Dam deformation information prediction model training method, prediction method and electronic equipment

Abstract

The embodiment of the present disclosure discloses a dam deformation information prediction model training method, a prediction method and an electronic device. A specific implementation of the method includes: obtaining a historical dam image data sequence of a target dam; generating each historical dam deformation point information sequence based on the historical dam image data sequence, each historical dam deformation point information sequence corresponding to a dam deformation point; normalizing each historical dam deformation point information sequence and a dam feature information set corresponding to the target dam; seasonally encoding the historical time corresponding to each historical dam deformation point information to obtain seasonal coding information; generating a sample set based on each normalized historical dam deformation point information sequence, a dam feature information set and each seasonal coding information obtained; training an initial dam deformation information prediction model based on the sample set to obtain a dam deformation information prediction model. This implementation improves the prediction accuracy of the dam deformation amount.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computer technology, and in particular to a dam deformation information prediction model training method, a prediction method, and an electronic device. Background Art

[0002] InSAR technology, short for Interferometric Synthetic Aperture Radar, is a microwave measurement method that can obtain high-precision surface deformation. InSAR technology has high resolution, wide coverage, no need for ground equipment, and all-weather time-series monitoring capabilities. It is widely used in research fields such as geological disaster monitoring, subsidence monitoring, and urban development changes. It is one of the important tools in the field of earth science research and natural disaster monitoring.

[0003] The complex changes in climate and topography in the area where the dam is located, water erosion, and lack of maintenance of the dam will cause the strength of the dam to gradually decrease. Once a dam fails, it will threaten the safety of life and property of surrounding residents and cause economic losses. Therefore, it is necessary to predict the deformation of the dam. At present, when predicting the deformation of the dam, the commonly used method is to use LSTM to capture the medium- and long-term surface deformation feature dependency to predict the dam deformation.

[0004] However, the inventors have found that when the above method is used to predict the dam deformation, the following technical problems often occur: the non-time-series deformation characteristics cannot be captured, resulting in poor prediction accuracy of the dam deformation.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0006] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] Some embodiments of the present disclosure propose a dam deformation information prediction model training method, a dam deformation information prediction method, a dam deformation information prediction model training device, a dam deformation information prediction device, an electronic device and a computer-readable medium to solve one or more of the technical problems mentioned in the above background technology section.

[0008] In a first aspect, some embodiments of the present disclosure provide a method for training a dam deformation information prediction model, the method comprising: obtaining a historical dam image data sequence of a target dam within a preset historical time period; generating each historical dam deformation point information sequence based on the historical dam image data sequence, wherein each historical dam deformation point information sequence corresponds to a dam deformation point, each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time, and each historical time corresponding to each historical dam deformation point information sequence is the same; normalizing each historical dam deformation point information sequence and the dam feature information set corresponding to the target dam to obtain each historical dam deformation point after normalization. An information sequence and a normalized dam feature information set, wherein the dam feature information set corresponds to each dam deformation point corresponding to each historical dam deformation point information sequence; for each historical dam deformation point information in each historical dam deformation point information sequence, seasonal coding is performed on the historical time corresponding to the historical dam deformation point information to obtain seasonal coding information; a sample set is generated based on each normalized historical dam deformation point information sequence, the normalized dam feature information set and each seasonal coding information obtained; based on the sample set, an initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model.

[0009] In a second aspect, some embodiments of the present disclosure provide a dam deformation information prediction model training device, the device comprising: an image data acquisition unit, configured to acquire a historical dam image data sequence of a target dam within a preset historical time period; a deformation point information generation unit, configured to generate each historical dam deformation point information sequence based on the above historical dam image data sequence, wherein each historical dam deformation point information sequence corresponds to a dam deformation point, each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time, and each historical time corresponding to each historical dam deformation point information sequence is the same; a normalization processing unit, configured to perform normalization processing on each of the above historical dam deformation point information sequences and the dam feature information set corresponding to the above target dam, to obtain each historical dam deformation point information sequence after normalization processing. A dam deformation point information sequence and a normalized dam feature information set, wherein the dam feature information set corresponds to each dam deformation point corresponding to each historical dam deformation point information sequence; a coding unit configured to perform season coding on the historical time corresponding to each historical dam deformation point information in each historical dam deformation point information sequence to obtain season coding information; a sample generating unit configured to generate a sample set based on each normalized historical dam deformation point information sequence, the normalized dam feature information set and each obtained season coding information; a training unit configured to train an initial dam deformation information prediction model based on the sample set to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model.

[0010] In a third aspect, some embodiments of the present disclosure provide a method for predicting dam deformation information, the method comprising: obtaining each dam deformation point information sequence and a real-time dam feature information set corresponding to each dam deformation point corresponding to a target dam; normalizing the above-mentioned each dam deformation point information sequence and the above-mentioned real-time dam feature information set to obtain each normalized dam deformation point information sequence and the real-time dam feature information set; for each dam deformation point information sequence in the above-mentioned each dam deformation point information sequence, performing the following steps: seasonally encoding each time corresponding to the above-mentioned dam deformation point information sequence to obtain each seasonal encoding information; converting the above-mentioned real-time dam feature information sequence into a real-time dam feature information set; The real-time dam feature information corresponding to the above-mentioned dam deformation point information sequence in the feature information set is determined as the target real-time dam feature information; based on the above-mentioned dam deformation point information sequence and the above-mentioned target real-time dam feature information, fused dam deformation point information is generated; the combined respective fused dam deformation point information is input into a pre-trained dam deformation information prediction model to obtain respective predicted dam deformation information corresponding to the target time, wherein the above-mentioned target time corresponds to the next time of the end time, the above-mentioned end time is the time corresponding to the last dam deformation point information in any dam deformation point information sequence, and the above-mentioned dam deformation information prediction model is trained by the method described in any implementation method of the above-mentioned first aspect.

