CN107230197B - Tropical cyclone objective strength determination method based on satellite cloud image and RVM - Google Patents
Tropical cyclone objective strength determination method based on satellite cloud image and RVM Download PDFInfo
- Publication number
- CN107230197B CN107230197B CN201710406752.1A CN201710406752A CN107230197B CN 107230197 B CN107230197 B CN 107230197B CN 201710406752 A CN201710406752 A CN 201710406752A CN 107230197 B CN107230197 B CN 107230197B
- Authority
- CN
- China
- Prior art keywords
- deviation angle
- objective
- gradient
- strength
- center
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims description 25
- 239000011159 matrix material Substances 0.000 claims abstract description 61
- 230000004927 fusion Effects 0.000 claims abstract description 37
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Chemical compound O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 13
- 238000012360 testing method Methods 0.000 claims description 10
- 238000007499 fusion processing Methods 0.000 claims description 4
- 238000010801 machine learning Methods 0.000 claims 1
- 230000003068 static effect Effects 0.000 abstract 1
- 238000011160 research Methods 0.000 description 8
- 238000002474 experimental method Methods 0.000 description 6
- 238000000354 decomposition reaction Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 235000002911 Salvia sclarea Nutrition 0.000 description 1
- 244000182022 Salvia sclarea Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20016—Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30192—Weather; Meteorology
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Processing (AREA)
- Radiation Pyrometers (AREA)
- Image Analysis (AREA)
Abstract
The invention constructs a Tropical Cyclone (TC) objective strong model based on a static satellite cloud picture and a relevant vector machine (Relevance Vector Machine, RVM). Mainly comprises the following two aspects: (1) First, the cloud images of the infrared and water vapor channels are fused by using the Laplacian pyramid algorithm. And constructing a deviation angle-gradient co-occurrence matrix by taking the TC center as a reference point. The invention utilizes a plurality of statistical parameters in the symbiotic matrix in combination with information such as TC kernel scale, center latitude and the like to construct a characteristic factor closely related to TC intensity, and utilizes RVM to establish a TC objective strength-determining model. (2) Based on the fusion satellite cloud image, each point is sequentially used as a reference point, a deviation angle-gradient co-occurrence matrix is constructed, and the minimum value, the median value and the mean value of the co-occurrence matrix statistical parameter matrix are calculated. The invention utilizes a plurality of statistical parameters in the symbiotic matrix parameter array and combines the information of TC kernel scale, center latitude and the like to construct the characteristic factors closely related to TC intensity, and utilizes RVM to build a TC objective strength-determining model.
Description
Technical Field
The invention belongs to the field of image processing technology and weather prediction. In particular to a tropical cyclone objective strength determination method based on satellite cloud pictures and a related vector machine (Relevance Vector Machine, RVM) aiming at improving the tropical cyclone strength determination precision.
Background
Tropical Cyclone (TC) is a common disaster among many natural disasters endangering china, and its activities are accompanied by storm, storm and storm surge, and even cause natural geological disasters such as landslide and debris flow. Accurate prediction of TC strength and path is critical to prevent and mitigate the disaster effects that it brings.
In recent years, the development of various domestic and foreign observation means and numerical forecasting techniques promotes the continuous improvement of the TC path forecasting level, but the forecasting capability of TC strength is very slow. The main reasons for this are on the one hand the insufficient capacity of TC to fix and on the other hand the unclear mechanisms affecting TC strength, which are in turn closely linked. The relevant study carried out by the scholars shows that: the lower accuracy of TC determination prevents the progress of TC intensity forecasting. In 1973, dvorak proposed the use of infrared satellite cloud patterns to determine TC intensities, and in 1975 formed the satellite cloud pattern based TC set-strength theory. Dvorak improved the TC intensity estimation technique twice in 1977 and 1984. Since 1987, the North Pacific aircraft detection method was disabled due to cost issues. Because of the greater subjectivity of the Dvorak technique, zehr proposed an objective strength determination method that could improve TC strength determination accuracy in 1989. In 1998, velden et al scholars proposed an objective intensity determination scheme based on stationary satellite infrared digital images to estimate TC intensity. The original Dvorak technology often depends on subjective experience in strength determination, and is easy to misjudge the TC cloud shape, so that the accuracy of TC strength determination is reduced. In 2007, olander and Velden improved on the Dvorak technique and proposed a TC objective strengthening method using stationary satellite infrared cloud images.
