Atmospheric turbulence intensity detection method, system and device based on SLM indoor turbulence simulation test platform and ConvNeXt convolutional neural network, and storage medium
Through the SLM-based indoor turbulence simulation platform and ConvNeXt convolutional neural network, the problems of large errors and limited detection range in atmospheric turbulence intensity detection were solved, and simple and accurate turbulence intensity prediction was achieved.
Patent Information
- Application Number
- CN202510775413.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-11
- Publication Date
- 2025-09-23
AI Technical Summary
Existing technologies have problems with large errors and limited detection range in detecting atmospheric turbulence intensity, especially when the turbulence intensity is weak, making it difficult to measure accurately. Traditional methods cannot effectively provide turbulence intensity within a certain distance range.
An indoor turbulence simulation test platform based on SLM and ConvNeXt convolutional neural network are used to generate a turbulence phase screen and obtain instantaneous spot images by rotation and translation. The ConvNeXt network is trained with sparse classification cross entropy and L2 regularization loss function to improve the convolution kernel size and feature fusion, thereby achieving accurate prediction of atmospheric turbulence intensity.
It enables easy acquisition of spot data sets under different turbulence intensities, improves the precision and accuracy of atmospheric turbulence intensity prediction, simplifies the detection process, and has a certain degree of accuracy.
Smart Images
Figure CN120688352A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of atmospheric turbulence intensity detection using deep learning, and in particular to an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method, system, equipment and storage medium using a ConvNeXt convolutional neural network. Background Art
[0002] At present, the detection of atmospheric turbulence intensity is mainly based on the inversion of atmospheric refractive index structure constants by the variance of the centroid drift of the light spot. The basic principle is: under the assumption of Kolmogorov uniform and isotropic turbulence, using Rytov perturbation theory, according to the second-order moment function, we can get in, is the drift variance of the spot centroid, C n is the atmospheric refractive index structure constant, an important parameter used to describe the atmospheric turbulence intensity, D is the laser emission aperture, and L is the transmission distance. Find C n However, this method requires a large amount of spot data to obtain Furthermore, the actual laser itself has a certain degree of deviation, which will produce a certain error in the calculated center of mass variance, especially when the turbulence intensity is weak, resulting in a certain error between the obtained result and the true value. In addition, according to Kolmogorov's theory of local uniform isotropy, temperature fluctuations or refractive index fluctuations at small scales satisfy the "2 / 3" law of the structure function. Therefore, when turbulence is fully developed, the atmospheric temperature structure constant can be expressed as follows:
[0003]
[0004] In formula (1), represent the inner scale and outer scale of atmospheric turbulence respectively, T is the air temperature, is the position vector, and is the vector size.
[0005]
[0006] Where (2) is atmospheric pressure P (unit: hPa) and T is air temperature (unit: K). Based on the above theory, a temperature pulsometer can be used to directly measure the atmospheric refractive index structure constant. Measuring turbulence intensity with a temperature pulsometer has certain advantages as a traditional turbulence detection method, but its disadvantage is that its detection range is too limited and it cannot effectively provide turbulence intensity within a certain distance range, making it difficult to apply in many current scenarios.
[0007] To this end, how to provide an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method, system, equipment and storage medium based on ConvNeXt convolutional neural network that can achieve more accurate detection of atmospheric turbulence intensity and has a relatively simple process is a problem that technical personnel in this field urgently need to solve. Summary of the Invention
[0008] In view of this, the present invention proposes an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method, system, equipment and storage medium based on ConvNeXt convolutional neural network.
[0009] In order to achieve the above object, the present invention adopts the following technical solutions:
[0010] An indoor turbulence simulation test platform based on SLM and a ConvNeXt convolutional neural network for detecting atmospheric turbulence intensity, comprising:
[0011] Step 1: Use an indoor turbulence simulation test platform based on an SLM to obtain instantaneous spot images of random turbulence effects at different turbulence intensities, and preprocess them to obtain training sets, validation sets, and test sets. The SLM in the indoor turbulence simulation test platform is loaded with a phase screen generated based on the modified Von Karman atmospheric turbulence spectrum model, and the instantaneous spot images are obtained by rotating and translating the static phase screen.