[0011] In a fourth aspect, some embodiments of the present disclosure provide a dam deformation information prediction device, the device comprising: an information acquisition unit, configured to acquire each dam deformation point information sequence and a real-time dam feature information set corresponding to each dam deformation point corresponding to a target dam; an information normalization processing unit, configured to normalize the above-mentioned each dam deformation point information sequence and the above-mentioned real-time dam feature information set to obtain each normalized dam deformation point information sequence and the real-time dam feature information set; an execution unit, configured to perform the following steps for each dam deformation point information sequence in the above-mentioned each dam deformation point information sequence: seasonally encode each time corresponding to the above-mentioned dam deformation point information sequence to obtain each seasonal encoding; code information; determining the real-time dam feature information corresponding to the dam deformation point information sequence in the real-time dam feature information set as the target real-time dam feature information; generating fused dam deformation point information based on the dam deformation point information sequence and the target real-time dam feature information; an input unit is configured to input the combined fused dam deformation point information into a pre-trained dam deformation information prediction model to obtain each predicted dam deformation information corresponding to the target time, wherein the target time corresponds to the next time of the end time, the end time is the time corresponding to the last dam deformation point information in any dam deformation point information sequence, and the dam deformation information prediction model is trained by the method described in any implementation of the first aspect.

[0012] In a fifth aspect, some embodiments of the present disclosure provide an electronic device, comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by one or more processors, the one or more processors implement the method described in any implementation manner of the above-mentioned first aspect.

[0013] In a sixth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner of the first aspect above is implemented.

[0014] The above-mentioned embodiments of the present disclosure have the following beneficial effects: the dam deformation information prediction model obtained by the dam deformation information prediction model training method of some embodiments of the present disclosure improves the prediction accuracy of the dam deformation amount. Specifically, the reason for the poor prediction accuracy of the dam deformation amount is that it is impossible to capture the non-time series deformation characteristics, resulting in poor prediction accuracy of the dam deformation amount. Based on this, the dam deformation information prediction model training method of some embodiments of the present disclosure first obtains the historical dam image data sequence of the target dam within a preset historical time period. Thus, the original image data sequence of deformation-related information for determining the dam deformation point can be obtained in advance. Then, based on the above-mentioned historical dam image data sequence, each historical dam deformation point information sequence is generated. Among them, each historical dam deformation point information sequence corresponds to a dam deformation point. Each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time. Each historical time corresponding to each historical dam deformation point information sequence is the same. Thus, the original image data sequence can be used to generate a deformation-related information sequence for each dam deformation point within a preset historical time period. Next, normalize the above-mentioned historical dam deformation point information sequences and the dam feature information set corresponding to the above-mentioned target dam to obtain normalized historical dam deformation point information sequences and normalized dam feature information sets. Among them, the dam feature information set corresponds to each dam deformation point corresponding to the above-mentioned historical dam deformation point information sequences. Thus, the dimensions of each data are unified. It should be noted that the dam feature information set corresponding to the target dam can be used as a physical attribute-related feature of the dam, that is, a non-time series feature. Secondly, for each historical dam deformation point information in the above-mentioned historical dam deformation point information sequences, the historical time corresponding to the above-mentioned historical dam deformation point information is seasonally encoded to obtain seasonal coding information. Thus, the seasonal feature can be determined based on the time series feature corresponding to the deformation-related information of the deformation point. The seasonal feature can also be used as a non-time series feature. Then, based on the normalized historical dam deformation point information sequences, the normalized dam feature information set and the obtained seasonal coding information, a sample set is generated. Therefore, the sample set for training the model can be formed by using the information sequence of each historical dam deformation point in time series, the non-time series dam feature information set and each seasonal coding information. Finally, based on the sample set, the initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model. Therefore, the sample set including time series features and non-time series features can be used to train the dam deformation information prediction model for dam deformation prediction. Also because the samples for training the dam deformation information prediction model take into account both time series features and non-time series features, the dam deformation information prediction model can capture more comprehensive features, thereby improving the prediction accuracy of the dam deformation variable. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0016] Figure 1 is a flowchart of some embodiments of the dam deformation information prediction model training method disclosed in the present invention;

[0017] Figure 2 is a schematic diagram of dam deformation points according to the dam deformation information prediction model training method disclosed in the present invention;

[0018] Figure 3 is a schematic diagram of the model structure of a dam deformation information prediction model according to the dam deformation information prediction model training method disclosed in the present invention;

[0019] Figure 4 is a flow chart of some embodiments of the dam deformation information prediction method according to the present disclosure;

[0020] Figure 5 It is a schematic diagram of the structure of some embodiments of the dam deformation information prediction model training device disclosed in the present invention;

[0021] Figure 6 is a schematic structural diagram of some embodiments of the dam deformation information prediction device disclosed in the present invention;

[0022] Figure 7 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0023] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0024] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0025] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0026] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0027] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0028] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0029] Figure 1 The process 100 of some embodiments of the dam deformation information prediction model training method according to the present disclosure is shown. The dam deformation information prediction model training method comprises the following steps:

[0030] Step 101, obtaining a historical dam image data sequence of a target dam within a preset historical time period.

[0031] In some embodiments, the execution subject (e.g., server) of the dam deformation information prediction model training method may obtain a historical dam image data sequence of the target dam within a preset historical time period from a database or server via a wired connection or a wireless connection. The target dam may be a dam whose deformation is to be predicted. The preset historical time period may be a pre-set historical time period. The historical dam image data sequence may be an image data sequence of the target dam collected within the preset historical time period. Each historical dam image data corresponds to a historical collection time point. For example, the historical dam image data in the historical dam image data sequence may be SAR image data. The number of each historical dam image data included in the historical dam image data sequence may be greater than or equal to a preset number. For example, the preset number may be 25.

[0032] It should be noted that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0033] Step 102: Generate a deformation point information sequence of each historical dam based on the historical dam image data sequence.

[0034] In some embodiments, the execution subject may generate each historical dam deformation point information sequence based on the historical dam image data sequence. Each historical dam deformation point information sequence corresponds to a dam deformation point. The dam deformation point may be a point-shaped deformation object on the dam. As an example, the dam deformation point may refer to Figure 2 . Each dam deformation point may correspond to image coordinates and geographic location coordinates. The geographic location coordinates may be three-dimensional location coordinates. Each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time. The historical time may be expressed as a time period or as a time point. When the historical time is expressed as a time period, it may be a historical time period consisting of two historical acquisition time points corresponding to two corresponding frames of adjacent historical dam image data. When the historical time is expressed as a time point, it may be a historical acquisition time point corresponding to the latter of the two corresponding frames of adjacent historical dam image data. The historical times corresponding to each historical dam deformation point information sequence are the same. That is, the historical dam deformation point information sequences of different dam deformation points are organized on the same time dimension. The historical dam deformation point information may include deformation variables and historical time.

