propagation such as scattering during the propagation process, it is necessary to combine the wavefront correction technology to achieve focusing. For example, in the field of optics, the binarized wavefront correction technology based on point-by-point trial or genetic algorithm can use digital micromirror devices to realize the focusing of light passing through the scatterer, but it cannot flexibly change the focusing position; if the focusing position needs to be changed, it needs to be re- Carry out the entire wavefront correction process. The binary wavefront correction technology based on transmission matrix measurement can realize focusing at multiple positions after obtaining the scattering transmission characteristic matrix by multiple measurements, but it has the disadvantages of long calculation time and easy to be affected by noise. The present invention can effectively realize the focusing based on the binarized spatial modulation, and has the advantages of simple implementation, strong scalability of the model construction method, short calculation time, and less influence by noise.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a diffraction focusing method and device based on binarized spatial modulation, which has the advantages of simple implementation, strong scalability of the model building method, short calculation time, and less influence by noise.

In order to achieve the above-mentioned technical purpose, the present invention has adopted the following technical solutions:

In a first aspect, the present invention provides a diffraction focusing method, comprising the following steps:

Randomly generate several binarized arrays and configure the modulation state of the binarized spatial modulator, spatially modulate the incident wave surface and obtain a number of diffracted wave intensities, where each diffracted wave intensity is obtained within the preset diffracted wave intensity collection range ;

Build an evaluation model, and train the evaluation model according to the selected position to be focused, the random binarization array and the corresponding diffracted wave intensity; wherein, the evaluation model is used to reflect the binarization array and the diffracted wave intensity The mapping relationship of the diffracted wave intensity at the position to be focused;

Build a strategy model, and use a well-trained evaluation model to train the strategy model, wherein the strategy model is used to generate an optimal array, and the optimal array can maximize the output of the evaluation model when the optimal array is used as the input of the evaluation model change;

The optimal array is binarized and the modulation state of the binarized spatial modulator is configured to realize focusing at the position to be focused.

In some of the embodiments, the random generation of several binarized arrays and the configuration of the modulation state of the binarized spatial modulator, the spatial modulation of the incident wave front and the acquisition of several diffracted wave intensities, including:

Step 1. Randomly generate a binarized array;

Step 2, changing the modulation state of the binarized spatial modulator according to the binarized array;

Step 3: Obtain the diffracted wave intensity of the incident wave front corresponding to the modulation state in the effective diffraction region after the spatial modulator collected by the sensor;

Step 4: Repeat steps 1 to 3 to obtain a plurality of diffracted wave intensities and a binarized array corresponding to the diffracted wave intensities.

In some of the embodiments, the building an evaluation model, and the evaluation model is trained according to the selected position to be focused, the random binarization array and the corresponding diffracted wave intensity, including:

Selecting the position to be focused, respectively acquiring the diffracted wave intensity at the position to be focused in each diffracted wave intensity, so as to obtain several training pairs, where the training pair is a binary array and the diffracted wave intensity at the position to be focused;

The constructed evaluation model is trained by several trainings, wherein the input of the evaluation model is a binarized array, and the output of the evaluation model is the diffracted wave intensity at the position to be focused in the diffracted wave intensity.

In some of the embodiments, the building a policy model, and using a fully trained evaluation model to train the policy model, including:

Build a policy model and randomly generate an input array;

Take the input array as the input of the strategy model, take the output of the strategy model as the input of the evaluation model, take the maximization of the output of the evaluation model as the training target, and update the strategy model according to the output of the evaluation model , until the policy model outputs an optimal array, which can maximize the output of the evaluation model when the optimal array is used as the input of the evaluation model.

In some of these embodiments, the performing binarization on the optimal array and configuring the modulation state of the binarized spatial modulator to achieve focusing at the position to be focused includes:

The optimal array is obtained, and the optimal array is subjected to binarization processing, and the modulation state of the binarized spatial modulator is configured according to the optimal array after the binarization processing, so as to realize the focusing of the position to be focused.

In some of these embodiments, the evaluation model and the policy model are machine learning models, and the machine learning models include but are not limited to neural network models.

In some of these embodiments, when the selected position to be focused changes, the training of the evaluation model, the training of the policy model, the binarization of the model output optimal array, and the configuration of the binarized spatial modulator are performed again, that is, Can.

