CN106886023B

CN106886023B - A kind of Radar Echo Extrapolation method based on dynamic convolutional neural networks

Info

Publication number: CN106886023B
Application number: CN201710110183.6A
Authority: CN
Inventors: 李骞; 施恩; 顾大权
Original assignee: PLA University of Science and Technology
Current assignee: PLA University of Science and Technology
Priority date: 2017-02-27
Filing date: 2017-02-27
Publication date: 2019-04-02
Anticipated expiration: 2037-02-27
Also published as: CN106886023A

Abstract

The invention discloses a kind of Radar Echo Extrapolation methods based on dynamic convolutional neural networks, it include: offline convolution neural metwork training: to given training image collection, training sample set is obtained by data prediction, initialize dynamic convolution neural network model, and using training sample set training dynamic convolutional neural networks, output valve is calculated by network propagated forward, the process of back-propagating update network parameter restrains dynamic convolutional neural networks.Online Radar Echo Extrapolation: test sample collection is converted for test chart image set by data prediction, trained dynamic convolutional neural networks are tested using test sample collection, by in input image sequence last width radar return image and network propagated forward in the probability vector phase convolution that obtains, the Radar Echo Extrapolation image predicted.

Description

A kind of Radar Echo Extrapolation method based on dynamic convolutional neural networks

Technical field

The invention belongs to surface weather observation technical fields in Atmospheric Survey, more particularly to one kind to be based on dynamic convolutional Neural The Radar Echo Extrapolation method of network.

Background technique

Nowcasting refers mainly to the weather forecast of 0~3 hour high-spatial and temporal resolution, and Main Prediction object includes strong drop The diastrous weathers such as water, strong wind, hail.Currently, many forecast systems all use Numerical Prediction Models, but due to numerical forecast In the presence of return slow (spin-up) has been forecast, Nowcasting ability is limited.New Generation Doppler Weather Radar has very high Sensitivity and resolution ratio, the spatial resolution of data information can reach 200~1000m, and temporal resolution can reach 2 ~15min.In addition, Doppler radar also has reasonable operating mode, comprehensive condition monitoring and fault warning, advanced Real-time Calibration System and Radar meteorology product algorithm abundant, the reliability of Nowcasting can be greatly improved.Nowadays, New Generation Doppler Weather Radar has become one of most effective tool of nowcasting, is faced using Doppler radar Nearly forecast is based primarily upon Radar Echo Extrapolation technology, i.e., according to current time radar observation result, thus it is speculated that radar return future Position and intensity, to realize the track prediction to strong convection system.

Traditional Radar Echo Extrapolation method is centroid tracking method and the cross-correlation technique based on maximum correlation coefficient (Tracking Radar Echoes by Correlation, TREC), but all there is certain deficiency, mass center in conventional method Tracing is only applicable to echo compared with strong, the lesser storm monomer of range, unreliable for the forecast of a wide range of precipitation；TREC is general Echo is considered as linear change, and echo variation is increasingly complex in reality, while such method is vulnerable in vector field Unordered vector disturbance.In addition, existing method is low to the utilization rate of Radar Data, and history Radar Data includes local weather system The important feature of system variation, has very high researching value.

For improve Radar Echo Extrapolation timeliness, and from a large amount of history Radar Data study radar return variation Machine learning method is introduced into Radar Echo Extrapolation by rule.Convolutional neural networks (Convolutional Neural Network, CNN) important branch as deep learning, it is widely used in image procossing, the fields such as pattern-recognition.The network is most Big feature is using part connection, weight is shared, down-sampling method, has to the deformation of input picture, translation and overturning There is stronger adaptability.For strong temporal correlation existing between radar return image, the dynamic convolution based on input is designed Neural network, which can dynamically change weighting parameter according to the radar echo map of input, and then predict extrapolated image.Benefit With history Radar Data training dynamic convolutional neural networks, network is made more fully to extract echo character, study echo variation Rule, for improving Radar Echo Extrapolation accuracy, optimization nowcasting effect is of great significance.

Summary of the invention

Goal of the invention: when the technical problem to be solved by the present invention is to be directed to the extrapolation of existing Radar Echo Extrapolation method Imitate it is short, it is insufficient to Radar Data utilization rate, propose a kind of Radar Echo Extrapolation method based on dynamic convolutional neural networks, it is real Now to the high plane such as radar echo intensity show CAPPI (Constant AltitudePlan Position Indicator, CAPPI) the outside forecast of image, comprising the following steps:

Step 1, the offline convolutional neural networks of training: input training image collection carries out data prediction to training image collection, Training sample set is obtained, designs dynamic convolution neural network structure, and initialize network training parameter；It is assembled for training using training sample Practice dynamic convolutional neural networks, the orderly image sequence of input obtains width prediction by dynamic convolutional neural networks propagated forward Image calculates forecast image and compares the error between label, updates the weighting parameter of network by backpropagation and biasing is joined Number repeats this process until reaching trained termination condition, obtains convergent dynamic convolutional neural networks；

Step 2, online Radar Echo Extrapolation: input test image set carries out data prediction to test chart image set, obtains Test sample collection, then by the dynamic convolutional neural networks obtained in test sample collection input step 1, by before network to biography Calculating probability vector is broadcast, and last width radar return image in input image sequence is mutually rolled up with obtained probability vector Product, the Radar Echo Extrapolation image predicted.

Step 1 of the present invention the following steps are included:

Step 1-1, data prediction: input training image collection, the every piece image concentrated to training image standardize Change processing, converts every piece image to 280 × 280 floating number image, obtains floating number image collection, to floating number image Set is divided, and construction includes the training sample set of TrainsetSize group sample；

Step 1-2 initializes dynamic convolutional neural networks: design dynamic convolution neural network structure is configured to generate The sub-network of probability vector reconstructs the probabilistic forecasting layer for image extrapolation, provides for the offline neural metwork training stage dynamic State convolutional neural networks initialization model；

Step 1-3 initializes dynamic convolution Neural Network Training Parameter: enabling e-learning rate λ=0.0001, training stage The sample size BatchSize=10 inputted every time, most large quantities of frequency of training of training sample setCurrently crowd frequency of training BatchNum=1, the maximum number of iterations of network training IterationMax=40, current iteration number IterationNum=1；

Step 1-4 reads training sample: by the way of batch training, training the training sample obtained from step 1-1 every time It concentrates and reads BatchSize group training sample, every group of training sample is { x₁,x₂,x₃,x₄, y }, it altogether include 5 width images, wherein {x₁,x₂,x₃,x₄It is used as input image sequence, y is corresponding control label；

Step 1-5, propagated forward: the input image sequence feature that extraction step 1-4 is obtained in a sub-network obtains level Probability vector HPV and vertical probability vector VPV；In probabilistic forecasting layer, by the last piece image in input image sequence according to It is secondary with VPV, HPV phase convolution, obtain the output forecast image of propagated forward；

Backpropagation: step 1-6 reversely acquires the error term of probability vector, according to probability vector in probabilistic forecasting layer Error term from rear to the preceding layer-by-layer error term for calculating each network layer in subnet network layers, and then calculate error in each network layer Item utilizes the parameter of obtained gradient updating dynamic convolutional neural networks to the gradient of weighting parameter and offset parameter；

Off-line training step control: step 1-7 carries out whole control to the offline neural metwork training stage, is divided into following Three kinds of situations:

If training sample is concentrated there are still original training sample, i.e. BatchNum < BatchMax, then step is returned Rapid 1-4 continues to read BatchSize group training sample, carries out network training；

If training sample, which is concentrated, is not present original training sample, i.e. BatchNum=BatchMax, and current net Network the number of iterations is less than maximum number of iterations, i.e. IterationNum < IterationMax then enables BatchNum=1, returns Step 1-4 continues to read BatchSize group training sample, carries out network training；

If training sample, which is concentrated, is not present original training sample, i.e. BatchNum=BatchMax, and network changes Generation number reaches maximum number of iterations, i.e. IterationNum=IterationMax, then terminates offline neural metwork training rank Section, obtains trained dynamic convolution neural network model.

Step 1-1 data prediction of the present invention the following steps are included:

Step 1-1-1, sampling: the image that training image is concentrated is sequentially arranged, and constant duration is distributed, when Between between be divided into 6 minutes, altogether include N_TrainWidth image determines TrainsetSize by following formula:

Wherein, Mod (N_Train, 4) and indicate N_TrainTo 4 modulus, [N] indicates the maximum integer for being not more than N, acquires After TrainsetSize, training image is retained by sampling and concentrates preceding 4 × TrainsetSize+1 width image, when sampling, which passes through, deletes Except training image concentrates last image to meet the requirements amount of images；

Normalized images: step 1-1-2 carries out image transformation, normalization operation, by original point to the image that sampling obtains The color image that resolution is 2000 × 2000 is converted into the floating number image that resolution ratio is 280 × 280；

Step 1-1-3 constructs training sample set:

Training sample set is constructed using the floating number image set that step 1-1-2 is obtained, by every four in floating number image set Adjacent image, i.e. { 4N+1,4N+2,4N+3,4N+4 } width image are as one group of list entries, [4 × (N+1)+1] width figure It is cut as passing through, the part that the central resolution ratio of reservation is 240 × 240 is as the control label for corresponding to sample, for N group sampleIts make is as follows:

In above formula, G_4N+1Indicate floating number image set in 4N+1 width image, N is positive integer, and have N ∈ [0, TrainsetSize-1], Crop () indicates trimming operation, and the portion that original image center size is 240 × 240 is retained after cutting Point, finally obtain the training sample set comprising TrainsetSize group training sample；

Wherein, step 1-1-2 the following steps are included:

Image conversion: step 1-1-2-1 for the computation complexity for reducing convolutional neural networks training, step 1-1-1 is adopted The image that sample obtains is converted into gray level image, retains the part that original image center resolution ratio is 560 × 560 by cutting, will Image resolution ratio boil down to 280 × 280 after cutting obtains the grayscale image that resolution ratio is 280 × 280；

Step 1-1-2-2, data normalization: by each of the grayscale image obtained in step 1-1-2-1 pixel Value is mapped to [0~1] from [0~255], by obtaining the floating number image that resolution ratio is 280 × 280 after normalization.

