CN103198447B - A kind of wind arrow field real-time metrics method based on satellite cloud picture - Google Patents

Abstract

The invention relates to a real-time measurement method of a wind vector field based on a satellite cloud image, which belongs to the field of image processing and motion estimation. It solves the problems of complex calculation, low real-time performance and low efficiency of search algorithm in the existing model. Convert the detected earth temperature data into grayscale data, convert the grayscale data into latitude and longitude data, and then convert the latitude and longitude data into image coordinates, and the satellite cloud image data is displayed in the form of an image; compare three consecutive satellite cloud images, block matching In the future, choose different search methods according to real-time requirements to determine the longitude, latitude and direction of the wind vector field; integrate data preprocessing and search results, and further obtain the gray level, temperature and isobaric surface of each wind vector; The position, size and direction of the Yaba on the satellite cloud map and the isobaric surface where it is located can be used to observe atmospheric circulation and medium and long-term weather forecasts. The invention avoids complex calculation and strong real-time performance during image processing of large data volume of satellite cloud images, and improves the execution efficiency of the system.

Description

A real-time measurement method of wind vector field based on satellite cloud images

technical field

The invention relates to a strong real-time and high-efficiency measurement method of a wind vector field based on a satellite cloud image, and belongs to the field of image processing and motion estimation.

Background technique

Satellite cloud images play an important role in mastering atmospheric circulation, mid- and long-term weather forecasts, and research on disastrous weather. It is made by the infrared detector on the geosynchronous satellite to detect the temperature data above the earth and then convert it into grayscale data. Cloud trace wind is very important for global weather and typhoon analysis and for providing numerical pre-initial wind field data. Yunjifeng has now become an important satellite product. It can be used as supplementary data for conventional wind measurements of land observation networks, and it is the main or only source of wind information in areas with few or no stations such as oceans, plateaus, and deserts. Cloud trace wind data have been widely used in synoptic studies such as typhoon, heavy rain and flood disasters, and mesoscale weather analysis. Domestic scholars have also clearly pointed out that cloud trace wind can clearly and vividly display the details of weather system development and changes. It has a wide application prospect in forecasting, and has important indication significance for the analysis and prediction of rainstorm fall area, typhoon range and tropical cyclone movement forecast.

At present, motion estimation technology has become a very important part in the field of digital image processing. Scientists have proposed many kinds of video motion estimation algorithms, such as Bayesian method, pixel recursive method, optical flow method and block matching method and so on. The block matching method is simple and effective, meets the real-time requirements of image processing, and requires a relatively small amount of calculation. Among them, the adaptive cross search ARPS (adaptive rood pattern search) algorithm can greatly improve the efficiency of motion estimation compared with the full search algorithm, and maintain a certain accuracy in the estimation quality.

Contents of the invention

The purpose of the present invention is to provide a real-time measurement method of wind and vector field based on satellite cloud images, which is used to solve the problems of complex calculation of existing models, difficult implementation of software and hardware, poor real-time performance, and low efficiency of search algorithms.

To achieve the above object, the present invention takes the following technical solutions:

A kind of wind vector field real-time measurement method based on satellite cloud image, the specific realization process of described wind vector field real-time measurement method is:

Step 1. Preprocessing of satellite detection data, converting the detected earth temperature data into grayscale data, then converting the grayscale data into latitude and longitude data, and then converting the latitude and longitude data into image coordinates, and displaying the satellite cloud image data in the form of images ;

Step 2, compare the satellite cloud images of three continuous times, after block matching (using the SAD matching criterion to match), select different search methods according to real-time requirements to determine the latitude and longitude of the wind vector field and direction;

Step 3, comprehensive data preprocessing and search results, and further obtain the gray level, temperature and isobaric surface of each wind vector;

Step 4. Finally, by observing the position, size and direction of the wind vector field on the satellite cloud image and the isobaric surface where it is located, the atmospheric circulation and mid- and long-term weather forecast can be observed.