[0012] Step 2: Input the training set into the ConvNeXt convolutional neural network, train the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network, and obtain the optimal weight parameters of the atmospheric turbulence intensity prediction model based on the validation set;
[0013] Step 3: Input the test set into the atmospheric turbulence intensity prediction model under the optimal weight parameters to obtain the atmospheric turbulence intensity prediction results.
[0014] Optionally, in step 1, an indoor turbulence simulation test platform based on SLM includes: a helium-neon laser, a beam expander, a first polarizer, a transmissive spatial light modulator (SLM), a second polarizer, an attenuator, and a CCD camera;
[0015] After the HeNe laser is expanded by a beam expander, the Gaussian light amplitude state is adjusted by the first polarizer, and phase modulation is performed by a transmissive spatial light modulator (SLM) to simulate turbulence. The turbulence amplitude state is adjusted again by the second polarizer, and after the light intensity is weakened by the attenuation plate, the light spot is recorded by a CCD camera.
[0016] Optionally, in step 1, preprocessing includes: CCD-based light environment correction and power correction and noise reduction processing, resolution compression processing and normalization preprocessing.
[0017] Optionally, in step 1, the modified Von karman atmospheric turbulence spectrum model is as follows:
[0018]
[0019] Among them, Φ n is the power spectral density; is the atmospheric refractive index structure constant; κ is the wave number; κ m is the cutoff wavenumber related to the inner scale; κ0 is the cutoff wavenumber related to the outer scale.
[0020] Optionally, in step 2, the loss function of the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network is selected as a sparse classification cross entropy plus L2 regularization loss function.
[0021] The present invention also provides an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection system based on ConvNeXt convolutional neural network, comprising:
[0022] The module for acquiring training, validation, and test sets for atmospheric turbulence intensity prediction uses an indoor turbulence simulation test platform based on an SLM to acquire instantaneous spot images of random turbulence effects at different turbulence intensities, and performs preprocessing to obtain training, validation, and test sets. The SLM in the indoor turbulence simulation test platform is loaded with a phase screen generated based on a modified Vonkarman atmospheric turbulence spectrum model, and instantaneous spot images are acquired by rotating and translating a static phase screen.
[0023] Atmospheric turbulence intensity prediction model training module: used to input the training set into the ConvNeXt convolutional neural network, train the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network, and obtain the optimal weight parameters of the atmospheric turbulence intensity prediction model based on the validation set;
[0024] Atmospheric turbulence intensity prediction module: used to input the test set into the atmospheric turbulence intensity prediction model under the optimal weight parameters to obtain the atmospheric turbulence intensity prediction results.
[0025] The present invention further provides an electronic device, comprising:
[0026] memory for storing computer programs;
[0027] The processor is used to implement the steps of an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method based on ConvNeXt convolutional neural network when executing a computer program.
[0028] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform based on SLM and a ConvNeXt convolutional neural network are implemented.
[0029] Through the above technical solutions, it can be seen that compared with the existing technology, the present invention proposes an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method, system, equipment and storage medium using a ConvNeXt convolutional neural network. By constructing an indoor turbulence simulation test platform based on SLM, a transmissive spatial light modulator (SLM) is loaded using a phase screen generated based on a modified Von karman atmospheric turbulence spectrum model and rotated for phase modulation, and by rotating and translating the static phase screen, instantaneous spot images of random turbulence effects under different turbulence intensities are obtained. This makes it possible to easily obtain the spot data set required for the atmospheric turbulence intensity prediction model according to different application requirements, effectively improving the prediction accuracy of the subsequently trained model. Furthermore, the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network was trained and improved. The stem layer was modified to increase the convolution kernel size from 7 to 16, and the step size remained unchanged. The results showed that the iterative effect of the convolution kernel of 16 was more stable. Secondly, different convolution kernels (1*1, 3*3, 5*5 sizes) were added to the black layer to extract features respectively and then perform feature fusion (enriching feature information). Sparse classification cross entropy and L2 regularization were used as the loss function of the atmospheric turbulence intensity prediction model. The sparse classification cross entropy was used to measure the accuracy of the model prediction and suppress overfitting. L2 regularization prevented the influence of hidden units caused by excessive weights. Finally, a simple, convenient and accurate atmospheric turbulence intensity prediction was achieved. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0031] Figure 1 Schematic diagram of the method of the present invention.
