A satellite cloud image classification method and classification system

Abstract

The present invention provides a classification method and classification system for satellite cloud images, the classification method comprising: obtaining spectral features and texture features of satellite cloud images according to historical satellite observation data and classification data of similar satellites, taking the two as original features of the image, and generating data matrices for a training set and a test set; using the data matrix to train a satellite cloud image classification model on the training set; using the trained satellite cloud image classification model to encode the data on the test set to obtain encoding coefficients; and inputting the encoding coefficients into an SVM multi-classifier for classification. The satellite cloud image classification method of the present invention constructs a model based on historical satellite observation data, classifies and annotates real-time satellite observation data, and overcomes the shortcomings of the prior art.

Description

Technical Field

[0001] The present invention relates to the field of satellite cloud image processing, and in particular to a classification method and a classification system for satellite cloud images. Background Art

[0002] The formation of clouds is inseparable from the complex water cycle system on the earth. Different formation methods and conditions make the shapes of clouds ever-changing. They are divided into different categories and are located at different altitudes in the atmosphere. The condition of clouds is closely related to the daily weather forecasts made by meteorological personnel and the expectations of major disastrous weather. It has a great impact on the aviation field, affecting people's lives and the national economy. In recent years, the meteorological department has attached great importance to the prediction and identification of clouds in order to prevent and reduce disasters and reduce economic losses.

[0003] Meteorological satellites allow people to use satellite cloud images to obtain a wider range of time and space information. Through continuous and large-scale observation of the earth's surface and clouds, the satellite cloud images produced contain very rich meteorological information. The thermal and dynamic processes in the atmosphere are comprehensively displayed through the early evolution of meteorological information. Satellite cloud images enable meteorological departments to obtain a large amount of detailed meteorological information, which facilitates the meteorological department to analyze the changes in weather phenomena. However, the current application of satellite data is still far from enough, and countries are constantly improving the observation capabilities of their own satellites.

[0004] At present, the detection of cloud types in satellite cloud images is mainly based on manual identification, which is highly subjective and cannot guarantee the reliability and uniqueness of the results; the work of meteorological personnel is heavy, but there is a large shortage of relevant professional talents; and the satellite cloud images are developing rapidly, covering a wide range and in large quantities, while the number of aviation meteorological personnel is limited and the manual energy is insufficient, so the detection is time-consuming and laborious, and the efficiency is extremely low. In addition, aviation safety flight is completed by the collaboration of multiple departments. If aviation meteorological personnel cannot quickly and accurately detect the specific cloud shape of satellite cloud images, the subsequent work of predicting weather phenomena will be difficult to carry out, and it will be impossible to communicate weather conditions with staff from other departments in a timely manner, and it will not be possible to ensure that each department obtains the latest meteorological information in a timely manner, which is not conducive to the implementation of subsequent work.

[0005] In view of this, the present invention is proposed. Summary of the invention

[0006] In view of this, the present invention discloses a classification method and a classification system for satellite cloud images, which utilizes the image classification capability of discriminative dictionary learning, builds a model based on historical satellite observation data, and classifies clouds for satellite observation data, making it closer to the actual situation, reducing manpower consumption, and realizing cloud classification in a larger observation area.

[0007] Specifically, the present invention is achieved through the following technical solutions:

[0008] In a first aspect, the present invention discloses a satellite cloud image classification method, comprising the following steps:

[0009] According to historical satellite observation data and classification data of similar satellites, the spectral characteristics and texture characteristics of satellite cloud images are obtained, and the two are used as the original features of the image to generate data matrices for the training set and the test set;

[0010] Using the data matrix to train the satellite cloud image classification model on the training set;

[0011] Using the trained satellite cloud image classification model to encode the data on the test set to obtain encoding coefficients;

[0012] The coding coefficients are input into a SVM multi-classifier for classification.

