CN111929683A - Landslide deformation accumulation area prediction model generation method and landslide deformation accumulation area prediction method - Google Patents
Landslide deformation accumulation area prediction model generation method and landslide deformation accumulation area prediction method Download PDFInfo
- Publication number
- CN111929683A CN111929683A CN202010738244.5A CN202010738244A CN111929683A CN 111929683 A CN111929683 A CN 111929683A CN 202010738244 A CN202010738244 A CN 202010738244A CN 111929683 A CN111929683 A CN 111929683A
- Authority
- CN
- China
- Prior art keywords
- historical data
- landslide deformation
- data
- landslide
- prediction model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
- G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
- G01S13/9021—SAR image post-processing techniques
- G01S13/9023—SAR image post-processing techniques combined with interferometric techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/86—Combinations of radar systems with non-radar systems, e.g. sonar, direction finder
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/885—Radar or analogous systems specially adapted for specific applications for ground probing
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Biophysics (AREA)
- Electromagnetism (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a landslide deformation gathering area prediction model generation method, which comprises the steps of obtaining synthetic aperture radar interferometric InSAR historical data and digital elevation model DEM historical data of a target area; obtaining a combined image data sample set of the target area according to the InSAR historical data and the DEM historical data; and obtaining a landslide deformation gathering area prediction model of the target area according to the combined image data sample set and a preset initial depth learning model. The landslide deformation accumulation area prediction model of the target area obtained based on the method can provide valuable reference data for developing applications such as landslide hidden danger identification, the landslide deformation accumulation area prediction model integrates multi-source and multi-scale characteristics of InSAR and DEM, the landslide deformation accumulation area prediction model is used for predicting the landslide deformation accumulation area of the target area, and prediction precision can be effectively improved. The invention also relates to a landslide deformation gathering area prediction method.
Description
Technical Field
The invention relates to the field of remote sensing geological and landslide geological disaster prevention and control, in particular to a landslide deformation accumulation area prediction model generation method and a landslide deformation accumulation area prediction method.
Background
China is a country with very frequent landslide hazards. In recent years, landslide disasters causing serious casualties mostly occur in high-altitude areas with complex conditions, and the landslide hazard identification method has the characteristics of concealment, high position, high destructive power and the like, is difficult to effectively eliminate the landslide hazard based on the traditional investigation method, and has urgent requirements on large-area, fast-aging and high-precision landslide hazard identification technology. However, in the past, the definition of the landslide deformation accumulation area is mainly performed in a manner of manual visual interpretation, and the method has high precision, but is time-consuming and labor-consuming and difficult to meet the application requirement of large-range and quick definition.
Disclosure of Invention
The invention aims to solve the technical problem of providing a landslide deformation accumulation area prediction model generation method and a landslide deformation accumulation area prediction method aiming at the defects of the prior art.
The technical scheme for solving the technical problems is as follows:
a method for generating a prediction model of a landslide deformation accumulation region, the method comprising:
acquiring synthetic aperture radar interferometric synthetic aperture radar (InSAR) historical data and Digital Elevation Model (DEM) historical data of a target area;
obtaining a combined image data sample set of the target area according to the InSAR historical data and the DEM historical data;
and obtaining a landslide deformation gathering area prediction model of the target area according to the combined image data sample set and a preset initial deep learning model.
The invention has the beneficial effects that: the method comprises the steps of obtaining InSAR historical data and DEM historical data of a target area, obtaining a combined image data sample set of the target area according to the InSAR historical data and the DEM historical data, and obtaining a landslide deformation aggregation area prediction model of the target area according to the combined image data sample set and a preset initial deep learning model. The landslide deformation accumulation area prediction model based on the target area can provide valuable reference data for carrying out applications such as landslide hidden danger identification, the landslide deformation accumulation area prediction model integrates multi-source and multi-scale characteristics of InSAR and DEM, and the landslide deformation accumulation area prediction model is used for predicting the landslide deformation accumulation area of the target area, so that the prediction precision can be effectively improved.
On the basis of the technical scheme, the invention can be further improved as follows.
Further, the obtaining a combined image data sample set of the target region according to the InSAR historical data and the DEM historical data specifically includes:
obtaining surface deformation phase historical data of the target area according to the InSAR historical data;
obtaining surface deformation rate historical data of the target area according to the surface deformation phase historical data;
and obtaining the surface elevation historical data, the gradient data and the slope data of the target area according to the surface deformation phase historical data and the DEM historical data.