At present, TC strength determination mainly comprises microwave data based on polar satellites and cloud image data based on stationary satellites. In the aspect of polar orbit satellite microwave data, such as scholars Lu Yi, the application research of TRMM/TMI data in TC intensity estimation is utilized. Neeru Jaiswal combines data mining techniques with data obtained from a QuickScat satellite ocean wind scatterometer to predict TC intensities. Although microwave data can detect TC structures below the cloud top, TC intensity changes and their associated internal convective structure evolution are often not fully captured due to the susceptibility to strong precipitation interference and low time resolution of polar satellites. The time resolution of the stationary satellite data is high, and the stationary satellite data is more applied to TC strength at present. Many students have conducted related research efforts, focusing mainly on the use of visible and infrared channel data intensity, with the use of single channel data, especially infrared channel data, being the most abundant, such as the Dvorak technology. Pinros et al used infrared satellite cloud mapping to study TC structure and intensity changes. Fetanat et al have studied a TC objective strength determination technology based on an infrared cloud picture in 2013, and the method needs to extract the bright temperature information of azimuth angle at a TC center. Kuaff et al scholars in 2015 proposed an improved method of TC intensity estimation based on TC information and infrared satellite cloud images. Scholars such as Zhao put forward a multi-element linear model based on infrared satellite cloud pictures in 2016, estimate TC intensity of North Pacific ocean, and experimental results show that: the model has better intensity fixing precision for strong typhoons, but has larger intensity fixing error for weaker TC. Gholamreza and Abdolah extract azimuth bright temperature data from the TC center, and objectively estimate TC intensity based on a K-class mean algorithm. Miguel et al estimate TC intensities using the bias angle variance and Sigmoid function.
When only an infrared channel is considered, the TC center is reliably positioned only when the TC has a clear eye area or the spiral rain belt is obvious in characteristics, but when the cloud top of the area where the TC is positioned is shielded by the coiled cloud and the structural characteristics cannot be observed, the TC center is difficult to position. In contrast, the infrared channel can clearly provide typhoons and richer cloud system detail information by combining with water vapor channel data, so that accuracy of TC center positioning is improved, and the accuracy of TC center positioning directly influences quality of subsequent TC strength determination. In addition, the water vapor channel data is more clear on the characterization of TC cloud type, water vapor characteristics and the like (which are closely related to TC strength), and particularly can highlight the characteristics of a strong convection area. Recently, few researches are combined with infrared and water vapor channel bright temperature difference data for TC strength setting, and a good strength setting effect is achieved. Such as Olander and Velden, uses infrared in combination with moisture channel data to estimate TC intensity. Zhuge Xiaoyong et al in 2015 proposed a TC intensity estimation method based on a combination of infrared and water vapor images of stationary satellites. However, there is still a great room for improvement in three aspects of the method, firstly, the analysis area of the existing method is fixed as a kernel area with a fixed radius around the TC center, but the dimensions of different TC kernel areas are changed, and the adoption of a uniform kernel dimension tends to affect the fixed precision and the application range. Secondly, the fixed strength technology adopts linear regression to model, and generally, the TC strength and the influence factor are relatively complex nonlinear relations, and the nonlinear relations between the TC strength and the influence factor cannot be well constructed by simply adopting the linear regression method for modeling. Third, in the existing studies, there is little simultaneous influence of center latitude on TC intensity. Related researches show that the latitude position of the TC center has a great influence on the TC intensity, and only the brightness temperature data is considered to lead to a certain error on the TC intensity.
