CN114660598A - InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method - Google Patents
InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method Download PDFInfo
- Publication number
- CN114660598A CN114660598A CN202210115680.6A CN202210115680A CN114660598A CN 114660598 A CN114660598 A CN 114660598A CN 202210115680 A CN202210115680 A CN 202210115680A CN 114660598 A CN114660598 A CN 114660598A
- Authority
- CN
- China
- Prior art keywords
- afsa
- cnn
- svm
- insar
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005065 mining Methods 0.000 title claims abstract description 85
- 238000001514 detection method Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000005305 interferometry Methods 0.000 claims abstract description 8
- 230000001629 suppression Effects 0.000 claims abstract description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 14
- 239000013598 vector Substances 0.000 claims description 5
- 241000251468 Actinopterygii Species 0.000 description 17
- 239000010410 layer Substances 0.000 description 16
- 230000000694 effects Effects 0.000 description 11
- 230000006870 function Effects 0.000 description 11
- 238000012544 monitoring process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 238000011176 pooling Methods 0.000 description 5
- 239000003245 coal Substances 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 235000013305 food Nutrition 0.000 description 3
- 230000019637 foraging behavior Effects 0.000 description 3
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000009849 deactivation Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
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
- G01S13/88—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
- G01S13/9094—Theoretical aspects
Abstract
The invention relates to an InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method, which comprises the following steps: acquiring a Sentinel-1A radar satellite image, processing the Sentinel-1A radar satellite image by a differential radar interferometry method to obtain an InSAR interferogram, and selecting a subsidence basin as a sample data set in the InSAR interferogram; establishing a CNN-AFSA-SVM model, and using the sample data set to train and classify the CNN-AFSA-SVM model to obtain a trained CNN-AFSA-SVM model; and processing the Sentinel-1A radar satellite image of the area to be detected by a differential radar interferometry method to obtain a target InSAR interferogram, inputting the target InSAR interferogram into the trained CNN-AFSA-SVM model, removing repeated search frames by adopting a non-maximum suppression method, and finally automatically outputting a mining subsidence basin detection result image. According to the method, the mining subsidence basin is effectively and automatically detected in the large-amplitude wide InSAR interferogram by constructing the CNN-AFSA-SVM model, and the accuracy and the automation degree of the detection of the mining subsidence basin are improved.
Description
Technical Field
The invention belongs to the technical field of mine detection, and particularly relates to an automatic detection method for mining subsidence basins by fusing InSAR and CNN-AFSA-SVM.
Background
Mining mineral resources tend to break the original stress equilibrium state of the overburden, causing movement and deformation of the rock and earth. Therefore, a series of mine geological environment disasters are caused, and the life and property safety of residents in a mining area is threatened. Therefore, efficient and accurate mining activity monitoring in a mining area is developed, on one hand, the method is very important for preventing geological environment disasters caused by mining, and on the other hand, scientific basis can be provided for government departments to supervise illegal mining activities.
At present, the government mainly monitors illegal mining activities according to a carpet traditional investigation method, and some illegal mining monitoring methods also utilize microseismic and information network modes. These approaches are inefficient and have a small range for monitoring a wide range of illegal mining activities. Despite certain preventive measures, it is often prohibited in the case of illegal mining. According to the statistics of the national safety production supervision and management bureau, China produces coal mine yield accounting for 35% of the world, but accounts for 80% of the death number caused by coal mining in the world, and most of mine disasters are caused by underground illegal mining. Therefore, there is a need for efficient and accurate monitoring of illegal mining activities in mine areas.
Synthetic aperture radar (InSAR) technology is widely applied to monitoring the ground surface subsidence of a mining area with the advantages of low cost, all weather, all time, high precision and the like. Ground subsidence can be formed after underground coal mining, and the surface deformation is characterized by a series of concentric circles or concentric ellipses on an InSAR interferogram, and the concentric circles or the concentric ellipses are called mining subsidence basins. Thus, the learner may monitor illegal mining problems by detecting mining subsidence basins on the InSAR interferograms. However, with the continuous development of the InSAR technology, the image amplitude of the InSAR technology is also continuously increased, the mining subsidence basin is searched in a large-scale interferogram, and the human error and the energy consumption are large by only visual search. Therefore, we need to study how to automatically find mining subsidence basins in large-breadth InSAR interferograms. In recent years, development of computer hardware and large-scale data collection help Convolutional Neural Networks (CNNs) have achieved excellent results in computer vision such as image classification and image detection. Convolutional neural networks have achieved good application in various industries, but CNNs have not been applied to mining subsidence basin detection at present.