Classifying the earth surface deformation phase historical data, the earth surface deformation rate historical data, the earth surface elevation historical data, the slope historical data and the slope historical data according to the year to obtain a combined image sample set of the target area.
The beneficial effect of adopting the further scheme is that: the prediction precision of the landslide deformation gathering area can be effectively improved by the combined image sample set of the target area constructed based on the InSAR data and the DEM data.
The obtaining of the landslide deformation aggregation area prediction model of the target area according to the combined image data sample set and a preset initial deep learning model specifically comprises:
preprocessing the combined image sample data in the combined image data sample set;
dividing the preprocessed combined image data sample set into a training set and a verification set;
training the preset initial deep learning model by using the training set to obtain the trained preset initial deep learning model;
and verifying the trained preset initial deep learning model by using the verification set to obtain an accuracy value and an error value, and selecting the preset initial deep learning model of which the accuracy value is greater than a preset accuracy threshold value and the error value is less than a preset error threshold value as a landslide deformation aggregation area prediction model of the target area.
The beneficial effect of adopting the further scheme is that: the initial deep learning model is trained and verified based on the combined image sample set of the target area constructed by InSAR data and DEM data, and the prediction precision of the landslide deformation gathering area can be effectively improved.
Further, the preprocessing the combined image sample data in the combined image data sample set specifically includes:
clipping the combined image sample data to obtain sliced image data;
carrying out augmentation processing on the slice image data through rotation, overturning and zooming;
and carrying out normalization processing on the slice image data after the augmentation processing.
Further, the initial deep learning model uses an Encoder-Decoder model, and comprises an Encoder Encoder and a Decoder Decoder, wherein the Encoder comprises a type-I convolution unit, a type-II convolution unit and an attention unit, the type-I convolution unit comprises two convolution combination layers and a pooling layer, the type-II convolution unit comprises three convolution combination layers and an upsampling layer, the convolution combination layers comprise a convolution layer, a normalization layer and an activation layer, and the attention unit comprises a global average pooling layer, a full-link layer and two activation layers;
the decoder comprises two branches, wherein the first branch receives a first characteristic layer output by the encoder, and first characteristic information is obtained after the first characteristic layer is processed by the convolution unit of the same type;
the second branch receives a second feature map output by the encoder, and second feature information is obtained after the second feature map is processed by the second type convolution unit;
fusing the first characteristic information and the second characteristic information, and then processing through an attention unit to obtain third characteristic information;
and inputting the third feature information into a softmax classifier to obtain a binary image.
Another technical solution of the present invention for solving the above technical problems is as follows:
a landslide deformation convergence zone prediction method, the method comprising:
acquiring current combined image data of a target area;
and inputting the current combined image data into a landslide deformation gathering area prediction model generated by pre-training to obtain landslide deformation gathering area data in the target area, wherein the landslide deformation gathering area prediction model is generated by training according to the landslide deformation gathering area prediction model generation method in the technical scheme.
The invention further provides a computer-readable storage medium, where instructions are stored, and when the instructions are run on a terminal device, the terminal device is caused to execute the steps of the landslide deformation aggregation area prediction model generation method or the landslide deformation aggregation area prediction method in the above technical solutions.
The present invention also provides a computer apparatus comprising: the method comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the processor executes the computer program, the steps of the landslide deformation aggregation zone prediction model generation method in the technical scheme are realized.
In addition, the present invention also provides a computer apparatus comprising: the method comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the processor executes the computer program, the steps of the landslide deformation gathering area prediction method in the technical scheme are realized.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments of the present invention or in the description of the prior art will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for generating a prediction model of a landslide deformation accumulation region according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of a method for predicting a landslide deformation accumulation area according to another embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an initial deep learning model according to another embodiment of the present invention;
fig. 4 is a diagram illustrating results obtained by using a landslide deformation accumulation region prediction method according to another embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
As shown in fig. 1, a flow diagram of a method for generating a prediction model of a landslide deformation accumulation region according to an embodiment of the present invention is shown, where the method for generating a prediction model of a landslide deformation accumulation region includes:
110. and acquiring synthetic aperture radar interferometric InSAR historical data and digital elevation model DEM historical data of a target area.