Disclosure of Invention
The invention aims to provide a tropical cyclone objective strength determining method based on satellite cloud pictures and RVMs. Firstly, fusing cloud images of infrared and water vapor channels by using a Laplacian pyramid fusion algorithm, so as to obtain a fused satellite cloud image. And then constructing a deviation angle-gradient co-occurrence matrix by taking the TC center as a reference point. The invention utilizes a plurality of statistical parameters in the symbiotic matrix in combination with information such as TC kernel scale, center latitude and the like to construct a characteristic factor closely related to TC intensity, and utilizes RVM to establish a TC objective strength-determining model. Based on the fusion satellite cloud image, each point is sequentially used as a reference point to construct a deviation angle-gradient co-occurrence matrix parameter array, and the minimum value, the median value and the mean value of the statistical parameter array of the co-occurrence matrix are calculated. The invention utilizes a plurality of statistical parameters (minimum value, median value and mean value) in the symbiotic matrix parameter array and combines the information such as TC kernel and center latitude to construct the characteristic factors closely related to TC intensity, and utilizes RVM to establish a TC objective intensity-determining model.
The invention discloses a tropical cyclone objective strength determining method based on satellite cloud pictures and RVMs, which comprises the following specific steps:
step 5, constructing an optimal characteristic factor capable of representing TC intensity, and testing an optimal TC kernel scale based on the optimal characteristic factor;
step 7, under the optimal kernel scale, constructing a TC objective strength-determining model based on RVM by utilizing characteristic factors and combining with TC center latitude, and carrying out objective estimation on TC strength;
step 8, calculating a bright-warm gradient matrix of the intercepted fusion cloud picture on the basis of the step 2, and calculating to obtain a deviation angle matrix by taking each point as a reference point;
step 11, constructing a TC objective strength-determining model based on RVM by utilizing characteristic factors under the optimal kernel scale, and objectively estimating the TC strength;
1. the cloud image data used by the invention is from the national weather stationary satellite FY-2, and has 5 channels: an infrared 1 channel, an infrared 2 channel, a water vapor channel, an infrared 4 channel and a visible light channel. The fusion processing of the infrared 1-channel satellite cloud image and the water vapor channel cloud image is carried out on the basis of the Laplacian pyramid image fusion algorithm in the step 1, and the steps are as follows:
and 4, carrying out inverse transformation, namely reconstruction, on the image to finally obtain a fused image.
In the step 3, the invention adopts regional gradient to get large to the fusion rule of the top layer image, and other layers of fusion rules adopt regional energy to get large, and the specific method is as follows:
the average gradient represents the sharpness of the image and also reflects the features of minor details and texture variations of the image. First, an average gradient of an m×n size region centered on each pixel in a top-level image is calculated:
wherein I is x Is the first difference in the x direction of pixel f (x, y), and I y The first order difference in the y direction of the pixel f (x, y) is expressed as follows:
ΔI x ＝f(x，y)-f(x-1，y) (2)
ΔI y ＝f(x，y)-f(x，y-1) (3)
thus, G (i, j) is just each pixel L in the top-level image N (i, j) the regional average gradient corresponding to L l Is a first layer image (l is more than or equal to 0 and less than or equal to N) after the Laplacian pyramid decomposition. Then, the top-level image fusion results are:
in LF N (i, j) is the top-level fused image, LA l And LB l The first layer images of the source images A and B after the Laplacian pyramid decomposition are respectively.
For other first layer images (0.ltoreq.l < N) after the Laplacian pyramid decomposition, calculating the regional energy:
where p=1, q=1,
then the other layer image fusion results are:
obtaining a fused image LF of each layer of the Laplacian pyramid l And (l is more than or equal to 0 and less than or equal to N), and finally obtaining a fusion image through reconstruction.