[0035] In some optional implementations of some embodiments, the execution subject may generate each historical dam deformation point information sequence based on the historical dam image data sequence through the following steps:

[0036] The first step is to preprocess the historical dam image data sequence to obtain the preprocessed historical dam image data sequence. In practice, the execution subject can perform at least one of the following processing on the historical dam image data sequence: image format conversion, radiation correction, study area clipping, orbit parameter refinement, to obtain the preprocessed historical dam image data sequence. Image format conversion can refer to converting the original image data into raster data.

[0037] The second step is to perform image registration and geocoding on the preprocessed historical dam image data sequence to obtain the processed historical dam image data sequence. In practice, the above-mentioned execution subject can use the feature point matching method to perform image registration on the preprocessed historical dam image data sequence. The above-mentioned execution subject can refer to DEM to perform geocoding on the preprocessed historical dam image data sequence.

[0038] The third step is to filter the processed historical dam image data sequence to obtain the filtered historical dam image data sequence. In practice, the above execution subject can perform joint pixel filtering on the processed historical dam image data sequence based on JSInSAR technology to improve phase coherence.

[0039] The fourth step is to generate each historical dam deformation point information sequence based on the filtered historical dam image data sequence. Among them, each historical dam deformation point information sequence corresponds to a dam deformation point. Each dam deformation point corresponds to location coordinate information, and each historical dam deformation point information corresponds to historical time. The historical times corresponding to each historical dam deformation point information in the historical dam deformation point information sequence are arranged in ascending order. In practice, the above-mentioned execution subject can use PS-InSAR technology to detect the deformation variable of the filtered historical dam image data sequence to obtain each historical dam deformation point information sequence corresponding to each dam deformation point. Specifically, the interference phase of the point data can be generated by identifying each PS point and each DS point. Then, the flat ground phase and terrain phase components are removed by secondary difference. Then, the deformation rate can be estimated and the elevation correction can be performed to obtain the unwrapped phase and residual phase. Secondly, the influence of atmospheric phase and noise can be removed by spatiotemporal filtering to obtain the nonlinear deformation phase. Finally, the high-precision time series deformation can be solved and the deformation results can be geocoded.

[0040] Step 103, normalizing each historical dam deformation point information sequence and the dam feature information set corresponding to the target dam to obtain each normalized historical dam deformation point information sequence and the normalized dam feature information set.

[0041] In some embodiments, the execution subject may normalize the historical dam deformation point information sequences and the dam feature information set corresponding to the target dam to obtain normalized historical dam deformation point information sequences and normalized dam feature information sets. The dam feature information set corresponds to each dam deformation point corresponding to each historical dam deformation point information sequence. That is, one dam feature information corresponds to one dam deformation point. The dam feature information set may be pre-acquired.

[0042] Optionally, the dam characteristic information in the above dam characteristic information set includes height and slope.

[0043] In some optional implementations of some embodiments, the execution subject may normalize the historical dam deformation point information sequences and the dam feature information set corresponding to the target dam through the following steps to obtain normalized historical dam deformation point information sequences and normalized dam feature information sets:

[0044] The first step is to normalize the above-mentioned historical dam deformation point information sequences to obtain normalized historical dam deformation point information sequences. In practice, the above-mentioned execution entity can use Z-score normalization to normalize the above-mentioned historical dam deformation point information sequences to obtain normalized historical dam deformation point information sequences.

[0045] The second step is to normalize the above dam feature information set to obtain the normalized dam feature information set. In practice, the above execution entity can use Z-score normalization to normalize the above dam feature information set to obtain the normalized dam feature information set.

[0046] In the third step, for each dam feature information in the normalized dam feature information set, perform the following steps:

[0047] First, the height and slope included in the dam feature information are subjected to feature cross processing to obtain a slope height feature. In practice, the execution entity may determine the product of the normalized height and slope as the slope height feature.

[0048] Then, the slope height feature is added to the dam feature information to update the normalized dam feature information. Thus, the slope height feature can reflect the influence relationship between the height of the dam deformation point and the slope, which helps to distinguish high slopes, gentle slopes and other terrains, so as to further mine the dam feature information.

[0049] Step 104: for each historical dam deformation point information in each historical dam deformation point information sequence, seasonal coding is performed on the historical time corresponding to the historical dam deformation point information to obtain seasonal coding information.

[0050] In some embodiments, for each historical dam deformation point information in the above-mentioned historical dam deformation point information sequences, the above-mentioned execution entity may perform season coding on the historical time corresponding to the above-mentioned historical dam deformation point information to obtain season coding information.

[0051] In some optional implementations of some embodiments, the execution subject may perform seasonal coding on the historical time corresponding to the historical dam deformation point information to obtain seasonal coding information through the following steps:

[0052] The first step is to determine the historical time corresponding to the above historical dam deformation point information as the target historical time.

[0053] The second step is to select the preset season coding information corresponding to the above-mentioned target historical time from the preset season coding information set as the season coding information corresponding to the above-mentioned target historical time. The above-mentioned preset season coding information set includes at least one preset season coding information corresponding to each season. The preset season coding information set can be each season coding information obtained after encoding each season into at least one season coding information. For example, there are four seasons in total, and each season can be evenly divided into two time periods before and after the corresponding time, and then each time period is encoded into a binary vector using a one-hot method. For example, the preset season coding information set can be expressed as: {T1 = [1,0,0,0,0,0,0,0], T2 = [0,1,0,0,0,0,0,0], T3 = [0,0,1,0,0,0,0,0], T4 = [0,0,0,1,0,0,0], T5 = [0,0,0,0,1,0,0,0], T6 = [0,0,0,0,0,1,0,0], T7 = [0,0,0,0,0,0,1,0], T8 = [0,0,0,0,0,0,0,1]}. Among them, T1 and T2 correspond to the first season. T3 and T4 correspond to the second season. T5 and T6 correspond to the third season. T7 and T8 correspond to the fourth season. The preset season coding information corresponding to the above-mentioned target historical time can be the preset season coding information corresponding to the time period in which the above-mentioned target historical time is located. Therefore, the encoding sampling frequency of seasonal features can be expanded, so that more fine-grained seasonal features can be captured.

[0054] Step 105, generating a sample set based on the normalized historical dam deformation point information sequence, the normalized dam feature information set and the obtained seasonal coding information.

[0055] In some embodiments, the execution entity may generate a sample set based on each normalized historical dam deformation point information sequence, a normalized dam feature information set, and each obtained seasonal coding information.