120. And obtaining a combined image data sample set of the target area according to the InSAR historical data and the DEM historical data.
130. And obtaining a landslide deformation gathering area prediction model of the target area according to the combined image data sample set and a preset initial deep learning model.
It should be appreciated that synthetic aperture radar interferometry InSAR is a new and rapidly developed surface detection technique in recent years to monitor the deformation of the surface by processing the received surface echoes at different time phases. The Digital Elevation Model (DEM) is a Digital expression of topographic surface form information, and is a full Digital expression with spatial position characteristics and topographic attribute characteristics. In this embodiment, the InSAR historical data and the DEM historical data of the target region may be obtained through the historical surface data information of the target region. According to the landslide deformation aggregation region prediction model generation method provided by the embodiment, the InSAR historical data and the DEM historical data of the target region are obtained, the combined image data sample set of the target region is obtained according to the InSAR historical data and the DEM historical data, and the landslide deformation aggregation region prediction model of the target region is obtained according to the combined image data sample set and the preset initial deep learning model. The landslide deformation accumulation area prediction model based on the target area can provide valuable reference data for carrying out applications such as landslide hidden danger identification, the landslide deformation accumulation area prediction model integrates multi-source and multi-scale characteristics of InSAR and DEM, and the landslide deformation accumulation area prediction model is used for predicting the landslide deformation accumulation area of the target area, so that the prediction precision can be effectively improved.
Further, step 120 specifically includes:
121. and obtaining surface deformation phase historical data of the target area according to InSAR historical data.
122. And obtaining the surface deformation rate historical data of the target area according to the surface deformation phase historical data.
123. And obtaining the earth surface elevation historical data, the gradient data and the slope data of the target area by the earth surface deformation phase historical data and the DEM historical data.
124. And classifying the earth surface deformation phase historical data, the earth surface deformation rate historical data, the earth surface elevation historical data, the slope historical data and the slope historical data according to the year to obtain a combined image sample set of the target area.
It should be understood that, in this embodiment, the Stacking-InSAR multi-temporal analysis technology may be used to process Sentinel-1 radar data to obtain a surface deformation phase diagram of a target area, and based on a formula, the surface deformation phase diagram is obtained
Where d denotes the amount of deformation of the line of sight LOS direction,
for phase change, λ is the signalAnd (5) obtaining the earth surface deformation rate data of the target area.
And processing the DEM data by taking the earth surface deformation phase data as a reference to obtain earth surface elevation data with the same spatial range and resolution, and respectively calculating the gradient and the slope direction according to the DEM data.
At present, gradient, slope direction and elevation data are extracted through DEM data through existing software such as ARCGIS and the like, and the gradient, slope direction and elevation data can also be extracted through a third-order inverse distance square weight difference algorithm or a fitting algorithm.
Classifying the obtained surface deformation phase data, surface deformation rate data, surface elevation data, gradient data and slope data of each historical period according to the year, for example, grouping all data of 2019 years into one group to obtain a combined image sample of 2019 years, wherein the combined image samples of all the years form a combined image sample set.
Further, step 130 specifically includes:
131. and preprocessing the combined image sample data in the combined image data sample set.
132. And dividing the preprocessed combined image data sample set into a training set and a verification set.
133. And training the preset initial deep learning model by using a training set to obtain the trained preset initial deep learning model.
134. And verifying the trained preset initial deep learning model by using the verification set to obtain an accuracy value and an error value, and selecting the preset initial deep learning model of which the accuracy value is greater than a preset accuracy threshold value and the error value is less than a preset error threshold value as a landslide deformation aggregation area prediction model of the target area.
It should be understood that, in the embodiment of the present invention, the combined image sample data is preprocessed, and a specific process is described later. The accuracy of the finally obtained landslide deformation gathering area prediction model can be higher by adopting the preprocessed combined image sample data.
In the embodiment of the invention, the processed combined image sample data is divided into two groups, one group is a training set, the other group is a verification set, the combined image data samples in the sample set are input into a preset initial deep learning model to obtain a binary image, wherein a value 1 represents a landslide deformation gathering area, a value 0 represents a non-landslide deformation gathering area, and the binary image can be further generated based on the mask file so as to represent the boundary information and the area of the landslide deformation gathering area. And inputting the combined image data samples in the verification set into a preset initial deep learning model, comparing the obtained prediction result with the manual delineation data, and if the prediction precision reaches more than 85% and the error value is less than 5%, determining that the initial deep learning model can be used as a landslide deformation gathering area prediction model of the target area.
