阅读说明：本技术 云图图像预测准确性的评估方法、装置、设备及存储介质 (Method, device and equipment for evaluating accuracy of cloud image prediction and storage medium ) 是由 周康明 姚广 于 2019-11-06 设计创作，主要内容包括：本申请提供一种云图图像预测准确性的评估方法、装置、设备及存储介质,实现方案包括：获取至少两幅实际云图图像,以及与实际云图图像对应的预测云图图像；根据预设规则分别确定实际云图图像、预测云图图像的质心；根据实际云图图像的质心确定实际移动信息,根据预测云图图像的质心确定预测移动信息；根据实际移动信息、预测移动信息评估预测云图图像是否准确。本申请提供的方法、装置、设备及存储介质中,根据至少两幅云图图像中的质心,确定这些云图图像中质心的移动信息,再比对实际云图图像中的质心移动信息以及预测云图图像中的质心移动信息,能够确定出预测云图图像中云的移动与实际云图图像中云的移动差异,从而评估预测云图是否准确。(The application provides a method, a device, equipment and a storage medium for evaluating cloud image prediction accuracy, and the implementation scheme comprises the following steps: acquiring at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images; respectively determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule; determining actual movement information according to the mass center of the actual cloud picture image, and determining predicted movement information according to the mass center of the predicted cloud picture image; and evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information. According to the method, the device, the equipment and the storage medium, the moving information of the mass centers in the cloud picture images is determined according to the mass centers in at least two cloud picture images, and then the mass center moving information in the actual cloud picture image is compared with the mass center moving information in the predicted cloud picture image, so that the difference between the moving of the cloud in the predicted cloud picture image and the moving of the cloud in the actual cloud picture image can be determined, and whether the predicted cloud picture is accurate or not is evaluated.)

1. A method for evaluating accuracy of cloud image prediction is characterized by comprising the following steps:

acquiring at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images;

respectively determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule;

determining actual movement information according to the mass center of the actual cloud picture image, and determining predicted movement information according to the mass center of the predicted cloud picture image;

and evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information.

2. The method of claim 1, wherein the preset rules comprise:

determining pixel information of each pixel point in the image according to the cloud image;

and determining the centroid of the cloud image according to the pixel information of each point.

3. The method of claim 2, wherein determining pixel information for each pixel point in the image from the cloud image comprises:

and reading the pixel value of each channel corresponding to each pixel point in the cloud image, and determining the average value of the pixel values of the pixel points as the pixel information.

4. The method of claim 2, wherein determining the centroid of the cloud image from the pixel information for each point comprises determining the centroid of the cloud image according to:

wherein I represents a row index of the centroid and J represents a column index of the centroid; i denotes a row index of a pixel in the cloud image N, j denotes a column index of a pixel in the cloud image N, N_i，jAnd representing the pixel information of the pixel points with indexes i and j.

5. The method of claim 1, wherein determining actual movement information from a centroid of the actual cloud image and determining predicted movement information from a centroid of the predicted cloud image comprises:

determining an actual movement vector according to the centroid of the actual cloud picture image and the time information of each actual cloud picture image;

and determining a predicted movement vector according to the centroid of the predicted cloud picture image and the time information of each predicted cloud picture image.

6. The method of claim 5, wherein the evaluating whether the predicted cloud image is accurate based on the actual movement information and the predicted movement information comprises:

determining a movement distance difference and a movement direction included angle according to the actual movement vector and the predicted movement vector;

and evaluating whether the predicted cloud image is accurate or not according to the moving distance difference and the moving direction difference.

7. The method of any one of claims 1-6, wherein the predicted cloud image corresponds to temporal information of the actual cloud image.

8. An apparatus for evaluating accuracy of cloud image prediction, comprising:

the cloud image prediction system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring at least two actual cloud image images and a prediction cloud image corresponding to the actual cloud image images;

the mass center determining module is used for respectively determining the mass centers of the actual cloud picture image and the predicted cloud picture image according to a preset rule;

the movement information determining module is used for determining actual movement information according to the mass center of the actual cloud picture image and determining predicted movement information according to the mass center of the predicted cloud picture image;

and the evaluation module is used for evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information.