2. And (2) intercepting the fusion cloud pictures under different kernel scales, namely firstly positioning a TC center by using TC annual survey data provided by Shanghai typhoon research institute, and then intercepting the fusion cloud pictures by expanding outwards from the TC center with each 50km in the radial direction within the range of 200km from the center to the TC center.
3. The fused cloud image deviation angle-gradient co-occurrence matrix is constructed in the step 4, and the shape of a developed TC is similar to a circle, namely an axisymmetric graph. For any graph, if it is to be determined whether it is an axisymmetric graph, it can be determined from the gradient direction of a point and the included angle between the gradient direction of the point and the radial line of the reference point, where the included angle between the gradient direction of the point and the radial line is a deviation angle. If the pattern is closer to an axisymmetric pattern, the probability that the deviation angle tends to be 0 ° is greater. TC gradually tends to axisymmetric pattern from early to mature to final death, with no increase in TC intensity, especially at the stage where TC intensity reaches maximum.
In 1984, hong Jiguang proposed a gray-gradient co-occurrence matrix, for a total of 15 statistical parameters. Based on the gray-gradient co-occurrence matrix principle, the invention provides a deviation angle-gradient co-occurrence matrix. The elements of the offset angle-gradient co-occurrence matrix are defined as the total number of pixels that together have an offset angle i and a gradient j in the normalized offset angle matrix and the normalized gradient image. The following deviation angle-gradient co-occurrence matrix was normalized:
TABLE 1 deviation angle-gradient co-occurrence matrix 15 statistical parameters
4. The construction in the step 5 can represent the best characteristic factor of TC intensity, and as 15 statistical parameters are shared in the deviation angle-gradient co-occurrence matrix, the characteristic factor which is most suitable for constructing a TC objective intensity-determining model can be found out through a modeling error test, and the steps are as follows:
5. In the above steps 5-7, based on the 9 feature factors obtained in the experiment in step 4, a TC fixed strength model is respectively built under different kernel scales based on RVMs, the kernel scale with the smallest measured error is 200km finally, as shown in table 2, and based on the 9 feature factors, a TC objective fixed strength model is built based on RVMs in combination with TC center latitude.
TABLE 2 fixed strength errors for different kernel radial dimensions with TC center as reference point
6. In the above steps 8-12, based on the fused satellite cloud image, each point is sequentially used as a reference point, a deviation angle array is calculated, so as to construct a deviation angle-gradient co-occurrence matrix statistical parameter array, the minimum value, the median value and the mean value of each statistical parameter array are calculated, then the minimum value, the median value and the mean value of the parameter T6 are respectively combined with the central wind speed to construct a strong model based on RVM, and the error result is shown in Table 3. When the mean value of the co-occurrence matrix parameter array is used as a modeling characteristic factor, the average absolute error and average error of the TC intensity are minimum, so that the mean value of the co-occurrence matrix parameter array is more suitable for constructing a TC objective intensity model. Then, constructing TC fixed strength models under different kernel scales based on RVM respectively, and finally measuring that the kernel scale with the minimum error is 200km as shown in Table 4, and constructing TC objective fixed strength models based on RVM by combining TC center latitude on the basis of 9 characteristic factors.
