Multi-light-spot centroid extraction method and device based on Gaussian mixture model clustering

Abstract

The invention discloses a multi-light-spot centroid extraction method and device based on Gaussian mixture model clustering, and belongs to the technical field of optical information measurement, and the method comprises the steps: carrying out the noise removal of a current effective sub-aperture image through employing a denoising algorithm; the number of light spots in the sub-apertures is determined, effective sub-aperture image pixel data is converted into cloud point data, and each light spot generates a corresponding Gaussian component through discrete sampling; initializing Gaussian mixture model parameters, obtaining an initial mean vector of each Gaussian component, calculating an initial covariance matrix of each cluster according to the initial clustering center of each cluster, and initializing a mixture coefficient of each Gaussian component based on the size of each cluster; updating Gaussian mixture model parameters through an expectation maximization algorithm; and after the Gaussian mixture model is converged, outputting a mean vector of the last round of maximization step based on probability weighted average calculation as the centroid of the cluster corresponding to each Gaussian component. According to the invention, accurate positioning of a plurality of sub-light spots under the condition of sub-light spot crosstalk is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical information measurement, and in particular relates to a method and device for extracting the centroid of multiple light spots based on Gaussian mixture model clustering. Background Art

[0002] The Hartmann wavefront sensor is a wavefront measurement device composed of a microlens array and a photodetector. With the continuous expansion of wavefront detection applications and the increasing complexity of application scenarios, the Hartmann wavefront sensor's detection targets are no longer limited to the traditional measurement of wavefront distortion from a single point light source. To adapt to this change, wide-field-of-view, multi-line-of-sight wavefront information reconstruction technology for multiple beacons and multiple targets is becoming a research hotspot in the field of wavefront sensing.

[0003] When a single Hartmann wavefront sensor performs multi-beacon optical detection, the dense spatial distribution of multiple sub-spots within a wide field of view leads to crosstalk between them. Existing technologies still rely on the traditional centroid method for centroid calculation. When crosstalk occurs between sub-spots, precise segmentation of the sub-spots is difficult, making it difficult to accurately extract the centroid of the spot. Existing centroid extraction methods are clearly inadequate for the precise segmentation and positioning of multiple sub-spots, severely limiting the accuracy of wavefront reconstruction. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: A multi-spot centroid extraction method based on Gaussian mixture model clustering, comprising: Step 1, extracting effective sub-aperture image; Step 2: Use a denoising algorithm to remove noise from the current effective sub-aperture image; Step 3: Determine the number of light spots within the sub-aperture, and convert the effective sub-aperture image pixel data into cloud point data. Each light spot generates a corresponding Gaussian component through discrete sampling. Step 4: Initialize the Gaussian mixture model parameters, obtain the initial mean vector of each Gaussian component, calculate the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initialize the mixing coefficient of each Gaussian component based on the size of each cluster; Step 5, update the Gaussian mixture model parameters through the expectation maximization algorithm; Step 6: Determine whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. If the Gaussian mixture model has not converged, continue to update the Gaussian mixture model parameters. After the Gaussian mixture model converges, output the mean vector calculated based on the probability-weighted average in the last maximization step as the centroid of the cluster corresponding to each Gaussian component. The mean vector corresponds to the centroid of the corresponding light spot. Step 7: Determine whether all valid sub-apertures have been traversed. If not, extract the next valid sub-aperture image and repeat steps 2-6 until all valid sub-apertures have been traversed.