9. An apparatus for evaluating accuracy of cloud image prediction, comprising:

a memory;

a processor; and

a computer program;

wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of any of claims 1-7.

10. A computer-readable storage medium, having stored thereon a computer program,

the computer program is executed by a processor to implement the method according to any one of claims 1 to 7.

Technical Field

The present disclosure relates to image processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for evaluating accuracy of cloud image prediction.

Background

Currently, radar detection means are mainly utilized in the field of weather prediction. And analyzing and predicting the future weather based on the radar detection data.

Disclosure of Invention

The application provides a method, a device, equipment and a storage medium for evaluating the prediction accuracy of a cloud image, so as to evaluate whether the predicted cloud image is accurate.

The first aspect of the present application provides a method for evaluating accuracy of cloud image prediction, including:

acquiring at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images;

respectively determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule;

determining actual movement information according to the mass center of the actual cloud picture image, and determining predicted movement information according to the mass center of the predicted cloud picture image;

and evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information.

Another aspect of the present application is to provide an apparatus for evaluating accuracy of cloud image prediction, including:

the cloud image prediction system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring at least two actual cloud image images and a prediction cloud image corresponding to the actual cloud image images;

the mass center determining module is used for respectively determining the mass centers of the actual cloud picture image and the predicted cloud picture image according to a preset rule;

the movement information determining module is used for determining actual movement information according to the mass center of the actual cloud picture image and determining predicted movement information according to the mass center of the predicted cloud picture image;

and the evaluation module is used for evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information.

Another aspect of the present application is to provide an apparatus for evaluating accuracy of cloud image prediction, including:

a memory;

a processor; and

a computer program;

wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method for evaluating cloud image prediction accuracy as described in the first aspect above.

Yet another aspect of the present application is to provide a computer-readable storage medium, on which a computer program is stored, the computer program being executed by a processor to implement the method for evaluating the accuracy of cloud image prediction as described in the above first aspect.

The technical effects of the method, the device, the equipment and the storage medium for evaluating the cloud image prediction accuracy are as follows:

the application provides a method, a device, equipment and a storage medium for evaluating cloud image prediction accuracy, which comprise the following steps: acquiring at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images; respectively determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule; determining actual movement information according to the mass center of the actual cloud picture image, and determining predicted movement information according to the mass center of the predicted cloud picture image; and evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information. According to the method, the device, the equipment and the storage medium, the moving information of the mass centers in the cloud picture images is determined according to the mass centers in at least two cloud picture images, and then the mass center moving information in the actual cloud picture image is compared with the mass center moving information in the predicted cloud picture image, so that the difference between the moving of the cloud in the predicted cloud picture image and the moving of the cloud in the actual cloud picture image can be determined, and whether the predicted cloud picture is accurate or not is evaluated.

Drawings

Fig. 1 is a flowchart illustrating a method for evaluating accuracy of cloud image prediction according to an exemplary embodiment of the present application;

fig. 2 is a flowchart illustrating a method for evaluating accuracy of cloud image prediction according to another exemplary embodiment of the present application;

FIG. 3A is a schematic diagram illustrating the recording of pixel values in a matrix format according to an exemplary embodiment of the present application;

FIG. 3B is a diagram illustrating motion vectors according to an exemplary embodiment of the present application;

fig. 4 is a block diagram of an apparatus for evaluating accuracy of cloud image prediction according to an exemplary embodiment of the present application;

fig. 5 is a block diagram of an apparatus for evaluating accuracy of cloud image prediction according to another exemplary embodiment of the present application;

fig. 6 is a block diagram of an apparatus for evaluating accuracy of cloud image prediction according to an exemplary embodiment of the present application.