Therefore, a new mining subsidence basin automatic detection method fusing InSAR and CNN-AFSA-SVM is needed to solve the technical problems.
Disclosure of Invention
The invention aims to solve the problems and provide an automatic detection method for mining subsidence basins, which integrates InSAR and CNN-AFSA-SVM.
The invention realizes the purpose through the following technical scheme:
an automatic detection method for mining subsidence basins fusing InSAR and CNN-AFSA-SVM comprises the following steps:
acquiring a Sentinel-1A radar satellite image, processing the Sentinel-1A radar satellite image by a differential radar interferometry method to obtain an InSAR interferogram, and selecting a subsidence basin as a sample data set in the InSAR interferogram;
establishing a CNN-AFSA-SVM model, and using the sample data set to train and classify the CNN-AFSA-SVM model to obtain a trained CNN-AFSA-SVM model;
processing the Sentinel-1A radar satellite image of the area to be detected by a differential radar interferometry method to obtain a target InSAR interferogram, inputting the target InSAR interferogram into the trained CNN-AFSA-SVM model, removing repeated search frames by adopting a non-maximum suppression method, and finally outputting a mining subsidence basin detection result image.
As a further optimization scheme of the present invention, the sample data set includes a positive sample data set and a negative sample data set, and the mining subsidence basin is selected as the positive sample data set and the non-mining subsidence basin is selected as the negative sample data set in the InSAR interferogram.
As a further optimization scheme of the invention, the specific steps of training and classifying the CNN-AFSA-SVM model by using the sample data set are as follows:
extracting feature vectors of the positive sample data set and the negative sample data set by using a CNN model;
and after the feature vectors are input into an SVM classifier, finding out an optimal punishment factor c and a kernel function parameter g through an AFSA algorithm, substituting the optimal punishment factor c and the kernel function parameter g into the SVM classifier, carrying out training and classification tests by using the SVM classifier, and carrying out training and classification tests on the test model precision to obtain a trained CNN-AFSA-SVM model.
As a further optimization scheme of the invention, the CNN model is specifically a ResNet50 convolutional neural network model, and the CNN-AFSA-SVM model is specifically a Resnet50-AFSA-SVM model.
The invention has the beneficial effects that:
the invention replaces the original softmax classifier of the CNN model with the SVM classifier with strong classification capability, and introduces an artificial fish swarm algorithm to construct an improved CNN-AFSA-SVM mining subsidence basin monitoring model, the model can effectively detect a mining subsidence basin in a large-amplitude wide InSAR interferogram, adopts non-maximum inhibition to remove repeated search frames, improves the detection precision of the mining subsidence basin, provides scientific basis for monitoring mining activities and preventing and controlling geological disasters, and also provides important reference significance for landslide and other geological disasters.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a schematic diagram of the AlexNet network structure of the present invention;
FIG. 3 is a schematic diagram of the VGGNet network structure of the present invention;
FIG. 4 is a schematic diagram of the ResNet50 network structure of the present invention;
FIG. 5 is a map of the geographical location of a Huainan mine;
FIG. 6 is a partial sample data set diagram of the present invention;
FIG. 7 is a diagram of the results of the ResNet50-AFSA-SVM model test according to the present invention.
Detailed Description
The present application will now be described in further detail with reference to the drawings, it should be noted that the following detailed description is given for illustrative purposes only and is not to be construed as limiting the scope of the present application, as those skilled in the art will be able to make numerous insubstantial modifications and adaptations to the present application based on the above disclosure.
Example 1
As shown in fig. 1 to 7, an automatic detection method for mining subsidence basins by fusing InSAR and CNN-AFSA-SVM comprises the following steps:
acquiring a Sentinel-1A radar satellite image, processing the Sentinel-1A radar satellite image by a differential radar interferometry method to obtain an InSAR interferogram, and selecting a subsidence basin as a sample data set in the InSAR interferogram;
establishing a CNN-AFSA-SVM model, and using the sample data set to train and classify the CNN-AFSA-SVM model to obtain a trained CNN-AFSA-SVM model;
and processing the Sentinel-1A radar satellite image of the area to be detected by a differential radar interferometry method to obtain a target InSAR interferogram, inputting the target InSAR interferogram into the trained CNN-AFSA-SVM model, removing repeated search frames by adopting a non-maximum suppression method, and finally outputting a mining subsidence basin detection result image.