The artificial delineation data refers to a ground surface deformation phase diagram, a landslide deformation gathering area in a target area artificially delineated needs to be removed in the artificial delineation process, and non-landslide deformation areas such as ground settlement and ground collapse and non-effective deformation caused by external factors such as atmospheric interference, DEM and inaccurate base line or interference factors such as low coherence and noise of radar data are removed. The binarized mask image is a target image to be processed, which is completely or partially blocked by a selected image, graphic or object, for example, a background area in the target image can be blocked, so as to control an image processing area or a processing process. In this embodiment, the binarized mask image may be represented as a binarized matrix for distinguishing a landslide deformation accumulation region from a non-landslide deformation accumulation region in the image, for example, the element value of the landslide deformation accumulation region is 1, and the element value of the non-landslide deformation accumulation region is 0.
Further, step 131 includes:
1311. and cutting the combined image sample data to obtain the slice image data.
1312. And carrying out augmentation processing on the slice image data through rotation, overturning and scaling.
1313. And normalizing the slice image data after the augmentation processing.
Further, step 1313 specifically includes:
dividing the slice image data into eight wave bands;
and respectively carrying out normalization processing on the eight wave band data in the slice image data.
It should be understood that, in the embodiment, the combined image data may be randomly cropped, or other cropping methods may be adopted, which may be determined according to specific situations, and the embodiment is not limited thereto.
In this embodiment, a random point is set in the combined image data, the coordinates are (x, y), a 256 × 256 clipping frame is generated along a random angle r with the random point as a vertex, the combined image data is clipped by the clipping frame, and the clipping x and y clipping of one combined image data are completed n times to obtain a plurality of pieces of 256 × 256 slice image data. The value ranges of the random point coordinates (x, y) are 0 to w-256 and 0 to h-256, respectively, and w and h are the length and width of the combined image data, respectively.
And carrying out augmentation processing on the slice image data, wherein the augmentation processing method comprises the steps of rotating, turning, dithering, zooming and the like on the slice image data, so that the original slice image data is augmented to be 32 times of the original slice image data.
The data after the augmentation processing is normalized, and as the slice image data consists of eight wave bands, the numerical value range of each wave band is different, different normalization processing methods can be adopted.
For example, since the slice image data is composed of eight bands and the numeric value ranges of each band are different, different normalization processing methods are required, where 1 to 5 bands are processed using the formula X '((Xabs-avg)/std, 6 bands are processed using the formula X' ═ X/90.0, X '═ X/360.0, and X' ═ X/255.0, 7 bands are processed using the formula X '═ X/360.0, and 8 bands are processed using the formula X' ═ X/360.0.
And combining the earth surface deformation phase data, the earth surface deformation rate data, the earth surface elevation data, the slope data and the circled true value result into combined image data with eight wave bands in sequence.
The steps can be repeated to obtain combined image data in the target area, representative images are selected from the combined image data to be subjected to normalization, data augmentation and other processing, and a training sample set and a test sample set are constructed. Further, the initial deep learning model uses an Encoder-Decoder model, and comprises an Encoder Encoder and a Decoder, wherein the Encoder comprises a first type convolution unit, a second type convolution unit and an attention unit, the first type convolution unit comprises two convolution combination layers and a pooling layer, the second type convolution unit comprises three convolution combination layers and an upsampling layer, the convolution combination layers comprise a convolution layer, a normalization layer and an activation layer, and the attention unit comprises a global average pooling layer, a full connection layer and two activation layers.
The decoder comprises two branches, wherein the first branch receives the first characteristic layer output by the encoder and obtains first characteristic information after the first characteristic layer is processed by a type of convolution unit.
And the second branch receives the second characteristic layer output by the encoder, and second characteristic information is obtained after the second characteristic layer is processed by the second type convolution unit.
And fusing the first characteristic image and the second characteristic image, and then obtaining a third characteristic image through attention unit processing.
And after the third feature map is input into the softmax classifier, outputting a mask file.