[0056] In some optional implementations of some embodiments, the execution subject may generate a sample set based on each normalized historical dam deformation point information sequence, the normalized dam feature information set and each obtained seasonal coding information:

[0057] In the first step, for each of the historical dam deformation point information sequences in the above normalized historical dam deformation point information sequences, the following steps are performed:

[0058] In the first sub-step, the dam deformation point corresponding to the above historical dam deformation point information sequence is determined as the target dam deformation point.

[0059] In the second sub-step, the dam feature information corresponding to the target dam deformation point in the normalized dam feature information set is determined as the target dam feature information.

[0060] The third sub-step is to combine the above historical dam deformation point information sequence and the above target dam characteristic information into a sample.

[0061] In the second step, the combined samples are determined as a sample set. Thus, the dam characteristic information and the historical dam deformation point information sequence can be integrated from the dimension of the dam deformation point to form a sample.

[0062] Step 106: Based on the sample set, the initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model.

[0063] In some embodiments, the above-mentioned execution entity may train the initial dam deformation information prediction model based on the sample set, and obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model. Among them, the initial dam deformation information prediction model may be an initial neural network model with undetermined model weights. The initial neural network model may include an encoding layer, a decoding layer, and an output layer. The encoding layer may include at least one multi-head attention mechanism layer, at least one residual connection and normalization layer, and at least one forward propagation layer. The decoding layer may include at least one multi-head attention mechanism layer, at least one residual connection and normalization layer, and a forward propagation layer. The output layer can be used to calculate the loss function, so that the optimizer back-propagation can be used to update the model parameters until the model converges, and then the model weights can be saved. As an example, the model structure of the dam deformation information prediction model can refer to Figure 3 In practice, the above-mentioned execution subject can use the batch training model training method, take the sample set as input data, and take the future dam deformation prediction information as output data to train the initial dam deformation information prediction model, and obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model. By providing an attention mechanism, the model can capture the long-term feature dependency on the historical dam deformation point information sequence, thereby improving the prediction accuracy of the dam deformation variable.

[0064] The output of the attention module in the multi-head attention layer can be expressed as:

[0065]

[0066] Attention Weights=softmax(Attention Scores)

[0067] Output=Attention Weights·V.

[0068] Among them, Output can represent the output of the multi-head attention mechanism layer. Q can represent the query matrix. K can represent the key matrix. V can represent the value matrix. k The dimensions of the key can be represented.

[0069] In practice, m attention mechanism modules are calculated in parallel, and their outputs are spliced ​​together to form a multi-head attention module, which is used to capture complex patterns of different positions and relationships in the sequence. The number of heads m of the multi-head attention mechanism can be reasonably adjusted according to the amount of input data and the situation of computing hardware. The larger the m, the larger the capacity of the model, the stronger the fitting ability, but the corresponding computing resources required. Basic layer components such as feedforward neural network layer and normalization layer are added in series after the multi-head attention mechanism layer to adjust the number of output channels and accelerate model training, forming the basic module of the network to capture the complex feature dependencies of the input data. According to the specific task scenario and computing resources, a suitable number of basic modules are selected to be connected in series in sequence to form the encoder of the model. This structure is responsible for feature extraction and encoding of the input feature sequence, mapping the input sequence to a high-dimensional representation space to capture the key information in the input sequence. The decoding layer can be constructed by the self-attention module, the encoding-decoding attention mechanism, and the feedforward neural network. The decoding layer can use the context information provided by the encoder to generate each transformed feature vector for input to the output layer.

[0070] The loss function of the model can be expressed as follows:

[0071]

[0072]

[0073] Among them, w i It can be a time weight factor, which represents the data weight of the historical dam deformation point information at the lth time point. smooth (θ) can represent the smoothing regularization term. t It can represent the deformation point information of the dam at time point t. The deformation point information of the dam can include the deformation variable. λ can represent the regularization strength coefficient. Through this regularization term, the deformation difference between adjacent time nodes can be constrained not to be too large, thereby eliminating the excessive interference of abnormal deformation points on the whole.

[0074] In some optional implementations of some embodiments, the execution subject may train the initial dam deformation information prediction model based on the sample set through the following steps to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model:

[0075] The first step is to divide the sample set to obtain an initial training sample set, a validation sample set, and a test sample set. In practice, the execution entity may divide the sample set according to a preset ratio to obtain an initial training sample set, a validation sample set, and a test sample set. The preset ratio may be 7:2:1.

[0076] In the second step, each initial training sample that meets preset relevant conditions is selected from the above initial training sample set as a screening training sample set.

[0077] The third step is to generate a training sample set based on the above-screened training sample set and the above-mentioned seasonal coding information.

[0078] The fourth step is to train the initial dam deformation information prediction model based on the training sample set, and obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model.

[0079] In some optional implementations of some embodiments, the execution subject may select each initial training sample that meets preset related conditions from the initial training sample set as the screening training sample set through the following steps:

[0080] In the first step, for every two initial training samples in the above initial training samples, perform the following steps:

[0081] In the first sub-step, the two initial training samples are respectively determined as the first initial training sample and the second initial training sample.

[0082] The second sub-step is to determine the mean of each first field value included in the first initial training sample as the first mean value. It should be noted that each first field value may include normalized height, slope, and each historical dam deformation point information.

[0083] The third sub-step is to determine the mean of the second field values ​​included in the second initial training sample as the second mean.

[0084] The fourth sub-step is to perform the following steps for each first field value included in the first initial training sample and based on each second field value included in the second initial training sample:

[0085] First, a difference between the first field value and the first mean is determined as a first difference.

[0086] Secondly, the square of the first difference is determined as a first square value.

[0087] Next, a difference between the second field value and the second mean is determined as a second difference.

[0088] Then, the square of the second difference value is determined as the second square value.

[0089] Finally, the product of the first difference and the second difference is determined as the first correlation component information.

[0090] The fifth sub-step is to determine each determined first related component information as the first related information.

[0091] In a sixth sub-step, the sum of the determined first square values ​​is determined as a first square sum.

[0092] The seventh sub-step is to determine the sum of the determined second square values ​​as the second square sum.

[0093] In the eighth sub-step, second relevant information is generated according to the product of the first square sum and the second square sum. In practice, the execution subject may determine the second relevant information as the product of the first square sum and the second square sum to the power of one half.

[0094] In a ninth sub-step, a ratio of the first relevant information to the second relevant information is determined as a first relevant value.

[0095] The tenth sub-step is to determine the correlation value between the first initial training sample and the second initial training sample based on the first correlation value.

