CN112802063A - Satellite cloud picture prediction method and device and computer readable storage medium - Google Patents
Satellite cloud picture prediction method and device and computer readable storage medium Download PDFInfo
- Publication number
- CN112802063A CN112802063A CN202110157314.2A CN202110157314A CN112802063A CN 112802063 A CN112802063 A CN 112802063A CN 202110157314 A CN202110157314 A CN 202110157314A CN 112802063 A CN112802063 A CN 112802063A
- Authority
- CN
- China
- Prior art keywords
- image
- satellite
- motion vector
- convolution
- satellite cloud
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Abstract
The invention discloses a method and a device for predicting a satellite cloud picture and a computer readable storage medium, wherein the method comprises the following steps: acquiring a first number of continuous historical satellite clouds in a first preset time interval; performing convolution operation on each historical satellite cloud image to obtain a convolution image; determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point; and determining a second number of continuous prediction satellite clouds in a second preset time interval according to the convolution images and the first motion vector. The method and the device enable the generated predicted satellite cloud picture to be more accurate.
Description
Technical Field
The present invention relates to the field of image recognition technologies, and in particular, to a method and an apparatus for predicting a satellite cloud map, and a computer-readable storage medium.
Background
The photovoltaic power generation industry has higher requirements on the solar radiation intensity, and the solar radiation intensity of certain regions can be effectively acquired. The current effective method is to establish some ground observation stations in the area and install an irradiation acquisition instrument at the observation stations to realize the real-time recording of irradiation intensity. The solar irradiation intensity calculation needs not only to calculate the irradiation intensity at the current moment, but also to predict the irradiation intensity at the future moment.
At present, many ground large-scale photovoltaic power stations are provided with light power prediction systems. With the rapid development of meteorological satellite imaging technology, the optical power prediction system obtains a large number of satellite cloud images with high space-time resolution, including visible light and near infrared satellite cloud images. Due to the fact that the observation range of the foundation cloud picture is small, the situation that cloud clusters in the satellite cloud picture move fast possibly exists, the cloud clusters move out of the observation range of the satellite cloud picture within the prediction time, and therefore the satellite cloud picture prediction result is inaccurate.
Disclosure of Invention
The invention mainly aims to provide a satellite cloud picture prediction method, a satellite cloud picture prediction device and a computer readable storage medium, and aims to solve the problems of information security, privacy disclosure and other risks existing in the existing identification mode.
In order to achieve the above object, the present invention provides a method for predicting a satellite cloud picture, including the following steps:
acquiring a first number of continuous historical satellite clouds in a first preset time interval;
performing convolution operation on each historical satellite cloud image to obtain a convolution image;
determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point;
and determining a second number of continuous prediction satellite clouds in a second preset time interval according to the convolution images and the first motion vector.
In an embodiment, the step of determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point includes:
and inputting the convolution images of the adjacent time points as input values into a preset optical flow estimation sub-network to obtain the first motion vector.
In one embodiment, the step of determining a second number of consecutive predicted satellite clouds in a second preset time interval according to the convolved image and the first motion vector comprises:
inputting the first motion vector and the convolution image as input values into a trajectory gating recursion unit to obtain the convolution image marked with the first motion vector;
determining a second number of consecutive predicted satellite clouds from the convolved image labeled with the first motion vector.
In an embodiment, the step of performing convolution operation on each historical satellite cloud image to obtain a convolution image includes:
performing down-sampling operation on each historical satellite cloud picture according to a preset number of sampling layers to obtain a down-sampling image;
carrying out convolution processing on the downsampled image sequentially through each convolution layer according to the number of sampling layers to obtain a convolution image output by each sampling layer;
the step of determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point comprises:
respectively determining a motion vector corresponding to the convolution image according to the convolution image of the adjacent time point of each sampling layer;
and determining the first motion vector of the corresponding convolution image according to the motion vector of each sampling layer.
In one embodiment, the step of determining a second number of consecutive predicted satellite clouds in a second preset time interval according to the convolved image and the first motion vector comprises:
determining the convolution image in a second preset time interval according to the convolution image in the first preset time interval of each sampling layer and the first motion vector;
performing upsampling operation on each convolution image in a second preset time interval of each sampling layer to obtain a second number of upsampled images;
and determining a predicted satellite cloud picture according to the up-sampling image of each sampling layer.
In one embodiment, the method for predicting the satellite cloud picture comprises the following steps:
acquiring a first number of continuous historical satellite clouds in a first preset time interval;
inputting the continuous historical satellite cloud pictures into a prediction model;
determining a second number of consecutive predicted satellite clouds by the prediction model.
In an embodiment, after the step of determining the second number of consecutive predicted satellite clouds by the prediction model, the method further includes:
determining gray values of pixel points of the real satellite cloud picture at different time points to determine a gray value function;
determining a gray scale change factor and a smooth factor of the real satellite cloud picture according to the gray scale value function;
and determining a real motion vector of the real satellite cloud picture according to the gray scale change factor, the smoothing factor and a preset energy function.