TABLE 3 error results of the statistical parameter matrix T6 with minimum, median and mean values, respectively, as TC intensity characteristic factors
TABLE 4 fixed strength errors for different radial kernel dimensions with each point as a reference point
Drawings
Figure 1 is based on a satellite cloud image and a TC objective strong-determining model with a TC center as a reference point by RVM;
figure 2 is a TC objective intensity model based on satellite cloud images and RVMs with each point in turn as a reference point;
FIG. 3 error curves for different statistical parameters as TC intensity characteristic factors;
FIG. 4 is an error curve of statistical parameters of different dimensions as TC intensity characteristic factors;
figure 5 is an error histogram of RVM strength model for Gauss as the kernel function;
figure 6 is an error histogram of RVM strength model for Cauchy;
figure 7 is an error histogram of RVM strength model for Cauchy;
figure 8 is an error histogram of RVM constant intensity model with kernel function Gauss
Detailed Description
The invention uses 132 TCs obtained by FY-2 satellite scanning during 2005 to 2014, including tropical storms, strong hot band storms, typhoons, strong typhoons, and super strong typhoons. Because the yearbook information is 3 hours or 6 hours apart, 2744 infrared 1-channel cloud pictures and 2744 vapor channel cloud pictures with the same time of the yearbook information are finally obtained through selection. The invention provides a TC objective strength determination method based on satellite cloud pictures and RVMs, which comprises the following two aspects: the following steps are sequentially performed with the TC center as a reference point and each point as a reference point as shown in fig. 1 and 2. The steps performed in fig. 1:
step 5, constructing an optimal characteristic factor capable of representing TC intensity, and testing an optimal TC kernel scale based on the optimal characteristic factor;
step 7, under the optimal kernel scale, constructing a TC objective strength-determining model based on RVM by utilizing characteristic factors and combining with TC center latitude, and carrying out objective estimation on TC strength;
the steps performed in fig. 2:
step 5, under the optimal kernel scale, constructing a TC objective strength-determining model based on RVM by utilizing characteristic factors and combining with TC center latitude, and carrying out objective estimation on TC strength;
the performance of this method is analyzed in detail by four sets of experiments:
experiment 1: with the TC center as a reference point, the invention constructs a fixed intensity model of 9 statistical parameters and center wind speed in the co-occurrence matrix based on RVM, experimental test results of different kernel functions are shown in table 5, and when the RVM selects a Gauss kernel function, the absolute error and relative error histogram results of TC fixed intensity are shown in figure 5.
TABLE 5 error comparison of RVM constant intensity models for different kernel functions
Experiment 2: taking a TC center as a reference point, the invention constructs 9 statistical parameters in the symbiotic matrix based on RVM and combines a fixed intensity model of center latitude and center wind speed, the center latitude data is derived from annual-image data, experimental test results of different kernel functions are shown in table 6, and when the RVM selects a Cauchy kernel function, the absolute error and relative error histogram results of TC fixed intensity are shown in figure 6.
TABLE 6 error comparison of RVM constant intensity models for different kernel functions
Experiment 3: with each point sequentially serving as a reference point, the invention constructs a fixed intensity model of 9 statistical parameters and a central wind speed in the symbiotic matrix parameter array based on RVMs, experimental test results of different kernel functions are shown in a table 7, and when the RVMs select the Cauchy kernel functions, the results of absolute error and relative error histogram of TC fixed intensity are shown in a figure 7.
TABLE 7 error comparison of RVM constant intensity models for different kernel functions
Experiment 4: with each point sequentially serving as a reference point, 9 statistical parameters in the symbiotic matrix parameter array are built based on RVMs, and a fixed intensity model of the center latitude and the center wind speed is combined, the center latitude data are derived from annual-image data, experimental test results of different kernel functions are shown in a table 8, and when the RVMs select Gauss kernel functions, the absolute error and relative error histogram results of TC fixed intensity are shown in fig. 8.
