A cloud particle image multi-scale aggregation change analysis method

Abstract

The present application relates to a kind of cloud particle image multiscale aggregation change analysis method.The method is first to cloud particle image is carried out grey normalization processing, and extract foreground pixel proportion, local area information entropy and grey distribution variance such as structure feature;According to structure feature, determine the structure complexity of image and grey distribution state, and combined with structure element configuration library adaptive selection target structure element configuration group;Using target structure element configuration group is carried out multiscale morphological processing to cloud particle image, extracts connected region change feature and calculates aggregation change intensity;According to aggregation change intensity, determine the aggregation sensitive scale, obtain cloud particle image multiscale aggregation change feature.The present application can realize the multiscale representation of cloud particle space aggregation structure, improve the accuracy and adaptability of aggregation change analysis.

Description

Technical Field

[0001] This invention belongs to the field of image processing and cloud particle observation data analysis technology, and particularly relates to a multi-scale aggregation change analysis method for cloud particle images based on multi-scale morphological processing and connected region change feature extraction. Background Technology

[0002] Cloud particle images are an important data source for studying cloud microphysical processes. Cloud particle images acquired by airborne cloud particle imaging detection equipment (such as 2D-S probes and CPI probes) can reflect the size, shape, concentration, and spatial distribution of cloud particles, providing important data support for cloud microphysical parameter inversion, precipitation formation mechanism research, and the evaluation of the effectiveness of weather modification operations.

[0003] Currently, the analysis of cloud particle images mainly focuses on particle size distribution statistics, particle concentration calculation, particle morphology classification, and phase identification. Related methods typically analyze individual particles, extracting their geometric, textural, or morphological features to achieve quantitative analysis of cloud particle physical properties. With the continuous increase in the scale of cloud particle observation data, how to further mine the spatial organization and aggregation characteristics of particle populations from a large number of cloud particle images has gradually become an important issue in cloud microphysics research.

[0004] Cloud particles are not completely randomly distributed within clouds; they often exhibit varying degrees of aggregation in different regions and at different developmental stages. This aggregation is not only closely related to microphysical processes such as cloud particle growth, collision, and breakup, but also reflects the evolutionary characteristics of the cloud's internal structure. Therefore, analyzing the aggregation changes in cloud particle images helps reveal the spatial organization patterns of cloud particles at the population scale, providing a basis for cloud structure research and subsequent cloud particle data selection. In existing technologies, when analyzing image structural features using morphological methods, pre-defined fixed structural elements are typically used to process the image. However, different cloud particle images exhibit significant differences in particle density, spatial distribution, structural complexity, and grayscale distribution, making it difficult for fixed structural elements to meet the analytical needs of different types of cloud particle images. When the scale of the structural elements is not chosen appropriately, local aggregation structures may be over-smoothed or fail to be effectively represented, thus affecting the accuracy of the aggregation feature extraction results. Meanwhile, most existing methods focus on the morphological processing results themselves, lacking a systematic analysis of the changing patterns of connected regions under different scale conditions. It is difficult to obtain key scale information that can characterize the spatial aggregation features of cloud particles, and it is also difficult to achieve effective analysis of the multi-scale aggregation and change features of cloud particle images.

[0005] Therefore, in order to effectively characterize the spatial aggregation features of cloud particle images and improve the adaptability and accuracy of the analysis process for different types of cloud particle images, it is necessary to design a multi-scale aggregation change analysis method that can adaptively select structural elements based on the structural features of cloud particle images. By constructing the correspondence between connected regions between multi-scale morphological processing results, aggregation change features under different scale conditions are extracted, and the aggregation-sensitive scale and aggregation change features of cloud particle images are further determined. This solves the problems of existing methods, such as fixed structural elements, insufficient scale adaptability, and difficulty in effectively characterizing the spatial aggregation change laws of cloud particles. Summary of the Invention