In step 1, when converting the gray scale data into longitude and latitude data, the line connecting the satellite and the center of the earth is used as the x-axis, the direction pointed by the North Pole is used as the z-axis, and the y-axis is established according to the right-hand spiral rule; the earth is an ideal ellipse The sphere and satellite detection data files are all 2288×2288 gray value matrix, and each element of the matrix corresponds to a detection point (or sampling point) on or off the earth; the sub-satellite point of the satellite is at 86.5 degrees east longitude, At 0 degrees north latitude, the matrix element corresponding to the sub-satellite point is located at the intersection of the 1145th row and the 1145th column of the matrix.

The specific process of determining the longitude, latitude and direction of the wind vector field in step 2 is as follows:

Step 2 (1), compare the satellite cloud images of three consecutive periods:

Take three consecutive static cloud images, and solve the cloud image wind vector at the middle moment, the previous moment is the reference image, and the next moment is the corrected image; the size of the matching block is 16×16, and the search area is 64×64, ensuring that each pixel is It can be searched without duplication; when scanning, the template is translated by 1 pixel each time and the search area is summed for absolute error matching;

Step two (2), block matching motion estimation, utilize SAD matching criterion to carry out:

Sum absolute errors:

$\mathrm{<span\; class="notranslate">\; SAD</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Among them, u, v represent the horizontal and vertical offsets between the prediction block in the reference image and the current block in the current image, -p≤u, v≤p; m, n represent the horizontal and vertical offsets of a pixel in the current block Coordinates; f _k (m, n) represents the gray value of a pixel in the current block, f _k-1 (m+u, n+v) represents the gray value of the corresponding pixel in the predicted block; p represents the maximum search in one direction Distance, M, N represent the size of the macroblock; compare the SAD values of different horizontal and vertical offsets, and the smallest of all SAD values is the matching block;

Step two (3), applying adaptive search to solve the wind vector field, the main steps of the adaptive cross search algorithm are as follows:

1) If the current macroblock is the first macroblock of the current frame, use it as the starting point of the search and skip to step 5);

2) If the current macroblock is in the top row of the frame, take the motion vector of the left macroblock as the candidate search starting point; if it is in the leftmost column of the frame, take the motion vector of the above macroblock as the candidate search starting point, and jump to the 4th )step;

3) Otherwise, take the average value of the motion vectors of the macroblocks above and to the left as a candidate search starting point;

4) Calculate the SAD values with the candidate search starting point and the first macroblock as the starting point respectively, take the smaller one as the initial value of the minimum SAD, and denote it as M _SAD , and the corresponding point as the actual search starting point;

5) Carry out a spiral search in the next circle, calculate the SAD value of each point, take 1 for the search step of the first circle, and take 2 for the search steps of other circles;

If it is greater than M _SAD during the calculation, exit the calculation and search for the next point, otherwise calculate SAD completely;

If the current SAD<M _SAD , assign it to M _SAD , and set the M _SAD update flag F of this circle to 1;

6) When the search in this circle ends, if F=1, go to step 5); otherwise, continue to step 7);

7) End the spiral search, if the current best matching point is the initial search point, continue to step 8); otherwise, do a further search around the 4 points that have not been searched (3 points when it is in the second circle) , then continue to step 8) step;

8) Carry out a small search mode search centering on the best matching point obtained in step 7), find the macroblock closest to the matching block, and determine the longitude, latitude and direction of the wind vector field from this macroblock.

In steps 3 and 4, the cloud top temperature of the infrared cloud image is roughly estimated and calculated according to the vertical profile of the atmospheric temperature, and the estimated value of the pressure height represented by the cloud trace wind vector can be obtained, and the logarithmic linear interpolation method is used to establish The relationship between temperature T and pressure P:

T＝a+blnP (2)

in Among them, (t ₁ , p ₁ ) and (t ₂ , p ₂ ) are two known temperature and pressure points.

In the geocentric coordinate system, the earth can be regarded as an ideal ellipsoid. The gray value matrix near the earth detected by the satellite is known. The known sub-satellite point is at 86.5 degrees east longitude and 0 degrees north latitude. The corresponding matrix element is located at the intersection of row 1145 and column 1145 of the matrix.

The two-dimensional wind vector has four indicators: the latitude and longitude of the starting point, the direction and size of the wind vector. The motion vector of the block is solved by using the block matching algorithm, and then the wind vector field can be obtained. At the same time, the number of all non-zero wind vectors is calculated in the range from 40 degrees south latitude to 40 degrees north latitude and from 46 degrees to 126 degrees east longitude.

The beneficial effects of the present invention are:

The method of the invention realizes the adaptive cross search algorithm using the block matching algorithm to efficiently solve the motion vector, and then obtains the wind vector field, which is very important for the analysis of global weather and typhoon and the provision of numerical forecast. The method of the invention has the advantages of strong real-time performance, high efficiency and the like.

The method of the invention proposes a method for determining the wind vector field of the satellite cloud image in real time and efficiently for the satellite cloud image detected by the geosynchronous orbit satellite, which is the prerequisite work for mastering atmospheric circulation and mid- and long-term weather forecast. The invention avoids the problems of complicated calculation and poor real-time performance during the image processing of large data volume of the satellite cloud image, improves the execution efficiency of the system, and has better processing effect. In the wind vector block matching process based on the satellite cloud image, the present invention uses an adaptive cross search algorithm to solve efficiently, and detects whether the matching is successful by setting a threshold, thereby achieving the purpose of efficiently obtaining the wind vector of the satellite cloud image in real time.

The window size and search range can be determined adaptively during block matching. In order to take both real-time and accuracy into account, the motion vector of the known macroblock is used to predict the motion vector of the current macroblock, and the large search mode is used to perform matching calculations at the center and eight surrounding points of the search area to find the minimum point of the value function, thereby realizing Adaptive search mode.

In this paper, the initial motion vector of the current block is likely to be close to its final motion vector through the prediction of the search starting point, and then according to the local characteristics of the image, the image is simply and effectively classified and the appropriate search mode is selected, so that it can be automatically searched according to the type of motion. Adaptive search, and finally use the search termination criterion to ensure that the search result has sufficient precision when it ends near the predicted starting point, so as to realize fast, uniform and high-precision motion vector search.

The specific advantages are mainly manifested in the following aspects:

1. Through the method of block matching to model the satellite cloud image of cloud guide wind, and use the efficient adaptive cross search method to solve the problem;

2. The block matching method is simple and effective, meets the real-time requirements of image processing, and requires a relatively small amount of calculation.

3. Adaptive cross search can greatly improve the efficiency of motion estimation, and maintain a certain accuracy in estimation quality.

4. It has wide application prospects in numerical weather analysis and forecasting, and has important indication significance for the analysis and prediction of rainstorm falling areas, typhoon range and tropical cyclone movement forecast.

Description of drawings

Fig. 1 is the workflow schematic diagram of the inventive method; Fig. 2 is the satellite cloud image of the earth's surface; Fig. 3 is the matching schematic diagram of cloud guide block; Fig. 4 is the self-adaptive cross search method flow chart; Fig. 5 is temperature and air pressure relation figure; Figure 6-1 is the cloud guide map before correction; Figure 6-2 is the cloud guide map after correction; Figure 7 is the upper left schematic diagram of cloud guide before correction after zooming in; Figure 8 is the upper left schematic diagram of cloud guide after correction .

Detailed ways

In order to describe the advantages of the technical solutions of the present invention more clearly, the specific implementation of the present invention will be further elaborated below in conjunction with the accompanying drawings. Apparently, the described embodiments are only some of the embodiments of the present invention, not all of them. On this basis, the embodiment of the present invention can be extended, and more optimization schemes can be obtained under the condition that the overall architecture is consistent. According to the embodiments of the present invention, those skilled in the art can realize all other embodiments of the present invention on the basis of no creative work, and all belong to the protection scope of the present invention.

Fig. 1 is a schematic diagram of a specific embodiment of the present invention, as shown in Fig. 1, the process includes the following steps:

Step 1: Preprocessing of satellite detection data, converting the detected earth temperature data into grayscale data, then converting the grayscale data into latitude and longitude data, and then converting the latitude and longitude data into image coordinates, and displaying the satellite cloud image data in the form of images .

Step 2: As shown in Figure 3, assuming that the cloud block only moves in translation, A is the position of the cloud image at the moment t ₁ of the cloud block, and C is the position of the cloud image at the time t ₂ of the cloud block. Taking the x direction as an example, module A is at Δt=t The displacement of ₂ -t ₁ time interval can be described as X=x ₀ +x', where x ₀ represents an integer multiple of pixel displacement air volume, and x' is a sub-pixel displacement component, that is, |x'|>1 pixel scale.

Use the correlation method to find the matching module of the target module A in the search area of the second cloud image to obtain the module B, and calculate the integer multiple pixel displacement x ₀ according to the position difference between the two. If the A module is completely matched with the B module, it is considered that there is no sub-pixel displacement, and the translation speed of the cloud block in the x direction is V=x ₀ /Δt, otherwise, the difference between the two modules is considered to be caused by the sub-pixel displacement of the cloud block, The frequency spectrum of modules A and B can be further analyzed by Fourier phase analysis method, the sub-pixel displacement can be calculated according to the phase difference, and the translation velocity V′ _x =(x ₀ +x′)/Δt can be obtained after combining x ₀ and x'. Similarly, the displacement and velocity in the y direction can be calculated as Y=y ₀ +y′, V′ _y =(y ₀ +y′)/Δt.

Step 3: Sum Absolute Error Matches, sum Absolute Error:

$\mathrm{<span\; class="notranslate">\; SAD</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Among them, u, v represent the horizontal and vertical offsets between the prediction block in the reference image and the current block in the current image, -p≤u, v≤p; m, n represent the horizontal and vertical offsets of a pixel in the current block coordinates; f _k (m, n) represents the gray value of a certain pixel in the current block, and f _k-1 (m+u, n+v) represents the gray value of the corresponding pixel in the predicted block. p represents the maximum search distance in one direction, and M and N represent the macroblock size. Comparing the SAD values of different horizontal and vertical offsets, the smallest of all SAD values is the matching block.

Step 4: Through the prediction of the search starting point, the initial motion vector of the current block may be close to its final motion vector, and then simply and effectively classify the image according to the local characteristics of the image and select the appropriate search mode, so that it can be based on the type of motion Carry out adaptive search, and finally use the search termination criterion to ensure that the search result has sufficient accuracy when it ends near the predicted starting point, so as to achieve fast, uniform, and high-precision motion vector search, as shown in Figure 4, Figure 6-1 and 6 -2 shown.

Step 5: As shown in Figure 5, the relationship between temperature and air pressure is shown. Through this figure, the logarithmic linear interpolation formula is

T=a+blnP(2)

Substitute (t ₁ , lnp ₁ ) and (t ₂ , lnp ₂ ) into formula (2) to get

t ₁ =a+blnp ₁ , t ₂ =a+blnp ₂ (3)

Solving these two equations gives

$\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; ln</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

From formula (2) can get

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{\frac{\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; a</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

This method can find the isobaric surface where the non-zero wind vector is located.

Step 6: Finally, by observing the position, size and direction of the wind vector field on the satellite cloud image and the isobaric surface where it is located, the atmospheric circulation and medium and long-term weather forecast can be observed.

Claims (2)

1., based on a wind arrow field real-time metrics method for satellite cloud picture, the specific implementation process of described wind arrow field real-time metrics method is:

The global warming data detected are converted to gradation data, then gradation data are converted to longitude and latitude data, then longitude and latitude data are converted to image coordinate, satellite cloud picture data are shown with image format by step one, satellite sounding data prediction;

Step 2, contrast three continuous times satellite cloud picture, after Block-matching, select different searching methods according to real-time demand, determine wind arrow field longitude and latitude and direction;

Step 3, integrated data pre-service and Search Results, obtain the gray scale of each wind arrow further, temperature and place isopressure surface thereof;