[0032] Figure 2 Schematic diagram of the indoor turbulence simulation test platform based on SLM of the present invention.
[0033] Figure 3 Schematic diagram of the ConvNeXt convolutional neural network structure of the present invention.
[0034] Figure 4 This is a schematic diagram of the actual external turbulence detection platform built during actual detection in the present invention. DETAILED DESCRIPTION
[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0036] Example 1:
[0037] Example 1 of the present invention discloses an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method using a ConvNeXt convolutional neural network, such as Figure 1 As shown, including:
[0038] Step 1: Use an indoor turbulence simulation test platform based on SLM to obtain instantaneous spot images of random turbulence effects under different turbulence intensities, and preprocess them to obtain training sets, validation sets, and test sets; among them, the SLM in the indoor turbulence simulation test platform is loaded with a phase screen generated based on the modified Von karman atmospheric turbulence spectrum model, and the instantaneous spot image is obtained by rotating and translating the static phase screen.
[0039] Indoor turbulence simulation test platform based on SLM, such as Figure 2 As shown, it includes: a HeNe laser (wavelength 632.8 nm, power 1 mW), a beam expander, a first polarizer, a transmissive spatial light modulator SLM (HOLOEYE, LC2012, resolution 768 × 1024, pixel size 36 μm), a second polarizer, an attenuator, and a CCD camera (Thorlabs BC207VIS / M, resolution 2448 × 2048, pixel size 3.45 μm);
[0040] After the HeNe laser is expanded by a beam expander, the Gaussian light amplitude state is adjusted by the first polarizer, and phase modulation is performed by a transmissive spatial light modulator (SLM) to simulate turbulence. The turbulence amplitude state is adjusted again by the second polarizer, and after the light intensity is weakened by the attenuation plate, the light spot is recorded by a CCD camera.
[0041] Preprocessing includes: CCD-based light environment correction, power correction and noise reduction, resolution compression (compressed to 256×256 resolution) and normalization preprocessing.
[0042] SLM turbulence simulation requires input of atmospheric turbulence phase screens of varying intensities. This paper uses a modified Vonkarman atmospheric turbulence spectrum model to generate a phase screen to load the SLM, which is randomly generated and acquired using MATLAB. The modified Von Karman atmospheric turbulence spectrum model takes into account the effects of both the outer and inner atmospheric scales, and experiments have shown that the modified turbulence spectrum is more accurate. The modified Von Karman atmospheric turbulence spectrum model is as follows:
[0043]
[0044] Among them, Φ n is the power spectral density; is the atmospheric refractive index structure constant; κ is the wave number; κ m is the cutoff wavenumber related to the inner scale; κ0 is the cutoff wavenumber related to the outer scale.
[0045] The SLM operating software uses a method of rotating and translating a static phase screen to obtain dynamically changing turbulence effects at different turbulence intensities. Each translation or rotation of the turbulence phase screen captures an image. Therefore, the spot image captured by the CCD camera is a spot image of the random turbulence effect at the same turbulence intensity, which is a transient spot image.
[0046] Compared with actual outdoor turbulence, the indoor turbulence simulation conducted by the present invention can provide stable turbulence conditions and can collect instantaneous spot images of random turbulence effects under different turbulence intensities on demand. Based on this, a huge spot database is established for training subsequent atmospheric turbulence intensity prediction models.
[0047] Step 2: Input the training set to Figure 3 The ConvNeXt convolutional neural network shown in FIG is trained to obtain an atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network, and the optimal weight parameters of the atmospheric turbulence intensity prediction model are obtained based on the validation set.
[0048] The training batch of the atmospheric turbulence intensity prediction model based on ConvNeXt convolutional neural network is set to 4, and the learning rate is set to 5×10 -4 , weight decay is set to 5×10 -2 The loss function for the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network is a sparse categorical cross entropy plus L2 regularization loss function. Sparse categorical cross entropy can measure the accuracy of model predictions and suppress overfitting, while L2 regularization prevents the influence of hidden units caused by excessive weights.