[0006] To address the shortcomings of the existing technology, this invention proposes a multi-scale aggregation change analysis method for cloud particle images, comprising the following steps: Step 1: Perform grayscale normalization on the original cloud particle image to obtain a grayscale normalized image; Step 2: Extract structural features from the grayscale normalized image. Structural features include the proportion of foreground pixels, local region information entropy, and grayscale distribution variance. Step 3: Set the scale and shape of the structural elements, build structural element configuration groups, and build a structural element configuration library based on multiple structural element configuration groups; the scale of the structural elements includes a variety of different sizes, and the shape of the structural elements includes a variety of different shapes; Step 4: Determine the structural complexity and grayscale distribution state of the grayscale normalized image based on structural features; structural complexity includes high complexity, medium complexity, and low complexity; grayscale distribution state includes uniform state and non-uniform state. Step 5: Determine the scale category and expansion characteristics of the structural element configuration group to be used based on the structural complexity and grayscale distribution. Step 6: Select the corresponding structural element configuration group from the structural element configuration library according to the scale category and extension characteristics of the structural element configuration group to be used; Step 7: Use the selected structuring element configuration group to perform morphological processing on the cloud particle image to be processed, and obtain the morphologically processed image dataset; Step 8: Extract connected region change features based on the correspondence between connected regions of images with different morphological processing, and analyze the multi-scale aggregation change of cloud particle images based on the connected region change features.

[0007] Furthermore, the foreground pixel ratio is obtained by statistically analyzing the ratio of the number of foreground pixels to the total number of pixels in the grayscale normalized image; the local region information entropy is obtained by extracting a fixed number of local sampling windows at preset positions in the grayscale normalized image, and the sampling windows include at least the image center region and multiple regions distributed at different positions; the grayscale distribution variance is obtained by statistically calculating the image grayscale values.

[0008] Furthermore, the setting of the scale and shape of structural elements, the construction of structural element configuration groups, and the construction of a structural element configuration library based on multiple structural element configuration groups specifically involve: pre-setting at least N structural elements of different sizes and shapes, and labeling all structural elements with three scale categories: small scale, medium scale, and large scale according to their size; further labeling all structural elements with isotropic expansion characteristics and directional expansion characteristics according to their shape characteristics; combining the structural elements to construct multiple structural element configuration groups, each of which includes the same number of structural elements with the same scale category and expansion characteristics but differing in their size or shape parameters; finally, classifying and organizing all structural element configuration groups, using the scale category and expansion characteristics labeled on the structural elements in the configuration group as the tags for that configuration group, and constructing a structural element configuration library.

[0009] Furthermore, the process of determining the structural complexity and grayscale distribution state of the grayscale normalized image based on structural features includes: when the proportion of foreground pixels and the local region information entropy are both higher than their respective thresholds, it is determined to be high complexity; when the proportion of foreground pixels and the local region information entropy are both lower than their respective thresholds, it is determined to be low complexity; the rest are determined to be medium complexity; when the grayscale distribution variance is lower than a preset corresponding threshold, it is determined to be uniform; when the grayscale distribution variance is higher than the corresponding threshold, it is determined to be non-uniform.

[0010] Furthermore, the process of determining the scale category and expansion characteristics of the structural element configuration group to be adopted based on the structural complexity and grayscale distribution includes: when the structural complexity is high and the grayscale distribution is uniform, the structural element configuration group to be adopted is determined to be a small-scale category with isotropic expansion characteristics; when the structural complexity is high and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is determined to be a small-scale category with directional expansion characteristics; when the structural complexity is medium and the grayscale distribution is uniform, the structural element configuration group to be adopted is determined to be... The structural element configuration group is classified as medium-scale and has uniform directional expansion characteristics. When the structural complexity is medium and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is classified as medium-scale and has directional expansion characteristics. When the structural complexity is low and the grayscale distribution is uniform, the structural element configuration group to be adopted is classified as large-scale and has uniform directional expansion characteristics. When the structural complexity is low and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is classified as large-scale and has directional expansion characteristics.

[0011] Furthermore, the process of selecting a corresponding structural element configuration group from the structural element configuration library based on the scale category and extension characteristics of the structural element configuration group to be used includes: filtering structural element configuration groups with corresponding scale categories from the structural element configuration library based on the scale category of the structural element configuration group to be used; filtering structural element configuration groups with corresponding extension characteristics from the structural element configuration library based on the extension characteristics of the structural element configuration group to be used; when there are multiple structural element configuration groups that simultaneously satisfy both scale category and extension characteristics among the filtered structural element configuration groups, determining the structural element configuration group to be used for the current cloud particle image to be processed based on the structural activity coefficient; when only one structural element configuration group simultaneously satisfies both scale category and extension characteristics, determining that structural element configuration group to be selected; when no structural element configuration group simultaneously satisfies both scale category and extension characteristics, determining the structural element configuration group to be used for the current cloud particle image to be processed based on the structural activity coefficient among the structural element configuration groups that satisfy either the scale category or the extension characteristics.