Step 4, eventually through observation the position of wind arrow field on satellite cloud picture, size and Orientation and place isopressure surface thereof can observe general circulation and medium-term and long-term weather forecast;

In step one, when gradation data being converted to longitude and latitude data, using satellite and the earth's core line as x-axis, with arctic direction for z-axis, according to right-hand screw rule, set up y-axis; If the earth is desirable ellipsoid, satellite sounding data file is the gray-scale value matrix of 2288 × 2288, on the corresponding earth of each element of matrix or an extraterrestrial sensing point (or claim sampled point); The substar of satellite east longitude 86.5 degree, north latitude 0 degree, matrix element corresponding to substar is positioned at the 1145th row and the 1145th row intersection of matrix;

It is characterized in that: in step 2, determine that the detailed process in wind arrow field longitude and latitude and direction is:

Step 2 (1), contrast three continuous time satellite cloud pictures:

Get continuous three static cloud atlas, solve intermediate time cloud atlas wind vector, with the previous moment for reference picture, a later moment is correcting image; The size of match block is 16 × 16, and region of search is 64 × 64, ensures that each pixel can both searchedly arrive, and does not repeat; Template translation 1 location of pixels is carried out summation absolute error with the field of search during scanning to mate at every turn;

Step 2 (2), block-based motion estimation, utilize SAD matching criterior to carry out:

Summation absolute error:

$SAD(u,v)=\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}|f_k(m,n)-f_{k-1}(m+u,n+v)|---\left(1\right)$

Wherein, u, v represent the prediction block in reference picture and the skew of current block in horizontal and vertical direction in present image ,-p≤u, v≤p; M, n represent the horizontal and vertical coordinate of certain pixel in current block; f _k(m, n) represents the gray-scale value of certain pixel of current block, f _k-1the gray-scale value of the respective pixel of (m+u, n+v) representative prediction block; P represents one direction maximum search distance, and M, N represent macroblock size; The relatively sad value of varying level and vertical shift, minimumly in all sad values is match block;

The search of step 2 (3), application self-adapting solves wind arrow field, and the key step of self-adaptation Cross Search algorithm is as follows:

1) if current macro is first macro block of present frame, then it can be used as search starting point, jump to the 5th) step;

2) if current macro is positioned at the far top row of frame, the motion vector alternatively search starting point of left side macro block is got; If be positioned at the left column of frame, the motion vector getting macro block is above candidate search starting point, jumps to the 4th) step;

3) otherwise, above getting and the mean value alternatively search starting point of left side macroblock motion vector;

4) sad value that to calculate with candidate search starting point and first macro block be respectively starting point, gets the initial value of smaller as minimum SAD, is designated as M _sAD, corresponding point is as actual search starting point;

5) in next circle, carry out spiral search, calculate the sad value of every bit, get 1 to first lap step-size in search, the step-size in search of other circle gets 2;

If be just greater than M in the calculation _sAD, exit calculating, under search a bit, otherwise calculate SAD completely;

If current SAD<M _sAD, then it is assigned to M _sAD, this circle of juxtaposition M _sADupgrading mark F is 1;

6) when the search of this circle terminates, if F=1, the 5th is forwarded to) step; Otherwise continue the 7th) step;

7) terminate spiral search, if current best match point is initiating searches point, continue the 8th) step; Otherwise, do search further, then continue the 8th for 4 that do not search for around it) and step;

8) by the 7th) carry out little search pattern search centered by the optimal match point that obtains of step, find the immediate macro block with match block, determine wind arrow field longitude and latitude and direction by this macro block.

2. the wind arrow field real-time metrics method based on satellite cloud picture according to claim 1, is characterized in that:

In step 3 and four, carry out rough estimate by infrared cloud image cloud-top temperature according to atmospheric vertical temperature profile and calculate cloud-top height, can show that Cloud-motion wind vows the barometer altitude estimated value of representative, and utilize log-linear interpolation method, set up the relation between temperature T and pressure P:

T＝a+blnP (2)

Wherein wherein, (t ₁, p ₁), (t ₂, p ₂) be two temperature pressure points.