CNN model principle: a typical CNN model consists of convolutional layers, pooling layers, fully-connected layers, and Softmax classification functions. The CNN has strong self-adaptive extraction characteristics, and a parameter sharing and interlayer connection sparsity mechanism introduced in a hidden layer can greatly reduce the number of model parameters. The invention adopts 3 classic convolution neural network models, and the principle is as follows:
AlexNet convolutional neural network model: the deep learning network comprises 6 hundred million and 3000 million links, 6000 million parameters and 65 million neuron nodes. The AlexNet structural model is shown in FIG. 2.
The main difference between the AlexNet architecture and the traditional convolutional neural network is its increased network depth, resulting in an increased number of model tunable parameters, and the regularization techniques used, such as random inactivation and data enhancement. Random deactivation techniques are applied after the first two fully-connected layers in the AlexNet architecture, reducing overfitting and better generalizing unknown examples; another significant feature of AlexNet is the use of ReLU nonlinear activation after each convolution and full-connection layer, which greatly improves training efficiency compared to the traditionally used hyperbolic tangent function.
VGGNet convolutional neural network model: the VGGNet successfully constructed 16-19 layers of the deep convolutional neural network by repeatedly stacking 3 × 3 small convolutional kernels and 2 × 2 max pooling layers, whose network structure is shown in fig. 3, by exploring the relationship between the depth of the convolutional neural network and its performance.
VGGNet has 5 convolutions with 2-3 convolutional layers in each segment, and a maximum pooling layer is connected to the end of each segment for reducing the picture size. The same number of convolution kernels per segment, the more the later segments have the greater number of convolution kernels: 64-128-256-512-wherein it is often the case that multiple identical 3 x 3 convolutional layers are stacked together-is a very useful design.
ResNet50 convolutional neural network model: the network structure is mainly composed of a plurality of residual modules. Assuming that x is the input data and f (x) represents the residual mapping, the characteristic output h (x) of the network residual module is:
H(x)＝F(x)+x (1)
when F (x) is equal to 0, indicating that the convolution layer is subjected to identity mapping; when F (x) is greater than 0, the convolutional layer learns new characteristic information, so that gradient transmission in reverse transmission is ensured, the problems of gradient loss and network degradation in the network training process are effectively solved, and the ResNet50 network structure is shown in fig. 4.
The ResNet50 network structure includes 49 convolutional layers and 1 full-link layer, and the network operation process includes 6 stages. The first stage comprises convolution, batch regularization, an activation function and a maximum pooling operation, the CONV (convolution) BLOCK in the second to fifth stages represents a convolution residual BLOCK, the ID (identity) BLOCK represents an identity residual BLOCK, and the sixth stage comprises global average pooling, a full connection layer and a softmax classifier.
AFSA algorithm principle: the AFSA algorithm is also called as an artificial fish swarm algorithm, is firstly proposed in the research of an optimization mode of an animal autonomous body, and the basic principle of the AFSA is as follows: the total artificial fish can be described as Z ═ X1,X2,X3…,Xi,…,XMM denotes the total number of AF, Xi＝(x1,x2,…,xn) Denotes the individual status of AF, xiFor the variable to be optimized, the food concentration of the current position of the artificial fish is Y ═ f (X), wherein Y is the objective function value, and the distance between the individual artificial fish is expressed as d ═ Xi-XjAnd l, Visual represents the sensing distance of the artificial fish, Step represents the maximum Step length of the movement of the artificial fish, and delta is a congestion factor, wherein delta is more than 0 and less than 1.
Foraging behavior: the current state of the artificial fish is XiRandomly selecting one state as X in its sensing rangej，Xj＝Xi+ rand (·) × Visual, rand (·) represents any random number between 0 and 1. Comparing the two food concentration functions Y when Y isi＜YjThen, go one step forward in the direction; otherwise, a state X is selectedjA comparison is made. After repeatedly trying the maximum trial times for several times, if the advancing condition of the artificial fish is still not met, randomly advancing one step, wherein the formula is as follows:
clustering behavior: the number n of artificial fishes in the current visual field range0And position X of artificial fish in the center of the swarmcFood concentration Y of central artificial fishc. When in use
Then go further in that direction, otherwise, foraging behavior is performed, as follows:
and (3) rear-end collision behavior: current field of view range YjThe smallest artificial fish is XjWhen is coming into contact with
If not, the foraging behavior is carried out, and the formula is as follows:
firstly, randomly generating artificial fish in a parameter interval, calculating a food concentration function (an objective function), and recording an optimal value. And secondly, comparing the state of each artificial fish subjected to the 3 behaviors with the optimal value, replacing the state if the state is superior to the optimal value, and determining the state of each artificial fish as the optimal state after gen (total number of iterations) iterations in the test.