TABLE 8 error comparison of RVM constant intensity models for different kernel functions
And constructing a TC objective strong-determination model based on the fusion cloud image and the RVM by taking the TC center as a reference point. Firstly, fusing cloud images of infrared and water vapor channels by using a Laplacian pyramid fusion algorithm to obtain a fused satellite cloud image. And then taking the TC center as a circle center and taking every 50km of radial direction as an interval from the TC center to expand outwards in the range of 200km from the center to intercept the fusion cloud picture. Calculating a brightness temperature gradient matrix of the intercepted fusion cloud picture, taking a TC center as a reference point, calculating a deviation angle matrix, and then constructing a deviation angle-gradient co-occurrence matrix. The invention utilizes a plurality of statistical parameters in the symbiotic matrix and combines the information such as TC kernel and center latitude to construct the characteristic factors closely related to TC intensity, and utilizes RVM to establish a TC objective strength-determining model. Experimental results show that the optimal radial scale distance is 200km by comparing the fixed strength errors of all the kernel scales. When the strength-fixing model is constructed by utilizing the 9 best statistical parameters of the co-occurrence matrix, the strength-fixing error of the RVM model is minimum. After the center latitude is added, the fixed strength error of the RVM model is reduced. The RVM model has better high-dimensional nonlinear processing capacity and strength estimation capacity, and can effectively estimate TC strength.
And constructing a TC objective strong-determination model based on the fusion cloud image and the RVM by taking each point as a reference point in sequence. And (3) taking the TC center as a circle center and taking every 50km of radial direction as an interval to expand and intercept the fused cloud image outwards from the TC center within the range of 200km from the center by utilizing the fused cloud image obtained in the last part. Calculating a bright-warm gradient matrix of the intercepted fusion cloud picture, sequentially taking each point as a reference point, calculating a deviation angle matrix, then constructing a deviation angle-gradient symbiotic matrix, and calculating the minimum value, the median value and the mean value of a statistical parameter matrix of the symbiotic matrix. The invention utilizes a plurality of statistical parameters (minimum value, median value and mean value) in the symbiotic matrix parameter array and combines the information such as TC kernel and center latitude to construct the characteristic factors closely related to TC intensity, and utilizes RVM to establish a TC objective intensity-determining model. Experimental results show that the optimal radial scale distance is 200km by comparing the fixed strength errors of all scales. The mean value of the symbiotic matrix statistical parameter array is more suitable for TC strength determination. When the strength-fixing model is constructed by utilizing the 9 best statistical parameters of the co-occurrence matrix, the strength-fixing error of the RVM model is minimum. After the center latitude is added, the fixed strength error of the RVM model is reduced. The RVM model has better high-dimensional nonlinear processing capacity and strength estimation capacity, and can effectively estimate TC strength. Compared with TC fixed strength errors of two schemes with a TC center as a reference point and each point as a reference point in sequence, the errors of a TC fixed strength model are smaller when each point is used as the reference point in sequence. The scheme that each point is sequentially used as a reference point is described, so that more comprehensive, richer and more accurate characteristic information can be extracted, and the accuracy of TC objective strength determination can be improved.
Claims (5)
1. The tropical cyclone TC objective strength determination method based on satellite cloud image and relevant vector machine RVM is the application of satellite cloud image data and machine learning algorithm in the meteorological field of TC strength determination, and comprises the following steps:
step 1, fusion processing is carried out on an infrared 1-channel cloud image and a water vapor channel cloud image in a satellite cloud image based on a Laplacian pyramid image fusion algorithm, so that a fusion cloud image is obtained;
step 2, taking the TC center as a circle center and taking every 50 kilometers of radial direction as an interval from the TC center to extend outwards and intercept a fusion cloud picture within the range of 200 kilometers from the center;
step 3, calculating a brightness temperature gradient matrix of the intercepted fusion cloud picture, and calculating to obtain a deviation angle matrix by taking a TC center as a reference point; constructing a deviation angle-gradient co-occurrence matrix of the fusion cloud picture, wherein the co-occurrence matrix comprises 15 statistical parameters;
step 4, constructing an optimal characteristic factor capable of representing TC strength based on 15 statistical parameters of the deviation angle-gradient co-occurrence matrix, and testing an optimal TC kernel scale based on the optimal characteristic factor, wherein the TC kernel scale is respectively 50km,100km,150km and 200km;
step 5, constructing a TC objective strength-determining model based on RVM by utilizing the optimal characteristic factors under the optimal kernel scale, and carrying out objective estimation on TC strength;
and 6, constructing a TC objective strength determination model based on RVM by utilizing the optimal characteristic factors and combining with TC center latitude under the optimal kernel scale, and carrying out objective estimation on TC strength.