It should be understood that in this embodiment, noting the force unit means using an attention mechanism. The attention mechanism in machine learning is similar to the selective visual attention mechanism of human beings in nature, the core objective of the attention mechanism is to select information which is more critical to the current task from a plurality of information, in the embodiment, a weight matrix can be generated through an attention unit and multiplied by the original input characteristics, and effective characteristics are screened by adjusting the weight value of a channel
As shown in fig. 3, a schematic structural diagram of an initial deep learning model provided by another embodiment of the present invention, in this embodiment, the initial deep learning model is built in a tensrflow environment, the initial deep learning model mainly includes an encor and a decor, where the encor includes a type of convolution unit with five down-sampling stages, the Encoder includes a type of convolution unit, and a type of attention unit, the type of convolution unit includes two convolution combination layers and a pooling layer, the type of convolution unit includes three convolution combination layers and an up-sampling layer, the convolution combination layers include a convolution layer, a normalization layer and an activation layer, and the attention unit includes a global average pooling layer, a full-connection layer and two activation layers.
The Decoder comprises two branches, wherein the first branch receives a feature layer with the size of 1/2 and 1/4 of the original image output by the Encoder, and first feature information rich in shallow spatial information is obtained after processing through a CBR convolution combination layer; the second branch receives the feature map layer with the size of 1/8 and 1/16 of the original image output by the Encoder, and second feature information rich in deep semantic information is obtained after the feature map layer is processed by a second type of convolution unit. After the first characteristic information and the second characteristic information are fused, different weights can be given to different areas of the image after the first characteristic information and the second characteristic information are fused by an attention unit CAB (context-based adaptive binary arithmetic coding) process, so that the initial deep learning model can pay more attention to a landslide deformation gathering area and a non-landslide deformation gathering area in the image, and further screening of effective information sensitive to landslide deformation information is realized. The attention unit CAB mainly comprises a global average pooling layer GAP, a full connection layer FC, a ReLu activation layer and a Sigmoid activation layer, and is used for generating a weight layer and multiplying the weight layer by original input features so as to realize screening of effective features, obtain a feature map with the size of 1/2 of an original image, then up-sample the feature map to the size of the original image to obtain a fusion feature map, calculate classification loss of each pixel point of the fusion feature map by using a softmax classifier, determine the category of each pixel point, namely the binary value of each pixel point, and finally obtain a binary image.
As shown in fig. 2, a flow chart of a method for predicting a landslide deformation aggregation area according to another embodiment of the present invention is schematically shown, and the method for predicting a landslide deformation aggregation area includes:
210. and acquiring current combined image data of the target area.
220. And inputting the current combined image data into a landslide deformation aggregation area prediction model generated by pre-training to obtain landslide deformation aggregation area data in the target area, wherein the landslide deformation aggregation area prediction model is generated by training according to the landslide deformation aggregation area prediction model generation method in the technical scheme.
It should be understood that the combined image data of the target area is input into a landslide deformation gathering area prediction model generated by pre-training for prediction, a binary image with the same resolution and spatial range as the original image can be obtained, wherein a value of 1 represents a landslide deformation gathering area, a value of 0 represents a non-landslide deformation gathering area, and boundary information and area of the landslide deformation gathering area are obtained according to the binary image.
Fig. 4 is a schematic diagram of a result obtained by using a landslide deformation convergence region prediction method according to another embodiment of the present invention, in which fig. a is a phase diagram of surface deformation of a target region, fig. b is a schematic diagram of a prediction result of a landslide deformation convergence region of the target region, and fig. c is a vector boundary of the landslide deformation convergence region.
The invention further provides a computer-readable storage medium, where instructions are stored, and when the instructions are run on a terminal device, the terminal device is caused to execute the steps of the landslide deformation aggregation area prediction model generation method or the landslide deformation aggregation area prediction method in the above technical solutions.
The present invention also provides a computer apparatus comprising: the method comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the processor executes the computer program, the steps of the landslide deformation aggregation zone prediction model generation method in the technical scheme are realized.
In addition, the present invention also provides a computer apparatus comprising: the method comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the processor executes the computer program, the steps of the landslide deformation gathering area prediction method in the technical scheme are realized.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method for generating a prediction model of a landslide deformation accumulation area is characterized by comprising the following steps:
acquiring synthetic aperture radar interferometric synthetic aperture radar (InSAR) historical data and Digital Elevation Model (DEM) historical data of a target area;
obtaining a combined image data sample set of the target area according to the InSAR historical data and the DEM historical data;
and obtaining a landslide deformation gathering area prediction model of the target area according to the combined image data sample set and a preset initial deep learning model.