Step 1-2 of the present invention the following steps are included:

Step 1-2-1 constructs sub-network:

Sub-network is made of 10 network layers, be followed successively by from front to back convolutional layer C1, down-sampling layer S1, convolutional layer C2, under Sample level S2, convolutional layer C3, down-sampling layer S3, convolutional layer C4, down-sampling layer S5, convolutional layer C5 and classifier layer F1；

Step 1-2-2 constructs probabilistic forecasting layer:

In probabilistic forecasting layer construct dynamic convolutional layer DC1 and dynamic convolutional layer DC2, by sub-network output vertical probability to Measure convolution kernel of the VPV as dynamic convolutional layer DC1, convolution kernel of the level probability vector HPV as dynamic convolutional layer DC2；

Wherein, step 1-2-1 the following steps are included:

Step 1-2-1-1 constructs convolutional layer: determining the following contents: the output characteristic pattern quantity OutputMaps of convolutional layer, Convolution kernel k and offset parameter bias.For convolution kernel, it is thus necessary to determine that width KernelSize (point of convolution kernel of convolution kernel Resolution is KernelSize × KernelSize), the quantity KernelNumber of convolution kernel (value be the convolutional layer input with it is defeated The product of characteristic pattern quantity out), and convolution kernel is constructed according to Xavier initial method；For offset parameter, quantity with should The output characteristic pattern quantity of layer is identical.For convolutional layer l_C,l_C∈ { C1, C2, C3, C4, C5 }, the output characteristic pattern width of this layer ForValue codetermined by its input feature vector figure resolution ratio and the size of convolution kernel, i.e.,Indicate convolutional layer l_CUpper one layer of convolutional layer Output characteristic pattern width；

For convolutional layer C1, C1 layers of output characteristic pattern quantity OutputMaps is enabled^C1=12, export the width of characteristic pattern OutputSize^C1=272, convolution kernel width KernelSize^C1=9, offset parameter bias^C1It is initialized as zero, C1 layers of volume Product core k^C1Quantity KernelNumber^C1=48, the initial value of each parameter is in convolution kernelRand () is for generating random number；

For convolutional layer C2, C2 layers of output characteristic pattern quantity OutputMaps are enabled^C2=32, export the width of characteristic pattern OutputSize^C2=128, convolution kernel width KernelSize^C2=9, offset parameter is initialized as zero, C2 layers of convolution kernel k^C2Quantity KernelNumber^C2=384, the initial value of each parameter is in convolution kernel

For convolutional layer C3, C3 layers of output characteristic pattern quantity OutputMaps are enabled^C3=32, export the width of characteristic pattern OutputSize^C3=56, convolution kernel width KernelSize^C3=9, offset parameter is initialized as zero, C3 layers of convolution kernel k^C3 Quantity KernelNumber^C3=1024, the initial value of each parameter is in convolution kernel

For convolutional layer C4, C4 layers of output characteristic pattern quantity OutputMaps are enabled^C4=32, export the width of characteristic pattern OutputSize^C4=20, convolution kernel width KernelSize^C4=9, offset parameter is initialized as zero, C4 layers of convolution kernel k^C4 Quantity KernelNumber^C4=1024, the initial value of each parameter is in convolution kernel

For convolutional layer C5, this layer of OutputMaps is enabled^C5=32, OutputSize^C5=4, KernelSize^C5=7, partially Set the convolution kernel k that parameter is initialized as zero, C5 layers^C5Quantity KernelNumber^C5=1024, each ginseng in convolution kernel Several initial values are

Step 1-2-1-2, construct down-sampling layer: in down-sampling layer do not include need training parameter, by down-sampling layer S1, The sampling core of S2, S3 and S4 are initialized asFor down-sampling layer l_S,l_S∈ { S1, S2, S3, S4 }, output Characteristic pattern quantityIt is consistent with the output characteristic pattern quantity of one layer of convolutional layer thereon, output characteristic pattern is wide DegreeIt is the 1/2 of the output characteristic pattern width of one layer of convolutional layer thereon, formula is expressed as follows:

Step 1-2-1-3, structural classification device layer: classifier layer is made of a full articulamentum F1, F1 layers of weighting parameter For horizontal weighting parameter matrix W H and vertical weighting parameter matrix W V, size is 41 × 512, is enabled every in weighting parameter matrix The initial value of one parameter isOffset parameter is Horizontal offset parameter BH and vertical off setting parameter BV, It is initialized as 41 × 1 one-dimensional null vector.

Step 1-5 of the present invention the following steps are included:

Step 1-5-1, sub-network calculate probability vector: passing through the alternate treatment of convolutional layer and down-sampling layer in a sub-network Extract input image sequence characteristic, handled in classifier layer by Softmax function, obtain level probability vector HPV with Vertical probability vector VPV；

Step 1-5-2, calculate probabilistic forecasting layer and export image: the HPV and VPV that step 1-5-1 is obtained are as probabilistic forecasting The convolution kernel of layer obtains the defeated of propagated forward by the last piece image in input image sequence successively with VPV, HPV phase convolution Forecast image out.

Step 1-5-1 of the present invention the following steps are included:

Step 1-5-1-1, judges network layer type: indicating the network layer in the sub-network being presently in, l initial value with l For C1, the type of network layer l is judged, if l ∈ { C1, C2, C3, C4, C5 }, then l is convolutional layer, step 1-5-1-2 is executed, if l ∈ { S1, S2, S3, S4 }, then l is down-sampling layer, executes step 1-5-1-4；

Step 1-5-1-2 handles convolutional layer: having l=l at this time_C,l_C∈ { C1, C2, C3, C4, C5 }, first calculating l_CLayer J-th of output characteristic patternBy l_CThe input feature vector figure convolution nuclear phase convolution corresponding with this layer respectively of layer, convolution results are asked L is added with, summed result_CJ-th of offset parameter of layerIt handles, obtains using ReLU activation primitiveIt calculates public Formula is as follows:

Wherein,For l_CI-th of input feature vector figure convolution kernel corresponding with j-th of output characteristic pattern of layer, n is current The output characteristic pattern number of the previous down-sampling layer of convolutional layer, Indicate l_CI-th of input of layer Characteristic pattern, while being also l_C- 1 layer of i-th of output characteristic pattern, * representing matrix convolution, if l_C=C1, then l_C- 1 layer is input Layer；

All output characteristic patterns are successively calculated, l is obtained_CThe output characteristic pattern of layerL is updated to l+1, and returns to step Rapid 1-5-1-1 judges network type, carries out the operation of next network layer；

Step 1-5-1-3 handles down-sampling layer: having l=l at this time_S,l_S∈ { S1, S2, S3, S4 }, step 1-5-1-2 is obtained The output characteristic pattern of the convolutional layer arrived respectively withPhase convolution, then sampled with step-length for 2, sampling obtains l_SLayer Output characteristic patternCalculation formula is as follows:

In above formula, Sample () indicates that step-length is 2 sampling processing, l_S- 1 indicates the previous convolution of current down-sampling layer Layer,Indicate l_SThe output characteristic pattern of layerIn j-th of output characteristic pattern, obtain l_SThe output characteristic pattern of layerAfterwards, more by l It is newly l+1, and return step 1-5-1-1 judges network type, carries out the operation of next network layer；

Step 1-5-1-4 calculates F1 layers of probability vector: if network layer l is classifier layer, i.e. l=F1 is become by matrix Change, by the 32 width resolution ratio of C5 be 4 × 4 output characteristic pattern with column sequential deployment, obtain the F1 layer that resolution ratio is 512 × 1 Export feature vector a^F1, calculate separately horizontal weighting parameter matrix W H and a^F1Apposition, vertical weighting parameter matrix W V and a^F1's Apposition sums calculated result with Horizontal offset parameter BH, vertical off setting parameter BV respectively, after the processing of Softmax function To level probability vector HPV and vertical probability vector VPV, specific formula for calculation is as follows:

By its vertical probability vector VPV transposition, final vertical probability vector is obtained；

Step 1-5-2 the following steps are included:

Step 1-5-2-1 predicts DC1 layers of vertical direction: by last width input picture of input layer and vertical probability to VPV phase convolution is measured, the DC1 layer that resolution ratio is 240 × 280 is obtained and exports characteristic pattern a^DC1；

Step 1-5-2-2 predicts DC2 layers of vertical direction: by DC1 layers of output characteristic pattern a^DC1With vertical probability vector HPV phase Convolution, obtains the output forecast image of propagated forward, and resolution ratio is 240 × 240.