[0049] Step 3: Input the test set into the atmospheric turbulence intensity prediction model under the optimal weight parameters to obtain the atmospheric turbulence intensity prediction results.
[0050] Experimental results show that the atmospheric turbulence intensity prediction model proposed in the present invention has a test set accuracy of 92.6% in 8 categories (9e-13, 5e-13, 1e-13, 5e-14, 1e-14, 5e-15, 1e-15, 5e-16).
[0051] Example 2:
[0052] Example 2 of the present invention discloses an actual outdoor turbulence detection platform constructed by using an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method based on ConvNeXt convolutional neural network for actual detection. Figure 4 As shown, including:
[0053] The laser beam emitted by a helium-neon laser (wavelength 632.8nm, power 1mW) is expanded by a beam expander and then transmitted through a real atmospheric environment. The beam is focused by an astronomical telescope, attenuated by an attenuator, and finally recorded by a CCD camera (Thorlabs BC207VIS / M, resolution 2448×2048, pixel size 3.45μm). A 500W mobile power bank is used to maintain power.
[0054] The actual needs are obtained by building an actual outdoor turbulence detection platform, and the final atmospheric turbulence intensity is identified through the established convolutional neural network model.
[0055] Example 3:
[0056] Embodiment 3 of the present invention discloses an atmospheric turbulence intensity detection system based on an indoor turbulence simulation test platform based on an SLM and a ConvNeXt convolutional neural network, comprising:
[0057] The module for acquiring training, validation, and test sets for atmospheric turbulence intensity prediction uses an indoor turbulence simulation test platform based on an SLM to acquire instantaneous spot images of random turbulence effects at different turbulence intensities, and performs preprocessing to obtain training, validation, and test sets. The SLM in the indoor turbulence simulation test platform is loaded with a phase screen generated based on a modified Vonkarman atmospheric turbulence spectrum model, and instantaneous spot images are acquired by rotating and translating a static phase screen.
[0058] Atmospheric turbulence intensity prediction model training module: used to input the training set into the ConvNeXt convolutional neural network, train the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network, and obtain the optimal weight parameters of the atmospheric turbulence intensity prediction model based on the validation set;
[0059] Atmospheric turbulence intensity prediction module: used to input the test set into the atmospheric turbulence intensity prediction model under the optimal weight parameters to obtain the atmospheric turbulence intensity prediction results.
[0060] Example 4:
[0061] Embodiment 4 of the present invention discloses an electronic device, including:
[0062] memory for storing computer programs;
[0063] The processor is used to implement the steps of an indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection method based on ConvNeXt convolutional neural network when executing a computer program.
[0064] Example 5:
[0065] Embodiment 5 of the present invention discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform based on SLM and a ConvNeXt convolutional neural network are implemented.
[0066] The embodiments of the present invention disclose a method, system, device, and storage medium for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform and a ConvNeXt convolutional neural network. By constructing an indoor turbulence simulation test platform based on the SLM, a transmissive spatial light modulator (SLM) is loaded using a phase screen generated based on a modified Von Karman atmospheric turbulence spectrum model and rotated for phase modulation. By rotating and translating the static phase screen, instantaneous spot images of random turbulence effects under different turbulence intensities are obtained. This allows for easy acquisition of the spot data set required for the atmospheric turbulence intensity prediction model corresponding to different application requirements, effectively improving the prediction accuracy of the subsequently trained model. Furthermore, the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network was trained and improved. The stem layer was modified to increase the convolution kernel size from 7 to 16, and the step size remained unchanged. The results showed that the iterative effect of the convolution kernel of 16 was more stable. Secondly, different convolution kernels (1*1, 3*3, 5*5 sizes) were added to the black layer to extract features respectively and then perform feature fusion (enriching feature information). Sparse classification cross entropy and L2 regularization were used as the loss function of the atmospheric turbulence intensity prediction model. The sparse classification cross entropy was used to measure the accuracy of the model prediction and suppress overfitting. L2 regularization prevented the influence of hidden units caused by excessive weights. Finally, a simple, convenient and accurate atmospheric turbulence intensity prediction was achieved.