2. The method for generating the landslide deformation aggregation zone prediction model according to claim 1, wherein the obtaining of the combined image data sample set of the target zone according to the InSAR historical data and the DEM historical data specifically comprises:
obtaining surface deformation phase historical data of the target area according to the InSAR historical data;
obtaining surface deformation rate historical data of the target area according to the surface deformation phase historical data;
obtaining surface elevation historical data, gradient data and slope data of the target area according to the surface deformation phase historical data and the DEM historical data;
classifying the earth surface deformation phase historical data, the earth surface deformation rate historical data, the earth surface elevation historical data, the slope historical data and the slope historical data according to the year to obtain a combined image sample set of the target area.
3. The method for generating the landslide deformation gathering area prediction model according to claim 1, wherein the obtaining the landslide deformation gathering area prediction model of the target area according to the combined image data sample set and a preset initial deep learning model specifically comprises:
preprocessing the combined image sample data in the combined image data sample set;
dividing the preprocessed combined image data sample set into a training set and a verification set;
training the preset initial deep learning model by using the training set to obtain the trained preset initial deep learning model;
and verifying the trained preset initial deep learning model by using the verification set to obtain an accuracy value and an error value, and selecting the preset initial deep learning model of which the accuracy value is greater than a preset accuracy threshold value and the error value is less than a preset error threshold value as a landslide deformation aggregation area prediction model of the target area.
4. The method according to claim 3, wherein the preprocessing of the combined image sample data in the combined image data sample set includes:
clipping the combined image sample data to obtain sliced image data;
carrying out augmentation processing on the slice image data through rotation, overturning and scaling;
and carrying out normalization processing on the slice image data after the augmentation processing.
5. The landslide deformation accumulation zone prediction model generation method according to any one of claims 1 to 4,
the initial deep learning model uses an Encoder-Decoder model and comprises an Encoder Encoder and a Decoder, wherein the Encoder comprises a first-class convolution unit, a second-class convolution unit and an attention unit, the first-class convolution unit comprises two convolution combination layers and a pooling layer, the second-class convolution unit comprises three convolution combination layers and an upsampling layer, the convolution combination layers comprise a convolution layer, a normalization layer and an activation layer, and the attention unit comprises a global average pooling layer, a full connection layer and two activation layers;
the decoder comprises two branches, wherein the first branch receives a first characteristic layer output by the encoder, and first characteristic information is obtained after the first characteristic layer is processed by the convolution unit of the same type;
the second branch receives a second feature map output by the encoder, and second feature information is obtained after the second feature map is processed by the second type convolution unit;
fusing the first characteristic information and the second characteristic information, and then processing through an attention unit to obtain third characteristic information;
and inputting the third feature information into a softmax classifier to obtain a binary image.
6. A landslide deformation convergence zone prediction method, comprising:
acquiring current combined image data of a target area;
inputting the current combined image data into a landslide deformation gathering area prediction model generated by pre-training to obtain landslide deformation gathering area data in the target area, wherein the landslide deformation gathering area prediction model is generated by training according to the landslide deformation gathering area prediction model generation method of any one of claims 1-5.
7. A computer-readable storage medium having stored therein instructions which, when run on a terminal device, cause the terminal device to perform the steps of the landslide deformation accumulation zone prediction model generation method of any one of claims 1-5.
8. A computer-readable storage medium having stored therein instructions which, when run on a terminal device, cause the terminal device to perform the steps of the landslide deformation convergence zone prediction method of claim 6.
9. A computer device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the landslide deformation focus prediction model generation method according to any one of claims 1-5 when executing the computer program.