In a second aspect, the present invention also provides a diffraction focusing device based on binarization spatial modulation, comprising:

Wave generation source, binarized spatial modulator, wave intensity detection device, and computing and storage device required to perform the above-described binary spatial modulation-based diffraction focusing method, wherein the computing and storage device executes an array generation, model building and training, and reading the intensity of the intensity detection device and controlling the modulation state of the binarized spatial modulation device, the wave generating source is used to generate incident waves.

Compared with the prior art, the diffraction focusing method and device provided by the present invention can optimize the modulation state of the binarized spatial modulator by constructing an evaluation model and a strategy model, and can make continuous The phase or amplitude modulated wave is controlled to complete the effective focusing at the specified position. In addition, the random diffraction intensity signal does not need to specify a specific focusing position when collecting the random diffraction intensity signal. Therefore, when changing different focusing positions, it is not necessary to re-collect the random diffraction intensity signal. , the process is saved, and the implementation method is simple, the model construction method is highly scalable, and still has high efficiency when the number of spatial modulation units is large.

Description of drawings

1 is a flowchart of an embodiment of a diffraction focusing method provided by the present invention;

Fig. 2 is the schematic diagram of the device in the embodiment;

FIG. 3 is a focusing effect diagram of scattered light at a selected focusing position in an embodiment.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

The diffraction focusing method and device based on binarization spatial modulation involved in the present invention can be used in an optical scattering focusing system. In an optical scattering focusing system, complex and unknown amplitude and phase modulation of the incident light occurs due to the scatterer. The outgoing light wave is diffracted into speckle and cannot be effectively focused. Therefore, it is necessary to spatially divide the incident light into independent units, keep the units that can interfere constructively at the selected focus position, and remove the units that cause destructive interference. The choice of such unit division and whether to keep it can be determined by the space The light modulator implementation, combined with the method of the present invention, can optimize the retention state of each cell to achieve focusing at the selected focus position.

According to the content of the present invention, a light scattering focusing method can be provided, please refer to FIG. 1 , including the following steps:

S100: Randomly generate several binarized arrays and configure the modulation state of the binarized spatial modulator, perform spatial modulation on the incident wave wavefront and obtain several diffracted wave intensities, wherein the diffracted wave intensities are all within a preset diffracted wave intensity collection range obtained within.

During specific implementation, the step S100 specifically includes:

Step 1. Randomly generate a binarized array, wherein, in this embodiment, the binarized array is a binarized array randomly composed of -1, 1. It should be understood that the binarized array is not limited to -1, 1, in other embodiments, it can also be composed of, for example, -0.1, 0.1, which is not limited in this embodiment of the present invention;

Step 2, changing the modulation state of the binarized spatial modulator according to the binarized array;

Step 3, acquiring the diffracted wave intensity of the incident wave front corresponding to the modulation state in the effective region behind the binarized spatial modulator collected by the sensor;

Step 4: Repeat steps 1 to 3 to obtain a plurality of diffracted wave intensities and a binarized array corresponding to the diffracted wave intensities.

Specifically, please refer to FIG. 2 , the semiconductor laser generates laser light with a center wavelength of 635 nm, which is incident on the surface of the DMD spatial light modulator 1 . The illumination area of the spatial light modulator 1 is divided into 100×100 independent modulation units, and each modulation unit can make the light irradiated thereon advance according to the optical path or deflect out of the optical path. Then, after being focused by the focusing lens 2, it is incident on the surface 3 of ordinary A4 printing paper and scattered, and the diameter of the focusing area is about 1.5mm. The center of the image sensor 4 is about 6 cm away from the focus area, and the imaging area size is 5.7 mm×4.8 mm. The computer 5 generates 2200 binary arrays S randomly composed of -1, 1, the size of the array is 100 × 100, and controls the modulation state of each modulation unit of the DMD respectively; the image sensor 4 is used to collect the scattered light intensity I, and the scattered light is analyzed. Intensity I is normalized.

S200. Build an evaluation model, and train the evaluation model according to the selected position to be focused, the random binarization array and the corresponding diffracted wave intensity; wherein, the evaluation model is used to reflect the binarization array and diffraction The mapping relationship of the diffracted wave intensity at the position to be focused in the wave intensity.