[0012] Furthermore, when multiple structural element configuration groups that simultaneously satisfy scale category and expansion characteristics exist among the selected structural element configuration groups, the structural element configuration group for the current cloud particle image to be processed is determined based on the structural activity coefficient as follows: the structural activity coefficient is obtained by weighted summation of the normalized foreground pixel ratio and the normalized local region information entropy; when the structural activity coefficient is higher than a preset high activity threshold, a target structural element configuration group whose structural element size parameter span is within a preset low span range is selected; when the structural activity coefficient is lower than a preset low activity threshold, a target structural element configuration group whose structural element size parameter span is within a preset high span range is selected; when the structural activity coefficient is between a preset high activity threshold and a preset low activity threshold, a target structural element configuration group whose structural element size parameter span is within a preset middle span range is selected; the structural element size parameter span is the difference between the largest and smallest structural element sizes within the same structural element configuration group.

[0013] Furthermore, the process of performing morphological processing on the cloud particle image to be processed using the selected structuring element configuration group to obtain a morphologically processed image dataset includes: performing morphological processing on the cloud particle image to be processed using multiple structuring elements in the target structuring element configuration group to obtain multiple morphologically processed images corresponding to different structuring elements; identifying each morphologically processed image according to the size parameters, shape parameters, and morphological processing methods corresponding to each structuring element; and associating and storing the morphologically processed images corresponding to different structuring elements and their corresponding identifiers to construct a morphologically processed image dataset.

[0014] Furthermore, the process of extracting connected region change features based on the correspondence between connected regions of images processed by different morphological methods includes: grouping the morphologically processed images in the morphologically processed image dataset according to the morphological processing method, and sorting the morphologically processed images in each group according to the order of the structural element size parameter from small to large; generating connected regions for the cloud particle image to be processed and each morphologically processed image to obtain the baseline connected region and the processed connected regions under different structural element conditions; calculating the region overlap ratio between the processed connected region and the baseline connected region within a preset local analysis region; when the region overlap ratio is higher than a preset overlap threshold, establishing an overlap mapping relationship between the processed connected region and the corresponding baseline connected region; calculating the mean change in the area of ​​connected regions and the mean change in the number of connected regions of each morphologically processed image relative to the cloud particle image to be processed based on the overlap mapping relationship; determining the intensity of connected region aggregation change under different structural element conditions based on the weighted result of the mean change in the area of ​​connected regions and the mean change in the number of connected regions, and using the intensity of connected region aggregation change as the connected region change feature.

[0015] Furthermore, the process of analyzing the multi-scale aggregation change of cloud particle images based on connected region change features includes: obtaining the aggregation change intensity of connected regions corresponding to different structural elements under the same morphological processing method; arranging the aggregation change intensity of connected regions in ascending order of structural element size parameters; calculating the difference in aggregation change intensity corresponding to adjacent structural elements; determining the structural element size corresponding to the maximum value of the aggregation change intensity or the difference in aggregation change intensity; determining the structural element size as the aggregation sensitive scale of the cloud particle image to be processed, and determining the aggregation dominant scale feature and aggregation change degree feature of the cloud particle image based on the aggregation sensitive scale and the corresponding connected region aggregation change intensity.

[0016] Compared with traditional cloud particle image enhancement methods, this invention has the following advantages, thereby solving the corresponding technical problems: 1. This invention automatically determines the complexity of image structure and grayscale distribution by extracting structural features such as the foreground pixel ratio, local region information entropy, and grayscale distribution variance of cloud particle images. Compared to traditional methods that use fixed parameters to process different types of cloud particle images, this invention can dynamically adjust subsequent analysis strategies based on the image's own characteristics, improving the method's adaptability to different cloud particle distribution scenarios.

[0017] 2. This invention constructs a structuring element configuration library containing different scale categories and extended characteristics, and adaptively selects matching structuring element configuration groups based on image structural features. Compared with traditional single structuring element processing methods, it can select a more suitable analysis scale for cloud particle images with different particle densities, different spatial distribution states, and different aggregation characteristics, thereby improving the ability of morphological processing results to represent actual structural information.

[0018] 3. This invention achieves multi-scale morphological processing of cloud particle images by setting small-scale, medium-scale, and large-scale structural elements and combining isotropic and directional expansion characteristics. This method can not only effectively characterize local fine particle aggregation structures but also reflect large-scale particle cluster features, thereby improving the ability to depict the spatial organization structure of cloud particles.