10. A computer device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor when executing the computer program implementing the steps of the landslide deformation convergence zone prediction method of claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010738244.5A CN111929683B (en) | 2020-07-28 | 2020-07-28 | Landslide deformation accumulation area prediction model generation method and landslide deformation accumulation area prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010738244.5A CN111929683B (en) | 2020-07-28 | 2020-07-28 | Landslide deformation accumulation area prediction model generation method and landslide deformation accumulation area prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111929683A true CN111929683A (en) | 2020-11-13 |
| CN111929683B CN111929683B (en) | 2021-06-25 |
Family
ID=73314725
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010738244.5A Active CN111929683B (en) | 2020-07-28 | 2020-07-28 | Landslide deformation accumulation area prediction model generation method and landslide deformation accumulation area prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111929683B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065455A (en) * | 2021-03-30 | 2021-07-02 | 中国水利水电科学研究院 | Landslide risk inspection method and system based on deep learning |
| CN113192086A (en) * | 2021-05-11 | 2021-07-30 | 中国自然资源航空物探遥感中心 | Generation method of geological disaster hidden danger deformation intensity distribution diagram and storage medium |
| CN113537240A (en) * | 2021-07-09 | 2021-10-22 | 北方工业大学 | Deformation region intelligent extraction method and system based on radar sequence image |
| CN113790667A (en) * | 2021-11-18 | 2021-12-14 | 中大检测(湖南)股份有限公司 | Dam deformation detection method based on radar |
| CN113887515A (en) * | 2021-10-28 | 2022-01-04 | 中国自然资源航空物探遥感中心 | Remote sensing landslide identification method and system based on convolutional neural network |
| CN114239379A (en) * | 2021-11-12 | 2022-03-25 | 云南电网有限责任公司昆明供电局 | Transmission line geological disaster analysis method and system based on deformation detection |
| CN116299438A (en) * | 2023-01-13 | 2023-06-23 | 中国南方电网有限责任公司超高压输电公司昆明局 | Ground surface deformation monitoring method and related equipment based on interference radar |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104515988A (en) * | 2014-12-16 | 2015-04-15 | 中国安全生产科学研究院 | Side slope safety monitoring and prewarning method based on ground-based synthetic aperture radar |
| CN108363886A (en) * | 2018-03-08 | 2018-08-03 | 杭州鲁尔物联科技有限公司 | Deformation prediction method and system based on deep learning |
| CN110333494A (en) * | 2019-04-10 | 2019-10-15 | 马培峰 | A kind of InSAR timing deformation prediction method, system and relevant apparatus |
| CN111223183A (en) * | 2019-11-14 | 2020-06-02 | 中国地质环境监测院 | Landslide terrain detection method based on deep neural network |
| KR20200061018A (en) * | 2018-11-23 | 2020-06-02 | 주식회사 유텔 | Synthetic aperture radar system |
| CN111337923A (en) * | 2020-04-10 | 2020-06-26 | 中国水利水电第四工程局有限公司 | Method for establishing landslide deformation time course model through time sequence InSAR data |
-
2020
- 2020-07-28 CN CN202010738244.5A patent/CN111929683B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104515988A (en) * | 2014-12-16 | 2015-04-15 | 中国安全生产科学研究院 | Side slope safety monitoring and prewarning method based on ground-based synthetic aperture radar |
| CN108363886A (en) * | 2018-03-08 | 2018-08-03 | 杭州鲁尔物联科技有限公司 | Deformation prediction method and system based on deep learning |
| KR20200061018A (en) * | 2018-11-23 | 2020-06-02 | 주식회사 유텔 | Synthetic aperture radar system |
| CN110333494A (en) * | 2019-04-10 | 2019-10-15 | 马培峰 | A kind of InSAR timing deformation prediction method, system and relevant apparatus |
| CN111223183A (en) * | 2019-11-14 | 2020-06-02 | 中国地质环境监测院 | Landslide terrain detection method based on deep neural network |
| CN111337923A (en) * | 2020-04-10 | 2020-06-26 | 中国水利水电第四工程局有限公司 | Method for establishing landslide deformation time course model through time sequence InSAR data |
Non-Patent Citations (1)
| Title |
|---|
| 李振洪 等: "卫星雷达遥感在滑坡灾害探测和监测中的应用：挑战与对策", 《武汉大学学报(信息科学版)》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065455A (en) * | 2021-03-30 | 2021-07-02 | 中国水利水电科学研究院 | Landslide risk inspection method and system based on deep learning |
| CN113065455B (en) * | 2021-03-30 | 2021-09-17 | 中国水利水电科学研究院 | Landslide risk inspection method and system based on deep learning |
| CN113192086A (en) * | 2021-05-11 | 2021-07-30 | 中国自然资源航空物探遥感中心 | Generation method of geological disaster hidden danger deformation intensity distribution diagram and storage medium |