[0067] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.
[0068] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform and a ConvNeXt convolutional neural network, characterized in that: include: Step 1: Using an indoor turbulence simulation test platform based on an SLM to obtain instantaneous spot images of random turbulence effects under different turbulence intensities, and preprocessing them to obtain training sets, validation sets, and test sets; wherein the SLM in the indoor turbulence simulation test platform is loaded with a phase screen generated based on a modified Von Karman atmospheric turbulence spectrum model, and the instantaneous spot images are obtained by rotating and translating a static phase screen; Step 2: Input the training set into the ConvNeXt convolutional neural network, train to obtain an atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network, and obtain the optimal weight parameters of the atmospheric turbulence intensity prediction model based on the validation set; Step 3: Input the test set into the atmospheric turbulence intensity prediction model under the optimal weight parameters to obtain the atmospheric turbulence intensity prediction result.
2. The method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform and a ConvNeXt convolutional neural network according to claim 1, characterized in that: In step 1, the SLM-based indoor turbulence simulation test platform includes: a He-Ne laser, a beam expander, a first polarizer, a transmissive spatial light modulator (SLM), a second polarizer, an attenuator, and a CCD camera; After the HeNe laser is expanded by the beam expander, the amplitude state of the Gaussian light is adjusted by the first polarizer, and phase modulation is performed by the transmissive spatial light modulator SLM to simulate the generation of turbulence; the amplitude state of the turbulence is adjusted again by the second polarizer, and after the light intensity is weakened by the attenuation plate, the light spot is recorded by the CCD camera.
3. The method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform and a ConvNeXt convolutional neural network according to claim 1, characterized in that: In step 1, the preprocessing includes: CCD-based light environment correction, power correction and noise reduction processing, resolution compression processing and normalization preprocessing.
4. The method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform and a ConvNeXt convolutional neural network according to claim 1, characterized in that: In step 1, the modified Von karman atmospheric turbulence spectrum model is as follows: Among them, Φ n is the power spectral density; is the atmospheric refractive index structure constant; κ is the wave number; κ m is the cutoff wavenumber related to the inner scale; κ0 is the cutoff wavenumber related to the outer scale.
5. The method for detecting atmospheric turbulence intensity based on an indoor turbulence simulation test platform and a ConvNeXt convolutional neural network according to claim 1, characterized in that: In step 2, the loss function of the atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network selects the sparse classification cross entropy plus L2 regularization loss function.
6. An indoor turbulence simulation test platform based on SLM and an atmospheric turbulence intensity detection system based on ConvNeXt convolutional neural network, utilizing the atmospheric turbulence intensity detection method based on SLM and ConvNeXt convolutional neural network according to any one of claims 1 to 5, characterized in that: include: The module for acquiring training, validation, and test sets for atmospheric turbulence intensity prediction is used to obtain instantaneous spot images of random turbulence effects at different turbulence intensities using an indoor turbulence simulation test platform based on an SLM, and to perform preprocessing to obtain training, validation, and test sets. The SLM in the indoor turbulence simulation test platform is loaded with a phase screen generated based on a modified Vonkarman atmospheric turbulence spectrum model, and the instantaneous spot images are obtained by rotating and translating a static phase screen. An atmospheric turbulence intensity prediction model training module is configured to input the training set into a ConvNeXt convolutional neural network, train an atmospheric turbulence intensity prediction model based on the ConvNeXt convolutional neural network, and obtain the optimal weight parameters of the atmospheric turbulence intensity prediction model based on the validation set; Atmospheric turbulence intensity prediction module: used to input the test set into the atmospheric turbulence intensity prediction model under the optimal weight parameters to obtain the atmospheric turbulence intensity prediction result.
7. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of an indoor turbulence simulation test platform based on SLM and a method for detecting atmospheric turbulence intensity using a ConvNeXt convolutional neural network as described in any one of claims 1 to 5 when executing the computer program.
8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an indoor turbulence simulation test platform based on SLM and a ConvNeXt convolutional neural network for detecting atmospheric turbulence intensity as described in any one of claims 1 to 5.