[0019] 4. This invention establishes structural relationships under different scale conditions based on the correspondence between connected regions between different morphological processing results, through region overlap mapping, and extracts the intensity of connected region aggregation changes by combining changes in the area and quantity of connected regions. Compared with methods that only use single-scale features for analysis, this invention can more accurately reflect the variation law of cloud particle spatial structure under different scale conditions and improve the reliability of aggregation feature extraction.

[0020] 5. This invention constructs a sequence of aggregation change intensity corresponding to different structural element scales, and determines the aggregation-sensitive scale corresponding to when the aggregation change intensity or the increment of aggregation change intensity reaches its maximum value, thereby achieving automatic identification of key aggregation scales in cloud particle images. This method can reveal the main response scales of cloud particle spatial aggregation structure from a multi-scale perspective, providing a new technical means for cloud particle swarm structure analysis. Attached Figure Description

[0021] Figure 1 A flowchart of a method for multi-scale aggregation and change analysis of cloud particle images. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. The method includes the following steps: Step 1: Perform grayscale normalization on the original cloud particle image to obtain a grayscale normalized image; Step 2: Extract structural features from the grayscale normalized image. Structural features include the proportion of foreground pixels, local region information entropy, and grayscale distribution variance. Step 3: Set the scale and shape of the structural elements, build structural element configuration groups, and build a structural element configuration library based on multiple structural element configuration groups; the scale of the structural elements includes a variety of different sizes, and the shape of the structural elements includes a variety of different shapes; Step 4: Determine the structural complexity and grayscale distribution state of the grayscale normalized image based on structural features; structural complexity includes high complexity, medium complexity, and low complexity; grayscale distribution state includes uniform state and non-uniform state. Step 5: Determine the scale category and expansion characteristics of the structural element configuration group to be used based on the structural complexity and grayscale distribution. Step 6: Select the corresponding structural element configuration group from the structural element configuration library according to the scale category and extension characteristics of the structural element configuration group to be used; Step 7: Use the selected structuring element configuration group to perform morphological processing on the cloud particle image to be processed, and obtain the morphologically processed image dataset; Step 8: Extract connected region change features based on the correspondence between connected regions of images with different morphological processing, and analyze the multi-scale aggregation change of cloud particle images based on the connected region change features.

[0023] Furthermore, the foreground pixel ratio is obtained by statistically analyzing the ratio of the number of foreground pixels to the total number of pixels in the grayscale normalized image; the local region information entropy is obtained by extracting a fixed number of local sampling windows at preset positions in the grayscale normalized image, and the sampling windows include at least the image center region and multiple regions distributed at different positions; the grayscale distribution variance is obtained by statistically calculating the image grayscale values.

[0024] Furthermore, the setting of the scale and shape of structural elements, the construction of structural element configuration groups, and the construction of a structural element configuration library based on multiple structural element configuration groups specifically involve: pre-setting at least N structural elements of different sizes and shapes, and labeling all structural elements with three scale categories: small scale, medium scale, and large scale according to their size; further labeling all structural elements with isotropic expansion characteristics and directional expansion characteristics according to their shape characteristics; combining the structural elements to construct multiple structural element configuration groups, each of which includes the same number of structural elements with the same scale category and expansion characteristics but differing in their size or shape parameters; finally, classifying and organizing all structural element configuration groups, using the scale category and expansion characteristics labeled on the structural elements in the configuration group as the tags for that configuration group, and constructing a structural element configuration library.

[0025] Furthermore, the process of determining the structural complexity and grayscale distribution state of the grayscale normalized image based on structural features includes: when the proportion of foreground pixels and the local region information entropy are both higher than their respective thresholds, it is determined to be high complexity; when the proportion of foreground pixels and the local region information entropy are both lower than their respective thresholds, it is determined to be low complexity; the rest are determined to be medium complexity; when the grayscale distribution variance is lower than a preset corresponding threshold, it is determined to be uniform; when the grayscale distribution variance is higher than the corresponding threshold, it is determined to be non-uniform.