[0013] In a second aspect, the present invention discloses a classification system for satellite cloud images, comprising:

[0014] Generation module: used to obtain the spectral characteristics and texture characteristics of satellite cloud images based on historical satellite observation data and classification data of similar satellites, and use the two as the original features of the image to generate data matrices for training sets and test sets;

[0015] Training module: used for training the satellite cloud image classification model on the training set using the data matrix;

[0016] Test encoding module: used to encode the data on the test set using the trained satellite cloud image classification model to obtain encoding coefficients;

[0017] Classification module: used for inputting the coding coefficients into the SVM multi-classifier for classification.

[0018] In a third aspect, the present invention discloses a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the classification method described in the first aspect.

[0019] In a fourth aspect, the present invention discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the classification method described in the first aspect when executing the program.

[0020] The satellite cloud image classification method and classification system proposed in the present invention utilize computer technology and machine learning theory to process extremely large satellite cloud image data, and the machine learning algorithm can fully extract the rich weather information contained in the cloud image to identify the type of cloud, thereby realizing fast and accurate automatic classification of satellite cloud images, which greatly reduces the pressure on meteorological staff.

[0021] In summary, by utilizing the image classification capabilities of discriminative dictionary learning and building a model based on historical satellite observation data, cloud classification is performed on satellite observation data, making it closer to the actual situation, reducing manpower consumption, and achieving cloud classification in a larger observation area. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0023] Figure 1 A schematic diagram of a flow chart of a satellite cloud image classification method provided by an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of obtaining texture features of satellite cloud images provided by an embodiment of the present invention;

[0025] Figure 3 A structural diagram of a satellite cloud image classification system provided by an embodiment of the present invention;

[0026] Figure 4 A flowchart of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0028] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0029] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0030] The present invention discloses a classification method for satellite cloud images, comprising the following steps:

[0031] According to historical satellite observation data and classification data of similar satellites, the spectral characteristics and texture characteristics of satellite cloud images are obtained, and the two are used as the original features of the image to generate data matrices for the training set and the test set;

[0032] Using the data matrix to train the satellite cloud image classification model on the training set;

[0033] Using the trained satellite cloud image classification model to encode the data on the test set to obtain encoding coefficients;

[0034] The coding coefficients are input into a SVM multi-classifier for classification.

[0035] like Figure 1 As shown, the present invention discloses a method step flow chart of satellite cloud image cloud classification based on discriminant dictionary learning. In this embodiment, the satellite cloud image cloud classification method based on discriminant dictionary learning specifically includes the following steps:

[0036] Step 1: Based on historical satellite observation data and similar satellite classification data, select about 50% of each category as training samples and the remaining 50% of the data as test samples to construct training data sets and test sets:

[0037] The image is intercepted from the original image after satellite positioning, and the satellite cloud atlas T = {T1, T2, ..., T n}, construct the matching label data L = {L1, L2, ..., L n In the data set T, about 50% of each class is selected as training samples, and the remaining 50% of the data is used as test samples.

[0038] Step 2, obtain the spectral characteristics and texture characteristics of the satellite cloud image, use the two as the original characteristics of the image, and normalize the elements of the original characteristics of the image to [0, 1], including:

[0039] (2a) Spectral features are extracted from the training set and the test set. For the five channels of the sample, nine spectral features are extracted for each channel, which are:

[0040] ①Minimum grayscale value

[0041] ② Minimum brightness or albedo

[0042] ③ Maximum grayscale value

[0043] ④ Maximum value of brightness or albedo

[0044] ⑤ Grayscale mean

[0045] ⑥ Mean brightness temperature or albedo

[0046] ⑦ Grayscale corresponding to the maximum value of histogram frequency

[0047] ⑧ Brightness temperature or albedo corresponding to the maximum value of histogram frequency