Determining a second motion vector of the predicted satellite cloud picture through the prediction model;
determining an error value from the true motion vector and the second motion vector;
and if the error value is larger than a preset threshold value, correcting the prediction model according to the real satellite cloud picture and the real motion vector.
In an embodiment, before the step of inputting the continuous historical satellite cloud image into a prediction model, the method further includes:
acquiring a training set, wherein the training set comprises a continuous satellite training cloud picture and a real satellite cloud picture;
training a preset neural network model through a training set;
calculating a loss value of the neural network model by adopting a preset loss function;
when the loss value is smaller than a preset threshold value, judging that the neural network model is converged;
and saving the converged neural network model as the prediction model.
In order to achieve the above object, the present invention further provides a satellite cloud image prediction apparatus, which includes a memory, a processor, and a satellite cloud image prediction program stored in the memory and executable on the processor, wherein the satellite cloud image prediction program, when executed by the processor, implements the steps of the satellite cloud image prediction method described above.
To achieve the above object, the present invention further provides a computer readable storage medium storing a prediction program of a satellite cloud image, wherein the prediction program of the satellite cloud image is executed by a processor to implement the steps of the prediction method of the satellite cloud image as described above.
According to the satellite cloud picture prediction method, the satellite cloud picture prediction device and the computer readable storage medium, after a first number of continuous historical satellite cloud pictures within a first preset time interval are obtained, convolution operation is conducted on each historical satellite cloud picture to obtain a convolution image; and determining a first motion vector of a cloud layer in the convolution image according to the convolution image of the adjacent time point, and determining a second number of continuous prediction satellite cloud images in a second preset time interval according to the convolution image and the first motion vector. Through the first motion vector of the cloud layer in the convolution image, the change trend of the cloud layer can be determined, and the generated predicted satellite cloud image is more accurate.
Drawings
Fig. 1 is a schematic hardware configuration diagram of a satellite cloud prediction apparatus according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a method for predicting a satellite cloud according to a first embodiment of the present invention;
FIG. 3 is a flowchart illustrating a second embodiment of a satellite cloud prediction method according to the present invention;
fig. 4 is a detailed flowchart of step S40 of the method for predicting a satellite cloud according to the third embodiment of the present invention;
FIG. 5 is a flowchart illustrating a method for predicting a satellite cloud according to a fourth embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a prediction model of the satellite cloud image prediction method according to the present invention;
fig. 7 is a detailed flowchart of step S40 of the method for predicting a satellite cloud according to the sixth embodiment of the present invention;
FIG. 8 is a flowchart illustrating a method for predicting a satellite cloud according to a seventh embodiment of the present invention;
FIG. 9 is a schematic diagram of a real satellite cloud of the method for predicting a satellite cloud according to the present invention;
FIG. 10 is a schematic diagram of a predicted satellite cloud image predicted by a trajectory gating recursion unit of the satellite cloud image prediction method of the present invention;
FIG. 11 is a schematic diagram of a predicted satellite cloud image predicted by a stream estimation sub-network plus a trajectory gating recursion unit according to the satellite cloud image prediction method of the present invention;
fig. 12 is a flowchart illustrating an eighth embodiment of a satellite cloud prediction method according to the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The main solution of the embodiment of the invention is as follows: acquiring a first number of continuous historical satellite clouds in a first preset time interval; performing convolution operation on each historical satellite cloud image to obtain a convolution image; determining a first motion vector according to a convolution image corresponding to a historical satellite cloud image of an adjacent time point; and determining a second number of continuous prediction satellite clouds in a second preset time interval according to the convolution images and the first motion vector. Through the first motion vector of the cloud layer in the convolution image, the change trend of the cloud layer can be determined, and the generated predicted satellite cloud image is more accurate.
As an implementation, the prediction device of the satellite cloud map may be as shown in fig. 1.
The embodiment scheme of the invention relates to a satellite cloud picture prediction device, which comprises: a processor 101, e.g. a CPU, a memory 102, a communication bus 103. Wherein a communication bus 103 is used for enabling the connection communication between these components.