[0096] In the second step, based on the determined correlation values, each initial training sample that meets the preset correlation condition is selected from the above initial training samples as a screening sample set. The above preset correlation condition can be that the correlation value corresponding to the initial training sample is the TopN correlation value among the above correlation values. There is no limitation on the specific setting of N.

[0097] In some optional implementations of some embodiments, the execution subject may determine the correlation value of the first initial training sample and the second initial training sample based on the first correlation value through the following steps:

[0098] In the first step, the position coordinate information corresponding to the first initial training sample is determined as the first position coordinate information, wherein the position coordinate information may be the three-dimensional geographic coordinates of the dam deformation point.

[0099] In the second step, the position coordinate information corresponding to the second initial training sample is determined as the second position coordinate information.

[0100] In the third step, based on the first position coordinate information and the second position coordinate information, distance information is generated as a second correlation value. In practice, the execution entity may use the first position coordinate information as a row vector, the second position coordinate information as a column vector, and determine the dot product of the row vector and the column vector as the distance information.

[0101] The fourth step is to perform weighted processing on the above-mentioned first correlation value and the above-mentioned second correlation value to obtain the correlation value of the above-mentioned first initial training sample and the above-mentioned second initial training sample. In practice, the above-mentioned execution entity may determine the product of the above-mentioned first correlation value and the first weighting coefficient as the first numerical value. Then, the product of the above-mentioned second correlation value and the second weighting coefficient may be determined as the second numerical value. Finally, the sum of the above-mentioned first numerical value and the above-mentioned second numerical value may be determined as the correlation value. Here, there is no limitation on the specific setting of the first weighting coefficient and the second weighting coefficient. Thus, the determined correlation value can characterize the spatiotemporal joint correlation of the two dam deformation points.

[0102] In some optional implementations of some embodiments, the execution subject may generate a training sample set based on the above-mentioned screening training sample set and the above-mentioned seasonal coding information through the following steps:

[0103] In the first step, each historical time corresponding to any filtered training sample in the filtered training sample set is determined as a historical time set.

[0104] The second step is to construct a seasonal coding matrix based on the seasonal coding information corresponding to the historical time set. In practice, the execution subject can use the seasonal coding information as row vectors to form a seasonal coding matrix.

[0105] Step 3: For each screening training sample in the above screening training sample set, perform the following steps:

[0106] The first sub-step is to determine the arrangement sequence number of the above-mentioned screening training samples in the above-mentioned screening training sample set.

[0107] The second sub-step is to combine the above arrangement number with the above historical time number into the deformation arrangement position information for each historical time number corresponding to the above historical time set. The historical time number can be the arrangement number of the historical time in the historical time set. For example, for the first screening training sample in the screening training sample set and the first historical time in the historical time set, the combined deformation arrangement position information can be "11".

[0108] The third sub-step is to encode the combined deformation arrangement position information to obtain each arrangement position encoding vector. In practice, the above execution subject can perform binary encoding on each deformation arrangement position information.

[0109] The fourth sub-step is to construct an arrangement position coding matrix based on the above-mentioned arrangement position coding vectors. In practice, the above-mentioned execution subject can use the arrangement position coding vectors as row vectors to form an arrangement position coding matrix. The arrangement position coding matrix can represent the relative arrangement position of each historical dam deformation point information in the historical dam deformation point information sequence when it is used as input data.

[0110] The fifth sub-step is to generate an arrangement position adjustment coefficient according to the dimension value of the corresponding arrangement position coding vector and the dimension value of the screening sample. The dimension value of the corresponding arrangement position coding vector may be the vector dimension of the arrangement position coding vector, which may be preset. The dimension value of the screening sample may be the sample dimension of the screening sample. In practice, the above-mentioned execution subject may determine the product of the first preset value and the dimension value of the corresponding arrangement position coding vector as the multiplied dimension value. The first preset value may be 2. Then, the ratio of the multiplied dimension value to the dimension value of the screening sample may be determined as the dimension ratio. Next, the above-mentioned dimension ratio power of the second preset value may be determined as the product value. The second preset value may be 10000. Secondly, the reciprocal of the product value may be determined as the reciprocal of the product value. Then, in response to determining that the dimension value of the corresponding arrangement position coding vector is an odd number, the reciprocal of the above-mentioned product value is input as an independent variable into the cosine function to obtain the cosine value as the arrangement position adjustment coefficient. Secondly, in response to determining that the dimension value of the corresponding arrangement position coding vector is an even number, the inverse of the above product value can be input as an independent variable into the sine function to obtain the sine value as the arrangement position adjustment coefficient.

[0111] The sixth sub-step is to determine the product of the arrangement position adjustment coefficient and the arrangement position coding matrix as the adjusted arrangement position coding matrix.

[0112] The seventh sub-step is to determine the sum of the adjusted arrangement position coding matrix and the season coding matrix as the arrangement position coding information corresponding to the screened samples.

[0113] An eighth sub-step is to add the arrangement position encoding information to the screening training samples to update the screening training samples.

[0114] The fourth step is to determine the updated screened training samples as the training sample set.

[0115] The above-mentioned first step to the fourth step, as an inventive point of an embodiment of the present disclosure, solves the technical problem of "failure to capture periodic deformation characteristics, resulting in poor prediction accuracy of dam deformation variables". The factors that lead to poor prediction accuracy of dam deformation variables are often as follows: failure to capture fine-grained periodic deformation characteristics. If the above factors are solved, the effect of improving the prediction accuracy of dam deformation variables can be achieved. In order to achieve this effect, the present disclosure incorporates arrangement position coding information that combines temporal arrangement position and seasonal characteristics into the training samples, so that the periodic seasonal characteristics can be integrated into the model training by means of arrangement position coding, and then the model can process the order information of the input sequence and the time period information of the dam deformation through the arrangement position coding information included in the training samples to capture the periodic deformation characteristics, thereby improving the prediction accuracy of the dam deformation variables.