[0005] A device for extracting centroids of multiple light spots based on Gaussian mixture model clustering, comprising: Image acquisition module, extracting effective sub-aperture images; Denoising module, which uses a denoising algorithm to remove noise from the current effective sub-aperture image; The data conversion module determines the number of light spots within the sub-aperture and converts the effective sub-aperture image pixel data into cloud point data. Each light spot generates a corresponding Gaussian component through discrete sampling; Initialization module, which initializes the Gaussian mixture model parameters, obtains the initial mean vector of each Gaussian component, calculates the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initializes the mixing coefficient of each Gaussian component based on the size of each cluster; Update module, which updates the Gaussian mixture model parameters through the expectation maximization algorithm; The convergence judgment module determines whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. If the Gaussian mixture model has not converged, the Gaussian mixture model parameters are updated. After the Gaussian mixture model converges, the mean vector calculated based on the probability-weighted average of the last maximization step is output as the centroid of the cluster corresponding to each Gaussian component. The mean vector corresponds to the centroid of the corresponding light spot. The judgment module determines whether all valid sub-apertures have been traversed. If not, the next valid sub-aperture image is extracted and the operations of the denoising module, data conversion module, initialization module, update module and convergence judgment module are repeatedly executed in sequence until all valid sub-apertures have been traversed.

[0006] The present invention has the following beneficial effects: Compared to existing methods for calculating the centroid of multiple sub-spots based on the centroid method, this method converts image data into cloud point data and, based on a Gaussian mixture model, uses probability-weighted averaging to calculate the centroid of corresponding spot clusters. This method can accurately locate multiple sub-spots even under sub-spot crosstalk conditions, providing an efficient multi-spot centroid calculation technology solution for wavefront reconstruction in wide-field-of-view multi-beacon Hartmann wavefront sensors. This method has significant application value for wavefront measurement in complex scenarios, especially for wide-field-of-view multi-beacon Hartmann wavefront sensors. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 This is a flow chart of the multi-spot centroid extraction method based on Gaussian mixture model clustering of the present invention.

[0008] Figure 2 Schematic diagram of the Hartmann wavefront sensor detection of dual beacon light in an embodiment of the present invention.

[0009] Figure 3 This is a light spot array image collected by the photoelectric detector when dual beacon light is incident in the embodiment of the present invention.

[0010] Figure 4a This is the original image of sub-aperture No. 1 extracted in the embodiment of the present invention.

[0011] Figure 4b This is the image extracted after denoising by sub-aperture No. 1 in an embodiment of the present invention.

[0012] Figure 4c Schematic diagram of converting the extracted No. 1 sub-aperture into cloud point data in an embodiment of the present invention.

[0013] Figure 5 This is the centroid calculation result of all light spots in the present invention.

[0014] Figure 6a It is the centroid calculation error between the centroid calculation result of the upper left sight spot array and the actual position of the single spot.

[0015] Figure 6b It is the centroid calculation error between the centroid calculation result of the right lower sight spot array and the actual position of the single light spot. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is 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 intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0017] The present invention utilizes the Gaussian mixture model clustering algorithm to achieve intelligent segmentation of overlapping light spots by establishing a statistical probability model of light spot distribution, thereby realizing a multi-spot centroid extraction solution that effectively solves the light spot crosstalk problem under multi-beacon and wide field of view conditions.

[0018] The technical solution adopted by the present invention is: a multi-spot centroid extraction method based on Gaussian mixture model clustering, and the specific implementation steps are as follows: Step 1: Extract the effective sub-aperture image.

[0019] Step 2: Use a denoising algorithm to remove noise from the current effective sub-aperture image.

[0020] Step 3: Determine the number of spots within the sub-aperture The effective sub-aperture image pixel data is converted into cloud point data, and each spot generates the corresponding Gaussian component through discrete sampling.

[0021] Step 4: Initialize the Gaussian mixture model parameters, obtain the initial mean vector of each Gaussian component, that is, the initial cluster center of each cluster, calculate the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initialize the mixing coefficient of each Gaussian component based on the size of each cluster.