Step 1-6 of the present invention the following steps are included:

Step 1-6-1 calculates probabilistic forecasting layer error term: by the instruction of the step 1-5-2-2 forecast image obtained and input The control label practiced in sample asks poor, calculates the error term of DC2 layers, DC1 layers, finally acquires the error term of level probability vector δ^HPVWith the error term δ of vertical probability vector^VPV；

Step 1-6-2 calculates sub-network error term: according to the error term δ of level probability vector^HPVWith vertical probability vector Error term δ^VPV, from it is rear to preceding successively calculate classification layer F1, convolutional layer C5, C4, C3, C2, C1 and down-sampling layer S4, S3, S2, The error term of S1, the resolution ratio of any layer error term matrix acquired and the output resolution ratio of characteristic pattern of this layer are consistent；

Step 1-6-3 calculates gradient: the mistake of each network layer of sub-network is calculated according to the error term that step 1-6-2 is obtained Gradient value of the poor item to this layer of weighting parameter and offset parameter；

Step 1-6-4, undated parameter: by the ladder of the weighting parameter of the step 1-6-3 each network layer obtained and offset parameter Angle value is multiplied by the learning rate of dynamic convolutional neural networks, obtains the update item of each network layer weighting parameter and offset parameter, will be former Weighting parameter and offset parameter ask poor with the update item respectively, obtain updated weighting parameter and offset parameter.

Step 1-6-1 of the present invention the following steps are included:

Step 1-6-1-1 calculates dynamic convolutional layer DC2 error term: the forecast image and the group that step 1-5-2-2 is obtained The control label of sample asks poor, obtains the error term matrix delta that size is 240 × 240^DC2；

Step 1-6-1-2 calculates dynamic convolutional layer DC1 error term: by zero padding by DC2 layers of error term matrix delta^DC2 Expanding is 240 × 320, by level probability Vector rotation 180 degree, by the error term matrix after expansion and the level probability after overturning Vector phase convolution obtains DC1 layers of error term matrix delta^DC1, size is 240 × 280, and formula is as follows:

δ^DC1=Expand_Zero (δ^DC2) * rot180 (HPV),

Wherein, Expand_Zero () indicates that zero extended function, rot180 () indicate that angle is 180 ° of rotation letter 2 × 2 matrix zero is extended for 4 × 4 matrix by number, the matrix after zero expansion, the region and former square that central resolution ratio is 2 × 2 Consistent, zero pixel filling of remaining position of battle array；

Step 1-6-1-3 calculates probability vector error term: the error term of level probability vector HPV is calculated, by DC1 layers Export characteristic pattern and error term matrix delta^DC2Phase convolution obtains 1 × 41 row vector after convolution, which is the error term of HPV δ^HPV, formula is as follows:

δ^HPV=a^DC1*δ^DC2,

The error term for calculating vertical probability vector VPV, by the input feature vector figure of input layer and error term matrix delta^DC1Mutually roll up It is long-pending, 41 × 1 column vector is obtained after convolution, which is the error term δ of VPV^VPV, formula is as follows:

Wherein,For the last piece image in the input image sequence of training sample；

Step 1-6-2 the following steps are included:

Step 1-6-2-1 calculates classifier layer F1 error term: by the error term of the step 1-6-1-3 probability vector obtained δ^VPVAnd δ^HPVWeighting parameter matrix W V vertical with F1 layers and horizontal weighting parameter matrix W H carry out matrix multiple respectively, then by square The apposition of battle array is summed and is averaged, and F1 layers of error term δ is obtained^F1, formula is as follows:

Wherein, × representing matrix apposition, ()^TThe transposition for representing matrix, obtained δ^F1Size be 512 × 1；

Step 1-6-2-2 calculates convolutional layer C5 error term: by matrixing, the F1 layer that will be obtained in step 1-6-2-1 Error term δ^F1It is transformed to the matrix that 32 resolution ratio are 4 × 4Obtain C5 layers of error term δ^C5,Table Show that transformed 32nd resolution ratio is 4 × 4 matrix；Matrixing in this time matrixing and step 1-5-1-6 is each other Inverse transformation,

Step 1-6-2-3, judges network layer type: network layer (the l initial value in the sub-network being presently in is indicated with l For S4), the type of network layer l is judged, if l ∈ { S4, S3, S2, S1 }, then l is down-sampling layer, step 1-6-2-4 is executed, if l ∈ { C4, C3, C2, C1 }, then l is convolutional layer, executes step 1-6-2-5；

Step 1-6-2-4 calculates down-sampling layer error term: if l layers are down-sampling layer, there is l=l at this time_S,l_S∈{S1,S2, S3, S4 }, for l_SI-th of error term matrix of layerBy zero padding respectively by l_S+ 1 layer of each error term matrix It expands to width and isWherein l_SThe latter convolutional layer of+1 layer of current down-sampling layer of expression, j indicate l_S+ 1 layer J-th of error term, and haveAgain by corresponding convolution kernel180 degree is rotated, after then expanding Matrix and overturning after convolution nuclear phase convolution, and convolution results are summed, obtain l_SI-th of error term matrix of layerIt is public Formula is as follows:

Wherein,

All error term matrixes are successively calculated, l is obtained_SThe output characteristic pattern of layerL is updated to l-1, and returns to step Rapid 1-6-2-3 judges network type, calculates the error term of a upper network layer；

Step 1-6-2-5 calculates convolutional layer error term: if l layers are convolutional layer, there is l=l at this time_C,l_C∈{C1,C2,C3, C4, C5 }, it is not in l since the initial value of l in step 1-6-2-3 is S4_CThe case where=C5, for l_CThe i-th of layer A error term matrixFirst to l_CCorresponding i-th of error term matrix in+1 layerIt is up-sampled, it will when up-samplingIn each element error entry value average mark into sampling area, obtaining resolution ratio is Up-sampling matrix, then calculate activation primitive in l_CThe inner product of derivative and the up-sampling matrix acquired at layer character pair figure, Obtain l_CI-th of error term matrix of layerFormula is as follows:

Wherein, representing matrix inner product, ReLU'() indicate the derivative of ReLU activation primitive, form is as follows:

UpSample () indicates up-sampling function, the corresponding up-sampling of each of original image pixel after up-sampling Region in each of original pixel value mean allocation to sampling area pixel, successively calculates all error term matrixes, obtains To l_CThe output characteristic pattern of layer

Step 1-6-2-6, l layers are convolutional layer, i.e. l=l at this time_C, it is divided into two kinds of situations later:

If l ≠ C1, l is updated to l-1, and return step 1-6-2-3 judges network type, calculates a upper network layer Error term；

If l=C1, the calculating of step 1-6-2 sub-network error term terminates；

Step 1-6-3 the following steps are included:

Step 1-6-3-1 calculates convolutional layer error term to the gradient of convolution kernel: using l_CIndicate currently processed convolutional layer, l_C ∈ { C1, C2, C3, C4, C5 } successively calculates each convolutional layer error term to the gradient of convolution kernel, by l since C1 layers_C- 1 layer I-th of output characteristic patternWith l_CJ-th of error term matrix of layerPhase convolution, convolution results are the ladder of corresponding convolution kernel Angle valueFormula is as follows:

In above formula,WithRespectively indicate l_CThe output characteristic pattern number and l of layer_C- 1 layer of output characteristic pattern number, ▽ k are gradient value of the error term to convolution kernel；

Step 1-6-3-2 calculates each convolutional layer error term to the gradient of biasing: using l_CIndicate currently processed convolutional layer, l_C ∈ { C1, C2, C3, C4, C5 } successively calculates each convolutional layer error term to the gradient of biasing, by l since C1 layers_CJ-th of layer Error term matrixIn all elements sum, obtain j-th of this layer biasing gradient valueThe following institute of formula Show:

Wherein, Sum () expression sums to all elements of matrix；

Step 1-6-3-3 calculates F1 layers of error term to the gradient of weighting parameter: calculate separately level probability vector with it is vertical The error term δ of probability vector^HPV、δ^VPVWith F1 layers of error term δ^F1Inner product, calculated result be F1 layers of error term to weighting parameter WH, The gradient value of WV, formula are as follows:

▽ WH=(δ^HPV)^T×(δ^F1)^T,

▽ WV=δ^VPV×(δ^F1)^T,

In above formula, ▽ WH is gradient value of the error term to horizontal weighting parameter, and ▽ WV is error term to vertical weighting parameter Gradient value；

Step 1-6-3-4 calculates F1 layers of error term to the gradient of offset parameter: by level probability vector and vertical probability to The error term δ of amount^HPV、δ^VPVRespectively as F1 layers of error term to the gradient value of Horizontal offset parameter BH and vertical off setting parameter BV, Formula is as follows:

▽ BH=(δ^HPV)^T,

▽ BV=δ^VPV,

Wherein, ▽ BH is gradient value of the error term to Horizontal offset parameter, and ▽ BV is error term to vertical off setting parameter Gradient value；

Step 1-6-4 the following steps are included:

Step 1-6-4-1 updates each convolutional layer weighting parameter: each convolutional layer error term pair that step 1-6-3-1 is obtained The gradient of convolution kernel is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of convolution kernel, then by former convolution kernel and be somebody's turn to do Correction term asks poor, the convolution kernel updatedFormula is as follows:

Wherein, λ is the e-learning rate determined in step 1-3, λ=0.0001；

Step 1-6-4-2 updates each convolutional layer offset parameter: each convolutional layer error term pair that step 1-6-3-2 is obtained The gradient of biasing is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of offset parameter, then by former bias term and be somebody's turn to do Correction term asks poor, the bias term updatedFormula is as follows:

Step 1-6-4-3, update F1 layers of weighting parameter: the F1 layer error term that step 1-6-3-3 is obtained is to weighting parameter The gradient value of WH and WV is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of weighting parameter, then former weight is joined Number WH and WV asks poor with the correction term acquired respectively, the WH and WV updated, and formula is as follows:

WH=WH- λ ▽ WH,

WV=WV- λ ▽ WV；

Step 1-6-4-4, update F1 layers of offset parameter: the F1 layer error term that step 1-6-3-4 is obtained is to offset parameter The gradient value of BH and BV is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of offset parameter, then original biasing is joined Number BH and BV asks poor with the correction term acquired respectively, the BH and BV updated, and formula is as follows:

BH=BH- λ ▽ BH,

BV=BV- λ ▽ BV.