[0116] The above-mentioned embodiments of the present disclosure have the following beneficial effects: the dam deformation information prediction model obtained by the dam deformation information prediction model training method of some embodiments of the present disclosure improves the prediction accuracy of the dam deformation amount. Specifically, the reason for the poor prediction accuracy of the dam deformation amount is that it is impossible to capture the non-time series deformation characteristics, resulting in poor prediction accuracy of the dam deformation amount. Based on this, the dam deformation information prediction model training method of some embodiments of the present disclosure first obtains the historical dam image data sequence of the target dam within a preset historical time period. Thus, the original image data sequence of deformation-related information for determining the dam deformation point can be obtained in advance. Then, based on the above-mentioned historical dam image data sequence, each historical dam deformation point information sequence is generated. Among them, each historical dam deformation point information sequence corresponds to a dam deformation point. Each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time. Each historical time corresponding to each historical dam deformation point information sequence is the same. Thus, the original image data sequence can be used to generate a deformation-related information sequence for each dam deformation point within a preset historical time period. Next, normalize the above-mentioned historical dam deformation point information sequences and the dam feature information set corresponding to the above-mentioned target dam to obtain normalized historical dam deformation point information sequences and normalized dam feature information sets. Among them, the dam feature information set corresponds to each dam deformation point corresponding to the above-mentioned historical dam deformation point information sequences. Thus, the dimensions of each data are unified. It should be noted that the dam feature information set corresponding to the target dam can be used as a physical attribute-related feature of the dam, that is, a non-time series feature. Secondly, for each historical dam deformation point information in the above-mentioned historical dam deformation point information sequences, the historical time corresponding to the above-mentioned historical dam deformation point information is seasonally encoded to obtain seasonal coding information. Thus, the seasonal feature can be determined based on the time series feature corresponding to the deformation-related information of the deformation point. The seasonal feature can also be used as a non-time series feature. Then, based on the normalized historical dam deformation point information sequences, the normalized dam feature information set and the obtained seasonal coding information, a sample set is generated. Therefore, the sample set for training the model can be formed by using the information sequence of each historical dam deformation point in time series, the non-time series dam feature information set and each seasonal coding information. Finally, based on the sample set, the initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model. Therefore, the sample set including time series features and non-time series features can be used to train the dam deformation information prediction model for dam deformation prediction. Also because the samples for training the dam deformation information prediction model take into account both time series features and non-time series features, the dam deformation information prediction model can capture more comprehensive features, thereby improving the prediction accuracy of the dam deformation variable.

[0117] Further references Figure 4 , which shows a process 400 of some embodiments of the dam deformation information prediction method. The process 400 of the dam deformation information prediction method includes the following steps:

[0118] Step 401, obtaining each dam deformation point information sequence and real-time dam feature information set corresponding to each dam deformation point corresponding to the target dam.

[0119] In some embodiments, the execution subject (e.g., computing device) of the dam deformation information prediction method can obtain the information sequence of each dam deformation point corresponding to each dam deformation point of the target dam and the real-time dam feature information set from the database or server through a wired connection or a wireless connection. The real-time dam feature information set can represent the latest updated dam feature information of each dam deformation point.

[0120] Step 402, normalizing each dam deformation point information sequence and the real-time dam feature information set to obtain each dam deformation point information sequence after normalization and the real-time dam feature information set after normalization.

[0121] In some embodiments, the execution subject may normalize the above-mentioned dam deformation point information sequences and the above-mentioned real-time dam feature information set to obtain normalized dam deformation point information sequences and normalized real-time dam feature information sets. Figure 1 The corresponding step 103 in the embodiments will not be described in detail here.

[0122] Step 403: for each dam deformation point information sequence in each dam deformation point information sequence, perform the following steps:

[0123] Step 4031, seasonally encode each time corresponding to the dam deformation point information sequence to obtain seasonal coding information.

[0124] In some embodiments, the execution subject may perform seasonal coding on each time corresponding to the dam deformation point information sequence to obtain seasonal coding information. Figure 1 The corresponding step 104 in the embodiments will not be described in detail here.

[0125] Step 4032: determine the real-time dam feature information corresponding to the dam deformation point information sequence in the real-time dam feature information set as the target real-time dam feature information.

[0126] In some embodiments, the execution entity may determine the real-time dam feature information corresponding to the dam deformation point information sequence in the real-time dam feature information set as the target real-time dam feature information.

[0127] Step 4033, generating fused dam deformation point information based on the dam deformation point information sequence and the target real-time dam characteristic information.

[0128] In some embodiments, the execution entity may combine the dam deformation point information sequence, the target real-time dam feature information and the arrangement position coding information into fused dam deformation point information. Thus, the dam deformation point information sequence, dam feature information and arrangement position coding information may be fused from the dimension of dam deformation points.

[0129] Optionally, the execution subject may generate fused dam deformation point information based on the dam deformation point information sequence and the target real-time dam feature information through the following steps:

[0130] The first step is to generate the arrangement position coding information corresponding to the above dam deformation point information sequence based on the arrangement sequence number of the above dam deformation point information sequence in each of the above dam deformation point information sequences and the above season coding information. In practice, the specific method of generating the arrangement position coding information corresponding to the above dam deformation point information sequence and the technical effect brought about can be referred to Figure 1 The specific implementation method and technical effects brought about by "generating a training sample set based on the above-mentioned screening training sample set and the above-mentioned seasonal coding information" in the corresponding embodiments are not repeated here.

[0131] In the second step, the above-mentioned dam deformation point information sequence, the above-mentioned target real-time dam feature information and the above-mentioned arrangement position coding information are combined into fused dam deformation point information.

[0132] Step 404, inputting the combined fused dam deformation point information into a pre-trained dam deformation information prediction model to obtain each predicted dam deformation information corresponding to the target time.

[0133] In some embodiments, the execution entity may input the combined fusion dam deformation point information into a pre-trained dam deformation information prediction model to obtain the predicted dam deformation information corresponding to the target time. The target time corresponds to the next time of the end time. The end time is the time corresponding to the last dam deformation point information in any dam deformation point information sequence. The predicted dam deformation information may include a deformation amount.

[0134] from Figure 4 It can be seen that Figure 1 Compared with the description of some corresponding embodiments, Figure 4The process 400 of the dam deformation information prediction method in some corresponding embodiments embodies the steps of applying the dam deformation information prediction model. Therefore, the solutions described in these embodiments can improve the prediction accuracy of the dam deformation amount.

[0135] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a dam deformation information prediction model training device, and these device embodiments are Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0136] like Figure 5 As shown, some embodiments of the dam deformation information prediction model training device 500 include: an image data acquisition unit 501, a deformation point information generation unit 502, a normalization processing unit 503, an encoding unit 504, a sample generation unit 505 and a training unit 506. The image data acquisition unit 501 is configured to acquire a historical dam image data sequence of a target dam within a preset historical time period; the deformation point information generation unit 502 is configured to generate each historical dam deformation point information sequence based on the above historical dam image data sequence, wherein each historical dam deformation point information sequence corresponds to a dam deformation point, each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time, and each historical dam deformation point information sequence corresponds to the same historical time; the normalization processing unit 503 is configured to normalize each historical dam deformation point information sequence and the dam feature information set corresponding to the above target dam, and obtain each normalized historical dam deformation point information sequence and each normalized dam deformation point information sequence. A feature information set, wherein the dam feature information set corresponds to each dam deformation point corresponding to each historical dam deformation point information sequence; the encoding unit 504 is configured to perform season encoding on the historical time corresponding to each historical dam deformation point information in the historical dam deformation point information sequence to obtain season encoding information; the sample generation unit 505 is configured to generate a sample set based on each normalized historical dam deformation point information sequence, the normalized dam feature information set and each obtained season encoding information; the training unit 506 is configured to train an initial dam deformation information prediction model based on the sample set to obtain a trained initial dam deformation information prediction model as a dam deformation information prediction model.

[0137] It is understood that the units described in the device 500 are similar to those described in the reference Figure 1Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 500 and the units included therein, and will not be described in detail here.

[0138] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a dam deformation information prediction model training device, and these device embodiments are Figure 4 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0139] like Figure 6 As shown, some embodiments of the dam deformation information prediction model training device 600 include: an information acquisition unit 601 and an execution unit 602. The information acquisition unit 601 is configured to acquire each dam deformation point information sequence and a real-time dam feature information set corresponding to each dam deformation point corresponding to the target dam; the information normalization processing unit is configured to perform normalization processing on the above-mentioned each dam deformation point information sequence and the above-mentioned real-time dam feature information set to obtain each normalized dam deformation point information sequence and the real-time dam feature information set after normalization processing; the execution unit 602 is configured to perform the following steps for each dam deformation point information sequence in the above-mentioned each dam deformation point information sequence: seasonally encode each time corresponding to the above-mentioned dam deformation point information sequence to obtain each seasonally encoded information sequence; information; determining the real-time dam feature information corresponding to the dam deformation point information sequence in the real-time dam feature information set as the target real-time dam feature information; generating fused dam deformation point information based on the dam deformation point information sequence and the target real-time dam feature information; an input unit is configured to input the combined fused dam deformation point information into a pre-trained dam deformation information prediction model to obtain the predicted dam deformation information corresponding to the target time, wherein the target time corresponds to the next time of the end time, the end time is the time corresponding to the last dam deformation point information in any dam deformation point information sequence, and the dam deformation information prediction model is through Figure 1 The corresponding steps in the embodiments are obtained by training.

[0140] It is understood that the units described in the device 600 are similar to those described in the reference Figure 4 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 600 and the units included therein, and will not be described in detail here.

[0141] Reference below Figure 7 , which shows a structural schematic diagram of an electronic device 700 suitable for implementing some embodiments of the present disclosure. Figure 7The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0142] like Figure 7 As shown, the electronic device 700 may include a processing device 701 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the electronic device 700 are also stored. The processing device 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0143] Typically, the following devices may be connected to the I / O interface 705: an input device 706 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 707 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 708 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 709. The communication device 709 may allow the electronic device 700 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 7 The electronic device 700 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 7 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0144] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 709, or installed from the storage device 708, or installed from the ROM 702. When the computer program is executed by the processing device 701, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0145] It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0146] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0147] The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist independently without being installed in the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains a historical dam image data sequence of the target dam within a preset historical time period; based on the above-mentioned historical dam image data sequence, generates each historical dam deformation point information sequence, wherein each historical dam deformation point information sequence corresponds to a dam deformation point, and each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time, and each historical time corresponding to each historical dam deformation point information sequence is the same; normalizes the above-mentioned each historical dam deformation point information sequence and the dam feature information set corresponding to the above-mentioned target dam to obtain each historical dam deformation point information sequence after normalization. A dam deformation point information sequence and a normalized dam feature information set, wherein the dam feature information set corresponds to each dam deformation point corresponding to each historical dam deformation point information sequence; for each historical dam deformation point information in each historical dam deformation point information sequence, seasonal coding is performed on the historical time corresponding to the historical dam deformation point information to obtain seasonal coding information; a sample set is generated based on each normalized historical dam deformation point information sequence, the normalized dam feature information set and each seasonal coding information obtained; based on the sample set, an initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model.

[0148] Or the electronic device is configured to: obtain each dam deformation point information sequence and a real-time dam feature information set corresponding to each dam deformation point corresponding to the target dam; normalize the above-mentioned each dam deformation point information sequence and the above-mentioned real-time dam feature information set to obtain each dam deformation point information sequence after normalization and a real-time dam feature information set after normalization; for each dam deformation point information sequence in the above-mentioned each dam deformation point information sequence, perform the following steps: seasonally encode each time corresponding to the above-mentioned dam deformation point information sequence to obtain each seasonal encoding information; convert the above-mentioned real-time dam feature information sequence into a real-time dam feature information set; The real-time dam feature information corresponding to the above-mentioned dam deformation point information sequence in the information set is determined as the target real-time dam feature information; based on the above-mentioned dam deformation point information sequence and the above-mentioned target real-time dam feature information, fused dam deformation point information is generated; the combined fused dam deformation point information is input into the pre-trained dam deformation information prediction model to obtain the predicted dam deformation information corresponding to the target time, wherein the above-mentioned target time corresponds to the next time of the end time, and the above-mentioned end time is the time corresponding to the last dam deformation point information in any dam deformation point information sequence, and the above-mentioned dam deformation information prediction model is through Figure 1The corresponding steps in the embodiments are obtained by training.

[0149] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0150] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0151] The units described in some embodiments of the present disclosure may be implemented by software or hardware. The described units may also be set in a processor, for example, it may be described as: a processor includes an image data acquisition unit, a deformation point information generation unit, a normalization processing unit, an encoding unit, a sample generation unit, and a training unit. The names of these units do not constitute a limitation on the units themselves in some cases. For example, the image data acquisition unit may also be described as "a unit for acquiring a historical dam image data sequence of a target dam within a preset historical time period".