The memory 102 may be a high-speed RAM memory or a non-volatile memory (e.g., a disk memory). As shown in fig. 1, the memory 102, which is a computer-readable storage medium, may include therein a prediction program of a satellite cloud; and the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
acquiring a first number of continuous historical satellite clouds in a first preset time interval;
performing convolution operation on each historical satellite cloud image to obtain a convolution image;
determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point;
and determining a second number of continuous prediction satellite clouds in a second preset time interval according to the convolution images and the first motion vector.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
and inputting the convolution images of the adjacent time points as input values into a preset optical flow estimation sub-network to obtain the first motion vector.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
inputting the first motion vector and the convolution image as input values into a trajectory gating recursion unit to obtain the convolution image marked with the first motion vector;
determining a second number of consecutive predicted satellite clouds from the convolved image labeled with the first motion vector.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
performing down-sampling operation on each historical satellite cloud picture according to a preset number of sampling layers to obtain a down-sampling image;
carrying out convolution processing on the downsampled image sequentially through each convolution layer according to the number of sampling layers to obtain a convolution image output by each sampling layer;
the step of determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point comprises:
respectively determining a motion vector corresponding to the convolution image according to the convolution image of the adjacent time point of each sampling layer;
and determining the first motion vector of the corresponding convolution image according to the motion vector of each sampling layer.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
determining the convolution image in a second preset time interval according to the convolution image in the first preset time interval of each sampling layer and the first motion vector;
performing upsampling operation on each convolution image in a second preset time interval of each sampling layer to obtain a second number of upsampled images;
and determining a predicted satellite cloud picture according to the up-sampling image of each sampling layer.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
acquiring a first number of continuous historical satellite clouds in a first preset time interval;
inputting the continuous historical satellite cloud pictures into a prediction model;
determining a second number of consecutive predicted satellite clouds by the prediction model.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
determining gray values of pixel points of the real satellite cloud picture at different time points to determine a gray value function;
determining a gray scale change factor and a smooth factor of the real satellite cloud picture according to the gray scale value function;
and determining a real motion vector of the real satellite cloud picture according to the gray scale change factor, the smoothing factor and a preset energy function.
Determining a second motion vector of the predicted satellite cloud picture through the prediction model;
determining an error value from the true motion vector and the second motion vector;
and if the error value is larger than a preset threshold value, correcting the prediction model according to the real satellite cloud picture and the real motion vector.
In one embodiment, the processor 101 may be configured to call a prediction program of the satellite cloud map stored in the memory 102, and perform the following operations:
acquiring a training set, wherein the training set comprises a continuous satellite training cloud picture and a real satellite cloud picture;
training a preset neural network model through a training set;
calculating a loss value of the neural network model by adopting a preset loss function;
when the loss value is smaller than a preset threshold value, judging that the neural network model is converged;
and saving the converged neural network model as the prediction model.
Based on the hardware architecture of the satellite cloud image prediction device, the embodiment of the satellite cloud image prediction method is provided.
Referring to fig. 2, fig. 2 is a first embodiment of a method for predicting a satellite cloud image according to the present invention, where the method for predicting a satellite cloud image includes the following steps:
step S10, a first number of consecutive historical satellite clouds in a first predetermined time interval is obtained.
Specifically, the historical satellite cloud picture is an image of cloud cover and ground surface features on the earth observed from top to bottom by a meteorological satellite, and the historical satellite cloud picture is satellite cloud picture data before a predicted time point.
For example, if the first preset time interval may be 2 hours and the first number is 12, the historical satellite clouds are acquired every 10 minutes, and the historical satellite clouds of 2 hours are arranged according to the time sequence, so that the continuous historical satellite clouds of 12 time points in total can be obtained.
Step S20, performing convolution operation on each historical satellite cloud image to obtain a convolution image;
specifically, convolution operation is performed on each historical satellite cloud image to obtain a convolution image, each historical satellite cloud image can correspond to a plurality of convolution images, illustratively, downsampling operation is performed on each historical satellite cloud image respectively, when the number of sampling layers is 3, each historical satellite cloud image corresponds to 3 downsampling images, and the convolution operation is performed on the 3 downsampling images to obtain the convolution image corresponding to each historical satellite cloud image.
Step S30, determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point;
specifically, a first motion vector of one or more cloud layers in the historical satellite cloud image is determined according to a convolution image of the historical satellite cloud image of the adjacent time point, wherein the first motion vector comprises the motion speed and the motion direction of the cloud layer, and the motion direction can be the direction along the x axis and the y axis of the coordinate axis.
Step S40, determining a second number of consecutive predicted satellite clouds within a second predetermined time interval according to the convolved image and the first motion vector.
Specifically, a second number of continuous predicted satellite clouds in a second preset time interval is determined according to the convolution images and the first motion vectors corresponding to the convolution images, for example, continuous historical satellite clouds at 12 time points in 2 hours are determined, the convolution images and the first motion vectors of the historical satellite clouds are respectively determined, and continuous predicted satellite clouds at 24 time points in 4 hours after the determination can be determined according to the convolution images and the first motion vectors.
In the technical scheme of the embodiment, after a first number of continuous historical satellite clouds in a first preset time interval are obtained, convolution operation is performed on each historical satellite cloud image to obtain a convolution image; and determining a first motion vector of a cloud layer in the convolution image according to the convolution image of the adjacent time point, and determining a second number of continuous prediction satellite cloud images in a second preset time interval according to the convolution image and the first motion vector. Through the first motion vector of the cloud layer in the convolution image, the change trend of the cloud layer can be determined, and the generated predicted satellite cloud image is more accurate.