[0048] ⑨Standard deviation

[0049] (2b) The original grayscale level (0-225) of the image is used for texture feature extraction. The algorithm based on the fusion of Gabor transform and gray-level co-occurrence matrix is ​​used to complete the texture feature extraction of the cloud. The principle is as follows: Figure 2 As shown, take any point (x, y) in the image (N x N) and another point (x+D x，y +D y ), let the gray value of the point pair be (i, J). If the point (x, y) moves across the entire screen, different (i, J) values ​​will be obtained. Let the number of gray values ​​be L, then there are L combinations of (i, J) in total. 2 In the gray-level co-occurrence matrix, for the entire image, the number of occurrences of each value of (i, J) is counted, and then arranged into a square matrix. The total number of occurrences of (i, J) is used to normalize them to the probability of occurrence P(i, j), which is the gray-level co-occurrence matrix. The parameters of the gray-level co-occurrence matrix are θ and δ. θ reflects the direction of extracting texture features, and takes four directions of 0°, 45°, 90°, and 135°; δ is the distance between the two image pairs scanned, and its value determines the coarseness of the texture. For images with coarse textures, δ takes a larger value, and for images with fine textures, δ takes a smaller value.

[0050] For each sample's 5 channels, the pixel distance of each channel is δ=1, and the gray-level co-occurrence matrix is ​​calculated at four directions: 0°, 45°, 90°, and 135°. After regularization, four gray-level difference vector features are calculated, which are:

[0051] ① Contrast, describing the clarity of cloud texture K1:

[0052]

[0053] ②Correlation, which reflects the linear correlation degree K2 between the rows and columns of the gray-level co-occurrence matrix:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] ③ Entropy, measuring the amount of information K3 of the cloud image:

[0060]

[0061] ④ Inverse moment, this value is inversely proportional to the change in the texture of the cloud image in different regions, reflecting the homogeneity K4:

[0062]

[0063] In summary, 45 spectral features and 80 grayscale difference features were extracted, totaling 125 features. In the training set and test set, 145-dimensional features were extracted for all images, and these features were output as column vectors of the data set. The feature vectors were normalized to generate data matrices X and X for the training set and test set. For detailed principles, see Figure 2 shown.

[0064] Step 3, using the training set X generated by the texture feature values ​​and spectral feature values ​​of the cloud obtained in step 2, the training samples are used to train a discriminant dictionary learning model, and the discriminant dictionary learning model is as follows:

[0065]

[0066]

[0067]

[0068] Where λ1, λ2, λ3 are fixed scalars, and X = [X1, X2, ..., X k ] is the feature of the image, k represents the number of samples, D = [D1, D2, ..., D k ] represents a synthetic dictionary, P = [P1; P2, ...; P k ] represents the parsing dictionary, A = [A1, A2, ..., A k] represents the encoding matrix, the synthetic dictionary D1 and the parsing dictionary P1 represent the dictionary pair of the input data X1, the synthetic dictionary D2 and the parsing dictionary P2 represent the dictionary pair of the input data X2, Represents x in X i The complement of . i represents the i-th atom of D, represents the data fidelity term, and Indicates a discriminant. It is used to limit the energy of dictionary atoms of the comprehensive dictionary to avoid trivial solutions P k =0, making the model more stable. k ∈Γ, is a constraint on the parsing dictionary, which ensures that the parsing dictionary has a relatively small F-norm.

[0069] Where, the structural parsing sub-dictionary P k The ability to project samples from the i-th class (i≠k) into the null space enhances the discriminability of the structured dictionary P, resulting in a block diagonal structure for PX.

[0070]

[0071] Step 4, solve the dictionary model, and encode the test data on the test set using the discriminant dictionary obtained in step 3, including:

[0072] (4a) When solving the dictionary model, a cyclic alternating iteration method is used, and the iteration termination condition is the convergence of the objective function or reaching a fixed number of iterations. In each iteration, the variables P, A, and D are optimized alternately.