[0152] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0153] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A dam deformation information prediction model training method, comprising: Acquire a historical dam image data sequence of the target dam within a preset historical time period; Based on the historical dam image data sequence, each historical dam deformation point information sequence is generated, wherein each historical dam deformation point information sequence corresponds to a dam deformation point, each historical dam deformation point information in the historical dam deformation point information sequence corresponds to a historical time, each historical time corresponding to each historical dam deformation point information sequence is the same, and the historical dam deformation point information includes a deformation amount and a historical time; Normalizing the historical dam deformation point information sequences and the dam feature information set corresponding to the target dam to obtain normalized historical dam deformation point information sequences and normalized dam feature information sets, wherein the dam feature information set corresponds to the dam deformation points corresponding to the historical dam deformation point information sequences; For each historical dam deformation point information in each historical dam deformation point information sequence, seasonal coding is performed on the historical time corresponding to the historical dam deformation point information to obtain seasonal coding information, wherein seasonal coding is performed on the historical time corresponding to the historical dam deformation point information, including: Determining the historical time corresponding to the historical dam deformation point information as the target historical time; Selecting preset season coding information corresponding to the target historical time from a preset season coding information set as the season coding information corresponding to the target historical time, wherein the preset season coding information set includes at least one preset season coding information corresponding to each season, the preset season coding information set is each season coding information obtained after encoding each season into at least one season coding information, and the preset season coding information corresponding to the target historical time is the preset season coding information corresponding to the time period where the target historical time is located; Generate a sample set based on the normalized information sequence of each historical dam deformation point, the normalized dam feature information set and the obtained seasonal coding information; Based on the sample set, the initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model, wherein the training of the initial dam deformation information prediction model based on the sample set includes: The sample set is divided into an initial training sample set, a verification sample set and a test sample set; For every two initial training samples in each initial training sample, perform the following steps: Determine the two initial training samples as a first initial training sample and a second initial training sample respectively; Determine a mean of each first field value included in the first initial training sample as a first mean; Determine a mean of each second field value included in the second initial training sample as a second mean; For each first field value included in the first initial training sample, based on each second field value included in the second initial training sample, perform the following steps: Determine a difference between the first field value and the first mean as a first difference value; determining the square of the first difference as a first square value; Determine a difference between the second field value and the second mean as a second difference value; determining the square of the second difference as a second square value; determining a product of the first difference and the second difference as first relevant component information; Determine each determined first relevant component information as first relevant information; Determine the sum of the determined first square values ​​as a first square sum; Determining a sum of the determined second square values ​​as a second square sum; determining the first power of half of the product of the first square sum and the second square sum as second relevant information; determining a ratio of the first relevant information to the second relevant information as a first relevant value; Determine the position coordinate information corresponding to the first initial training sample as the first position coordinate information, wherein the position coordinate information is the three-dimensional geographic coordinates of the dam deformation point; Determining the position coordinate information corresponding to the second initial training sample as the second position coordinate information; Taking the first position coordinate information as a row vector, taking the second position coordinate information as a column vector, determining a dot product of the row vector and the column vector as distance information, and taking the distance information as a second correlation value; Performing weighted processing on the first correlation value and the second correlation value to obtain a correlation value of the first initial training sample and the second initial training sample; Based on the determined correlation values, selecting initial training samples that meet preset correlation conditions from the initial training samples as a screening training sample set, wherein the preset correlation condition is that the correlation value corresponding to the initial training sample is a TopN correlation value among the correlation values; Generate a training sample set based on the screened training sample set and the encoding information of each season; Based on the training sample set, the initial dam deformation information prediction model is trained to obtain the trained initial dam deformation information prediction model as the dam deformation information prediction model.

2. The method according to claim 1, wherein: The generating of each historical dam deformation point information sequence based on the historical dam image data sequence comprises: Preprocessing the historical dam image data sequence to obtain a preprocessed historical dam image data sequence; Performing image registration processing and geocoding processing on the pre-processed historical dam image data sequence to obtain a processed historical dam image data sequence; Performing filtering processing on the processed historical dam image data sequence to obtain a filtered historical dam image data sequence; Based on the historical dam image data sequence after filtering, each historical dam deformation point information sequence is generated, wherein each historical dam deformation point information sequence corresponds to a dam deformation point, each dam deformation point corresponds to position coordinate information, each historical dam deformation point information corresponds to historical time, and each historical time corresponding to each historical dam deformation point information in the historical dam deformation point information sequence is arranged in ascending order.

3. The method according to claim 1, wherein: The dam feature information in the dam feature information set includes height and slope; and the normalization processing of each historical dam deformation point information sequence and the dam feature information set corresponding to the target dam is performed to obtain each normalized historical dam deformation point information sequence and the normalized dam feature information set, including: Normalizing the historical dam deformation point information sequences to obtain normalized historical dam deformation point information sequences; Normalizing the dam feature information set to obtain a normalized dam feature information set; For each dam feature information in the normalized dam feature information set, perform the following steps: Performing feature cross processing on the height and slope included in the dam feature information to obtain a slope height feature; The slope height feature is added to the dam feature information to update the normalized dam feature information.

4. The method according to claim 3, wherein: The method generates a sample set based on each normalized historical dam deformation point information sequence, the normalized dam feature information set and each seasonal coding information obtained, including: For each of the historical dam deformation point information sequences after the normalization process, the following steps are performed: Determining the dam deformation point corresponding to the historical dam deformation point information sequence as the target dam deformation point; Determine the dam feature information corresponding to the target dam deformation point in the normalized dam feature information set as the target dam feature information; Combining the historical dam deformation point information sequence and the target dam characteristic information into a sample; The combined samples are determined as a sample set.

5. A method for predicting dam deformation information, comprising: Obtaining each dam deformation point information sequence and real-time dam feature information set corresponding to each dam deformation point corresponding to the target dam; Normalizing the deformation point information sequences of each dam and the real-time dam feature information set to obtain normalized deformation point information sequences of each dam and normalized real-time dam feature information set; For each dam deformation point information sequence in the dam deformation point information sequences, the following steps are performed: Seasonally coding each time corresponding to the dam deformation point information sequence to obtain seasonal coding information; Determine the real-time dam feature information corresponding to the dam deformation point information sequence in the real-time dam feature information set as target real-time dam feature information; Based on the dam deformation point information sequence and the target real-time dam characteristic information, generating fused dam deformation point information; The combined fused dam deformation point information are input into a pre-trained dam deformation information prediction model to obtain the predicted dam deformation information corresponding to the target time, wherein the target time corresponds to the next time of the end time, and the end time is the time corresponding to the last dam deformation point information in any dam deformation point information sequence, and the dam deformation information prediction model is trained by the method described in one of claims 1-4.

6. An electronic device comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 4 or 5.

Images