Step 2 of the present invention the following steps are included:

Step 2-1, data prediction: input test image set, the every piece image concentrated to test image standardize Change processing, converts every piece image to 280 × 280 floating number image, then divide to floating number image collection, constructs Test sample collection comprising TestsetSize group sample；

Step 2-2, read test sample: the TestsetSize group test sample input that step 2-1 is obtained is by training Dynamic convolutional neural networks in；

Step 2-3, propagated forward: extracting the image sequence characteristic of input in a sub-network, obtains level probability vector HPV_testWith vertical probability vector VPV_test；In probabilistic forecasting layer, by the last piece image in input image sequence successively with VPV_test、HPV_testPhase convolution obtains the final extrapolated image of dynamic convolutional neural networks.

Step 2-1 of the present invention the following steps are included:

Step 2-1-1, sampling: the image that test image is concentrated is sequentially arranged, and constant duration is distributed, when Between between be divided into 6 minutes, altogether include N_TestWidth image determines TestsetSize by following formula:

If Mod (N_Test, 4)=0

If Mod (N_Test, 4) ≠ 0

After acquiring TestsetSize, test image is retained by sampling and concentrates preceding 4 × TestsetSize+1 width image, is adopted Last image is concentrated to meet the requirements amount of images by deleting test image when sample；

Image normalization: step 2-1-2 carries out image transformation, normalization operation, by original point to the image that sampling obtains The color image that resolution is 2000 × 2000 is converted into the floating number image that resolution ratio is 280 × 280；

Step 2-1-3 constructs test sample collection: constructing test sample using the floating number image set that step 2-1-2 is obtained Collection, by four adjacent images every in floating number image set, i.e. { 4M+1,4M+2,4M+3,4M+4 } width image is defeated as one group Enter sequence, for [4 × (M+1)+1] width image by cutting, retaining central resolution ratio is 240 × 240 part as corresponding sample This control label, wherein for positive integer, and there is M ∈ [0, TestsetSize-1] to obtain comprising TestsetSize group test specimens This test sample collection；

Step 2-1-2 the following steps are included:

Step 2-1-2-1, image conversion: to reduce the computation complexity that convolutional neural networks are trained, by colored echo Intensity CAPPI image is converted into gray level image, then retains the part that original image center resolution ratio is 560 × 560 by cutting, By the image resolution ratio boil down to 280 × 280 after cutting, the grayscale image that resolution ratio is 280 × 280 is obtained；

Step 2-1-2-2, data normalization: by each of the grayscale image obtained in step 1-1-2-1 pixel Value is mapped to [0~1] from [0~255], by obtaining the floating number image that resolution ratio is 280 × 280 after normalization；

Step 2-3 the following steps are included:

Step 2-3-1 calculates sub-network probability vector: passing through the alternate treatment of convolutional layer and down-sampling layer in a sub-network The image sequence characteristic for extracting input, is then handled by Softmax function in classifier layer, obtains level probability vector HPV_testWith vertical probability vector VPV_test；

Step 2-3-2 calculates probabilistic forecasting layer and exports image: the VPV that step 2-3-1 is obtained_testAnd HPV_testAs probability The convolution kernel of prediction interval, by the last piece image in input image sequence successively with VPV_testAnd HPV_testPhase convolution, is moved The final extrapolated image of state convolutional neural networks；

Step 2-3-1 the following steps are included:

Step 2-3-1-1, judges network layer type: indicating the network layer in the sub-network being presently in p, judges network The type of layer p, p initial value is C1, if p ∈ { C1, C2, C3, C4, C5 }, then p is convolutional layer, step 2-3-1-2 is executed, if p ∈ { S1, S2, S3, S4 }, then p is down-sampling layer, executes step 2-3-1-4；

Step 2-3-1-2 handles convolutional layer: having p=p at this time_C,p_C∈ { C1, C2, C3, C4, C5 }, first calculating p_CLayer V-th of output characteristic patternBy p_CThe input feature vector figure convolution nuclear phase convolution corresponding with this layer respectively of layer, then by convolution knot Fruit summation, summed result add p_CV-th of biasing of layerIt is handled using ReLU activation primitive, obtains p_CThe v of layer A output characteristic patternCalculation formula is as follows:

Wherein,For p_CU-th of input feature vector figure convolution kernel corresponding with v-th of output characteristic pattern of layer, m are to work as The output characteristic pattern number of the previous down-sampling layer of preceding convolutional layer,Indicate p_CU-th of layer Input feature vector figure, while being also p_C- 1 layer of u-th of output characteristic pattern, * representing matrix convolution work as p_CWhen=C1, then p_C- 1 layer is Input layer.

All output characteristic patterns are successively calculated, p is obtained_CThe output characteristic pattern of layerP is updated to p+1, and returns to step Rapid 2-3-1-1 judges network type, carries out the operation of next network layer；

Step 2-3-1-3 handles down-sampling layer: having p=p at this time_S,p_S∈ { S1, S2, S3, S4 }, step 2-3-1-2 is obtained The output characteristic pattern of the convolutional layer arrived then with step-length is respectively 2 to be sampled with phase convolution, and sampling obtains p_S The output characteristic pattern calculation formula of layer is as follows:

Wherein, Sample () indicates that step-length is 2 sampling processing, p_S- 1 indicates the previous convolution of current down-sampling layer Layer,Indicate p_SThe output characteristic pattern of layerIn j-th of output characteristic pattern, obtain p_SThe output characteristic pattern of layerAfterwards, by p It is updated to p+1, and return step 2-3-1-1 judges network type, carries out the operation of next network layer；

Step 2-3-1-4 calculates F1 layers of probability vector: if network layer p is classifier layer, i.e. p=F1 is become by matrix Change, by the 32 width resolution ratio of C5 be 4 × 4 output characteristic pattern with column sequential deployment, obtain the F1 layer that resolution ratio is 512 × 1 Export feature vectorThen calculate separately horizontal parameters matrix W H, Vertical Parameters matrix W V withApposition, by calculate tie Fruit is summed with Horizontal offset parameter BH, vertical off setting parameter BV respectively, and summed result obtains level after the processing of Softmax function Probability vector HPV_test, vertical probability vector VPV_test, calculation formula is as follows:

By its vertical probability vector VPV_testTransposition obtains final vertical probability vector；

Step 2-3-2 the following steps are included:

Step 2-3-2-1 predicts DC1 layers of vertical direction: by last width input picture of input layer and vertical probability to Measure VPV_testPhase convolution obtains the DC1 layer that resolution ratio is 240 × 280 and exports characteristic pattern

Step 2-3-2-2 predicts DC2 layers of vertical direction: step 2-3-2-1 is obtainedWith level probability vector HPV_testPhase convolution obtains the final extrapolated image of dynamic convolutional neural networks, and resolution ratio is 240 × 240, by extrapolated image It compares with the control label of corresponding test sample, determines that dynamic convolutional neural networks carry out the accuracy of Radar Echo Extrapolation.

The utility model has the advantages that being based on dynamic convolutional neural networks, trained on a large amount of radar return image set, and utilize Trained network carries out Radar Echo Extrapolation, effectively increases the accuracy of Radar Echo Extrapolation.

Specifically the present invention has the advantage that the utilization rate of 1. Radar Datas is high compared with existing method, with traditional thunder It is only compared up to echo extrapolation technique using a small amount of Radar Data, present invention training on a large amount of radar return image set, number It is higher according to utilization rate；2. timeliness of extrapolating is long, the time interval between sample can be concentrated outer to improve by adjusting training sample Push away timeliness.

Detailed description of the invention

The present invention is done with reference to the accompanying drawings and detailed description and is further illustrated, it is of the invention above-mentioned or Otherwise advantage will become apparent.

Fig. 1 is flow chart of the present invention.

Fig. 2 is dynamic convolutional neural networks initialization model structure chart.

Fig. 3 is sub-network structural map.

Fig. 4 is probabilistic forecasting layer structural map.

Fig. 5 is that matrix zero expands schematic diagram.

Fig. 6 is the process schematic up-sampled to 2 × 2 matrix.

Specific embodiment

The present invention will be further described with reference to the accompanying drawings and embodiments.

As shown in Figure 1, the present invention the following steps are included:

Step 1, as shown in Fig. 2, the offline convolutional neural networks of training: input training image collection carries out training image collection Data prediction obtains training sample set, designs dynamic convolution neural network structure, and initialize network training parameter；It utilizes Training sample set trains dynamic convolutional neural networks, and the orderly image sequence of input passes through dynamic convolutional neural networks propagated forward A width forecast image is obtained, forecast image is calculated and compares the error between label, the weight of network is updated by backpropagation Parameter and offset parameter repeat this process until reaching trained termination condition, obtain convergent dynamic convolutional neural networks；

Step 1 of the present invention the following steps are included:

Step 1-4 reads training sample: by the way of batch training, training the training sample obtained from step 1-1 every time It concentrates and reads BatchSize group training sample, every group of training sample is { x₁,x₂,x₃,x₄, y }, it altogether include 5 width images, wherein {x₁,x₂,x₃,x₄It is used as input image sequence, x₁Indicate that piece image, y are corresponding control label；

Step 1-1-3 constructs training sample set:

Wherein, step 1-1-2 the following steps are included:

Step 1-2 of the present invention the following steps are included:

Step 1-2-1 constructs sub-network: sub-network is configured to structure as shown in Figure 3:

Step 1-2-2 constructs probabilistic forecasting layer: probabilistic forecasting layer is configured to structure as shown in Figure 4:

Wherein, step 1-2-1 the following steps are included:

Step 1-2-1-1 constructs convolutional layer: determining the following contents: the output characteristic pattern quantity OutputMaps of convolutional layer, Convolution kernel k and offset parameter bias.For convolution kernel, it is thus necessary to determine that width KernelSize (point of convolution kernel of convolution kernel Resolution is KernelSize × KernelSize), the quantity KernelNumber of convolution kernel (value be the convolutional layer input with it is defeated The product of characteristic pattern quantity out), and convolution kernel is constructed according to Xavier initial method；For offset parameter, quantity with should The output characteristic pattern quantity of layer is identical.For convolutional layer l_C,l_C∈ { C1, C2, C3, C4, C5 }, the output characteristic pattern width of this layer ForValue codetermined by its input feature vector figure resolution ratio and the size of convolution kernel, i.e.,

For convolutional layer C2, this layer of OutputMaps is enabled^C2=32, OutputSize^C2=128, KernelSize^C2=9, Offset parameter is initialized as zero, C2 layers of convolution kernel k^C2Quantity KernelNumber^C2=384, each ginseng in convolution kernel Several initial values are

For convolutional layer C3, this layer of OutputMaps is enabled^C3=32, OutputSize^C3=56, KernelSize^C3=9, partially Set the convolution kernel k that parameter is initialized as zero, C3 layers^C3Quantity KernelNumber^C3=1024, each ginseng in convolution kernel Several initial values are

For convolutional layer C4, this layer of OutputMaps is enabled^C4=32, OutputSize^C4=20, KernelSize^C4=9, partially Set the convolution kernel k that parameter is initialized as zero, C4 layers^C4Quantity KernelNumber^C4=1024, each ginseng in convolution kernel Several initial values are

Step 1-5 of the present invention the following steps are included:

Step 1-5-1 of the present invention the following steps are included:

Wherein,For l_CI-th of input feature vector figure convolution kernel corresponding with j-th of output characteristic pattern of layer, n is current The output characteristic pattern number of the previous down-sampling layer of convolutional layer,Indicate l_CI-th of input of layer Characteristic pattern, while being also l_C- 1 layer of i-th of output characteristic pattern, * representing matrix convolution, if l_C=C1, then l_C- 1 layer is input Layer；

Step 1-5-1-3 handles down-sampling layer: having l=l at this time_S,l_S∈ { S1, S2, S3, S4 }, step 1-5-1-2 is obtained The output characteristic pattern of the convolutional layer arrived respectively withPhase convolution, then sampled with step-length for 2, sampling obtains l_S The output characteristic pattern of layerCalculation formula is as follows:

Step 1-5-2 the following steps are included:

Step 1-6 of the present invention the following steps are included:

Step 1-6-1 of the present invention the following steps are included:

δ^DC1=Expand_Zero (δ^DC2) * rot180 (HPV),

Wherein, Expand_Zero () indicates that zero extended function, rot180 () indicate that angle is 180 ° of rotation letter Number, by 2 × 2 matrix zero be extended for 4 × 4 process as shown in figure 5, zero expand after matrix, central resolution ratio is 2 × 2 Region is consistent with original matrix, zero pixel filling of remaining position；

δ^HPV=a^DC1*δ^DC2,

Step 1-6-2 the following steps are included:

Step 1-6-2-2 calculates convolutional layer C5 error term: by matrixing, the F1 layer that will be obtained in step 1-6-2-1 Error term δ^F1It is transformed to the matrix that 32 resolution ratio are 4 × 4Obtain C5 layers of error term Indicate that transformed 32nd resolution ratio is 4 × 4 matrix；This time matrixing and the matrixing in step 1-5-1-6 are mutual For inverse transformation,

Wherein,

UpSample () indicates up-sampling function, and Fig. 6 is the process that is up-sampled of matrix to 2 × 2, after up-sampling Each of original image pixel corresponds to a up-sampling region, each of original pixel value mean allocation to sampling area picture In vegetarian refreshments, all error term matrixes are successively calculated, l is obtained_CThe output characteristic pattern of layer

If l=C1, the calculating of step 1-6-2 sub-network error term terminates；

Step 1-6-3 the following steps are included:

In above formula,WithRespectively indicate l_CLayer and l_C- 1 layer of output characteristic pattern Number, ▽ k are gradient value of the error term to convolution kernel；

Wherein, Sum () expression sums to all elements of matrix；

▽ WH=(δ^HPV)^T×(δ^F1)^T,

▽ WV=δ^VPV×(δ^F1)^T,

▽ BH=(δ^HPV)^T,

▽ BV=δ^VPV,

Step 1-6-4 the following steps are included:

Wherein, λ is the e-learning rate determined in step 1-3, λ=0.0001；

WH=WH- λ ▽ WH,

WV=WV- λ ▽ WV；

BH=BH- λ ▽ BH,

BV=BV- λ ▽ BV.

Step 2 of the present invention the following steps are included:

Step 2-1 of the present invention the following steps are included:

If Mod (N_Test, 4)=0

If Mod (N_Test, 4) ≠ 0

Step 2-1-2 the following steps are included:

Step 2-3 the following steps are included:

Step 2-3-1 the following steps are included:

Step 2-3-1-2 handles convolutional layer: having p=p at this time_C,p_C∈ { C1, C2, C3, C4, C5 }, first calculating p_CLayer V-th of output characteristic patternBy p_CThe input feature vector figure convolution nuclear phase convolution corresponding with this layer respectively of layer, then by convolution results Summation, summed result add p_CV-th of biasing of layerIt is handled using ReLU activation primitive, obtains p_CV-th of layer Export characteristic patternCalculation formula is as follows:

Wherein,For p_CU-th of input feature vector figure convolution kernel corresponding with v-th of output characteristic pattern of layer, m is current The output characteristic pattern number of the previous down-sampling layer of convolutional layer,Indicate p_CU-th of layer is defeated Enter characteristic pattern, while being also p_C- 1 layer of u-th of output characteristic pattern, * representing matrix convolution work as p_CWhen=C1, then p_C- 1 layer is defeated Enter layer.

Step 2-3-1-3 handles down-sampling layer: having p=p at this time_S,p_S∈ { S1, S2, S3, S4 }, step 2-3-1-2 is obtained The output characteristic pattern of the convolutional layer arrived respectively withPhase convolution, then sampled with step-length for 2, sampling obtains p_S The output characteristic pattern of layerCalculation formula is as follows:

Step 2-3-1-4 calculates F1 layers of probability vector: if network layer p is classifier layer, i.e. p=F1 is become by matrix Change, by the 32 width resolution ratio of C5 be 4 × 4 output characteristic pattern with column sequential deployment, obtain the F1 layer that resolution ratio is 512 × 1 Export feature vectorThen calculate separately horizontal parameters matrix W H, Vertical Parameters matrix W V withApposition, will calculate As a result it sums respectively with Horizontal offset parameter BH, vertical off setting parameter BV, summed result obtains water after the processing of Softmax function Flat probability vector HPV_test, vertical probability vector VPV_test, calculation formula is as follows:

Step 2-3-2 the following steps are included:

The present invention provides a kind of Radar Echo Extrapolation methods based on dynamic convolutional neural networks, implement the technology There are many method and approach of scheme, the above is only a preferred embodiment of the present invention, it is noted that for the art Those of ordinary skill for, various improvements and modifications may be made without departing from the principle of the present invention, these change It also should be regarded as protection scope of the present invention into retouching.The available prior art of each component part being not known in the present embodiment adds To realize.

Claims

1. a kind of Radar Echo Extrapolation method based on dynamic convolutional neural networks, which comprises the following steps:

Step 1, the offline convolutional neural networks of training: input training image collection carries out data prediction to training image collection, obtains Training sample set designs dynamic convolution neural network structure, and initializes network training parameter；It is dynamic using training sample set training The orderly image sequence of state convolutional neural networks, input obtains a width prognostic chart by dynamic convolutional neural networks propagated forward Picture calculates forecast image and compares the error between label, the weighting parameter and offset parameter of network are updated by backpropagation, This process is repeated until reaching trained termination condition, obtains convergent dynamic convolutional neural networks；

Step 2, online Radar Echo Extrapolation: input test image set carries out data prediction to test chart image set, is tested Sample set, then by the dynamic convolutional neural networks obtained in test sample collection input step 1, by network propagated forward meter Calculate probability vector, and by input image sequence last width radar return image and obtained probability vector phase convolution, obtain To the Radar Echo Extrapolation image of prediction；

Step 1 the following steps are included:

Step 1-1, data prediction: input training image collection, the every piece image concentrated to training image carry out at standardization Reason, converts every piece image to 280 × 280 floating number image, obtains floating number image collection, to floating number image collection It is divided, construction includes the training sample set of TrainsetSize group sample；

Step 1-2 initializes dynamic convolutional neural networks: design dynamic convolution neural network structure is configured to generating probability The sub-network of vector reconstructs the probabilistic forecasting layer for image extrapolation, provides dynamic volume for the offline neural metwork training stage Product neural network initialization model；

Step 1-3 initializes dynamic convolution Neural Network Training Parameter: enabling e-learning rate λ=0.0001, the training stage is each The sample size BatchSize=10 of input, most large quantities of frequency of training of training sample set Currently crowd frequency of training BatchNum=1, the maximum number of iterations IterationMax=40 of network training, current iteration number IterationNum=1；

Step 1-4 reads training sample: by the way of batch training, the training sample that training is obtained from step 1-1 every time is concentrated BatchSize group training sample is read, every group of training sample is { x₁,x₂,x₃,x₄, y }, it include altogether 5 width images, wherein { x₁,x₂, x₃,x₄It is used as input image sequence, y is corresponding control label；

Step 1-5, propagated forward: the input image sequence feature that extraction step 1-4 is obtained in a sub-network obtains level probability Vector HPV and vertical probability vector VPV；In probabilistic forecasting layer, by the last piece image in input image sequence successively with VPV, HPV phase convolution obtain the output forecast image of propagated forward；

Step 1-6, backpropagation: reversely acquiring the error term of probability vector in probabilistic forecasting layer, according to the mistake of probability vector The error term of poor Xiang Conghou each network layer into preceding layer-by-layer calculating subnet network layers, and then calculate error term pair in each network layer The gradient of weighting parameter and offset parameter utilizes the parameter of obtained gradient updating dynamic convolutional neural networks；

Off-line training step control: step 1-7 carries out whole control to the offline neural metwork training stage, is divided into following three kinds Situation:

If training sample is concentrated there are still original training sample, i.e. BatchNum < BatchMax, then return step 1-4 Continue to read BatchSize group training sample, carries out network training；

If training sample, which is concentrated, is not present original training sample, i.e. BatchNum=BatchMax, and current network changes Generation number is less than maximum number of iterations, i.e. IterationNum < IterationMax then enables BatchNum=1, return step 1-4 continues to read BatchSize group training sample, carries out network training；

If training sample, which is concentrated, is not present original training sample, i.e. BatchNum=BatchMax, and network iteration time Number reaches maximum number of iterations, i.e. IterationNum=IterationMax, then terminates the offline neural metwork training stage, obtain To trained dynamic convolution neural network model.