[0022] Step 5: Update the Gaussian mixture model parameters using the Expectation-Maximization (EM) algorithm. The specific process of step 5 is as follows: Step 5.1: Calculate data points using the Expectation Step Belong to The posterior probability of a Gaussian component is calculated as follows: γ i k = π k  ( x i | μ k , Σ k ) ∑ j = 1 K [ π j  ( x i | μ j , Σ j ) ] (1) Where, For data points Belong to The posterior probability of the Gaussian components, For the The mixing coefficient of Gaussian components, is the Gaussian probability density function, For the The mean vector of the Gaussian components, For the The covariance matrix of the Gaussian components, is the number of light spots within the sub-aperture, The value range is 1 to Integers between for Equal to 1 to When , the mixing coefficient of each Gaussian component is, for Equal to 1 to When , the mean vector of each Gaussian component is, for Equal to 1 to When , the covariance matrix of each Gaussian component is.

[0023] Step 5.2, the maximization step updates the Gaussian mixture model parameters based on the posterior probability 、 and , and the calculation formulas are: π k = 1 N ∑ i = 1 N γ i k μ k = ∑ i = 1 N ( γ i k x i ) ∑ i = 1 N γ i k Σ k = ∑ i = 1 N [ γ i k ( x i − μ k ) ( x i − μ k ) T ] ∑ i = 1 N γ i k (2) Where, is the total amount of cloud point data, is the transpose operation.

[0024] Step 6: Determine whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. If the Gaussian mixture model has not converged, set the Gaussian mixture model parameters updated in step 5.2 to 、 and Substitute into formula (1) in step 5.1, recalculate the posterior probability of each data point belonging to each Gaussian component, continue to update the Gaussian mixture model parameters, and after the Gaussian mixture model converges, calculate the mean vector based on the probability weighted average in the last round of maximization step The output is the centroid of the cluster corresponding to each Gaussian component, which corresponds to the centroid of the corresponding light spot. The convergence condition of the model used is that the change in the log-likelihood function between two adjacent iterations is less than the preset threshold. , preset threshold The value is set according to the actual situation. The calculation formula of the log-likelihood function is: ,in, is the log-likelihood function, From 1 to integer, is the total amount of cloud point data, From 1 to integer, is the number of light spots within the sub-aperture, is the set of Gaussian mixture model parameters.

[0025] Step 7: Determine whether all valid sub-apertures have been traversed. If not, extract the next valid sub-aperture image and repeat steps 2-6 until all valid sub-apertures have been traversed.

[0026] Furthermore, the denoising algorithm described in step 2 includes noise removal algorithms such as mean filtering, median filtering, non-local mean filtering (NLM), three-dimensional block matching filtering (BM3D), wavelet transform, adaptive threshold filtering, morphological filtering, deep convolutional neural network, etc., and can also be other methods that can remove image noise.

[0027] Furthermore, the number of spots within the sub-aperture is determined in step 3. This can be achieved by analyzing the connected area or the number of peaks of the sub-aperture image, or by using other methods that can determine the number of light spots.

[0028] Furthermore, the method for initializing the Gaussian mixture model parameters in step 4 and obtaining the initial mean vector of each Gaussian component includes a k-means++ algorithm initialization method, a pure random initialization method, a uniform distribution initialization method, or other methods for initializing the mean vector.

[0029] Furthermore, the method of initializing the mixing coefficients of each Gaussian component based on the size of each cluster in step 4 includes initializing the mixing coefficients (weights) based on the ratio of the number of cluster points of each Gaussian component, or other methods of initializing the mixing coefficients (weights), as long as the sum of the weights is ensured to be 1.

[0030] The embodiment of the present invention takes a dual light spot as an example to extract the centroid of multiple light spots. Figure 1 This is a flow chart of the multi-spot centroid extraction method based on Gaussian mixture model clustering described in the present invention. Figure 2 Schematic diagram of the detection of the dual beacon light by the Hartmann wavefront sensor in an embodiment of the present invention. Figure 2 As shown, two beacon light beams, after being transmitted through atmospheric turbulence, enter the Hartmann wavefront sensor separately. The Hartmann wavefront sensor focuses the beacon light sources through a microlens array, forming two sets of light spot arrays. The photosensitive surface of the Hartmann wavefront sensor's photodetector is located at the focal plane of the microlens array and is used to capture the light spot array image. In an embodiment of the present invention, the light beam wavelength is 635 nm, the spatial sampling frequency of the Hartmann wavefront sensor is 16×16, the number of pixels per subaperture is 24×24, the focal length of the microlens array is 20 mm, and the photodetector pixel size is 12.8 μm.

[0031] Figure 3 This is the spot array image collected by the photodetector when the simulated dual-beacon light is incident. Each square box in the figure corresponds to an effective sub-aperture for wavefront restoration. Figure 3 The first effective sub-aperture on the leftmost side of the first row is named effective sub-aperture No. 1. Each effective sub-aperture is numbered from left to right and from top to bottom. The rightmost effective sub-aperture in the bottom row is named effective sub-aperture No. 192 (a total of 192 effective sub-apertures). The position of the centroid of the dual spot within each sub-aperture is calculated by the following steps: Step 1: Extract the effective sub-aperture image No. 1, such as Figure 4a shown.

[0032] Step 2: The extracted effective sub-aperture image is subjected to noise removal processing by using the adaptive threshold algorithm and morphological filtering. The denoised sub-aperture image is as follows: Figure 4b shown.

[0033] Step 3: Determine the number of spots within the effective sub-aperture after denoising by calculating the area of the connected sub-aperture region. In this embodiment, the number of light spots =2, and at the same time, by sampling the pixels of the effective sub-aperture image, the pixel data of the effective sub-aperture image is converted into cloud point data, and each light spot generates the corresponding Gaussian component through discrete sampling. Figure 4c The figure shows the conversion of sub-aperture No. 1 into cloud point data.

[0034] Step 4: Initialize the clustering parameters of the Gaussian mixture model. Use the k-means++ algorithm to obtain the initial mean vector of each Gaussian component, that is, the initial cluster center of each cluster. Calculate the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initialize the mixing coefficient of each Gaussian component based on the size of each cluster.

[0035] Step 5: Update the model parameters using the Expectation-Maximization (EM) algorithm. The specific process of step 5 is as follows: Step 5.1: Calculate data points using the Expectation Step Belong to The posterior probability of a Gaussian component is calculated as follows: γ i k = π k  ( x i | μ k , Σ k ) ∑ j = 1 K [ π j  ( x i | μ j , Σ j ) ] (1) Step 5.2, the maximization step updates the Gaussian mixture model parameters based on the posterior probability 、 and , and the calculation formulas are: π k = 1 N ∑ i = 1 N γ i k μ k = ∑ i = 1 N ( γ i k x i ) ∑ i = 1 N γ i k Σ k = ∑ i = 1 N [ γ i k ( x i − μ k ) ( x i − μ k ) T ] ∑ i = 1 N γ i k (2) Step 6: Determine whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. The model convergence condition used in this embodiment is that the change in the log-likelihood function between two adjacent iterations is less than a preset threshold. ,Right now , where is the log-likelihood value under the current parameters, is the log-likelihood value under the latest model parameters, is the log-likelihood value under the previous model parameters, is the total amount of cloud point data, From 1 to integer, From 1 to integer, is the number of light spots within the sub-aperture. = 10 -6If the model parameters do not meet the convergence conditions, re-execute step 5. After the model converges, the mean vector calculated based on the probability weighted average of the last round of maximization step is The output is the centroid of the cluster corresponding to each Gaussian component, which corresponds to the centroid of the corresponding light spot.

[0036] Step 7, determine whether all effective sub-apertures have been traversed. If not, extract the next effective sub-aperture image and repeat steps 2-6 until 192 effective sub-apertures have been traversed. The centroid data of all light spots can be obtained. The calculation results of the centroid of all light spots are as follows: Figure 5 As shown, Figure 6a is the centroid calculation error between the calculation result of the center of mass position of the upper left sight spot array and the actual position of the single spot, Figure 6b is the calculation error between the center of mass position of the right lower sight spot array and the actual position of the single spot. Figure 6a and Figure 6b The average of the centroid calculation errors in is 0.1058 and 0.1127 pixels respectively.

[0037] The present invention further proposes a multi-spot centroid extraction device based on Gaussian mixture model clustering, comprising: Image acquisition module, extracting effective sub-aperture images; Denoising module, which uses a denoising algorithm to remove noise from the current effective sub-aperture image; The data conversion module determines the number of light spots within the sub-aperture and converts the effective sub-aperture image pixel data into cloud point data. Each light spot generates a corresponding Gaussian component through discrete sampling; Initialization module, which initializes the Gaussian mixture model parameters, obtains the initial mean vector of each Gaussian component, calculates the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initializes the mixing coefficient of each Gaussian component based on the size of each cluster; Update module, which updates the Gaussian mixture model parameters through the expectation maximization algorithm; The convergence judgment module determines whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. If the Gaussian mixture model has not converged, the Gaussian mixture model parameters are updated. After the Gaussian mixture model converges, the mean vector calculated based on the probability-weighted average of the last maximization step is output as the centroid of the cluster corresponding to each Gaussian component. The mean vector corresponds to the centroid of the corresponding light spot. The judgment module determines whether all valid sub-apertures have been traversed. If not, the next valid sub-aperture image is extracted and the operations of the denoising module, data conversion module, initialization module, update module and convergence judgment module are repeatedly executed in sequence until all valid sub-apertures have been traversed.

[0038] The present invention further proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps of the multi-spot centroid extraction method based on Gaussian mixture model clustering are implemented.

[0039] The present invention further proposes a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the multi-spot centroid extraction method based on Gaussian mixture model clustering are implemented.

[0040] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk drives, CD-ROMs, optical storage devices, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention may be implemented using various computer languages, such as Matlab and Python.

[0041] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0042] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0043] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0044] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0045] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

[0046] The above descriptions are merely embodiments of the present invention and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied to other related system fields, are also included in the scope of protection of the present invention.

[0047] The contents not described in detail in the specification of the present invention belong to the prior art known to those skilled in the art.

Claims

1. A multi-spot centroid extraction method based on Gaussian mixture model clustering, characterized in that: include: Step 1, extracting effective sub-aperture image; Step 2: Use a denoising algorithm to remove noise from the current effective sub-aperture image; Step 3: Determine the number of light spots within the sub-aperture, and convert the effective sub-aperture image pixel data into cloud point data. Each light spot generates a corresponding Gaussian component through discrete sampling. Step 4: Initialize the Gaussian mixture model parameters, obtain the initial mean vector of each Gaussian component, calculate the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initialize the mixing coefficient of each Gaussian component based on the size of each cluster; Step 5, update the Gaussian mixture model parameters through the expectation maximization algorithm; Step 6: Determine whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. If the Gaussian mixture model has not converged, continue to update the Gaussian mixture model parameters. After the Gaussian mixture model converges, output the mean vector calculated based on the probability weighted average in the last round of maximization step as the centroid of the cluster corresponding to each Gaussian component. Step 7: Determine whether all valid sub-apertures have been traversed. If not, extract the next valid sub-aperture image and repeat steps 2-6 until all valid sub-apertures have been traversed.

2. The method for extracting the centroid of multiple light spots based on Gaussian mixture model clustering according to claim 1, characterized in that: In step 2, the denoising algorithm includes mean filtering, median filtering, non-local mean filtering, three-dimensional block matching filtering, wavelet transform, adaptive threshold filtering, morphological filtering, deep convolutional neural network or other noise removal algorithms.

3. The method for extracting the centroid of multiple light spots based on Gaussian mixture model clustering according to claim 1, characterized in that: In step 3, the determination of the number of light spots within the sub-aperture is achieved by analyzing the connected area or the number of peaks of the sub-aperture image.

4. The method for extracting the centroid of multiple light spots based on Gaussian mixture model clustering according to claim 1, characterized in that: In step 4, the method of initializing the Gaussian mixture model parameters and obtaining the initial mean vector of each Gaussian component includes: k-means enhancement algorithm initialization method, pure random initialization method, uniform distribution initialization method or other methods of initializing the mean vector.

5. The method for extracting centroids of multiple light spots based on Gaussian mixture model clustering according to claim 1, characterized in that: In step 4, the initialization of the mixing coefficients of the Gaussian components based on the size of each cluster includes initializing the mixing coefficients based on the ratio of the number of cluster points of each Gaussian component or other methods of initializing the mixing coefficients.

6. The method for extracting centroids of multiple light spots based on Gaussian mixture model clustering according to claim 1, characterized in that: The step 5 comprises: Step 5.1, calculate the data points by the expected step Belong to The posterior probability of the Gaussian components; In step 5.2, the maximization step updates the Gaussian mixture model parameters according to the posterior probability.

7. The method for extracting centroids of multiple light spots based on Gaussian mixture model clustering according to claim 6, characterized in that: In step 5.2, the Gaussian mixture model parameters include: The mixing coefficient of Gaussian components , No. The mean vector of the Gaussian components , No. The covariance matrix of the Gaussian components .

8. The method for extracting centroids of multiple light spots based on Gaussian mixture model clustering according to claim 6, characterized in that: In step 5.1, the data points are calculated by the expected step Belong to The calculation formula for the posterior probability of a Gaussian component is: ； Where, For data points Belong to The posterior probability of the Gaussian components, For the The mixing coefficient of Gaussian components, is the Gaussian probability density function, For the The mean vector of the Gaussian components, For the The covariance matrix of the Gaussian components, is the number of light spots within the sub-aperture, The value range is 1 to Integers between for Equal to 1 to When , the mixing coefficient of each Gaussian component is, for Equal to 1 to When , the mean vector of each Gaussian component is, for Equal to 1 to When , the covariance matrix of each Gaussian component; In step 5.2, the Gaussian mixture model parameters 、 and as follows: ； Where, is the total amount of cloud point data, is the transpose operation, Is an integer from 1 to .

9. The method for extracting centroids of multiple light spots based on Gaussian mixture model clustering according to claim 8, characterized in that: In step 6, the convergence condition is that the change in the log-likelihood function between two adjacent iterations is less than the preset threshold ; The calculation formula of the log-likelihood function is ;in, is the log-likelihood function, From 1 to integer, is the total amount of cloud point data, From 1 to integer, is the number of light spots within the sub-aperture, is the set of Gaussian mixture model parameters.

10. A multi-spot centroid extraction device based on Gaussian mixture model clustering, characterized in that: include: Image acquisition module, extracting effective sub-aperture images; Denoising module, which uses a denoising algorithm to remove noise from the current effective sub-aperture image; The data conversion module determines the number of light spots within the sub-aperture and converts the effective sub-aperture image pixel data into cloud point data. Each light spot generates a corresponding Gaussian component through discrete sampling; Initialization module, which initializes the Gaussian mixture model parameters, obtains the initial mean vector of each Gaussian component, calculates the initial covariance matrix of each cluster based on the initial cluster center of each cluster, and initializes the mixing coefficient of each Gaussian component based on the size of each cluster; Update module, which updates the Gaussian mixture model parameters through the expectation maximization algorithm; The convergence judgment module determines whether the Gaussian mixture model has converged by calculating the change in the log-likelihood function between two adjacent iterations. If the Gaussian mixture model has not converged, the Gaussian mixture model parameters are updated. After the Gaussian mixture model converges, the mean vector calculated based on the probability-weighted average of the last maximization step is output as the centroid of the cluster corresponding to each Gaussian component. The mean vector corresponds to the centroid of the corresponding light spot. The judgment module determines whether all valid sub-apertures have been traversed. If not, the next valid sub-aperture image is extracted and the operations of the denoising module, data conversion module, initialization module, update module and convergence judgment module are repeatedly executed in sequence until all valid sub-apertures have been traversed.

Images