The quality of the SVM algorithm depends on the values of a penalty factor c and a kernel function parameter g, the optimal penalty factor c and the kernel function parameter g are searched by using an artificial fish school algorithm with strong optimizing capability and good global convergence, and an improved Resnet50_ AFSA _ SVM mining subsidence basin detection model based on the artificial fish school algorithm is constructed.
Evaluation criteria: the invention selects the values of the precision ratio P, the recall ratio R and the F1 to evaluate the precision of model detection, and the formula is as follows:
wherein, the indexes are shown in the following table 1:
TABLE 1 meaning table of each index
The precision ratio represents the proportion of the samples divided into positive examples that are actually positive examples; the recall rate represents the proportion of detected positive samples to the actual total number of positive samples. The value of F1 represents the comprehensive identification capability of positive and negative samples, and the higher the value of F1 is, the more robust the model is.
Engineering examples
The study area was: in order to verify the InSAR exploitation subsidence basin automatic detection method based on the convolutional neural network, a Huainan mining area with a plurality of subsidence basins and obvious subsidence is used as a test area. The Huainan mining area is located in the central north of Anhui province, and the geographical range is about 116 degrees 21 ' to 117 degrees 11 ' 59 ' of east longitude and 32 degrees 32 ' 45 ' to 33 degrees 00 ' 24 ' of north latitude, as shown in FIG. 5. The Huainan mining area takes Huaihe river as a boundary, takes north as a new penexing mining area and takes south as an old mining area, and the continuous large-scale mining of the Huainan coal field makes important contribution to the national economic construction and also causes remarkable ecological environment problems.
Constructing a sample data set: the invention downloads 7 scene 'Sentinel-1A' IW mode images from 11 and 16 days in 2017 to 27 days in 1 and 2018 in Huainan mining area. The DInSAR data processing is carried out to generate 6 interference images, VV polarization is carried out, and the detailed parameters of the images are shown in a table 2.
TABLE 2 Sentinel-1A interference pairs for construction of training samples
From the interferograms formed by the 6 interference pairs shown in table 2, mining subsidence basin targets and other non-mining subsidence basin targets are selected as a sample data set, 120 mining subsidence basins are used as a positive sample data set, 180 non-mining subsidence basins are used as a negative sample data set, and part of the sample data set is shown in fig. 6.
Test results and analysis: the invention takes 80% of the selected data set as a training set and 20% as a testing set. The method comprises the steps of respectively extracting image features of three models of AlexNet, VGG19 and ResNet50, inputting the extracted feature vectors into an SVM classifier, and performing classification and precision evaluation, wherein the precision of model detection is evaluated by selecting the values of precision P, recall rate R and F1, and the precision of three CNN _ SVM models is shown in Table 3.
TABLE 3 three CNN _ SVM model precisions table
The table shows that the detection accuracy of the AlexNet model, the VGG19 model and the ResNet model 50 in combination with the SVM model for mining subsidence basins is higher than 90%, and the accuracy of the ResNet model 50_ AFSA _ SVM model constructed by the method reaches 97.6% at most. And after the model training and testing are finished, the trained model is used for detecting the mining subsidence basin of the large-width InSAR interferogram. In the research, 2 scenes of "Sentinel-1A" IW mode images from 28 days 11 and 28 days 2018 and 10 days 12 and 2018 are selected to form interference pairs as detection objects, and the research results are shown in table 4. Wherein, the accuracy rate refers to the ratio of the number of the mining subsidence basins detected by the model to the number of all the mining subsidence basins in the interference pattern; the miss rate refers to the ratio of the number of undetected mining subsidence basins to the number of all mining subsidence basins in the interferogram.