[0073] ①Fix P and D, update A

[0074] ②Fix P and A, update D

[0075] ③Fix D and A, update P

[0076] (4b) Use the optimized parsing dictionary P to encode the real-time satellite cloud image and obtain the encoding coefficient

[0077] Step 5: Set the projection coefficient Substitute it into the SVM classifier obtained in step 3 and determine the category of the test sample according to the following formula. The category it belongs to is the category corresponding to the position of the largest element in z, as follows:

[0078]

[0079] Where j = 1…n, there are n test samples in total, z j for The j-th row vector of corresponds to the classification result of the j-th test data. i is the vector z j The i-th element of , k is the number of categories.

[0080] In summary, the method for cloud classification of satellite cloud images based on discriminative dictionary learning provided by the embodiment of the present invention utilizes the image classification capability of discriminative dictionary learning, builds a model based on historical satellite observation data, and classifies clouds on the satellite observation data, making it closer to the actual situation, reducing manpower consumption, and realizing cloud classification in a larger observation area.

[0081] like Figure 3 As shown, the present invention provides a classification system for satellite cloud images, comprising:

[0082] Generating module 101: used to obtain the spectral characteristics and texture characteristics of the satellite cloud image based on historical satellite observation data and classification data of similar satellites, and use the two as original features of the image to generate data matrices for the training set and the test set;

[0083] Training module 102: used to train the satellite cloud image classification model on the training set using the data matrix;

[0084] Test encoding module 103: used to encode the data on the test set using the trained satellite cloud image classification model to obtain encoding coefficients;

[0085] Classification module 104: used for inputting the coding coefficients into the SVM multi-classifier for classification.

[0086] The system is mainly composed of the above four modules. Through the construction of this system, cloud classification in the observation area of ​​satellite cloud images can be well realized.

[0087] In specific implementation, the above modules can be implemented as independent entities, or can be arbitrarily combined and implemented as the same or several entities. The specific implementation of the above units can refer to the previous method embodiments, which will not be repeated here.

[0088] Figure 4 This is a schematic diagram of the structure of a computer device disclosed in the present invention. Figure 4 As shown, the computer device 400 includes at least a memory 402 and a processor 401; the memory 402 is connected to the processor via a communication bus 403, and is used to store computer instructions executable by the processor 401, and the processor 301 is used to read computer instructions from the memory 402 to implement the steps of the satellite cloud image classification method described in any of the above embodiments.

[0089] For the above-mentioned device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial description of the method embodiments. The device embodiments described above are only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0090] Computer readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including, for example, semiconductor memory devices (e.g., EPROM, EEPROM and flash memory devices), magnetic disks (e.g., internal disks or removable disks), magneto-optical disks, and CD ROM and DVD-ROM disks. The processor and memory may be supplemented by, or incorporated in, special purpose logic circuits.

[0091] Finally, it should be noted that although this specification contains many specific implementation details, these should not be interpreted as limiting the scope of any invention or the scope of protection claimed, but are mainly used to describe the features of specific embodiments of specific inventions. Certain features described in multiple embodiments within this specification may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may work in certain combinations as described above and even initially claimed as such, one or more features from the claimed combination may be removed from the combination in some cases, and the claimed combination may point to a sub-combination or a variation of a sub-combination.

[0092] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that these operations be performed in the particular order shown or performed sequentially, or requiring that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

[0093] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0094] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the scope of protection of the present disclosure.