[0026] Furthermore, the process of determining the scale category and expansion characteristics of the structural element configuration group to be adopted based on the structural complexity and grayscale distribution includes: when the structural complexity is high and the grayscale distribution is uniform, the structural element configuration group to be adopted is determined to be a small-scale category with isotropic expansion characteristics; when the structural complexity is high and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is determined to be a small-scale category with directional expansion characteristics; when the structural complexity is medium and the grayscale distribution is uniform, the structural element configuration group to be adopted is determined to be... The structural element configuration group is classified as medium-scale and has uniform directional expansion characteristics. When the structural complexity is medium and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is classified as medium-scale and has directional expansion characteristics. When the structural complexity is low and the grayscale distribution is uniform, the structural element configuration group to be adopted is classified as large-scale and has uniform directional expansion characteristics. When the structural complexity is low and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is classified as large-scale and has directional expansion characteristics.

[0027] Furthermore, the process of selecting a corresponding structural element configuration group from the structural element configuration library based on the scale category and extension characteristics of the structural element configuration group to be used includes: filtering structural element configuration groups with corresponding scale categories from the structural element configuration library based on the scale category of the structural element configuration group to be used; filtering structural element configuration groups with corresponding extension characteristics from the structural element configuration library based on the extension characteristics of the structural element configuration group to be used; when there are multiple structural element configuration groups that simultaneously satisfy both scale category and extension characteristics among the filtered structural element configuration groups, determining the structural element configuration group to be used for the current cloud particle image to be processed based on the structural activity coefficient; when only one structural element configuration group simultaneously satisfies both scale category and extension characteristics, determining that structural element configuration group to be selected; when no structural element configuration group simultaneously satisfies both scale category and extension characteristics, determining the structural element configuration group to be used for the current cloud particle image to be processed based on the structural activity coefficient among the structural element configuration groups that satisfy either the scale category or the extension characteristics.

[0028] Furthermore, when multiple structural element configuration groups that simultaneously satisfy scale category and expansion characteristics exist among the selected structural element configuration groups, the structural element configuration group for the current cloud particle image to be processed is determined based on the structural activity coefficient as follows: the structural activity coefficient is obtained by weighted summation of the normalized foreground pixel ratio and the normalized local region information entropy; when the structural activity coefficient is higher than a preset high activity threshold, a target structural element configuration group whose structural element size parameter span is within a preset low span range is selected; when the structural activity coefficient is lower than a preset low activity threshold, a target structural element configuration group whose structural element size parameter span is within a preset high span range is selected; when the structural activity coefficient is between a preset high activity threshold and a preset low activity threshold, a target structural element configuration group whose structural element size parameter span is within a preset middle span range is selected; the structural element size parameter span is the difference between the largest and smallest structural element sizes within the same structural element configuration group.

[0029] Furthermore, the process of performing morphological processing on the cloud particle image to be processed using the selected structuring element configuration group to obtain a morphologically processed image dataset includes: performing morphological processing on the cloud particle image to be processed using multiple structuring elements in the target structuring element configuration group to obtain multiple morphologically processed images corresponding to different structuring elements; identifying each morphologically processed image according to the size parameters, shape parameters, and morphological processing methods corresponding to each structuring element; and associating and storing the morphologically processed images corresponding to different structuring elements and their corresponding identifiers to construct a morphologically processed image dataset.

[0030] Furthermore, the process of extracting connected region change features based on the correspondence between connected regions of images processed by different morphological methods includes: grouping the morphologically processed images in the morphologically processed image dataset according to the morphological processing method, and sorting the morphologically processed images in each group according to the order of the structural element size parameter from small to large; generating connected regions for the cloud particle image to be processed and each morphologically processed image to obtain the baseline connected region and the processed connected regions under different structural element conditions; calculating the region overlap ratio between the processed connected region and the baseline connected region within a preset local analysis region; when the region overlap ratio is higher than a preset overlap threshold, establishing an overlap mapping relationship between the processed connected region and the corresponding baseline connected region; calculating the mean change in the area of ​​connected regions and the mean change in the number of connected regions of each morphologically processed image relative to the cloud particle image to be processed based on the overlap mapping relationship; determining the intensity of connected region aggregation change under different structural element conditions based on the weighted result of the mean change in the area of ​​connected regions and the mean change in the number of connected regions, and using the intensity of connected region aggregation change as the connected region change feature.