TABLE 4 table of test results
The CNN _ SVM mining subsidence basin automatic detection model constructed by the method is applied to mining subsidence basin detection on InSAR interferograms in Huainan mining areas, the CNN _ SVM mining subsidence basin automatic detection model can detect most of mining subsidence basins, the accuracy of the ResNet50_ AFSA _ SVM model improved based on the AFSA algorithm is 88.3% at most, and the detection result is shown in FIG. 7. The invention discovers, through research, that: a small number of undetected mining subsidence basins are mining areas with very small subsidence areas or areas with poor interference quality. When the edge characteristics of the mining subsidence basin are not obvious, the detection effect of the CNN _ SVM model constructed by the invention is poor. In order to improve the detection effect of the method and the expandability of the method, the method is further discussed and analyzed.
The impact of adding a dataset on the method: the influence of the number of the data sets on the CNN _ SVM mining subsidence basin automatic detection model is discussed. Because the quantity of the interference images of the mining subsidence basin is limited, the mining subsidence basin map is subjected to operations such as translation, rotation and the like to expand a data set. Taking 180 mining subsidence basins as a positive sample data set, and 240 non-mining subsidence basins as a negative sample data set, and the test result of the ResNet50-AFSA-SVM model is shown in Table 5:
table 5 ResNet50_ AFSA _ SVM model adds dataset detection results
After the ResNet50-AFSA-SVM model increases the data set, the accuracy rate of detecting the mining subsidence basin is 88.3%, and compared with the increase of 86.0%, the accuracy rate of detecting the model is improved by increasing the data set, but the detection effect is still poor for some regions with unobvious characteristics of the mining subsidence basin.
Impact of varying dataset scale on the method: the invention takes 80% of the selected data set as a training set and 20% as a testing set. In order to discuss whether the change of the proportion of the training set and the test set influences the effect of detecting the mining subsidence basin by the CNN _ SVM method, the invention carries out 2 times of tests on the + 10% of the training set, the-10% of the test set, the-10% of the training set and the + 10% of the test set aiming at the ResNet50_ CNN model, and the test results are shown in Table 6:
TABLE 6 ResNet50-AFSA-SVM model Change dataset ratio test results
From the above table, it can be seen that: the influence of changing the proportion of the data set on the accuracy of detecting the mining subsidence basin by the ResNet50_ CNN model is small, and the detection effect of the mining subsidence basin with unobvious image characteristics is still poor.
Scalability study of the method: the mining subsidence basin automatic detection method fusing InSAR and CNN-AFSA-SVM provided by the invention is successfully applied to Huainan mining areas. Wu Luchuan and the like propose that landslide is identified and monitored in early stage based on InSAR technology and optical remote sensing, and the method provided by the invention is applied to detection of ground surface deformation characteristics caused by other geological disasters such as landslide and the like in the next step, so that monitoring, prevention and control of landslide and other geological disasters are realized.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is specific and detailed, but not to be understood as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.
Claims (4)
1. An automatic detection method for mining subsidence basins fusing InSAR and CNN-AFSA-SVM is characterized by comprising the following steps:
acquiring a Sentinel-1A radar satellite image, processing the Sentinel-1A radar satellite image by a differential radar interferometry method to obtain an InSAR interferogram, and selecting a subsidence basin as a sample data set in the InSAR interferogram;
establishing a CNN-AFSA-SVM model, and using the sample data set to train and classify the CNN-AFSA-SVM model to obtain a trained CNN-AFSA-SVM model;
and processing the Sentinel-1A radar satellite image of the area to be detected by a differential radar interferometry method to obtain a target InSAR interferogram, inputting the target InSAR interferogram into the trained CNN-AFSA-SVM model, removing repeated search frames by adopting a non-maximum suppression method, and finally outputting a mining subsidence basin detection result image.
2. The method for automatically detecting the mining subsidence basin fused with InSAR and CNN-AFSA-SVM as claimed in claim 1, wherein the method comprises the following steps: the sample data set comprises a positive sample data set and a negative sample data set, and the mining subsidence basin is selected from the InSAR interferogram as the positive sample data set, and the non-mining subsidence basin is selected as the negative sample data set.
3. The method for automatically detecting the mining subsidence basin fused with the InSAR and the CNN-AFSA-SVM as claimed in claim 2, wherein the method comprises the following steps: the specific steps of training and classifying the CNN-AFSA-SVM model by using the sample data set are as follows:
extracting feature vectors of the positive sample data set and the negative sample data set by using a CNN model;
and after the feature vector is input into an SVM classifier, finding out an optimal penalty factor c and a kernel function parameter g through an AFSA algorithm, substituting the optimal penalty factor c and the kernel function parameter g into the SVM classifier, and carrying out training and classification test by using the SVM classifier to obtain a trained CNN-AFSA-SVM model.