In some embodiments, step S200 specifically includes:

Selecting the position to be focused, respectively acquiring the diffracted wave intensity at the position to be focused in each diffracted wave intensity, so as to obtain several training pairs, where the training pair is the binarized array and the diffracted wave intensity of the position to be focused;

The constructed evaluation model is trained by several trainings, wherein the input of the evaluation model is a binarized array, and the output of the evaluation model is the diffracted wave intensity at the position to be focused in the diffracted wave intensity.

Wherein, the evaluation model is a machine learning model, and the machine learning model includes but is not limited to a neural network model.

Specifically, the position to be focused is selected in the imaging area of the image sensor, the sum of the diffraction signal intensities Ic of the position to be focused in each I is calculated respectively, and the corresponding S is formed into a training data pair; the position to be focused can be selected arbitrarily on the imaging plane of the image sensor. A location, such as the one shown in Figure 3, contains 5×5 adjacent pixels. The evaluation model uses a fully connected neural network as the evaluation model, including an input layer, two hidden layers and an output layer. The number of neurons in the output layer is 1, and the number of neurons in the other layers is 64. The activation function of the input layer and the hidden layer is ReLU, and the output layer does not use an activation function. Initialize the evaluation neural network, take S as the input, calculate the error between the output of the evaluation neural network and Ic, use the square sum error as the error function, and use the gradient descent method for training. The parameters of the evaluation neural network are optimized using the root mean square propagation method, and the learning rate is set to 2×10 ^-3 .

S300, constructing a strategy model, and using a well-trained evaluation model to train the strategy model, wherein the strategy model is used to generate an optimal array, and when the optimal array is used as the input of the evaluation model, the evaluation model can be maximize output.

In some embodiments, step S300 specifically includes:

Build a policy model and randomly generate an input array;

Take the input array as the input of the strategy model, take the output of the strategy model as the input of the evaluation model, take the maximization of the output of the evaluation model as the training target, and update the strategy model according to the output of the evaluation model , until the policy model outputs an optimal array, which can maximize the output of the evaluation model when the optimal array is used as the input of the fully trained evaluation model.

Wherein, the strategy model is a machine learning model, and the machine learning model includes but is not limited to a neural network model. When the strategy model is a neural network model, the method of gradient ascent can be used to update the model. Of course, in other embodiments, other methods that can realize the update of the strategy model can also be used, which is not done in this embodiment of the present invention. limited.

Specifically, a fully connected neural network is used as the strategy model and initialized, including an input layer, two hidden layers, and an output layer. The activation functions of the input and hidden layers are ReLU, and the output layer does not use an activation function; randomly generate a The array is the input, the output of the policy neural network is used as the input of the evaluation model, and the gradient ascent method is used to maximize the output of the evaluation model as the training target; the policy neural network is trained by the root mean square propagation method, and the learning rate is set to 2×10 ^{− 5} . After the training is completed, the output array of the policy neural network is the optimal array, and the optimal array can maximize the output of the evaluation model.

S400. Binarize the optimal array and configure the modulation state of the binarized spatial modulator, so as to realize focusing at the position to be focused.

Specifically, step S400 specifically includes:

The optimal array is obtained, and the optimal array is subjected to binarization processing, and the modulation state of the binarized spatial modulator is configured according to the optimal array after the binarization processing, so as to realize the focusing of the position to be focused.

Wherein, the binarization method is to change the positive value in the optimal array to 1, and the negative value to -1.

Based on the above-mentioned diffraction focusing method based on binarized spatial modulation, the present invention also provides a diffraction focusing device based on binarized spatial modulation, including a wave generating source, a binarized spatial modulator, and a wave intensity detection device. , and the computing and storage devices required to perform the diffraction focusing method based on binarization spatial modulation as described in the above-mentioned embodiments, wherein the computing and storage devices perform the generation of arrays and the construction and training of models, and read the intensity Detecting the intensity of the device and controlling the modulation state of the binarized spatial modulation device, the wave generating source is used to generate the incident wave.

Since the light scattering focusing method and device based on binarization spatial modulation have been described in detail above, they will not be repeated here.

The present invention has been described in detail above. The above are only examples of the present invention, and should not limit the scope of implementation of the present invention, that is, all equivalent changes and modifications made according to the scope of the application should still be covered by the present invention. within the range.