Detailed Description

At present, in recent decades, radar meteorologists at home and abroad do a lot of research work in the field of strong convection weather detection, radar data with high space-time resolution are utilized, physical mechanisms of storm occurrence, development, maturity and extinction are deeply analyzed, storm algorithms are provided, the basic idea is that each storm monomer is regarded as a whole with a three-dimensional continuous structure, relevant physical characteristic quantities are calculated, and an extrapolation technology is adopted during forecasting.

Due to the wide variety of extrapolation algorithms, after the radar cloud chart is drawn by adopting the extrapolation algorithm, the accuracy of the drawn radar cloud chart needs to be evaluated. How to accurately evaluate the drawn radar cloud chart is a technical problem that needs to be solved urgently by a person skilled in the art.

In the scheme provided by the embodiment of the application, the actual direction of the mass center is determined according to the actual cloud picture images of multiple frames, the predicted direction of the mass center is determined according to the cloud picture images drawn by the corresponding multiple frames, and the matching degree of the predicted cloud picture images and the actual cloud picture images is determined by comparing the actual direction of the mass center with the predicted direction of the mass center.

Fig. 1 is a flowchart illustrating a method for evaluating accuracy of cloud image prediction according to an exemplary embodiment of the present application.

As shown in fig. 1, the method for estimating the accuracy of radar cloud image prediction provided in this embodiment includes:

step 101, at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images are obtained.

The method provided by the embodiment may be executed by an electronic device with computing capability, for example, a computer. The method provided by the embodiment can be arranged in the electronic equipment in a software form.

Specifically, the actual cloud image and the predicted cloud image may be stored in the electronic device in advance, or the images may be stored in another device, and at this time, the electronic device may access the device in which the cloud image is stored, so as to read the cloud image.

Further, the cloud image can be in a video form or a picture form. For example, the cloud image may be stored in the form of a video for a period of time, such as an actual cloud image stored for 5 minutes, and a predicted cloud image corresponding to the period of time. The actual cloud image may be actually obtained through radar device detection, and may be, for example, an actual cloud image in a period of time from 17:00 to 17: 05. The cloud image in the period of 17:00-17:05 can be predicted in advance based on a preset algorithm, namely the predicted cloud image. For example, cloud images at 13:00 may be predicted for the period of time 17:00-17: 05.

In practical application, if the cloud image is stored in a picture form, a plurality of images can be stored. For example, a plurality of actual cloud images, which have a time sequence, may be actual cloud images that are acquired continuously within 5 minutes, for example. The predicted cloud image is stored in a picture form, and is similar to the predicted cloud image, and the description is omitted.

The user can select the cloud image to be analyzed, and operate the electronic device to execute the method provided by the embodiment. For example, the user may operate the electronic device, select the storage location of the actual cloud image and the storage location of the predicted cloud image, and then click a button for starting the analysis.

Specifically, the electronic device may obtain at least two actual cloud images. Specifically, at least two consecutive actual cloud images can be acquired. The walking direction of the cloud image centroid can be determined according to two adjacent cloud image images of the time information. In the scheme provided by this embodiment, two cloud images are taken as an example, and the setting can be performed according to the requirements in practical application.

Further, a predicted cloud image corresponding to the actual cloud image may also be obtained. The corresponding relation between the actual cloud image and the predicted cloud image can be determined according to the time information of the cloud image. For example, if an actual cloud image is used to show a cloud layer at a time of 17:01, the predicted cloud image at a time of 17:01 has a corresponding relationship with the actual cloud image.

In actual application, the cloud layer changes are not consistent at different times, so that the accuracy of the predicted cloud picture image can be accurately evaluated by acquiring the actual cloud picture image and the predicted cloud picture image which have the corresponding relation.

And 102, respectively determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule.

Wherein, the centroid refers to the mass center in the cloud image. The determination may be specifically performed according to pixel information of each pixel point in the cloud image.

Specifically, the color information of the cloud image can be used to indicate the thickness and density of the cloud layer. Therefore, a preset rule may be set to determine the centroid from the color information of the cloud image. The thicker point of the cloud layer can be determined in the cloud image as the centroid of the image.

Furthermore, the centroid of the image can be determined by combining the pixel value of each pixel point. For example, pixel values of each point in the cloud image at each channel may be read to form a plurality of matrices. And combining the matrixes into a matrix, wherein the pixel value of each point in the matrix can represent the reflectivity of the point to radar signals, and further represent the density of the cloud layer of the point. In practical application, the centroid can be determined according to the pixel value of each point in the matrix.

In practical application, in order to make the centroids of the actual cloud picture image and the predicted cloud picture image have comparability, the same rule is adopted when determining the centroids of the actual cloud picture image and the predicted cloud picture image.

And determining the centroid of each actual cloud image and each predicted cloud image.

And 103, determining actual movement information according to the mass center of the actual cloud picture image, and determining predicted movement information according to the mass center of the predicted cloud picture image.

Specifically, the actual movement information may be determined according to the centroid of at least two actual cloud images. The actual movement information may be specifically determined according to the time information of the actual cloud image. For example, sorting the actual cloud image in time order, as in A, B, can result in the actual centroids P, respectively_A、P_BThen from P_AMove to P_BI.e. the actual movement information.

Furthermore, the predicted movement information can be determined according to the centroid of the predicted cloud image, and the specific determination mode is similar to that described above. For example, there are two predicted cloud images A ', B ', the time information of A ' and A is t₁B' is identical to the time information of the graph B and is t₂. The actual t from can be found from the centroid of the image A, B₁To t₂And moving information of the cloud picture centroid. The predicted mean value t can also be obtained from the centroids of the images A', B₁To t₂And moving information of the cloud picture centroid.

In practical application, if the number of the cloud image is larger than two, the movement information between every two adjacent cloud images can be determined according to the time information, and thus a group of movement information can be obtained. For example in the temporal orderIncluding the actual cloud image C, D in addition to the actual cloud image A, B, the actual centroid P may also be obtained_C、P_DThen the actual movement information that can be obtained is the slave P_AMove to P_BFrom P_BMove to P_CFrom P_CMove to P_D。

And step 104, evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information.

In actual application, if the predicted movement information is consistent with the actual movement information, the predicted cloud picture image can be considered to be accurate, that is, the movement track of the cloud layer can be accurately predicted.

The predicted movement information and the actual movement information may be compared to determine a difference therebetween. The difference between the predicted movement information and the actual movement information may specifically be determined from two dimensions of the movement direction and the movement distance.

Specifically, the movement angle of the actual movement information and the movement angle of the predicted movement information may be determined, and the two angles may be compared to obtain the angle difference. The moving distance of the actual moving information and the moving distance of the predicted moving information can be determined, and the distance difference is obtained by comparing the two distances.

Further, if the distance difference and the angle difference are both within the allowable range, the predicted cloud image and the actual cloud image can be considered to be consistent, otherwise, the predicted cloud image and the actual cloud image are considered to be inconsistent. If the cloud images are consistent, the predicted cloud image can be considered to be accurate.

In actual application, if the number of the cloud image is more than 2, the corresponding actual movement information and the predicted movement information can be compared, and whether the predicted cloud image is accurate or not can be evaluated according to a comparison result. For example, the actual movement information is from P_AMove to P_BFrom P_BMove to P_CFrom P_CMove to P_DPredicting the movement information as the slave P_A' move to P_BFrom P_B' move to P_CFrom P_C' move to P_D". Then the slave P can be compared separately_AMove to P_BAnd from P_A' move to P_BWhether or not it is consistent with P_BMove to P_CAnd from P_B' move to P_CWhether or not it is consistent with P_CMove to P_DAnd from P_C' move to P_DAnd if the cloud images are consistent, and evaluating the accuracy of the predicted cloud image according to the comparison result.