Claims

1. A satellite cloud image classification method, characterized in that: The steps include: According to historical satellite observation data and classification data of similar satellites, the spectral characteristics and texture characteristics of satellite cloud images are obtained, and the two are used as the original features of the image to generate data matrices for the training set and the test set; The method of obtaining the spectral characteristics and texture characteristics of the satellite cloud image, taking the two as the original characteristics of the image, and generating the data matrix of the training set and the test set includes the following steps: For the n channels of the sample, spectral features are extracted for each channel; The texture feature extraction is performed using the gray level difference vector based on the gray level co-occurrence matrix and the first-order probability feature algorithm. The specific methods include: For each sample's n channels, the pixel distance of each channel is 1, and the gray level co-occurrence matrix is ​​calculated in four directions of 0°, 45°, 90°, and 135°. After regularization, four gray level difference vector features are calculated, which are: contrast, describing the cloud texture clarity K1; correlation, reflecting the linear correlation between the rows and columns of the gray level co-occurrence matrix K2; entropy, measuring the amount of information of the cloud image K3; inverse moment, which is inversely proportional to the change between different regions of the cloud image texture, reflecting the homogeneity K4; Perform 25n-dimensional feature extraction on all images, output these features as column vectors of the data set, and normalize the feature vectors to generate data matrices x and y for the training set and test set. Using the data matrix to train the satellite cloud image classification model on the training set; The satellite cloud image classification model obtained by training the data matrix on the training set is: In the above formula, λ1, λ2, λ3 are fixed scalars, where X = [X1, X2, ..., X k ] is the feature of the image, k represents the number of samples, D = [D1, D2, ..., D k ] represents a synthetic dictionary, P = [P1; P2, ...; P k ] represents the parsing dictionary, A = [A1, A2, ..., A k ] represents the encoding matrix, the synthetic dictionary D1 and the parsing dictionary P1 represent the dictionary pair of the input data X1, the synthetic dictionary D2 and the parsing dictionary P2 represent the dictionary pair of the input data X2, Represents X in x i The complement of i represents the i-th atom of D, represents the data fidelity term, and Denotes a discriminant; It is used to limit the energy of dictionary atoms of the comprehensive dictionary to avoid trivial solutions P k =0, making the model more stable; is a constraint on the parsing dictionary, which ensures that the parsing dictionary has a relatively small F norm; The structural parsing sub-dictionary P in the above formula k Projecting the samples from the i-th class (i≠k) into the null space enhances the discriminability of the structured dictionary P, resulting in a block diagonal structure for PX; Using the trained satellite cloud image classification model to encode the data on the test set to obtain encoding coefficients; The encoding method comprises: The cyclic alternating iteration method is adopted. The termination condition of the iteration is the convergence of the objective function or reaching a fixed number of iterations. In each iteration, the variables P, A, and D are optimized alternately. Use the optimized parsing dictionary P to encode the real-time satellite cloud image and obtain the encoding coefficient The coding coefficients are input into a SVM multi-classifier for classification.

2. The classification method according to claim 1, characterized in that: The method of constructing the training set includes: The image is intercepted from the original image after satellite positioning, and the satellite cloud atlas T={T1,T2,...,T n }, construct the matching label data L = {L1,L 2, ...,L n }, 50% of the data in each category are selected as training samples.

3. The classification method according to claim 1, characterized in that: The method of inputting the coding coefficient into the SVM multi-classifier for classification includes: The coding coefficient Substitute it into the SVM classifier and determine the category of the test sample according to the following formula. The category it belongs to is the category corresponding to the position of the largest element in z: Where j = 1...n, there are n test samples in total, z j for The j-th row vector of corresponds to the classification result of the j-th test data, z i is the vector z j The i-th element of , k is the number of categories.

4. A classification system using the satellite cloud image classification method according to any one of claims 1 to 3, characterized in that: include: Generation module: used to obtain the spectral characteristics and texture characteristics of satellite cloud images based on historical satellite observation data and classification data of similar satellites, and use the two as the original features of the image to generate data matrices for training sets and test sets; Training module: used for training the satellite cloud image classification model on the training set using the data matrix; Test encoding module: used to encode the data on the test set using the trained satellite cloud image classification model to obtain encoding coefficients; Classification module: used for inputting the coding coefficients into the SVM multi-classifier for classification.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed, the steps of the satellite cloud image classification method described in any one of claims 1 to 3 are implemented.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the satellite cloud image classification method according to any one of claims 1 to 3 are implemented.

Images