Referring to fig. 3, fig. 3 is a second embodiment of the method for predicting a satellite cloud map according to the present invention, wherein the step S30 includes:
step S31, inputting the convolution images of adjacent time points as input values into a preset optical flow estimation sub-network to obtain the first motion vector.
Specifically, the optical flow estimation subnetwork is used for extracting the motion vector features of the image, and the optical flow estimation subnetwork can be expressed by the following formula:
Ut,Vt＝γ(Xt,Ht-1)
where γ denotes the optical flow estimation subnetwork, Ht-1Represents the convolution image at time t-1; xtA convolution image representing time t; u shapet,VtRepresenting the motion vectors in the satellite cloud at time t.
And inputting the convolution images of the adjacent time points as input values into a preset optical flow estimation sub-network to extract a first motion vector of the cloud layer.
In the technical solution of this embodiment, a first motion vector of a convolution image is obtained by inputting convolution images of adjacent time points into a preset optical flow estimation sub-network. Extracting the first motion vector of the convolution image through the optical flow estimation sub-network facilitates prediction of the historical satellite cloud image.
Referring to fig. 4, fig. 4 is a third embodiment of the method for predicting a satellite cloud map according to the present invention, where the step S40 includes:
step S41, inputting the first motion vector and the convolved image as input values into a trajectory-gated recursion unit to obtain the convolved image labeled with the first motion vector;
step S42, determining a second number of consecutive predicted satellite clouds based on the convolved images marked with the first motion vector.
Specifically, the trajectory-gated recursion unit is configured to apply the first motion vector of each cloud layer to the convolved image, and the trajectory-gated recursion unit may be as follows:
M＝warp(I,U,V)
wherein, I is the representation of the input data, which is referred to as a convolution image, U, V are the motion vectors in the satellite cloud image, U is the velocity along the x-axis direction, V is the velocity along the y-axis direction, and the warp function represents the interaction of the motion vectors with the input convolution image.
The following formula:
wherein Z istRepresents an update gate; rtRepresents a reset gate; f represents an activation function; σ represents an activation function; denotes a convolution operation;
representing a dot product operation; w represents a weight parameter; l represents a network layer number; l represents the total number of network layers; t represents a time point; ht-1A convolution image representing the output at time t-1; xtA convolution image representing the input at time t; htA convolution image representing the output at time t; ht' represents the convolution image calculated in the trajectory-gated recursion unit at the time t. U shapet,Vt∈RL×H×WIs a flow field storing the local connection structure generated by the structure generation network gamma,
is the weight of the projection channel, which is achieved by a 1 × 1 convolution. warp (H)t-1,Ut,l,Vt,l) Function is run through bilinear sampling kernel from Ht-1In selecting Ut,l,Vt,lThe indicated position.
If M is warp (I, U, V) where M, I ∈ RC×H×WAnd U, V ∈ RH×WAnother formula for a trajectory-gated recursion unit can be determined, which is used to apply the first motion vector to the convolved image extracted by convolution, as follows:
wherein, VijIs the velocity vector of the (i, j) pixel point along the y-axis direction, UijThe velocity vector of the (i, j) pixel point along the x-axis direction is shown, and c is the number of sampling layers.
In the technical scheme of the embodiment, the track gating recursion unit is used for enabling the convolution image to interact with the first motion vector, the first motion vector is marked on the convolution image, the convolution image marked with the first motion vector is used for predicting the predicted satellite cloud image subsequently, and the accuracy of prediction is improved.
Referring to fig. 5, fig. 5 is a fourth embodiment of the method for predicting a satellite cloud image according to the present invention, where, based on any one of the first to third embodiments, the step S20 includes:
step S21, carrying out down-sampling operation on each historical satellite cloud picture according to a preset sampling layer number to obtain a down-sampled image;
step S22, sequentially carrying out convolution processing on the downsampled image by each convolution layer according to the number of sampling layers to obtain a convolution image output by each sampling layer;
the step S30 includes:
step S32, respectively determining a motion vector corresponding to the convolution image according to the convolution image of the adjacent time point of each sampling layer;
step S33, determining the first motion vector of the corresponding convolution image according to the motion vector of each sampling layer.
Specifically, downsampling operation is performed on the historical satellite cloud image according to the preset number of sampling layers to obtain downsampled images, convolution processing is performed on the downsampled images corresponding to each sampling layer to obtain convolution images, first motion vectors corresponding to the convolution images are determined according to the convolution images of adjacent time points of each sampling layer, and the first motion vectors of the corresponding convolution images are determined according to the first motion vectors of each sampling layer. As shown in fig. 6, in the coding structure, an optical flow estimation sub-network is represented by flow, a trajectory gating recursion unit is represented by a unit 1, a unit 2, and a unit 3, the number of sampling layers is 3, each sampling layer performs downsampling processing on a historical satellite cloud image, the obtained 36 downsampled images are respectively subjected to convolution processing to obtain 36 convolution images, and 36 first motion vectors can be determined according to the convolution images of adjacent time points.