For example, if the comparison results of most of the movement information are consistent, the predicted cloud image may be considered to be more accurate, and if the comparison results of most of the movement information are inconsistent, the predicted cloud image may be considered to be inaccurate.

The method provided by the present embodiment is used for evaluating cloud images, and is performed by a device provided with the method provided by the present embodiment, and the device is generally implemented in a hardware and/or software manner.

The method for evaluating the accuracy of cloud image prediction provided by the embodiment comprises the following steps: acquiring at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images; respectively determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule; determining actual movement information according to the mass center of the actual cloud picture image, and determining predicted movement information according to the mass center of the predicted cloud picture image; and evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information. In the method provided by the embodiment, the moving information of the centroid in the cloud image is determined according to the centroids in at least two cloud image images, and then the centroid moving information in the actual cloud image is compared with the centroid moving information in the predicted cloud image, so that the difference between the cloud moving in the predicted cloud image and the cloud moving in the actual cloud image can be determined, and whether the predicted cloud image is accurate or not is evaluated.

Fig. 2 is a flowchart illustrating a method for evaluating accuracy of cloud image prediction according to another exemplary embodiment of the present application.

As shown in fig. 2, the method for evaluating the accuracy of cloud image prediction provided by this embodiment includes:

step 201, at least two actual cloud image images and a predicted cloud image corresponding to the actual cloud image images are obtained.

Step 201 is similar to the specific principle and implementation of step 101.

And the time information of the predicted cloud picture image corresponds to the time information of the actual cloud picture image.

Specifically, each acquired predicted cloud image corresponds to the time information of one acquired actual cloud image. For example, if two actual cloud images A, B are obtained, a predicted cloud image a 'corresponding to image a and a predicted cloud image B' corresponding to image B may be obtained.

Further, in order to make the actual cloud image comparable to the predicted cloud image, the actual cloud image and the predicted cloud image corresponding to the time information may be compared. That is, the predicted cloud image and the actual cloud image having the corresponding relationship correspond to each other in time information, for example, the time information of the image a and the image a 'are identical, and the time information of the image B and the image B' are identical.

In practical application, if the time information of the actual cloud picture image is t₁The cloud image is used to represent t₁The state of the cloud layer. If the time information of the predicted cloud picture is t₂The cloud image is used to predict t₂The state of the cloud layer.

The actual cloud image and the predicted cloud image having the corresponding relationship have the same time information. Such as time information t of image a₁Time information t with image A₂The same is true.

Step 202, determining the centroids of the actual cloud picture image and the predicted cloud picture image according to a preset rule; wherein, the preset rule comprises: determining pixel information of each pixel point in the image according to the cloud image; and determining the centroid of the cloud image according to the pixel information of each point.

The cloud image center determining method includes the steps that a rule for determining a cloud image can be preset, and the centers of mass of an actual cloud image and a predicted cloud image are respectively determined based on the rule. In order to make the centroids of the actual cloud image and the predicted cloud image comparable, the same rule may be used to determine the centroids in the cloud image.

Specifically, the centroid in the image can be determined from the pixel values of the respective points in the cloud image. Specifically, the pixel information of each pixel point in the image can be determined, and then the centroid of the cloud image is determined according to the pixel information of each point.

Further, for a cloud image, such as an actual cloud image or a predicted cloud image, the pixel values of each channel corresponding to each pixel point in the cloud image may be read. For example, a pixel value of a red channel, a pixel value of a green channel, and a pixel value of a blue channel of a pixel. That is, three pixel values can be obtained for one pixel point.

In practical application, a pixel value matrix corresponding to the cloud image can be obtained. The dimensions of the matrix are (m, n, c), where m, n are the number of rows and columns of the matrix and c is the number of color channels.

After the pixel value of each channel corresponding to each pixel point in the cloud image is read, the average value of the pixel values of the pixel points can be determined as pixel information. That is, the multi-channel pixel value of a pixel point is averaged to obtain the pixel information of the point.

Specifically, if the pixel values of the cloud image are recorded in a matrix form, the average pixel information of each pixel point can be calculated in the dimension c, so as to obtain a pixel information matrix with the dimension (m, n).

Fig. 3A is a schematic diagram illustrating a recording of pixel values in a matrix form according to an exemplary embodiment of the present application.

As shown in the left side of fig. 3A, for the pixel value of each channel corresponding to each read pixel, it is assumed that three channels are included, and the pixel values are represented by a, b, and c.

As shown in the right side of fig. 3A, pixel information obtained by processing the matrix on the left side is represented by d. dmn ═ amn + bmn + cmn)/3, specifically, the pixel average value is calculated in the channel dimension, and the pixel information matrix with dimension (m, n) is obtained.

After the pixel information of each pixel point in the cloud image is determined, the centroid of the cloud image can be determined according to the pixel information of each point.

Specifically, the pixel information may reflect the reflection of the radar signal at each position in the image, and the higher the pixel value is, the higher the reflectivity is, and the lower the pixel value is, the lower the reflectivity is.

Further, the centroid of the cloud image may be determined according to the following equation:

wherein I represents a row index of the centroid and J represents a column index of the centroid; i denotes a row index of a pixel in the cloud image N, j denotes a column index of a pixel in the cloud image N, N_i，jAnd representing the pixel information of the pixel points with indexes i and j.

Specifically, the centroid position in the cloud image can be determined according to the pixel information of each pixel point in the cloud image and the index of each pixel point.

Further, if the pixel information of the cloud image is expressed in a matrix form, each value in the matrix is the pixel information of one pixel point, and therefore, the row sequence and the column sequence of each value in the matrix are consistent with the row index and the column index of the pixel point.

And step 203, determining an actual motion vector according to the mass center of the actual cloud picture image and the time information of each actual cloud picture image.

And step 204, determining a predicted movement vector according to the mass center of the predicted cloud image and the time information of each predicted cloud image.

In practical application, the actual movement vector of the centroid can be determined according to the centroids of the at least two cloud image images.

The cloud image may have time information, such as cloud layer information indicating which time, or cloud layer information used to predict which time. The cloud image can be sorted according to the time information, and then the centroid with the sorting information is obtained. E.g. the ordered centroid is P_A、P_B. The motion vector is P_BPoint of direction P_AThe vector of (2).

In particular, if the centroid is from P_AMove to P_BThen canAccording to P_ARow index i_AColumn index j_A，P_BRow index i_BColumn index j_BAnd determining a motion vector.

Further, the motion vector V ═ i (i)_B-i_A，j_B-j_A)。

In practical application, if more than 2 cloud images are obtained, a motion vector can be obtained according to every two centroids adjacent in the time dimension. For example, a set of centroid movement vectors, V₁、V₂、V₃。

The motion vectors may be determined in the manner described above for both the actual cloud image and the predicted cloud image. For example, an actual motion vector Vg and a predicted motion vector Vp may be obtained.

The execution timing of steps 203 and 204 is not limited, and the actual motion vector may be determined first, the predicted motion vector may be determined first, or both the two motion vectors may be determined simultaneously.

Step 205, determining the moving distance difference and the moving direction included angle according to the actual moving vector and the predicted moving vector.

In particular, the movement distance of the actual centroid may be determined from the actual movement vector, e.g., the centroid is moved from P_AMove to P_BThe moving distance of (2). The movement distance of the predicted centroid may also be determined from the predicted movement vector, e.g., the centroid moves from P_A' move to P_BDistance of travel.

Further, if the motion vector V is equal to (i)_B-i_A，j_B-j_A) Then the moving distance is

In practical application, the centroid moving distance in the actual cloud picture image and the centroid moving distance in the predicted cloud picture image can be determined according to the formula. The timing of determining the two movement distances may not be limiting.