2. the method according to claim 1, wherein step 1-1 data prediction the following steps are included:

Step 1-1-1, sampling: the image that training image is concentrated is sequentially arranged, and constant duration is distributed, between the time It is divided into 6 minutes, altogether includes N_TrainWidth image determines TrainsetSize by following formula:

Wherein, Mod (N_Train, 4) and indicate N_TrainTo 4 modulus,Expression is not more thanMaximum integer, acquire After TrainsetSize, training image is retained by sampling and concentrates preceding 4 × TrainsetSize+1 width image, when sampling, which passes through, deletes Except training image concentrates last image to meet the requirements amount of images；

Normalized images: step 1-1-2 carries out image transformation, normalization operation, by original resolution to the image that sampling obtains The floating number image that resolution ratio is 280 × 280 is converted into for 2000 × 2000 color image；

Step 1-1-3 constructs training sample set: training sample set is constructed using the floating number image set that step 1-1-2 is obtained, it will Every four adjacent images in floating number image set, i.e., { 4N+1,4N+2,4N+3,4N+4 } width image is as one group of input sequence Column, [4 × (N+1)+1] width image is by cutting, and the part that the central resolution ratio of reservation is 240 × 240 is as corresponding sample Label is compareed, for N group sampleIts make is as follows:

Wherein, step 1-1-2 the following steps are included:

Step 1-1-2-1, image conversion: converting gray level image for the step 1-1-1 image sampled, is retained by cutting The image resolution ratio boil down to 280 × 280 after cutting is divided in the part that original image center resolution ratio is 560 × 560 The grayscale image that resolution is 280 × 280；

Step 1-1-2-2, data normalization: by the value of each of the grayscale image obtained in step 1-1-2-1 pixel from [0~255] is mapped to [0~1], by obtaining the floating number image that resolution ratio is 280 × 280 after normalization.

3. according to the method described in claim 2, it is characterized in that, step 1-2 the following steps are included:

Step 1-2-1 constructs sub-network:

Sub-network is made of 10 network layers, is followed successively by convolutional layer C1, down-sampling layer S1, convolutional layer C2, down-sampling from front to back Layer S2, convolutional layer C3, down-sampling layer S3, convolutional layer C4, down-sampling layer S5, convolutional layer C5 and classifier layer F1；

Step 1-2-2 constructs probabilistic forecasting layer:

Dynamic convolutional layer DC1 and dynamic convolutional layer DC2 is constructed in probabilistic forecasting layer, the vertical probability vector that sub-network is exported Convolution kernel of the VPV as dynamic convolutional layer DC1, convolution kernel of the level probability vector HPV as dynamic convolutional layer DC2；

Wherein, step 1-2-1 the following steps are included:

Step 1-2-1-1 constructs convolutional layer: determining the following contents: the output characteristic pattern quantity OutputMaps of convolutional layer, convolution Core k and offset parameter bias, for convolution kernel, it is thus necessary to determine that the width KernelSize of convolution kernel, the quantity of convolution kernel KernelNumber, the quantity of convolution kernel are the product that the convolutional layer inputs and exports characteristic pattern quantity, the resolution ratio of convolution kernel For KernelSize × KernelSize, and convolution kernel is constructed according to Xavier initial method；For offset parameter, quantity It is identical as the output characteristic pattern quantity of this layer；For convolutional layer l_C,l_C∈ { C1, C2, C3, C4, C5 }, the output characteristic pattern of this layer Width is Value by convolutional layer l_CInput feature vector figure resolution ratio and convolution kernel widthIt codetermines, i.e.,

Indicate convolutional layer l_CIt is upper one layer volume The output characteristic pattern width of lamination；

For convolutional layer C1, C1 layers of output characteristic pattern quantity OutputMaps is enabled^C1The width of=12, C1 layers of output characteristic pattern OutputSize^C1=272, C1 layers of convolution kernel width KernelSize^C1=9, C1 layers of offset parameter bias^C1It is initialized as zero, C1 layers of convolution kernel k^C1Quantity KernelNumber^C1=48, the initial value of each parameter is in convolution kernelRand () is for generating random number；

For convolutional layer C2, C2 layers of output characteristic pattern quantity OutputMaps are enabled^C2The width of=32, C2 layers of output characteristic pattern OutputSize^C2=128, C2 layers of convolution kernel width KernelSize^C2=9, C2 layers of offset parameter are initialized as zero, C2 layers Convolution kernel k^C2Quantity KernelNumber^C2=384, the initial value of each parameter is in convolution kernel

For convolutional layer C3, C3 layers of output characteristic pattern quantity OutputMaps are enabled^C3The width of=32, C3 layers of output characteristic pattern OutputSize^C3=56, C3 layers of convolution kernel width KernelSize^C3=9, C3 layers of offset parameter are initialized as zero, C3 layers Convolution kernel k^C3Quantity KernelNumber^C3=1024, the initial value of each parameter is in convolution kernel

For convolutional layer C4, C4 layers of output characteristic pattern quantity OutputMaps are enabled^C4The width of=32, C4 layers of output characteristic pattern OutputSize^C4=20, C4 layers of convolution kernel width KernelSize^C4=9, C4 layers of offset parameter are initialized as zero, C4 layers Convolution kernel k^C4Quantity KernelNumber^C4=1024, the initial value of each parameter is in convolution kernel

For convolutional layer C5, C5 layers of output characteristic pattern quantity OutputMaps are enabled^C5The width of=32, C5 layers of output characteristic pattern OutputSize^C5=4, C5 layers of convolution kernel width KernelSize^C5=7, C5 layers of offset parameter are initialized as zero, C5 layers Convolution kernel k^C5Quantity KernelNumber^C5=1024, the initial value of each parameter is in convolution kernel

Step 1-2-1-2, construct down-sampling layer: in down-sampling layer do not include need training parameter, by down-sampling layer S1, S2, The sampling core of S3 and S4 is initialized asFor down-sampling layer l_S,l_S∈ { S1, S2, S3, S4 }, output are special Levy figure quantityIt is consistent with the output characteristic pattern quantity of one layer of convolutional layer thereon, exports characteristic pattern widthIt is the 1/2 of the output characteristic pattern width of one layer of convolutional layer thereon, formula is expressed as follows:

Step 1-2-1-3, structural classification device layer: classifier layer is made of a full articulamentum F1, and F1 layers of weighting parameter is water Equal rights value parameter matrix W H and vertical weighting parameter matrix W V, size are 41 × 512, enable each of weighting parameter matrix The initial value of parameter isOffset parameter is Horizontal offset parameter BH and vertical off setting parameter BV, just Beginning turns to 41 × 1 one-dimensional null vector.

4. according to the method described in claim 3, it is characterized in that, step 1-5 the following steps are included:

Step 1-5-1, sub-network calculate probability vector: being extracted in a sub-network by the alternate treatment of convolutional layer and down-sampling layer The image sequence characteristic of input is handled in classifier layer by Softmax function, and level probability vector HPV and vertical is obtained Probability vector VPV；

Step 1-5-2, calculate probabilistic forecasting layer and export image: the HPV and VPV that step 1-5-1 is obtained are as probabilistic forecasting layer Convolution kernel, by the last piece image in input image sequence successively with VPV, HPV phase convolution, the output for obtaining propagated forward is pre- Altimetric image.