Claims (8)

1. a diffraction focusing method and device based on binarization spatial modulation, is characterized in that, comprises the steps: Randomly generate several binarized arrays and configure the modulation state of the binarized spatial modulator, spatially modulate the incident wave surface and obtain a number of diffracted wave intensities, where each diffracted wave intensity is obtained within the preset diffracted wave intensity collection range ; Build an evaluation model, and train the evaluation model according to the selected position to be focused, the random binarization array and the corresponding diffracted wave intensity; wherein, the evaluation model is used to reflect the binarization array and the diffracted wave intensity The mapping relationship of the diffracted wave intensity at the position to be focused; Build a strategy model, and use a well-trained evaluation model to train the strategy model, wherein the strategy model is used to generate an optimal array, and the optimal array can maximize the output of the evaluation model when the optimal array is used as the input of the evaluation model change; The optimal array is binarized and the modulation state of the binarized spatial modulator is configured to realize focusing at the position to be focused. 2 . The diffraction focusing method based on binarized spatial modulation according to claim 1 , wherein the random generation of several binarized arrays and the configuration of the modulation state of the binarized spatial modulator are performed on the incident wave front. 3 . Spatially modulate and acquire several diffracted wave intensities, including: Step 1. Randomly generate a binarized array; Step 2, changing the modulation state of the binarized spatial modulator according to the binarized array; Step 3, acquiring the diffracted wave intensity of the incident wave front corresponding to the modulation state in the effective region behind the binarized spatial modulator collected by the sensor; Step 4: Repeat steps 1 to 3 to obtain a plurality of diffracted wave intensities and a binarized array corresponding to the diffracted wave intensities. 3. The diffraction focusing method based on binarization spatial modulation according to claim 2, wherein the construction of the evaluation model is based on the selected position to be focused, the random binarization array and the corresponding diffracted wave intensity , to train the evaluation model, including: Selecting the position to be focused, respectively acquiring the diffracted wave intensity at the position to be focused in each diffracted wave intensity, so as to obtain several training pairs, where the training pair is the binarized array and the diffracted wave intensity of the position to be focused; The constructed evaluation model is trained by several trainings, wherein the input of the evaluation model is a binarized array, and the output of the evaluation model is the diffracted wave intensity at the position to be focused in the diffracted wave intensity. 4. The diffraction focusing method based on binarization spatial modulation according to claim 3, is characterized in that, described constructing strategy model, adopts training complete evaluation model to carry out training to described strategy model, comprising: Build a policy model and randomly generate an input array; Take the input array as the input of the strategy model, take the output of the strategy model as the input of the evaluation model, take the maximization of the output of the evaluation model as the training target, and update the strategy model according to the output of the evaluation model , until the policy model outputs an optimal array, which can maximize the output of the evaluation model when the optimal array is used as the input of the fully trained evaluation model. 5 . The diffraction focusing method based on binarization spatial modulation according to claim 4 , wherein the optimal array is binarized and the modulation state of the binarized spatial modulator is configured to realize Focus at the position to be focused, including: The optimal array is obtained, and the optimal array is subjected to binarization processing, and the modulation state of the binarized spatial modulator is configured according to the optimal array after the binarization processing, so as to realize the focusing of the position to be focused. 6. The diffraction focusing method based on binarization spatial modulation according to any one of claims 1 to 5, wherein the evaluation model and the strategy model are machine learning models, and the machine learning models include but Not limited to neural network models. 7. The diffraction focusing method based on binarization spatial modulation according to any one of claims 1 to 5, wherein when the selected focusing position changes, the training of the evaluation model and the training of the strategy model are performed again , the binarization of the model output optimal array and the configuration of the binarized spatial modulator. 8. A diffraction focusing device based on binarized spatial modulation, characterized in that, comprising: A wave generating source, a binarized spatial modulator, a wave intensity detection device, and a computing and storage device required to perform the binarized spatial modulation-based diffraction focusing method according to any one of claims 1-7, Among them, the computing and storage device performs the generation of the array and the construction and training of the model, and reads the intensity measured by the intensity detection device and controls the modulation state of the binarized spatial modulation device, and the generation source of the wave is used to generate the incident wave. .

Images