Wherein can be according toAnd determining the moving distance difference by the moving distance of the centroid in the actual cloud picture image and the moving distance of the centroid in the predicted cloud picture image. For example, the moving distance of the centroid in the actual cloud image is d₁Predicting the moving distance of the centroid in the cloud picture image as d₂Then the difference in the distance of movement may be:

Δd＝|d₁-d₂|

specifically, the moving distance difference can be used to represent the distance difference between the moving of the centroid in the actual cloud image and the predicted cloud image. If the predicted cloud image is accurate, for example, the predicted cloud image is consistent with the actual cloud image, the moving distance difference should be small. Therefore, the difference in the moving distance can be used as an index for evaluating whether the predicted cloud image is accurate.

Furthermore, the included angle of the moving direction can be determined according to the actual moving vector and the predicted moving vector. The moving direction included angle can be used for representing the angle deviation of the centroid movement in the actual cloud picture image and the predicted cloud picture image.

During practical application, if the moving distance of the centroid in the actual cloud image is Vg and the moving distance of the centroid in the cloud image is Vp, the included angle of the moving direction is as follows:

if the predicted cloud image is accurate, for example, the predicted cloud image is consistent with the actual cloud image, the moving direction angle should be small. Therefore, the moving direction angle can be used as an index for evaluating whether the predicted cloud image is accurate or not.

Specifically, the timing for determining the moving direction angle and the moving distance difference is not limited.

And step 206, evaluating whether the predicted cloud image is accurate according to the moving distance difference and the moving direction difference.

Further, an allowable deviation value may be set, which may be an angle deviation value corresponding to the direction angle, and a distance deviation value corresponding to the movement distance difference. If the moving distance difference is within the allowable deviation range, for example, less than or equal to the distance deviation value, and the moving direction angle is within the allowable deviation range, for example, less than or equal to the angle deviation value, it can be estimated that the predicted cloud image is accurate. Otherwise, the cloud image may be evaluated as inaccurate.

Fig. 3B is a diagram illustrating a motion vector according to an exemplary embodiment of the present application.

As shown in fig. 3B, the left side of the figure is an actual motion vector 31 determined from two actual cloud images, and the right side of the figure is a predicted motion vector 32 determined from predicted cloud images corresponding to the two actual cloud images.

The two motion vectors can be compared to determine the motion distance difference and the motion direction angle, and whether the predicted cloud image is consistent with the actual cloud image or not is determined according to the two indexes, so that the accuracy of the predicted cloud image is evaluated.

Optionally, in step 202, after determining the pixel information of each pixel point in the cloud image, the method further includes:

the pixel information smaller than the preset threshold is set to 0.

The centroid of the cloud image can then be determined from the adjusted pixel information.

A preset threshold may be set and pixel information smaller than the preset threshold may be set to 0. Noise points can be filtered in this way, and the accuracy of determining the mass center is prevented from being influenced.

In addition, the pixel information which meets the requirement range in the actual cloud image and the predicted cloud image can be reserved by adjusting the preset threshold value. Furthermore, according to the method provided by the embodiment, the difference of the centroid movement in the actual cloud image and the predicted cloud image in different intensity ranges is determined.

Fig. 4 is a block diagram of an apparatus for evaluating accuracy of cloud image prediction according to an exemplary embodiment of the present application.

As shown in fig. 4, the apparatus for evaluating the accuracy of cloud image prediction provided in this embodiment includes:

an obtaining module 41, configured to obtain at least two actual cloud image images and a predicted cloud image corresponding to the actual cloud image images;

a centroid determining module 42, configured to determine centroids of the actual cloud image and the predicted cloud image according to a preset rule;

a movement information determining module 43, configured to determine actual movement information according to the centroid of the actual cloud image, and determine predicted movement information according to the centroid of the predicted cloud image;

and the evaluation module 44 is configured to evaluate whether the predicted cloud image is accurate according to the actual movement information and the predicted movement information.