2. The TC objective strength determination method based on satellite cloud images and RVMs as set forth in claim 1, wherein: the optimal TC kernel scale is 200km.
3. The TC objective strength determination method based on satellite cloud images and RVMs as set forth in claim 1, wherein: the element of the deviation angle-gradient co-occurrence matrix is defined as the total pixel number which has the deviation angle i and the gradient j in the normalized deviation angle matrix and the normalized gradient image; the following deviation angle-gradient co-occurrence matrix was normalized:
4. the TC objective strength determination method based on satellite cloud images and RVMs as set forth in claim 1, wherein: the optimal characteristic factors are a deviation angle mean value T6, an inverse differential moment T15, a small gradient advantage T1, a deviation angle entropy T11, differential moments T14 and T3, a mixed entropy T13, a correlation T10 and energy T5.
5. The TC objective strength determination method based on satellite cloud images and RVMs as set forth in claim 1, wherein: the optimal characteristic factors are the minimum value, the median value and the mean value of the deviation angle mean value T6 or the optimal characteristic factors are the deviation angle mean value T6, the inverse differential moment T15, the small gradient dominance T1, the deviation angle entropy T11, the differential moment T14, the differential moment T3, the mixed entropy T13, the correlation T10 and the energy T5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710406752.1A CN107230197B (en) | 2017-05-27 | 2017-05-27 | Tropical cyclone objective strength determination method based on satellite cloud image and RVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710406752.1A CN107230197B (en) | 2017-05-27 | 2017-05-27 | Tropical cyclone objective strength determination method based on satellite cloud image and RVM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107230197A CN107230197A (en) | 2017-10-03 |
| CN107230197B true CN107230197B (en) | 2023-05-12 |
Family
ID=59934635
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710406752.1A Active CN107230197B (en) | 2017-05-27 | 2017-05-27 | Tropical cyclone objective strength determination method based on satellite cloud image and RVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107230197B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325960B (en) * | 2018-11-20 | 2021-07-09 | 南京信息工程大学 | Infrared cloud chart cyclone analysis method and analysis system |
| CN110119494B (en) * | 2018-12-31 | 2023-03-31 | 三亚中科遥感研究所 | Tropical cyclone automatic strength determining method based on active and passive microwave remote sensing observation |
| CN110059423B (en) * | 2019-04-23 | 2023-01-31 | 中国人民解放军国防科技大学 | Tropical cyclone objective strength determining method based on multi-factor generalized linear model |
| CN111127365B (en) * | 2019-12-26 | 2023-08-29 | 重庆矢崎仪表有限公司 | HUD distortion correction method based on cubic spline curve fitting |
| CN111695473B (en) * | 2020-06-03 | 2023-12-19 | 中国人民解放军国防科技大学 | Tropical cyclone strength objective monitoring method based on long-short-term memory network model |
| CN111862005B (en) * | 2020-07-01 | 2023-11-17 | 自然资源部第二海洋研究所 | Method and system for precisely positioning tropical cyclone center by utilizing synthetic radar image |
| CN115082439B (en) * | 2022-07-22 | 2022-11-29 | 浙江大学 | Tropical cyclone strength determining method, medium and equipment fused with satellite cloud picture space-time information |
| CN116030359B (en) * | 2023-01-09 | 2024-11-01 | 中科星图维天信科技股份有限公司 | Tropical cyclone center positioning and strength fixing method based on maskrcnn deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6879944B1 (en) * | 2000-03-07 | 2005-04-12 | Microsoft Corporation | Variational relevance vector machine |
| CN102938075A (en) * | 2012-11-29 | 2013-02-20 | 浙江师范大学 | RVM (relevant vector machine) method for maximum wind radius and typhoon eye dimension modeling |