5. according to the method described in claim 4, it is characterized in that, step 1-5-1 the following steps are included:

Step 1-5-1-1, judges network layer type: the network layer in the sub-network being presently in is indicated with l, l initial value is C1, Judge the type of network layer l, if l ∈ { C1, C2, C3, C4, C5 }, then l is convolutional layer, step 1-5-1-2 is executed, if l ∈ S1, S2, S3, S4 }, then l is down-sampling layer, executes step 1-5-1-4；

Step 1-5-1-2 handles convolutional layer: having l=l at this time_C,l_C∈ { C1, C2, C3, C4, C5 }, first calculating l_CThe jth of layer A output characteristic patternBy l_CThe input feature vector figure convolution nuclear phase convolution corresponding with this layer respectively of layer, convolution results are summed, Summed result adds l_CJ-th of offset parameter of layerIt handles, obtains using ReLU activation primitiveCalculation formula is such as Shown in lower:

Wherein,For l_CI-th of input feature vector figure convolution kernel corresponding with j-th of output characteristic pattern of layer, n are current convolution The output characteristic pattern number of the previous down-sampling layer of layer, Indicate l_CI-th of input feature vector of layer Figure, while being also l_C- 1 layer of i-th of output characteristic pattern, * representing matrix convolution, if l_C=C1, then l_C- 1 layer is input layer；

All output characteristic patterns are successively calculated, l is obtained_CThe output characteristic pattern of layerL is updated to l+1, and return step 1- 5-1-1 judges network type, carries out the operation of next network layer；

Step 1-5-1-3 handles down-sampling layer: having l=l at this time_S,l_S∈ { S1, S2, S3, S4 }, step 1-5-1-2 is obtained The output characteristic pattern of convolutional layer respectively withPhase convolution, then sampled with step-length for 2, sampling obtains l_SLayer Export characteristic patternCalculation formula is as follows:

In above formula, Sample () indicates that step-length is 2 sampling processing, l_S- 1 indicates the previous convolutional layer of current down-sampling layer, Indicate l_SThe output characteristic pattern of layerIn j-th of output characteristic pattern, obtain l_SThe output characteristic pattern of layerAfterwards, l is updated to l + 1, and return step 1-5-1-1 judges network type, carries out the operation of next network layer；

Step 1-5-1-4 calculates F1 layers of probability vector: if network layer l is classifier layer, i.e. l=F1 will by matrixing For the output characteristic pattern that the 32 width resolution ratio of C5 are 4 × 4 with column sequential deployment, the output for obtaining the F1 layer that resolution ratio is 512 × 1 is special Levy vector a^F1, calculate separately horizontal weighting parameter matrix W H and a^F1Apposition, vertical weighting parameter matrix W V and a^F1Apposition, Calculated result is summed with Horizontal offset parameter BH, vertical off setting parameter BV respectively, obtains level after the processing of Softmax function Probability vector HPV and vertical probability vector VPV, specific formula for calculation are as follows:

HPV=Softmax (WH × a^F1+BH)

VPV=Softmax (WV × a^F1+ BV),

Step 1-5-2 the following steps are included:

Step 1-5-2-1 predicts DC1 layers of vertical direction: by last width input picture of input layer and vertical probability vector VPV Phase convolution obtains the DC1 layer that resolution ratio is 240 × 280 and exports characteristic pattern a^DC1；

Step 1-5-2-2 predicts DC2 layers of vertical direction: by DC1 layers of output characteristic pattern a^DC1It is rolled up with level probability vector HPV phase Product, obtains the output forecast image of propagated forward, and resolution ratio is 240 × 240.

6. according to the method described in claim 5, it is characterized in that, step 1-6 the following steps are included:

Step 1-6-1 calculates probabilistic forecasting layer error term: by the training sample of the step 1-5-2-2 forecast image obtained and input Control label in this asks poor, calculates the error term of DC2 layers, DC1 layers, finally acquires the error term δ of level probability vector^HPVWith The error term δ of vertical probability vector^VPV；

Step 1-6-2 calculates sub-network error term: according to the error term δ of level probability vector^HPVWith the mistake of vertical probability vector Poor item δ^VPV, from rear to preceding classification the layer F1, convolutional layer C5, C4, C3, C2, C1 and down-sampling layer S4, S3, S2, S1 of successively calculating Error term, the resolution ratio of any layer error term matrix acquired and the output resolution ratio of characteristic pattern of this layer are consistent；

Step 1-6-3 calculates gradient: the error term of each network layer of sub-network is calculated according to the error term that step 1-6-2 is obtained To the gradient value of this layer of weighting parameter and offset parameter；

Step 1-6-4, undated parameter: by the gradient value of the weighting parameter of the step 1-6-3 each network layer obtained and offset parameter It is multiplied by the learning rate of dynamic convolutional neural networks, obtains the update item of each network layer weighting parameter and offset parameter, by former weight Parameter and offset parameter ask poor with the update item respectively, obtain updated weighting parameter and offset parameter.

7. according to the method described in claim 6, it is characterized in that, step 1-6-1 the following steps are included:

Step 1-6-1-1 calculates dynamic convolutional layer DC2 error term: the forecast image that step 1-5-2-2 is obtained and this group of sample Control label ask poor, obtain size be 240 × 240 error term matrix delta^DC2；

Step 1-6-1-2 calculates dynamic convolutional layer DC1 error term: by zero padding by DC2 layers of error term matrix delta^DC2It expands It is 240 × 320, by level probability Vector rotation 180 degree, by the error term matrix after expansion and the level probability vector after overturning Phase convolution obtains DC1 layers of error term matrix delta^DC1, size is 240 × 280, and formula is as follows:

δ^DC1=Expand_Zero (δ^DC2) * rot180 (HPV),

Wherein, Expand_Zero () indicates zero extended function, and rot180 () indicates that angle is 180 ° of rotation function, by 2 × 2 matrix zero is extended for 4 × 4 matrix, the matrix after zero expansion, the region and original matrix phase one that central resolution ratio is 2 × 2 It causes, zero pixel filling of remaining position；

Step 1-6-1-3 calculates probability vector error term: the error term of level probability vector HPV is calculated, by DC1 layers of output Characteristic pattern and error term matrix delta^DC2Phase convolution obtains 1 × 41 row vector after convolution, which is the error term δ of HPV^HPV, public Formula is as follows:

δ^HPV=a^DC1*δ^DC2,

The error term for calculating vertical probability vector VPV, by the input feature vector figure of input layer and error term matrix delta^DC1Phase convolution, volume 41 × 1 column vector is obtained after product, which is the error term δ of VPV^VPV, formula is as follows:

Step 1-6-2 the following steps are included:

Step 1-6-2-1 calculates classifier layer F1 error term: by the error term δ of the step 1-6-1-3 probability vector obtained^VPVWith δ^HPVWeighting parameter matrix W V vertical with F1 layers and horizontal weighting parameter matrix W H carry out matrix multiple respectively, then by matrix Apposition is summed and is averaged, and F1 layers of error term δ is obtained^F1, formula is as follows:

Step 1-6-2-2 calculates convolutional layer C5 error term: by matrixing, by the mistake of the F1 layer obtained in step 1-6-2-1 Poor item δ^F1It is transformed to the matrix that 32 resolution ratio are 4 × 4Obtain C5 layers of error term δ^C5,It indicates to become The matrix that the 32nd resolution ratio after changing is 4 × 4；

Step 1-6-2-3, judges network layer type: the network layer in the sub-network being presently in is indicated with l, l initial value is S4, Judge the type of network layer l, if l ∈ { S4, S3, S2, S1 }, then l is down-sampling layer, step 1-6-2-4 is executed, if l ∈ C4, C3, C2, C1 }, then l is convolutional layer, executes step 1-6-2-5；

Step 1-6-2-4 calculates down-sampling layer error term: if l layers are down-sampling layer, there is l=l at this time_S,l_S∈{S1,S2,S3, S4 }, for l_SI-th of error term matrix of layerBy zero padding respectively by l_S+ 1 layer of each error term matrixIt opens up It opens up to width and isWherein l_SThe latter convolutional layer of+1 layer of current down-sampling layer of expression, j indicate l_S+ 1 layer of jth A error term, and haveAgain by corresponding convolution kernel180 degree is rotated, then by the square after expansion Battle array and the convolution nuclear phase convolution after overturning, and convolution results are summed, obtain l_SI-th of error term matrix of layerFormula is such as Shown in lower:

Wherein,

All error term matrixes are successively calculated, l is obtained_SThe output characteristic pattern of layerL is updated to l-1, and return step 1- 6-2-3 judges network type, calculates the error term of a upper network layer；

Step 1-6-2-5 calculates convolutional layer error term: if l layers are convolutional layer, there is l=l at this time_C,l_C∈{C1,C2,C3,C4, C5 }, it is not in l since the initial value of l in step 1-6-2-3 is S4_CThe case where=C5, for l_CI-th of layer Error term matrixFirst to l_CCorresponding i-th of error term matrix in+1 layerIt is up-sampled, it will when up-sampling In each element error entry value average mark into sampling area, obtaining resolution ratio isIt is upper Sampling matrix, then activation primitive is calculated in l_CThe inner product of derivative and the up-sampling matrix acquired at layer character pair figure, obtains l_C I-th of error term matrix of layerFormula is as follows:

UpSample () indicates up-sampling function, the corresponding up-sampling area of each of original image pixel after up-sampling Domain in each of original pixel value mean allocation to sampling area pixel, successively calculates all error term matrixes, obtains l_CThe output characteristic pattern of layer

If l ≠ C1, l is updated to l-1, and return step 1-6-2-3 judges network type, calculates the mistake of a upper network layer Poor item；

If l=C1, the calculating of step 1-6-2 sub-network error term terminates；

Step 1-6-3 the following steps are included:

Step 1-6-3-1 calculates convolutional layer error term to the gradient of convolution kernel: using l_CIndicate currently processed convolutional layer, l_C∈ { C1, C2, C3, C4, C5 } successively calculates each convolutional layer error term to the gradient of convolution kernel, by l since C1 layers_CThe i-th of -1 layer A output characteristic patternWith l_CJ-th of error term matrix of layerPhase convolution, convolution results are the gradient of corresponding convolution kernel ValueFormula is as follows:

l_C∈{C1,C2,C3,C4,C5},

In above formula,WithRespectively indicate l_CThe output characteristic pattern number and l of layer_C- 1 layer Characteristic pattern number is exported, ▽ k is gradient value of the error term to convolution kernel；