The device for evaluating the accuracy of cloud image prediction provided by the embodiment comprises: the acquisition module is used for acquiring at least two actual cloud picture images and a predicted cloud picture image corresponding to the actual cloud picture images; the mass center determining module is used for respectively determining the mass centers of the actual cloud picture image and the predicted cloud picture image according to a preset rule; the movement information determining module is used for determining actual movement information according to the mass center of the actual cloud picture image and determining predicted movement information according to the mass center of the predicted cloud picture image; and the evaluation module is used for evaluating whether the predicted cloud picture image is accurate or not according to the actual movement information and the predicted movement information. In the device provided by this embodiment, the movement information of the centroid in the cloud image is determined according to the centroids in at least two cloud image images, and then the centroid movement information in the actual cloud image is compared with the centroid movement information in the predicted cloud image, so that the difference between the movement of the cloud in the predicted cloud image and the movement of the cloud in the actual cloud image can be determined, thereby evaluating whether the predicted cloud image is accurate.

The specific principle and implementation of the device for evaluating the accuracy of cloud image prediction provided by this embodiment are similar to those of the embodiment shown in fig. 1, and are not described herein again.

Fig. 5 is a block diagram of an apparatus for evaluating accuracy of cloud image prediction according to another exemplary embodiment of the present application.

As shown in fig. 5, on the basis of the foregoing embodiment, in the apparatus for evaluating cloud image prediction accuracy provided in this embodiment, optionally, the preset rule includes:

determining pixel information of each pixel point in the image according to the cloud image;

and determining the centroid of the cloud image according to the pixel information of each point.

Optionally, the centroid determining module 42 includes a pixel information determining unit 421, configured to read a pixel value of each channel corresponding to each pixel point in the cloud image, and determine an average value of the pixel values of the pixel points as the pixel information.

Optionally, the centroid determining module 42 includes a centroid determining unit 422, configured to determine a centroid of the cloud image according to the following formula:

wherein I represents a row index of the centroid and J represents a column index of the centroid; i denotes a row index of a pixel in the cloud image N, j denotes a column index of a pixel in the cloud image N, N_i，jAnd representing the pixel information of the pixel points with indexes i and j.

Optionally, the movement information determining module 43 is specifically configured to:

determining an actual movement vector according to the centroid of the actual cloud picture image and the time information of each actual cloud picture image;

and determining a predicted movement vector according to the centroid of the predicted cloud picture image and the time information of each predicted cloud picture image.

Optionally, the evaluation module 44 includes:

an index determining unit 441, configured to determine a moving distance difference and a moving direction included angle according to the actual moving vector and the predicted moving vector;

an evaluating unit 442, configured to evaluate whether the predicted cloud image is accurate according to the moving distance difference and the moving direction difference.

Optionally, the predicted cloud image corresponds to time information of the actual cloud image.

Optionally, after determining the pixel information of each pixel point in the image according to the cloud image, the centroid determining module 42 is further configured to:

setting the pixel information smaller than a preset threshold value to 0.

The specific principle and implementation of the apparatus provided in this embodiment are similar to those of the embodiment shown in fig. 2, and are not described here again.

Fig. 6 is a block diagram of an apparatus for evaluating accuracy of cloud image prediction according to an exemplary embodiment of the present application.

As shown in fig. 6, the apparatus for evaluating accuracy of cloud image prediction provided in this embodiment includes:

a memory 61;

a processor 62; and

a computer program;

wherein the computer program is stored in the memory 61 and configured to be executed by the processor 62 to implement any one of the cloud image prediction accuracy assessment methods described above.

The present embodiments also provide a computer-readable storage medium, having stored thereon a computer program,

the computer program is executed by a processor to realize any one of the above methods for evaluating the accuracy of cloud image prediction.

The embodiment also provides a computer program, which includes a program code, and when the computer program is executed by a computer, the program code executes any one of the above methods for evaluating the prediction accuracy of a cloud image.

Those of ordinary skill in the art will understand that: all or a portion of the steps of implementing the above-described method embodiments may be performed by hardware associated with program instructions. The program may be stored in a computer-readable storage medium. When executed, the program performs steps comprising the method embodiments described above; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.

Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present application.