[0031] Furthermore, the process of analyzing the multi-scale aggregation change of cloud particle images based on connected region change features includes: obtaining the aggregation change intensity of connected regions corresponding to different structural elements under the same morphological processing method; arranging the aggregation change intensity of connected regions in ascending order of structural element size parameters; calculating the difference in aggregation change intensity corresponding to adjacent structural elements; determining the structural element size corresponding to the maximum value of the aggregation change intensity or the difference in aggregation change intensity; determining the structural element size as the aggregation sensitive scale of the cloud particle image to be processed, and determining the aggregation dominant scale feature and aggregation change degree feature of the cloud particle image based on the aggregation sensitive scale and the corresponding connected region aggregation change intensity.

[0032] Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are all within the scope of protection of the present invention.

Claims

1. A method for multi-scale aggregation and change analysis of cloud particle images, comprising the following steps: Step 1: Perform grayscale normalization on the original cloud particle image to obtain a grayscale normalized image; Step 2: Extract structural features from the grayscale normalized image. Structural features include the proportion of foreground pixels, local region information entropy, and grayscale distribution variance. Step 3: Set the scale and shape of the structural elements, build structural element configuration groups, and build a structural element configuration library based on multiple structural element configuration groups; the scale of the structural elements includes a variety of different sizes, and the shape of the structural elements includes a variety of different shapes; Step 4: Determine the structural complexity and grayscale distribution state of the grayscale normalized image based on structural features; structural complexity includes high complexity, medium complexity, and low complexity; grayscale distribution state includes uniform state and non-uniform state. Step 5: Determine the scale category and expansion characteristics of the structural element configuration group to be used based on the structural complexity and grayscale distribution. Step 6: Select the corresponding structural element configuration group from the structural element configuration library according to the scale category and extension characteristics of the structural element configuration group to be used; Step 7: Use the selected structuring element configuration group to perform morphological processing on the cloud particle image to be processed, and obtain the morphologically processed image dataset; Step 8: Extract connected region change features based on the correspondence between connected regions of images with different morphological processing, and analyze the multi-scale aggregation change of cloud particle images based on the connected region change features.

2. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The foreground pixel ratio is obtained by statistically analyzing the ratio of the number of foreground pixels to the total number of pixels in the grayscale normalized image; the local region information entropy is obtained by extracting a fixed number of local sampling windows at preset positions in the grayscale normalized image, and the sampling windows include at least the image center region and multiple regions distributed at different positions; the grayscale distribution variance is obtained by statistically calculating the image grayscale values.

3. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The process of setting the scale and shape of structural elements, constructing structural element configuration groups, and constructing a structural element configuration library based on multiple structural element configuration groups involves: pre-setting at least N structural elements of different sizes and shapes, and identifying all structural elements into three scale categories: small scale, medium scale, and large scale according to their size. Based on the shape characteristics of the structural elements, all structural elements are further labeled with isotropic and directional expansion characteristics. The structural elements are then combined to construct multiple structural element configuration groups. Each structural element configuration group includes the same number of structural elements with the same scale category and expansion characteristics, but with differences in structural element size parameters or shape parameters. Finally, all structural element configuration groups are classified and organized, and the scale category and expansion characteristics marked by the structural elements in the configuration group are used as the labels for that configuration group to construct a structural element configuration library.

4. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The process of determining the structural complexity and grayscale distribution state of a grayscale normalized image based on structural features includes: when the proportion of foreground pixels and the local region information entropy are both higher than their respective thresholds, it is determined to be high complexity; when the proportion of foreground pixels and the local region information entropy are both lower than their respective thresholds, it is determined to be low complexity; the rest are determined to be medium complexity; when the grayscale distribution variance is lower than a preset corresponding threshold, it is determined to be uniform; when the grayscale distribution variance is higher than the corresponding threshold, it is determined to be non-uniform.

5. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The process of determining the scale category and expansion characteristics of the structural element configuration group to be adopted based on structural complexity and grayscale distribution includes: when the structural complexity is high and the grayscale distribution is uniform, the structural element configuration group to be adopted is determined to be a small-scale category with isotropic expansion characteristics; when the structural complexity is high and the grayscale distribution is non-uniform, the structural element configuration group to be adopted is determined to be a small-scale category with directional expansion characteristics; when the structural complexity is medium and the grayscale distribution is uniform, the structural elements to be adopted are determined to be... The structural element configuration group is determined to be of medium scale and has uniform expansion characteristics. When the structural complexity is medium and the gray-scale distribution is non-uniform, the structural element configuration group to be adopted is determined to be of medium scale and has directional expansion characteristics. When the structural complexity is low and the gray-scale distribution is uniform, the structural element configuration group to be adopted is determined to be of large scale and has uniform expansion characteristics. When the structural complexity is low and the gray-scale distribution is non-uniform, the structural element configuration group to be adopted is determined to be of large scale and has directional expansion characteristics.

6. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The process of selecting a corresponding structural element configuration group from the structural element configuration library based on the scale category and extension characteristics of the structural element configuration group to be used includes: filtering structural element configuration groups with corresponding scale categories from the structural element configuration library based on the scale category of the structural element configuration group to be used; filtering structural element configuration groups with corresponding extension characteristics from the structural element configuration library based on the extension characteristics of the structural element configuration group to be used; when there are multiple structural element configuration groups that simultaneously satisfy both scale category and extension characteristics among the filtered structural element configuration groups, determining the structural element configuration group to be used for the current cloud particle image to be processed based on the structural activity coefficient; when only one structural element configuration group simultaneously satisfies both scale category and extension characteristics, determining that structural element configuration group to be selected; when no structural element configuration group simultaneously satisfies both scale category and extension characteristics, determining the structural element configuration group to be used for the current cloud particle image to be processed based on the structural activity coefficient among the structural element configuration groups that satisfy either scale category or extension characteristics.

7. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 6, characterized in that: When multiple structural element configuration groups that simultaneously satisfy scale category and expansion characteristics exist among the selected structural element configuration groups, the structural element configuration group used for the current cloud particle image to be processed is determined based on the structural activity coefficient as follows: the structural activity coefficient is obtained by weighted summation of the normalized foreground pixel ratio and the normalized local region information entropy; when the structural activity coefficient is higher than a preset high activity threshold, the target structural element configuration group whose structural element size parameter span is within a preset low span range is selected; when the structural activity coefficient is lower than a preset low activity threshold, the target structural element configuration group whose structural element size parameter span is within a preset high span range is selected; when the structural activity coefficient is between the preset high activity threshold and the preset low activity threshold, the target structural element configuration group whose structural element size parameter span is within a preset middle span range is selected. The span of the structural element size parameter is the difference between the largest and smallest structural element size within the same structural element configuration group.

8. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The process of performing morphological processing on the cloud particle image to be processed using the selected structuring element configuration group to obtain a morphologically processed image dataset includes: performing morphological processing on the cloud particle image to be processed using multiple structuring elements in the target structuring element configuration group to obtain multiple morphologically processed images corresponding to different structuring elements; labeling each morphologically processed image according to the size parameters, shape parameters, and morphological processing methods corresponding to each structuring element; and associating and storing the morphologically processed images corresponding to different structuring elements and their corresponding labels to construct a morphologically processed image dataset.

9. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 1, characterized in that: The process of extracting connected region change features based on the correspondence between connected regions of images processed by different morphological methods includes: grouping the morphologically processed images in the morphologically processed image dataset according to the morphological processing method, and sorting the morphologically processed images in each group according to the order of the structural element size parameter from small to large; generating connected regions for the cloud particle image to be processed and each morphologically processed image to obtain the baseline connected region and the processed connected regions under different structural element conditions; calculating the region overlap ratio between the processed connected region and the baseline connected region within a preset local analysis region; when the region overlap ratio is higher than a preset overlap threshold, establishing an overlap mapping relationship between the processed connected region and the corresponding baseline connected region; calculating the mean change in the area of ​​connected regions and the mean change in the number of connected regions of each morphologically processed image relative to the cloud particle image to be processed based on the overlap mapping relationship; determining the intensity of connected region aggregation change under different structural element conditions based on the weighted result of the mean change in the area of ​​connected regions and the mean change in the number of connected regions, and using the intensity of connected region aggregation change as the connected region change feature.

10. The method for multi-scale aggregation and change analysis of cloud particle images as described in claim 9, characterized in that: The process of analyzing the multi-scale aggregation change of cloud particle images based on connected region change features includes: obtaining the aggregation change intensity of connected regions corresponding to different structural elements under the same morphological processing method; arranging the aggregation change intensity of connected regions in ascending order of structural element size parameters; calculating the difference in aggregation change intensity corresponding to adjacent structural elements; determining the structural element size corresponding to the maximum value of the aggregation change intensity or the difference in aggregation change intensity; determining the aggregation sensitive scale of the cloud particle image to be processed, and determining the aggregation dominant scale feature and aggregation change degree feature of the cloud particle image based on the aggregation sensitive scale and the corresponding connected region aggregation change intensity.

Images