| CN113192086B (en) * | 2021-05-11 | 2022-01-28 | 中国自然资源航空物探遥感中心 | Generation method of geological disaster hidden danger deformation intensity distribution diagram and storage medium |
| CN113537240A (en) * | 2021-07-09 | 2021-10-22 | 北方工业大学 | Deformation region intelligent extraction method and system based on radar sequence image |
| CN113537240B (en) * | 2021-07-09 | 2023-09-05 | 北方工业大学 | Deformation zone intelligent extraction method and system based on radar sequence image |
| CN113887515A (en) * | 2021-10-28 | 2022-01-04 | 中国自然资源航空物探遥感中心 | Remote sensing landslide identification method and system based on convolutional neural network |
| CN114239379A (en) * | 2021-11-12 | 2022-03-25 | 云南电网有限责任公司昆明供电局 | Transmission line geological disaster analysis method and system based on deformation detection |
| CN113790667A (en) * | 2021-11-18 | 2021-12-14 | 中大检测(湖南)股份有限公司 | Dam deformation detection method based on radar |
| CN116299438A (en) * | 2023-01-13 | 2023-06-23 | 中国南方电网有限责任公司超高压输电公司昆明局 | Ground surface deformation monitoring method and related equipment based on interference radar |
| CN116299438B (en) * | 2023-01-13 | 2023-12-01 | 中国南方电网有限责任公司超高压输电公司昆明局 | Ground surface deformation monitoring method and related equipment based on interference radar |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111929683B (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111929683B (en) | Landslide deformation accumulation area prediction model generation method and landslide deformation accumulation area prediction method | |
| Yokoya et al. | Breaking limits of remote sensing by deep learning from simulated data for flood and debris-flow mapping | |
| Park et al. | Computer vision–based estimation of flood depth in flooded-vehicle images | |
| CN113610070A (en) | Landslide disaster identification method based on multi-source data fusion | |
| CN112287983B (en) | Remote sensing image target extraction system and method based on deep learning | |
| CN110991430B (en) | Ground feature identification and coverage rate calculation method and system based on remote sensing image | |
| Lv et al. | ShapeFormer: A shape-enhanced vision transformer model for optical remote sensing image landslide detection | |
| CN103871039A (en) | Generation method for difference chart in SAR (Synthetic Aperture Radar) image change detection | |
| Janalipour et al. | Evaluation of effectiveness of three fuzzy systems and three texture extraction methods for building damage detection from post-event LiDAR data | |
| Yao et al. | Performance comparison of landslide susceptibility mapping under multiple machine-learning based models considering InSAR deformation: a case study of the upper Jinsha River | |
| Xia et al. | A Deep Learning Application for Building Damage Assessment Using Ultra-High-Resolution Remote Sensing Imagery in Turkey Earthquake | |
| CN117911879B (en) | SAM-fused fine-granularity high-resolution remote sensing image change detection method | |
| Gokon et al. | Verification of a method for estimating building damage in extensive tsunami affected areas using L-band SAR data | |
| Kim et al. | Learning Structure for Concrete Crack Detection Using Robust Super‐Resolution with Generative Adversarial Network | |
| Chen et al. | Landslide Inventory Mapping Based on Independent Component Analysis and UNet3+: A Case of Jiuzhaigou, China | |
| Parape et al. | Optimization of structure elements for morphological hit-or-miss transform for building extraction from VHR airborne imagery in natural hazard areas | |
| CN113096129B (en) | Method and device for detecting cloud cover in hyperspectral satellite image | |
| Luo et al. | Real-time pedestrian detection method based on improved YOLOv3 | |
| Zhang et al. | Fast and Slow Changes Constrained Spatio-Temporal Subpixel Mapping | |
| Markov | Fractal methods for extracting artificial objects from the unmanned aerial vehicle images | |
| Dağdelen et al. | Automated building extraction from very high-resolution remote sensing data with deep learning approaches | |
| Dubois et al. | Automated building damage classification for the case of the 2010 Haiti earthquake | |
| Seydi et al. | A New Hyperspectral Change Detection (HCD-Net) Framework Based on Double Stream CNN and Attention Module | |
| CN117611877B (en) | LS-YOLO network-based remote sensing image landslide detection method | |
| Feng et al. | Unsupervised change detection of man-made objects using coherent and incoherent features of multi-temporal SAR images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |