Solar activity area automatic detection method based on density clustering algorithm

Abstract

The invention discloses a solar activity area automatic detection method based on a density clustering algorithm. The solar activity area automatic detection method comprises the steps that S1, a radial magnetic graph is acquired; s2, carrying out the preprocessing of the visual direction magnetic graph; s3, continuously carrying out selective DBSCAN clustering twice on the preprocessed radial magnetic graph to obtain a primary clustering result of the solar active area; and S4, integrating and filtering the initial clustering result of the solar activity area to obtain a final detection result of the solar activity area. The method has the advantages of high detection precision, high detection efficiency and high robustness.

Description

Technical Field

[0001] The invention belongs to the technical field of solar active region detection, and in particular relates to an automatic solar active region detection method based on a density clustering algorithm. Background Art

[0002] Solar active regions (ARs) are areas on the solar disk where solar active bodies such as sunspots, flares and prominences are active frequently. Solar activity exhibits complex spatial and temporal behaviors from the photosphere to the corona. Among them, sunspots are the first feature to be used as an indicator of AR intensity, because they appear as dark spots on the surface of the solar photosphere in visible light continuum observations, and are sometimes even visible to the naked eye. Since Galileo invented the telescope in the 17th century, people have begun to observe sunspots systematically. The temperature of sunspots is three orders of magnitude lower than that of the surrounding photosphere, but the magnetic flux is highly concentrated here. The number of sunspots is an indicator of the frequency of solar activity generated by the solar generator, while the area of ​​sunspots is a measure of how much magnetic flux the solar generator generates. Therefore, the number and area of ​​sunspots can be regarded as a scale of the long period of solar activity. With the iterative update of observation equipment and technology, we can define active regions through observation data of the solar magnetic field, which is more accurate and comprehensive than sunspots.

[0003] The AR Long-term Comprehensive Catalog provided by the National Oceanic and Atmospheric Administration (NOAA) is one of the widely used AR long-term catalogs. NOAA assigns a number to each AR containing at least one sunspot through manual or computer-assisted human operations based on white-light images. The catalog reports the location, area, and magnetic field configuration of sunspot groups, etc. However, with the advancement of observation technology and the increase in the quantity and quality of synchronous observation data, the traditional method of manually detecting or marking AR cannot meet the requirements of speed and accuracy when processing large amounts of data, especially when conducting statistical studies of long-term data and modeling solar dynamics models. Automated detection can objectively and consistently extract AR, avoiding the subjectivity of manual inspection, so a series of automated detection methods have been developed.

[0004] Some early automated detection methods were mainly based on image processing techniques, such as threshold segmentation and region growing methods. Although these methods have improved detection efficiency, they still have problems such as relying on empirical threshold settings and difficulty in separating complex magnetic field regions in high-resolution images. With the development of the application of machine learning methods in computer vision technology, machine learning methods have gradually been applied in the task of identifying solar active regions. A widely used machine learning method is convolutional neural networks (CNNs), which perform well in image feature extraction. CNNs are supervised learning algorithms, which are usually trained based on labeled data by adjusting parameters until the model reaches the desired performance. Although such supervised learning models have higher accuracy and generalization capabilities, they require a large amount of objective and consistent AR labeled data, and it is difficult to completely get rid of the subjectivity of human inspection. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides an automatic detection method for solar active regions based on a density clustering algorithm to solve the problem that the existing related detection methods are affected by human factors and rely on labeled data, thus leading to incomplete detection of active regions.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for automatically detecting solar active regions based on a density clustering algorithm, comprising the following steps:

[0007] S1, obtain the line-of-sight magnetic map;

[0008] S2, preprocessing the line-of-sight magnetic map;

[0009] S3, perform two consecutive selective DBSCAN clustering on the preprocessed line-of-sight magnetogram to obtain the preliminary results of solar active region clustering;

[0010] S4. Integrate and filter the solar active region clustering results to obtain the final solar active region detection results.

[0011] Furthermore, in step S2, the preprocessing is: performing grayscale threshold segmentation on the viewing magnetic map to obtain pixels with significant magnetic field intensity as input data points for DBSCAN clustering.