Step 1-6-3-2 calculates each convolutional layer error term to the gradient of biasing: using l_CIndicate currently processed convolutional layer, l_C∈ { C1, C2, C3, C4, C5 } successively calculates each convolutional layer error term to the gradient of biasing, by l since C1 layers_CJ-th of mistake of layer Poor item matrixIn all elements sum, obtain j-th of this layer biasing gradient valueFormula is as follows:

Wherein, Sum () expression sums to all elements of matrix；

Step 1-6-3-3 calculates F1 layers of error term to the gradient of weighting parameter: calculating separately level probability vector and vertical probability The error term δ of vector^HPV、δ^VPVWith F1 layers of error term δ^F1Inner product, calculated result be F1 layers of error term to weighting parameter WH, WV Gradient value, formula are as follows:

▽ WH=(δ^HPV)^T×(δ^F1)^T,

▽ WV=δ^VPV×(δ^F1)^T,

In above formula, ▽ WH is gradient value of the error term to horizontal weighting parameter, and ▽ WV is ladder of the error term to vertical weighting parameter Angle value；

Step 1-6-3-4 calculates F1 layers of error term to the gradient of offset parameter: by level probability vector and vertical probability vector Error term δ^HPV、δ^VPVRespectively as F1 layers of error term to the gradient value of Horizontal offset parameter BH and vertical off setting parameter BV, formula It is as follows:

▽ BH=(δ^HPV)^T,

▽ BV=δ^VPV,

Wherein, ▽ BH is gradient value of the error term to Horizontal offset parameter, and ▽ BV is gradient of the error term to vertical off setting parameter Value；

Step 1-6-4 the following steps are included:

Step 1-6-4-1 updates each convolutional layer weighting parameter: each convolutional layer error term that step 1-6-3-1 is obtained is to convolution The gradient of core is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of convolution kernel, then by former convolution kernel and the amendment Item asks poor, the convolution kernel updatedFormula is as follows:

Wherein, λ is the e-learning rate determined in step 1-3, λ=0.0001；

Step 1-6-4-2 updates each convolutional layer offset parameter: each convolutional layer error term that step 1-6-3-2 is obtained is to biasing Gradient be multiplied by the learning rates of dynamic convolutional neural networks, obtain the correction term of offset parameter, then by former bias term and the amendment Item asks poor, the bias term updatedFormula is as follows:

Step 1-6-4-3, update F1 layer weighting parameter: the F1 layer error term that step 1-6-3-3 is obtained to weighting parameter WH with The gradient value of WV is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of weighting parameter, then by former weighting parameter WH Ask poor with the correction term acquired respectively with WV, the WH and WV updated, formula is as follows:

WH=WH- λ ▽ WH,

WV=WV- λ ▽ WV；

Step 1-6-4-4, update F1 layer offset parameter: the F1 layer error term that step 1-6-3-4 is obtained to offset parameter BH with The gradient value of BV is multiplied by the learning rate of dynamic convolutional neural networks, obtains the correction term of offset parameter, then by former offset parameter BH Ask poor with the correction term acquired respectively with BV, the BH and BV updated, formula is as follows:

BH=BH- λ ▽ BH,

BV=BV- λ ▽ BV.

8. the method according to the description of claim 7 is characterized in that it is characterized in that, step 2 the following steps are included:

Step 2-1, data prediction: input test image set, the every piece image concentrated to test image carry out at standardization Reason, converts every piece image to 280 × 280 floating number image, then divide to floating number image collection, construction includes The test sample collection of TestsetSize group sample；

Step 2-2, read test sample: the TestsetSize group test sample input that step 2-1 is obtained is trained dynamic In state convolutional neural networks；

Step 2-3, propagated forward: extracting the image sequence characteristic of input in a sub-network, obtains level probability vector HPV_testWith Vertical probability vector VPV_test；In probabilistic forecasting layer, by the last piece image in input image sequence successively with VPV_test、 HPV_testPhase convolution obtains the final extrapolated image of dynamic convolutional neural networks.

9. according to the method described in claim 8, it is characterized in that, step 2-1 the following steps are included:

Step 2-1-1, sampling: the image that test image is concentrated is sequentially arranged, and constant duration is distributed, between the time It is divided into 6 minutes, altogether includes N_TestWidth image determines TestsetSize by following formula:

If Mod (N_Test, 4)=0

If Mod (N_Test, 4) ≠ 0

After acquiring TestsetSize, test image is retained by sampling and concentrates preceding 4 × TestsetSize+1 width image, when sampling Last image is concentrated to meet the requirements amount of images by deleting test image；

Image normalization: step 2-1-2 carries out image transformation, normalization operation, by original resolution to the image that sampling obtains The floating number image that resolution ratio is 280 × 280 is converted into for 2000 × 2000 color image；

Step 2-1-3 constructs test sample collection: test sample collection is constructed using the floating number image set that step 2-1-2 is obtained, it will Every four adjacent images in floating number image set, i.e., { 4M+1,4M+2,4M+3,4M+4 } width image is as one group of input sequence Column, [4 × (M+1)+1] width image is by cutting, and the part that the central resolution ratio of reservation is 240 × 240 is as corresponding sample Label is compareed, wherein for positive integer, and there is M ∈ [0, TestsetSize-1] to obtain comprising TestsetSize group test sample Test sample collection；

Step 2-1-2 the following steps are included:

Image conversion: step 2-1-2-1 converts gray level image for colored echo strength CAPPI image, then is protected by cutting Staying original image center resolution ratio is that 560 × 560 part obtains the image resolution ratio boil down to 280 × 280 after cutting The grayscale image that resolution ratio is 280 × 280；

Step 2-1-2-2, data normalization: by the value of each of the grayscale image obtained in step 1-1-2-1 pixel from [0~255] is mapped to [0~1], by obtaining the floating number image that resolution ratio is 280 × 280 after normalization；

Step 2-3 the following steps are included:

Step 2-3-1 calculates sub-network probability vector: being extracted in a sub-network by the alternate treatment of convolutional layer and down-sampling layer Then the image sequence characteristic of input is handled in classifier layer by Softmax function, level probability vector HPV is obtained_test With vertical probability vector VPV_test；

Step 2-3-2 calculates probabilistic forecasting layer and exports image: the VPV that step 2-3-1 is obtained_testAnd HPV_testAs probabilistic forecasting Layer convolution kernel, by the last piece image in input image sequence successively with VPV_testAnd HPV_testPhase convolution, obtains dynamic volume The final extrapolated image of product neural network；

Step 2-3-1 the following steps are included:

Step 2-3-1-1, judges network layer type: indicating the network layer in the sub-network being presently in p, judges network layer p Type, p initial value be C1, if p ∈ { C1, C2, C3, C4, C5 }, then p be convolutional layer, execute step 2-3-1-2, if p ∈ { S1, S2, S3, S4 }, then p is down-sampling layer, executes step 2-3-1-4；

Step 2-3-1-2 handles convolutional layer: having p=p at this time_C,p_C∈ { C1, C2, C3, C4, C5 }, first calculating p_CThe v of layer A output characteristic patternBy p_CThe input feature vector figure convolution nuclear phase convolution corresponding with this layer respectively of layer, then convolution results are asked P is added with, summed result_CV-th of biasing of layerIt is handled using ReLU activation primitive, obtains p_CV-th of layer is defeated Characteristic pattern outCalculation formula is as follows:

Wherein,For p_CU-th of input feature vector figure convolution kernel corresponding with v-th of output characteristic pattern of layer, m are current convolution The output characteristic pattern number of the previous down-sampling layer of layer, Indicate p_CU-th of input feature vector of layer Figure, while being also p_C- 1 layer of u-th of output characteristic pattern, * representing matrix convolution work as p_CWhen=C1, then p_C- 1 layer is input layer；

All output characteristic patterns are successively calculated, p is obtained_CThe output characteristic pattern of layerP is updated to p+1, and return step 2- 3-1-1 judges network type, carries out the operation of next network layer；

Step 2-3-1-3 handles down-sampling layer: having p=p at this time_S,p_S∈ { S1, S2, S3, S4 }, step 2-3-1-2 is obtained The output characteristic pattern of convolutional layer respectively withPhase convolution, then sampled with step-length for 2, sampling obtains p_SLayer Export characteristic patternCalculation formula is as follows:

Wherein, Sample () indicates that step-length is 2 sampling processing, p_S- 1 indicates the previous convolutional layer of current down-sampling layer, Indicate p_SThe output characteristic pattern of layerIn j-th of output characteristic pattern, obtain p_SThe output characteristic pattern of layerAfterwards, p is updated to P+1, and return step 2-3-1-1 judges network type, carries out the operation of next network layer；

Step 2-3-1-4 calculates F1 layers of probability vector: if network layer p is classifier layer, i.e. p=F1 will by matrixing For the output characteristic pattern that the 32 width resolution ratio of C5 are 4 × 4 with column sequential deployment, the output for obtaining the F1 layer that resolution ratio is 512 × 1 is special Levy vectorThen calculate separately horizontal parameters matrix W H, Vertical Parameters matrix W V withApposition, calculated result is distinguished With Horizontal offset parameter BH, vertical off setting parameter BV sum, summed result through Softmax function processing after obtain level probability to Measure HPV_test, vertical probability vector VPV_test, calculation formula is as follows:

Step 2-3-2 the following steps are included:

Step 2-3-2-1 predicts DC1 layers of vertical direction: by last width input picture of input layer and vertical probability vector VPV_testPhase convolution obtains the DC1 layer that resolution ratio is 240 × 280 and exports characteristic pattern

Step 2-3-2-2 predicts DC2 layers of vertical direction: step 2-3-2-1 is obtainedWith level probability vector HPV_test Phase convolution, obtains the final extrapolated image of dynamic convolutional neural networks, and resolution ratio is 240 × 240.