In the technical scheme of the embodiment, the historical satellite cloud image is subjected to downsampling operation to obtain downsampled images, the downsampled image of each sampling layer is obtained, and convolution processing is respectively carried out on the downsampled images of each downsampling layer according to the sequence of the number of sampling layers from small to large. And determining a plurality of first motion vectors according to the convolution images of the plurality of sampling layers for predicting the satellite cloud picture, so that the prediction result of the satellite cloud picture is more accurate.
Referring to fig. 7, fig. 7 is a fifth embodiment of the method for predicting a satellite cloud image according to the present invention, where, based on any one of the first embodiment and the fourth embodiment, the step S40 includes:
step S43, determining the convolution image in a second preset time interval according to the convolution image in a first preset time interval of each sampling layer and the first motion vector;
step S44, performing upsampling operation on each convolution image in a second preset time interval of each sampling layer to obtain a second number of upsampled images;
and step S45, determining a predicted satellite cloud picture according to the up-sampling image of each sampling layer.
Specifically, as shown in the decoding structure of fig. 6, according to a first number of convolution images and first motion vectors in a first preset time interval of each sampling layer, a second motion vector corresponding to a convolution image and a convolution image in a second preset time interval may be determined, an upsampling operation is performed on each convolution image in the second preset time interval of each sampling layer to obtain a second number of upsampled images, and a predicted satellite cloud image is determined according to each upsampled image. The convolved image and the second motion vector may also be determined from the first motion vector and the convolved image.
In the technical scheme of this embodiment, a convolution image in a second preset time interval is obtained according to the convolution image in the first preset time interval of each sampling layer and the first motion vector prediction, an up-sampling operation is performed on the convolution image according to a prediction time point, and a predicted satellite cloud image is determined according to the up-sampled image, so that the predicted satellite cloud image result obtained through prediction is more accurate.
Referring to fig. 8, fig. 8 is a sixth embodiment of a method for predicting a satellite cloud image according to the present invention, where the method for predicting a satellite cloud image includes:
step S50, acquiring a first number of continuous historical satellite clouds in a first preset time interval;
step S60, inputting the continuous historical satellite cloud pictures into a prediction model;
and step S70, determining a second number of continuous prediction satellite clouds through the prediction model.
Specifically, a first number of continuous historical satellite clouds in a first preset time interval are obtained, and the continuous historical satellite clouds are input into a prediction model, wherein the prediction model can comprise an optical flow estimation sub-network and a track gating recursion unit. As shown in fig. 9, 10, and 11, fig. 9 is a real satellite cloud image, fig. 10 is a predicted satellite cloud image predicted by a trajectory gating recursion unit, and fig. 11 is a predicted satellite cloud image predicted by an optical flow estimation sub-network plus the trajectory gating recursion unit, and it can be seen from the drawings that the blur phenomenon of the predicted satellite cloud image in fig. 10 is reduced and the cloud layer deformation is reduced compared with the predicted satellite cloud image in fig. 11.
Before inputting the continuous historical satellite cloud pictures into a prediction model, obtaining the prediction model by training, and obtaining a training set, wherein the training set comprises continuous satellite training cloud pictures and real satellite cloud pictures; training a preset neural network model through a training set; calculating a loss value of the neural network model by adopting a preset loss function; when the loss value is smaller than a preset threshold value, judging that the neural network model is converged; and saving the converged neural network model as the prediction model. The loss function preset here may be a softmax function.
In the technical scheme of the embodiment, the input continuous historical satellite cloud pictures of the first quantity are predicted through the trained prediction model to obtain the continuous prediction satellite cloud pictures of the second quantity, so that the prediction efficiency of the prediction satellite cloud pictures is improved.
Referring to fig. 12, fig. 12 is a seventh embodiment of the method for predicting a satellite cloud map according to the present invention, and based on the sixth embodiment, after step S70, the method further includes:
step S80, determining a gray value function according to the gray values of the pixel points of the real satellite cloud picture at different time points;
step S90, determining a gray scale change factor and a smooth factor of the real satellite cloud picture according to the gray scale value function;
step S100, determining a real motion vector of the real satellite cloud picture according to the gray scale change factor, the smoothing factor and a preset energy function;
step S110, determining a second motion vector of the predicted satellite cloud picture through the prediction model;
step S120, determining an error value according to the real motion vector and the second motion vector;
and step S130, if the error value is larger than a preset threshold value, correcting the parameters of the prediction model according to the real satellite cloud picture and the real motion vector.