4. The method for automatically detecting the mining subsidence basin by fusing the InSAR and the CNN-AFSA-SVM as claimed in claim 3, wherein the method comprises the following steps: the CNN model is a ResNet50 convolutional neural network model, and the CNN-AFSA-SVM model is a Resnet50-AFSA-SVM model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210115680.6A CN114660598A (en) | 2022-02-07 | 2022-02-07 | InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210115680.6A CN114660598A (en) | 2022-02-07 | 2022-02-07 | InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114660598A true CN114660598A (en) | 2022-06-24 |
Family
ID=82025623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210115680.6A Pending CN114660598A (en) | 2022-02-07 | 2022-02-07 | InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114660598A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819521A (en) * | 2023-02-22 | 2023-09-29 | 长安大学 | Mining subsidence area detection method based on Mask R-CNN model |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874889A (en) * | 2017-03-14 | 2017-06-20 | 西安电子科技大学 | Multiple features fusion SAR target discrimination methods based on convolutional neural networks |
| CN108304873A (en) * | 2018-01-30 | 2018-07-20 | 深圳市国脉畅行科技股份有限公司 | Object detection method based on high-resolution optical satellite remote-sensing image and its system |
| CN108801211A (en) * | 2018-04-08 | 2018-11-13 | 山东农业大学 | A kind of sinking land in coalmining areas INSAR image method for quickly identifying |
| CN109063750A (en) * | 2018-07-17 | 2018-12-21 | 西安电子科技大学 | SAR target classification method based on CNN and SVM decision fusion |
| CN109558859A (en) * | 2018-12-13 | 2019-04-02 | 中国地质大学（武汉） | Mining area distributed intelligence extracting method and system based on DInSAR and DCNN |
| JP2019185179A (en) * | 2018-04-03 | 2019-10-24 | 株式会社東芝 | Target recognition device, target recognition method, and program |
| CN110532953A (en) * | 2019-08-30 | 2019-12-03 | 南京大学 | SAR image glacier recognition methods based on textural characteristics auxiliary |
| CN111626092A (en) * | 2020-03-26 | 2020-09-04 | 陕西陕北矿业韩家湾煤炭有限公司 | Unmanned aerial vehicle image ground crack identification and extraction method based on machine learning |
| US10976429B1 (en) * | 2017-10-16 | 2021-04-13 | Hrl Laboratories, Llc | System and method for synthetic aperture radar target recognition utilizing spiking neuromorphic networks |
| CN113408538A (en) * | 2021-07-12 | 2021-09-17 | 安徽耀峰雷达科技有限公司 | SVM-based radar RD image weak target detection method and system, storage medium and electronic terminal |
-
2022
- 2022-02-07 CN CN202210115680.6A patent/CN114660598A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874889A (en) * | 2017-03-14 | 2017-06-20 | 西安电子科技大学 | Multiple features fusion SAR target discrimination methods based on convolutional neural networks |
| US10976429B1 (en) * | 2017-10-16 | 2021-04-13 | Hrl Laboratories, Llc | System and method for synthetic aperture radar target recognition utilizing spiking neuromorphic networks |
| CN108304873A (en) * | 2018-01-30 | 2018-07-20 | 深圳市国脉畅行科技股份有限公司 | Object detection method based on high-resolution optical satellite remote-sensing image and its system |
| JP2019185179A (en) * | 2018-04-03 | 2019-10-24 | 株式会社東芝 | Target recognition device, target recognition method, and program |
| CN108801211A (en) * | 2018-04-08 | 2018-11-13 | 山东农业大学 | A kind of sinking land in coalmining areas INSAR image method for quickly identifying |
| CN109063750A (en) * | 2018-07-17 | 2018-12-21 | 西安电子科技大学 | SAR target classification method based on CNN and SVM decision fusion |
| CN109558859A (en) * | 2018-12-13 | 2019-04-02 | 中国地质大学（武汉） | Mining area distributed intelligence extracting method and system based on DInSAR and DCNN |
| CN110532953A (en) * | 2019-08-30 | 2019-12-03 | 南京大学 | SAR image glacier recognition methods based on textural characteristics auxiliary |
| CN111626092A (en) * | 2020-03-26 | 2020-09-04 | 陕西陕北矿业韩家湾煤炭有限公司 | Unmanned aerial vehicle image ground crack identification and extraction method based on machine learning |
| CN113408538A (en) * | 2021-07-12 | 2021-09-17 | 安徽耀峰雷达科技有限公司 | SVM-based radar RD image weak target detection method and system, storage medium and electronic terminal |
Non-Patent Citations (5)
| Title |
|---|
| 刘志宏;李玉峰;: "基于特征融合卷积神经网络的SAR图像目标检测方法", 微处理机, vol. 41, no. 2, 15 April 2020 (2020-04-15) * |
| 徐苗: "基于人工鱼群算法的SVM参数优化", 山西电子技术, no. 01, 15 February 2019 (2019-02-15), pages 30 - 33 * |
| 徐苗: "基于人工鱼群算法的SVM参数优化", 山西电子技术, no. 1, 31 December 2019 (2019-12-31), pages 30 - 33 * |
| 王志勇 等: "基于HOG特征的InSAR矿区开采沉陷盆地检测方法", 中国矿业大学学报, vol. 50, no. 2, 15 March 2021 (2021-03-15), pages 404 - 410 * |
| 郑重;张敬东;杜建华;: "基于深度学习的遥感图像中地面塌陷识别方法研究", 现代商贸工业, no. 35, 15 December 2017 (2017-12-15) * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819521A (en) * | 2023-02-22 | 2023-09-29 | 长安大学 | Mining subsidence area detection method based on Mask R-CNN model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110442666B (en) | Mineral resource prediction method and system based on neural network model | |
| CN112434643A (en) | Classification and identification method for low-slow small targets | |
| Wang et al. | Distilling knowledge from an ensemble of convolutional neural networks for seismic fault detection | |
| CN113516012A (en) | Pedestrian re-identification method and system based on multi-level feature fusion | |
| CN109581339B (en) | Sonar identification method based on automatic adjustment self-coding network of brainstorming storm | |
| CN105787501A (en) | Vegetation classification method capable of automatically selecting features in power transmission line corridor area | |
| Cui et al. | Deep learning model for seabed sediment classification based on fuzzy ranking feature optimization | |
| US20230134531A1 (en) | Method and system for rapid retrieval of target images based on artificial intelligence | |
| Xu et al. | Deep learning of rock microscopic images for intelligent lithology identification: Neural network comparison and selection | |
| CN116665067B (en) | Ore finding target area optimization system and method based on graph neural network | |
| CN112949189A (en) | Modeling method for multi-factor induced landslide prediction based on deep learning | |
| Bhagwat et al. | A novel multilabel classification of remote sensing images using XGBoost | |
| Ma et al. | Rock thin sections identification based on improved squeeze-and-Excitation Networks model | |
| Ronaldo et al. | Effective Soil type classification using convolutional neural network | |
| CN114660598A (en) | InSAR and CNN-AFSA-SVM fused mining subsidence basin automatic detection method | |
| CN114722926A (en) | Convolution clustering method for scale vortex time sequence diagram in towed sensor array | |
| CN105354585A (en) | Improved cat swarm algorithm based target extraction and classification method | |
| Zalatan et al. | Recurrent graph convolutional networks for spatiotemporal prediction of snow accumulation using airborne radar | |
| Angelin | A roc curve based k-means clustering for outlier detection using dragon fly optimization | |
| CN113869136A (en) | Semi-supervised polarimetric SAR image classification method based on multi-branch network | |
| CN109887005B (en) | TLD target tracking method based on visual attention mechanism | |
| Liu et al. | Peaks Fusion assisted Early-stopping Strategy for Overhead Imagery Segmentation with Noisy Labels | |
| Du et al. | Enhancement of land-use change modeling using convolutional neural networks and convolutional denoising autoencoders | |
| CN114119582B (en) | Synthetic aperture radar image target detection method | |
| CN113033484B (en) | Urban classification method for unmanned aerial vehicle emergency network deployment |
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 | ||
| CB02 | Change of applicant information |
Address after: 232000 Tian Jia An District, Huainan City, Anhui Applicant after: Anhui University of Science and Technology Address before: 232000 tianjia'an District, Hefei City, Anhui Province Applicant before: Anhui University of Science and Technology |
|
| CB02 | Change of applicant information |