[0012] Furthermore, in step S3, the method for performing the first DBSCAN clustering is specifically as follows:

[0013] S31, calculate the ∈ neighborhood N of each input data point ∈ (p);

[0014] N ∈ (p)={q∈D|dist(p,q)≤∈}

[0015] In the formula, q represents the pixel point that meets the conditions, D represents the pixel point that is screened out after threshold segmentation, dist(p,q) represents the distance between pixel points p and q, and ∈ represents the selected distance threshold;

[0016] S32, identifying core points according to the calculated ∈ neighborhood;

[0017] For any pixel point p, it is determined whether |N∈(p)|≥minpts holds; wherein minpts represents the minimum number of points required to form a dense pattern;

[0018] If yes, the pixel point p is identified as the core point p, and the core point p is added to the core object set ψ = ψ∪{p};

[0019] If not, no operation is performed;

[0020] S33, determining whether the core object set ψ is empty;

[0021] If yes, then end the first DBSCAN clustering;

[0022] If not, proceed to step S35;

[0023] S34, randomly select a core point p from the core object set ψ, and update the current cluster core queue φ={p};

[0024] S35, according to the updated cluster core queue φ = {p}, update the cluster sample set C = C∪{p} and the unvisited sample set D = D-{p};

[0025] S36. Select a core point in the current cluster core queue φ, find its corresponding subset N∈(p), and then update the current cluster sample set C=C∪N ∈ (p), cluster core queue φ = φ ∪ (N ∈ (p)∩ψ)-{p} and the core object set ψ=ψ-(N ∈ (p)∩ψ);

[0026] S37, determining whether the current cluster core queue φ is empty;

[0027] If yes, the sample search of the current cluster sample set C is completed, the cluster sample set C and the core object set ψ are updated, a preliminary clustering result is obtained, and the process returns to step S33;

[0028] If not, return to step S36.

[0029] Furthermore, in step S3, in the preliminary clustering results, for clustering results exceeding a set size, a second minpts2 parameter greater than minpts is set to perform a second DBSCAN clustering, thereby obtaining a solar active region distance result.

[0030] Furthermore, the step S4 is specifically as follows:

[0031] S41, setting the integration parameter ratio and classifying the solar active region clustering results;

[0032] The parameter ratio is the ratio of the number of points with positive magnetic field intensity to the number of points with negative magnetic field intensity in each cluster of the solar active region clustering result, or the ratio of the number of points with negative magnetic field intensity to the number of points with positive magnetic field intensity; the types of clusters obtained by the classification include positive clusters, negative clusters and neutral clusters;

[0033] S42, calculating the distance between the closest points of any two clusters on the solar surface;

[0034] S43, defining a distance coefficient between every two clusters, and multiplying the distance coefficient by the distance between the closest points between the clusters on the solar surface, thereby obtaining a weighted distance between different clusters;

[0035] S44, when the weighted distance between different clusters is less than the preset minimum merging distance, the corresponding two clusters are merged, and steps S42 to S44 are repeated until no more clusters can be merged, thereby obtaining an integrated cluster;

[0036] S45. Among the integrated clusters, the neutral clusters whose pixel numbers are lower than the set pixel size threshold and the positive clusters or negative clusters whose pixel numbers are lower than four times the set pixel size threshold are discarded, and the positive and negative polarity regions corresponding to the remaining clusters are used as the finally identified solar active regions.

[0037] Further, in step S43, the distance coefficient of the central cluster is greater than the distance coefficients of the positive cluster and the negative cluster.

[0038] The beneficial effects of the present invention are:

[0039] 1. Improve detection accuracy:

[0040] Effective separation of complex magnetic field regions: Traditional image processing methods and some supervised learning algorithms are not effective in separating complex magnetic field regions in high-resolution images, and often result in misjudgment and missed judgment. The DBSCAN algorithm used in the present invention can effectively process and separate complex magnetic field regions, reducing misjudgment and missed judgment. For example, the DBSCAN algorithm can accurately divide the overly large clusters after preliminary clustering into smaller clusters by performing secondary clustering on the overly large clusters, thus avoiding the problem of incomplete recognition results caused by overly large clusters in traditional methods.

[0041] Sensitivity to small and weak magnetic field strength areas and diffuse magnetic field areas: The DBSCAN algorithm has a high sensitivity in detecting small or scattered active areas, and can identify small or diffuse ARs that other algorithms or manual detection methods may miss. By adjusting the ∈ and minpts parameters, the DBSCAN algorithm can identify areas with small magnetic field strength and classify them as ARs.

[0042] 2. Improve detection efficiency

[0043] Reduce the reliance on empirical thresholds: Traditional image processing methods usually rely on empirical threshold settings, which require a lot of experiments and adjustments. However, the present invention uses the DBSCAN algorithm for clustering, which does not require pre-setting thresholds, reduces manual intervention, and improves detection efficiency. After threshold segmentation, two DBSCAN clustering processes can quickly filter out noise points and form reasonable clusters.

[0044] Rapid processing of large-scale data: When processing large-scale observation data, traditional methods are insufficient in terms of speed and consistency. The DBSCAN algorithm adopted by the present invention can quickly and effectively process a large amount of observation image data every day by preprocessing the image and performing two clustering processes.

[0045] 3. Improve robustness

[0046] Robustness to noise points and outliers: The DBSCAN algorithm is highly robust to noise points and outliers and can adapt to data under different observation conditions. Through two clustering processes, the first clustering filter removes noise points after threshold segmentation, and the second clustering process segments over-large clusters and integrates clusters of reasonable size, and finally obtains stable AR detection results.

[0047] Consistency and objectivity: The method of the present invention avoids the subjectivity of manual inspection and can objectively and consistently extract AR through automated detection methods, ensuring the consistency of long-term data statistical research and modeling solar generator models. Compared with the NOAA catalog, the number of ARs detected by the present invention has a high correlation in annual change trends and can identify more ARs, especially in smaller and diffuse areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flow chart of the method for automatic detection of solar active regions based on a density clustering algorithm provided by the present invention.

[0049] Figure 2 A schematic diagram of distance coefficient setting provided by the present invention.

[0050] Figure 3 A schematic diagram of the process from image preprocessing to final clustering integration provided by the present invention.

[0051] Figure 4 A schematic diagram showing a comparison between the method provided by the present invention and a classical image processing method in detecting solar active regions.

[0052] Figure 5 A schematic diagram comparing the annual changes in the number of AR detected by the method provided by the present invention and the NOAA catalog. DETAILED DESCRIPTION

[0053] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0054] The embodiment of the present invention provides a method for automatically detecting solar active regions based on a density clustering algorithm. Figure 1 As shown, the following steps are included:

[0055] S1, obtain the line-of-sight magnetic map;

[0056] S2, preprocessing the line-of-sight magnetic map;

[0057] S3, performing two consecutive selective DBSCAN clustering on the preprocessed line-of-sight magnetogram to obtain the clustering results of the solar active regions;

[0058] S4. Integrate and filter the solar active region clustering results to obtain the final solar active region detection results.

[0059] In an example of an embodiment of the present invention, in step S1, the line-of-sight magnetic map obtained by the Helioseismic and Magnetic Imager (HMI) on the Solar Dynamics Observatory (SDO) is used to select one image per day, covering the period from May 2010 to December 2023.

[0060] In step S2 of the embodiment of the present invention, the preprocessing is: performing grayscale threshold segmentation on the viewing direction magnetic map to obtain pixels with significant magnetic field intensity as input data points for DBSCAN clustering.

[0061] In an example of this embodiment, the grayscale threshold is set to 150 Gauss to capture pixels with significant magnetic field strength. In the image after threshold segmentation, white and black patches can be used to identify positive and negative magnetic field regions, respectively, which are expressed as:

[0062] I threshold ={p∈I|B(p)>150Gauss}

[0063] Where I represents the original image, and B(p) represents the magnetic field intensity at pixel point p.

[0064] On this basis, the solar disk still shows a wide distribution of noise, which affects the identification of solar active regions.

[0065] In step S3 of the embodiment of the present invention, the input data points are clustered using the DBSCAN clustering method, which is used to identify clusters of arbitrary shapes in spatial data. Unlike other clustering methods that require pre-specified number of clusters, the DBSCAN clustering method classifies points according to the density of the points, thereby being able to effectively handle noise and outliers. Therefore, it can classify magnetic field areas into different active zones without declaring the number of active zones in advance.

[0066] Specifically, in step S3 of this embodiment, the method for performing the first DBSCAN clustering is specifically as follows:

[0067] S31, calculate the ∈ neighborhood N of each input data point ∈ (p);

[0068] N ∈ (p)={q∈D|dist(p,q)≤∈}

[0069] In the formula, q represents the pixel point that meets the conditions, D represents the pixel point that is screened out after threshold segmentation, dist(p,q) represents the distance between pixel points p and q, and ∈ represents the selected distance threshold;

[0070] S32, identifying core points according to the calculated ∈ neighborhood;

[0071] For any pixel point p, it is determined whether |N∈(p)|≥minpts holds; wherein minpts represents the minimum number of points required to form a dense pattern;

[0072] If yes, the pixel point p is identified as the core point p, and the core point p is added to the core object set ψ = ψ∪{p};

[0073] If not, no operation is performed;

[0074] Optionally, the following formula may be used when identifying core points:

[0075]

[0076] In the formula, Core point(p) A Boolean variable used to determine whether point p is a core point. True means that point p is indeed a core point, and False means that point p is not a core point.

[0077] S33, determining whether the core object set ψ is empty;

[0078] If yes, then end the first DBSCAN clustering;

[0079] If not, proceed to step S35;

[0080] S34, randomly select a core point p from the core object set ψ, and update the current cluster core queue φ={p};

[0081] S35, according to the updated cluster core queue φ = {p}, update the cluster sample set C = C∪{p} and the unvisited sample set D = D-{p};

[0082] S36. Select a core point in the current cluster core queue φ, find its corresponding subset N∈(p), and then update the current cluster sample set C=C∪N ∈ (p), cluster core queue φ = φ ∪ (N ∈ (p)∩ψ)-{p} and the core object set ψ=ψ-(N ∈ (p)∩ψ);

[0083] S37, determining whether the current cluster core queue φ is empty;

[0084] If yes, the sample search of the current cluster sample set C is completed, the cluster sample set C and the core object set ψ are updated, a preliminary clustering result is obtained, and the process returns to step S33;

[0085] If not, return to step S36.

[0086] The purpose of the first DBSCAN clustering is to form a preliminary clustering result by filtering out noise points.

[0087] In step S3 of this embodiment, in the preliminary clustering results, for the clustering results exceeding the set size threshold, a second minpts2 parameter greater than minpts is set to perform a second DBSCAN clustering, thereby obtaining the solar active region distance result.

[0088] Specifically, in the second DBSCAN clustering, a larger minpts parameter is set to perform more accurate region division for clustering results that are too large.

[0089] For example, taking the view-direction magnetic map obtained in step S1 as an example, in the second DBSCAN clustering, the size threshold is set to 3000 Mm 2 When a cluster exceeds this size, re-cluster with a larger minpts parameter to split into smaller accurate clusters.

[0090] In step S4 of the embodiment of the present invention, after the above steps are processed, relatively scattered clusters are obtained. On this basis, some relatively small clusters and / or only unipolar regions need to be integrated into a complete active region as the final solar active region identification result.

[0091] In this embodiment, step S4 is specifically as follows:

[0092] S41, setting the integration parameter ratio and classifying the solar active region clustering results;

[0093] The parameter ratio is the ratio of the number of points with positive magnetic field intensity to the number of points with negative magnetic field intensity in each cluster of the solar active region clustering result, or the ratio of the number of points with negative magnetic field intensity to the number of points with positive magnetic field intensity; the types of clusters obtained by the classification include positive clusters, negative clusters and neutral clusters;

[0094] For example, in this embodiment, the ratio parameter is set to 10 by the recall rate of the detected active area relative to NOAA; further, the number of points with positive magnetic field strength in the cluster exceeds the number of points with negative magnetic field strength by a ratio factor, and it is defined as a "positive cluster"; similarly, a "negative cluster" can be defined. For clusters that do not belong to these two categories, they are called the third category "neutral cluster";

[0095] S42, calculating the distance between the closest points of any two clusters on the solar surface;

[0096] Exemplarily, in this embodiment, by finding the closest point between two clusters on the image, and then calculating their distance on the sun surface as the distance value;

[0097] S43, defining a distance coefficient between every two clusters, and multiplying the distance coefficient by the distance between the closest points between the clusters on the solar surface, thereby obtaining a weighted distance between different clusters;

[0098] In this embodiment, the distance coefficient of the central cluster is greater than the distance coefficients of the positive cluster and the negative cluster.

[0099] For example, in this embodiment, an example of setting the distance coefficient is given, such as Figure 2 shown.

[0100] S44, when the weighted distance between different clusters is less than the preset minimum merging distance, the corresponding two clusters are merged, and steps S42 to S44 are repeated until no more clusters can be merged, thereby obtaining an integrated cluster;

[0101] Exemplarily, taking the line-of-sight magnetic map obtained in step S1 as an example, the preset minimum merging distance is 50 mm.

[0102] S45. Among the integrated clusters, the neutral clusters whose pixel numbers are lower than the set pixel size threshold and the positive clusters or negative clusters whose pixel numbers are lower than four times the set pixel size threshold are discarded, and the positive and negative polarity regions corresponding to the remaining clusters are used as the finally identified solar active regions.

[0103] In this embodiment, the pixel threshold of the size of the positive or negative cluster is set to four times that of the neutral cluster because these clusters show unipolar regions. In fact, these regions are usually decaying active regions, in which one polarity region is relatively dispersed or weakened, and then clusters of appropriate size are obtained, and the positive and negative polarity regions are reasonably matched. These clusters are determined as the final identified solar active regions.

[0104] For example, taking the visual magnetic map obtained in step S1 as an example, the size pixel threshold is set to 75 mm 2 .

[0105] In an example of this embodiment, taking the line-of-sight magnetic map obtained in step S1 as an example, Figure 3 The figure shows a schematic diagram of the process from preprocessing to obtaining the final clustering result based on the above steps, including the original image (a), the threshold segmentation result (b), the preliminary clustering result (c), the secondary clustering result (d) and the final detection result (e); specifically, from Figure 3 In (c), a relatively large cluster can be seen on the northwest side, which actually contains several active areas; Figure 3 The relatively large cluster on the northwest side of the solar disk in (c) has been split into several small clusters; Figure 3 (d) It can be seen that some clusters are relatively small and / or have only unipolar regions, which need to be integrated into a complete active region; Figure 3 (e) is the final solar active region identification result.

[0106] Further, Figure 3 An example image of solar active regions identified by NOAA on the same day is shown in (f) for comparison. In addition to NOAA AR 13140, 13141, and 13145, our method detected five more active regions, labeled 1 to 5. NOAA AR 13142, 13143, and 13144 are weak and scattered, and are therefore discarded by our method. Figure 3 (e) and Figure 3 In (f), the ARs detected by the DBSCAN algorithm (labeled 1, 3, 4, and 5) are not marked in the NOAA catalog, but these regions do exist in the solar magnetogram, indicating the advantage of this method in detecting small and diffuse active regions.

[0107] In one example of this embodiment, Figure 4The comparison between the detection effect of this method and classical image processing in solar active region is shown. It can be seen that the DBSCAN algorithm in this method can better segment adjacent complex AR regions, while the classical image processing method, due to algorithm limitations, often incorrectly connects different AR regions together, or causes the detected AR area to be smaller through morphological operations.

[0108] In one example of this embodiment, Figure 5 The figure shows the annual changes of the number of ARs detected by the method of the present invention and the NOAA catalog from 2010 to 2023 when the method of the present invention is used to detect solar active regions. In this process, 4984 full-disk magnetograms were processed by the method and 25178 ARs were detected. To remove duplication, we used a ±6° range constraint close to the central meridian of the solar disk, and finally detected 2863 ARs. The comparison with the annotated active regions in the same area of ​​NOAA proves the efficiency and consistency of the method in processing large-scale data.

[0109] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0110] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. An automatic detection method for solar active regions based on density clustering algorithm, It is characterized in that The following steps are involved: S1, obtain the line-of-sight magnetic map; S2, preprocessing the line-of-sight magnetic map; S3, performing two selective DBSCAN clustering operations on the preprocessed line-of-sight magnetogram to obtain the clustering results of the solar active regions; S4. Integrate and filter the solar active region clustering results to obtain the final solar active region detection results.

2. The method for automatically detecting solar active regions based on density clustering algorithm according to claim 1, It is characterized in that In the step S2, the preprocessing is: performing image grayscale threshold segmentation on the viewing magnetic map to obtain pixels with significant magnetic field intensity as input data points for DBSCAN clustering.

3. The method for automatically detecting solar active regions based on density clustering algorithm according to claim 2, It is characterized in that In step S3, the method for performing the first DBSCAN clustering is specifically as follows: S31, calculate the ∈ neighborhood N of each input data point ∈ (p); N ∈ (p)={q∈D∣dist(p,q)≤∈} In the formula, q represents the pixels that meet the conditions, and D represents the pixels that are screened out after threshold segmentation. dist(p,q) represents the distance between pixels p and q, ∈ represents the selected distance threshold; S32, identifying core points according to the calculated ∈ neighborhood; For any pixel point p, it is determined whether |N∈(p)|≥minpts holds; wherein minpts represents the minimum number of points required to form a dense pattern; If yes, the pixel point p is identified as the core point p, and the core point p is added to the core object set ψ = ψ∪{p}; If not, no operation is performed; S33, determining whether the core object set ψ is empty; If yes, then end the first DBSCAN clustering; If not, proceed to step S35; S34, randomly select a core point p from the core object set ψ, and update the current cluster core queue φ={p}; S35, according to the updated cluster core queue φ = {p}, update the cluster sample set C = C∪{p} and the unvisited sample set D = D-{p}; S36. Select a core point in the current cluster core queue φ, find its corresponding subset N∈(p), and then update the current cluster sample set C=C∪N ∈ (p), cluster core queue φ = φ ∪ (N ∈ (p)∩ψ)-{p} and the core object set ψ=ψ-(N ∈ (p)∩ψ); S37, determining whether the current cluster core queue φ is empty; If yes, the sample search of the current cluster sample set C is completed, the cluster sample set C and the core object set ψ are updated, a preliminary clustering result is obtained, and the process returns to step S33; If not, return to step S36.

4. The method for automatically detecting solar active regions based on density clustering algorithm according to claim 3, It is characterized in that In step S3, in the preliminary clustering results, for clustering results exceeding the set size, a second minpts2 parameter greater than minpts is set to perform a second DBSCAN clustering, thereby obtaining the solar active region distance result.

5. The method for automatically detecting solar active regions based on density clustering algorithm according to claim 1, It is characterized in that The step S4 is specifically as follows: S41, setting the integration parameter ratio and classifying the solar active region clustering results; The parameter ratio is the ratio of the number of points with positive magnetic field intensity to the number of points with negative magnetic field intensity in each cluster of the solar active region clustering result, or the ratio of the number of points with negative magnetic field intensity to the number of points with positive magnetic field intensity; the types of clusters obtained by the classification include positive clusters, negative clusters and neutral clusters; S42, calculating the distance between the closest points of any two clusters on the solar surface; S43, defining a distance coefficient between every two clusters, and multiplying the distance coefficient by the distance between the closest points of the two clusters on the solar surface, thereby obtaining a weighted distance between different clusters; S44, when the weighted distance between different clusters is less than the preset minimum merging distance, the corresponding two clusters are merged, and steps S42 to S44 are repeated until no more clusters can be merged, thereby obtaining an integrated cluster; S45. Among the integrated clusters, the neutral clusters whose pixel numbers are lower than the set pixel size threshold and the positive clusters or negative clusters whose pixel numbers are lower than four times the set pixel size threshold are discarded, and the positive and negative polarity regions corresponding to the remaining clusters are used as the finally identified solar active regions.

6. The method for automatically detecting solar active regions based on density clustering algorithm according to claim 5, It is characterized in that In the step S43, the distance coefficient of the central cluster is greater than the distance coefficients of the positive cluster and the negative cluster.

Images