Specifically, the second motion vector of the real satellite cloud picture is calculated through an optical flow algorithm, the real motion vector is extracted, and the motion speed and the motion direction of the cloud layer in the real satellite cloud picture are obtained.
The real motion vector of the optical flow in the real satellite cloud image can be represented as (u, v), and a gray value function I (x, y, t) is determined according to gray values of pixel points of the real satellite cloud image at different time points, wherein (x, y) is the gray value of the pixel point, and (x, y, t) is the gray value of the pixel point at the time point t.
Determining a gray scale change factor and a smooth factor of a real satellite cloud picture according to a gray scale value function based on a gray scale invariant hypothesis and an optical flow field smooth hypothesis of an optical flow algorithm, wherein the gray scale change factor is xibRepresenting smoothing factors by
Representing the calculation of u and v in the optical flow, as follows:
ξb＝uIx+vIy+It
wherein the content of the first and second substances,
for Laplace, Ix is the partial derivative for x, Iy is the partial derivative for y, It is the partial derivative for t, u is the velocity along the x-axis, and v is the velocity along the y-axis.
Two components u (x, y) and v (x, y) of the optical flow are solved based on the gray scale change factor and the smoothing factor, and an energy function is defined as follows,
substituting the specific gray scale change factor and the smoothing factor to obtain a real motion vector, wherein the following formula is shown as follows:
wherein the content of the first and second substances,
is the laplacian operator.
And determining the real motion vector of the real satellite cloud picture corresponding to the adjacent moment point by adopting an optical flow algorithm, and performing error calculation with a second motion vector output by the prediction model. The prediction model can comprise an optical flow estimation sub-network and a track gating recursion unit, and if the error value is larger than a preset threshold value, the optical flow sub-network parameters in the prediction model are updated according to a real satellite cloud image and the real motion vector.
In the technical scheme of the embodiment, an optical flow algorithm is adopted to obtain a real motion vector of a real satellite cloud picture, an error value of the real motion vector and a second motion vector is determined, and when the error value is greater than a preset threshold value, a first motion vector is extracted from an optical flow sub-network in a prediction model for supervised training learning, so that the predicted satellite cloud picture obtained through prediction is more accurate.
The invention further provides a satellite cloud picture prediction device, which includes a memory, a processor, and a satellite cloud picture prediction program stored in the memory and executable on the processor, and when the satellite cloud picture prediction program is executed by the processor, the steps of the satellite cloud picture prediction method according to the above embodiment are implemented.
The present invention further provides a computer-readable storage medium, which stores a prediction program of a satellite cloud image, and when the prediction program of the satellite cloud image is executed by a processor, the steps of the prediction method of the satellite cloud image according to the above embodiments are implemented.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a computer-readable storage medium (such as ROM/RAM, magnetic disk, optical disk) as described above, and includes several instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A satellite cloud picture prediction method is characterized by comprising the following steps:
acquiring a first number of continuous historical satellite clouds in a first preset time interval;
performing convolution operation on each historical satellite cloud image to obtain a convolution image;
determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point;
and determining a second number of continuous prediction satellite clouds in a second preset time interval according to the convolution images and the first motion vector.
2. The method of claim 1, wherein the step of determining the first motion vector according to the convolved images corresponding to the historical satellite clouds at the neighboring time points comprises:
and inputting the convolution images of the adjacent time points as input values into a preset optical flow estimation sub-network to obtain the first motion vector.
3. The method of predicting satellite clouds according to claim 1, wherein the step of determining a second number of consecutive predicted satellite clouds within a second preset time interval based on the convolved image and the first motion vector comprises:
inputting the first motion vector and the convolution image as input values into a trajectory gating recursion unit to obtain the convolution image marked with the first motion vector;
determining a second number of consecutive predicted satellite clouds from the convolved image labeled with the first motion vector.
4. The method for predicting satellite clouds according to claim 1, wherein the step of performing convolution operation on each historical satellite cloud image to obtain a convolution image comprises:
performing down-sampling operation on each historical satellite cloud picture according to a preset number of sampling layers to obtain a down-sampling image;
carrying out convolution processing on the downsampled image sequentially through each convolution layer according to the number of sampling layers to obtain a convolution image output by each sampling layer;
the step of determining a first motion vector according to the convolution image corresponding to the historical satellite cloud image of the adjacent time point comprises:
respectively determining a motion vector corresponding to the convolution image according to the convolution image of the adjacent time point of each sampling layer;
and determining the first motion vector of the corresponding convolution image according to the motion vector of each sampling layer.
5. The method of predicting satellite clouds according to claim 4, wherein said step of determining a second number of consecutive predicted satellite clouds within a second predetermined time interval based on said convolved image and said first motion vector comprises:
determining the convolution image in a second preset time interval according to the convolution image in the first preset time interval of each sampling layer and the first motion vector;
performing upsampling operation on each convolution image in a second preset time interval of each sampling layer to obtain a second number of upsampled images;
and determining a predicted satellite cloud picture according to the up-sampling image of each sampling layer.
6. A satellite cloud picture prediction method is characterized by comprising the following steps:
acquiring a first number of continuous historical satellite clouds in a first preset time interval;
inputting the continuous historical satellite cloud pictures into a prediction model;
determining a second number of consecutive predicted satellite clouds by the prediction model.
7. The method of predicting satellite clouds of claim 6, wherein said step of determining a second number of successive predicted satellite clouds by said prediction model is followed by further comprising:
determining gray values of pixel points of the real satellite cloud picture at different time points to determine a gray value function;
determining a gray scale change factor and a smooth factor of the real satellite cloud picture according to the gray scale value function;
and determining a real motion vector of the real satellite cloud picture according to the gray scale change factor, the smoothing factor and a preset energy function.
Determining a second motion vector of the predicted satellite cloud picture through the prediction model;
determining an error value from the true motion vector and the second motion vector;
and if the error value is larger than a preset threshold value, correcting the parameters of the prediction model according to the real satellite cloud picture and the real motion vector.
8. The method for predicting satellite clouds according to claim 6, wherein said step of inputting said continuous historical satellite clouds into a prediction model is preceded by the steps of:
acquiring a training set, wherein the training set comprises a continuous satellite training cloud picture and a real satellite cloud picture;
training a preset neural network model through a training set;
calculating a loss value of the neural network model by adopting a preset loss function;
when the loss value is smaller than a preset threshold value, judging that the neural network model is converged;
and saving the converged neural network model as the prediction model.
9. An apparatus for predicting a satellite cloud map, the apparatus comprising a memory, a processor, and a satellite cloud map prediction program stored in the memory and executable on the processor, wherein the satellite cloud map prediction program, when executed by the processor, implements the steps of the method for predicting a satellite cloud map according to any one of claims 1 to 8.
10. A computer-readable storage medium, wherein the computer-readable storage medium stores a prediction program of a satellite cloud image, and the prediction program of the satellite cloud image realizes the steps of the prediction method of the satellite cloud image according to any one of claims 1 to 8 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110157314.2A CN112802063A (en) | 2021-02-03 | 2021-02-03 | Satellite cloud picture prediction method and device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110157314.2A CN112802063A (en) | 2021-02-03 | 2021-02-03 | Satellite cloud picture prediction method and device and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112802063A true CN112802063A (en) | 2021-05-14 |
Family
ID=75814356
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110157314.2A Pending CN112802063A (en) | 2021-02-03 | 2021-02-03 | Satellite cloud picture prediction method and device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112802063A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113420939A (en) * | 2021-07-15 | 2021-09-21 | 上海眼控科技股份有限公司 | Cloud picture forecasting method, equipment and storage medium |
| CN113724287A (en) * | 2021-09-02 | 2021-11-30 | 北京华云星地通科技有限公司 | Satellite cloud picture prediction method and system |
| CN114139760A (en) * | 2021-09-29 | 2022-03-04 | 苏州浪潮智能科技有限公司 | Method, system, storage medium and equipment for predicting typhoon path |
| CN115661277A (en) * | 2022-10-20 | 2023-01-31 | 中山大学 | Typhoon cloud picture extrapolation method, system, equipment and medium based on variational self-coding |
| CN115984688A (en) * | 2022-12-20 | 2023-04-18 | 中国科学院空天信息创新研究院 | Continuous frame-oriented double-current cloud picture prediction method, storage medium and equipment |
| CN117239744A (en) * | 2023-11-15 | 2023-12-15 | 威派(武汉)高新技术有限公司 | Ultra-short-term photovoltaic power prediction method integrating wind cloud No. 4 meteorological satellite data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016205930A (en) * | 2015-04-20 | 2016-12-08 | 清水建設株式会社 | Weather information prediction device, weather information prediction method and program |
| CN107563397A (en) * | 2016-06-30 | 2018-01-09 | 中国电力科学研究院 | Cloud cluster method for calculation motion vector in a kind of satellite cloud picture |
| US20190220746A1 (en) * | 2017-08-29 | 2019-07-18 | Boe Technology Group Co., Ltd. | Image processing method, image processing device, and training method of neural network |
| CN111798484A (en) * | 2020-06-30 | 2020-10-20 | 武汉大学 | Continuous dense optical flow estimation method and system based on event camera |
-
2021
- 2021-02-03 CN CN202110157314.2A patent/CN112802063A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016205930A (en) * | 2015-04-20 | 2016-12-08 | 清水建設株式会社 | Weather information prediction device, weather information prediction method and program |
| CN107563397A (en) * | 2016-06-30 | 2018-01-09 | 中国电力科学研究院 | Cloud cluster method for calculation motion vector in a kind of satellite cloud picture |
| US20190220746A1 (en) * | 2017-08-29 | 2019-07-18 | Boe Technology Group Co., Ltd. | Image processing method, image processing device, and training method of neural network |
| CN111798484A (en) * | 2020-06-30 | 2020-10-20 | 武汉大学 | Continuous dense optical flow estimation method and system based on event camera |
Non-Patent Citations (1)
| Title |
|---|
| XINGJIAN SHI 等: "Deep Learning for Precipitation Nowcasting : A Benchmark and A New Model", 《ARXIV》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113420939A (en) * | 2021-07-15 | 2021-09-21 | 上海眼控科技股份有限公司 | Cloud picture forecasting method, equipment and storage medium |
| CN113724287A (en) * | 2021-09-02 | 2021-11-30 | 北京华云星地通科技有限公司 | Satellite cloud picture prediction method and system |
| CN114139760A (en) * | 2021-09-29 | 2022-03-04 | 苏州浪潮智能科技有限公司 | Method, system, storage medium and equipment for predicting typhoon path |
| CN114139760B (en) * | 2021-09-29 | 2023-11-07 | 苏州浪潮智能科技有限公司 | Method, system, storage medium and equipment for predicting typhoon path |
| CN115661277A (en) * | 2022-10-20 | 2023-01-31 | 中山大学 | Typhoon cloud picture extrapolation method, system, equipment and medium based on variational self-coding |
| CN115661277B (en) * | 2022-10-20 | 2023-06-02 | 中山大学 | Typhoon cloud picture extrapolation method, system, equipment and medium based on variation self-coding |
| CN115984688A (en) * | 2022-12-20 | 2023-04-18 | 中国科学院空天信息创新研究院 | Continuous frame-oriented double-current cloud picture prediction method, storage medium and equipment |
| CN115984688B (en) * | 2022-12-20 | 2023-07-04 | 中国科学院空天信息创新研究院 | Continuous frame-oriented double-flow cloud picture prediction method, storage medium and equipment |
| CN117239744A (en) * | 2023-11-15 | 2023-12-15 | 威派(武汉)高新技术有限公司 | Ultra-short-term photovoltaic power prediction method integrating wind cloud No. 4 meteorological satellite data |
| CN117239744B (en) * | 2023-11-15 | 2024-02-02 | 威派(武汉)高新技术有限公司 | Ultra-short-term photovoltaic power prediction method integrating wind cloud No. 4 meteorological satellite data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112802063A (en) | Satellite cloud picture prediction method and device and computer readable storage medium | |
| US11763466B2 (en) | Determining structure and motion in images using neural networks | |
| CN111915663B (en) | Image depth prediction neural network | |
| CN111862126B (en) | Non-cooperative target relative pose estimation method combining deep learning and geometric algorithm | |
| CN110363816B (en) | Mobile robot environment semantic mapping method based on deep learning | |
| CN113657388B (en) | Image semantic segmentation method for super-resolution reconstruction of fused image | |
| US11508157B2 (en) | Device and method of objective identification and driving assistance device | |
| KR20200087808A (en) | Method and apparatus for partitioning instances, electronic devices, programs and media | |
| CN114898315A (en) | Driving scene information determination method, object information prediction model training method and device | |
| CN113610087B (en) | Priori super-resolution-based image small target detection method and storage medium | |
| CN113554039B (en) | Method and system for generating optical flow graph of dynamic image based on multi-attention machine system | |
| Zhang et al. | Pqa-cnn: Towards perceptual quality assured single-image super-resolution in remote sensing | |
| Wang et al. | Object counting in video surveillance using multi-scale density map regression | |
| CN111444923A (en) | Image semantic segmentation method and device under natural scene | |
| EP4323952A1 (en) | Semantically accurate super-resolution generative adversarial networks | |
| CN113326734B (en) | Rotational target detection method based on YOLOv5 | |
| Jiang et al. | Semantic segmentation network combined with edge detection for building extraction in remote sensing images | |
| CN116310832A (en) | Remote sensing image processing method, device, equipment, medium and product | |
| CN115661694A (en) | Intelligent detection method, system, storage medium and electronic equipment for light-weight main transformer focusing on key characteristics | |
| CN111460941A (en) | Visual navigation feature point extraction and matching method in wearable navigation equipment | |
| CN112597825A (en) | Driving scene segmentation method and device, electronic equipment and storage medium | |
| CN117237677B (en) | Precipitation prediction correction method for overall similarity of strong precipitation space based on deep learning | |
| CN110414337B (en) | Target attitude detection system and detection method thereof | |
| Liu et al. | Positioning of Catenary Support Components | |
| CN117218344A (en) | Mars surface target segmentation method, device and equipment based on deep neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210514 |
|
| RJ01 | Rejection of invention patent application after publication |