| CN103839243A (en) * | 2014-02-19 | 2014-06-04 | 浙江师范大学 | Multi-channel satellite cloud picture fusion method based on Shearlet conversion |
-
2017
- 2017-05-27 CN CN201710406752.1A patent/CN107230197B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6879944B1 (en) * | 2000-03-07 | 2005-04-12 | Microsoft Corporation | Variational relevance vector machine |
| CN102938075A (en) * | 2012-11-29 | 2013-02-20 | 浙江师范大学 | RVM (relevant vector machine) method for maximum wind radius and typhoon eye dimension modeling |
| CN103839243A (en) * | 2014-02-19 | 2014-06-04 | 浙江师范大学 | Multi-channel satellite cloud picture fusion method based on Shearlet conversion |
Non-Patent Citations (2)
| Title |
|---|
| 王瑾,江吉喜.热带气旋强度的卫星探测客观估计方法研究.应用气象学报.2005,(03),全文. * |
| 鲁小琴 ; 雷小途 ; 余晖 ; 赵兵科 ; .基于卫星资料进行热带气旋强度客观估算.应用气象学报.2014,(01),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107230197A (en) | 2017-10-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107230197B (en) | Tropical cyclone objective strength determination method based on satellite cloud image and RVM | |
| CN112836610B (en) | Land use change and carbon reserve quantitative estimation method based on remote sensing data | |
| Chen et al. | Extraction of glacial lake outlines in Tibet Plateau using Landsat 8 imagery and Google Earth Engine | |
| CN109886217B (en) | Method for detecting wave height from offshore wave video based on convolutional neural network | |
| CN105528619B (en) | SAR remote sensing image variation detection method based on wavelet transformation and SVM | |
| CN110555841B (en) | SAR image change detection method based on self-attention image fusion and DEC | |
| CN111273378B (en) | Typhoon center positioning method based on wind stress disturbance | |
| CN111339827A (en) | SAR image change detection method based on multi-region convolutional neural network | |
| Zhai | Inversion of organic matter content in wetland soil based on Landsat 8 remote sensing image | |
| CN109002792B (en) | SAR image change detection method based on layered multi-model metric learning | |
| CN112419196B (en) | Unmanned aerial vehicle remote sensing image shadow removing method based on deep learning | |
| CN110186851A (en) | It is a kind of based on the semi-supervised Hyperspectral imaging heavy metal-polluted soil concentration evaluation method from Coded Analysis | |
| CN110321528B (en) | Hyperspectral image soil heavy metal concentration assessment method based on semi-supervised geospatial regression analysis | |
| Li et al. | Classification of tropical cyclone intensity based on deep learning and Yolo V5 | |
| CN107346549B (en) | Multi-class change dynamic threshold detection method utilizing multiple features of remote sensing image | |
| CN117710508A (en) | Near-surface temperature inversion method and device for generating countermeasure network based on improved condition | |
| Zhai et al. | Object-oriented land cover change detection combining optical and radar remote sensing data | |
| CN115546658B (en) | Night cloud detection method combining quality improvement and CNN improvement of data set | |
| CN107341798B (en) | High Resolution SAR image change detection method based on the overall situation-part SPP Net | |
| CN112729562B (en) | Sea ice distribution detection method based on improved U-shaped convolutional neural network | |
| CN114047563B (en) | All-weather assimilation method for infrared hyperspectrum | |
| CN109753896A (en) | A kind of unsupervised heterologous method for detecting change of remote sensing image based on general character autocoder | |
| Chen et al. | Generating Daily Gap-Free MODIS Land Surface Temperature Using the Random Forest Model and Similar Pixels Method | |
| Lian et al. | End-to-end building change detection model in aerial imagery and digital surface model based on neural networks | |
| CN111640117B (en) | Method for searching leakage source position of building |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |