CN112149502A - Unfavorable geology positioning forecasting method based on convolutional neural network - Google Patents
Unfavorable geology positioning forecasting method based on convolutional neural network Download PDFInfo
- Publication number
- CN112149502A CN112149502A CN202010838249.5A CN202010838249A CN112149502A CN 112149502 A CN112149502 A CN 112149502A CN 202010838249 A CN202010838249 A CN 202010838249A CN 112149502 A CN112149502 A CN 112149502A
- Authority
- CN
- China
- Prior art keywords
- positioning
- prediction
- layer
- neural network
- target
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 22
- 238000013277 forecasting method Methods 0.000 title claims abstract description 7
- 238000000034 method Methods 0.000 claims abstract description 75
- 238000012549 training Methods 0.000 claims abstract description 37
- 238000003062 neural network model Methods 0.000 claims abstract description 34
- 238000000605 extraction Methods 0.000 claims abstract description 30
- 238000013461 design Methods 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 41
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 22
- 238000001514 detection method Methods 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 11
- 230000008878 coupling Effects 0.000 claims description 9
- 238000010168 coupling process Methods 0.000 claims description 9
- 238000005859 coupling reaction Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 7
- 238000013434 data augmentation Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 230000001052 transient effect Effects 0.000 claims description 4
- 238000012512 characterization method Methods 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000013526 transfer learning Methods 0.000 claims description 3
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 238000003786 synthesis reaction Methods 0.000 claims description 2
- 230000003416 augmentation Effects 0.000 claims 1
- 238000010276 construction Methods 0.000 abstract description 8
- 230000008569 process Effects 0.000 abstract description 8
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 238000009430 construction management Methods 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000002547 anomalous effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008014 freezing Effects 0.000 description 1
- 238000007710 freezing Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Astronomy & Astrophysics (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a bad geological positioning forecasting method based on a convolutional neural network, which comprises the following steps: firstly, establishing a geophysical prospecting image data set; then constructing a target prediction positioning neural network model based on feature extraction, and training the target prediction positioning neural network model by adopting the image data set to obtain a trained target prediction positioning neural network model; and finally, inputting image result data obtained by a certain geophysical prospecting method into the trained target prediction positioning neural network model to perform actual unfavorable geological positioning prediction. The invention has the beneficial effects that: the method can accurately forecast the position scale and the property state of the bad geological body in the passing range in the underground engineering construction process of tunnels and the like, provides decision basis for the engineering design and construction management part, reduces the problems of low interpretability, dependence on expert experience and low prediction accuracy of the existing geological forecast geophysical prospecting method, and improves the safety of engineering construction.
Description
Technical Field
The invention relates to the technical field of unfavorable geological positioning and forecasting, in particular to an unfavorable geological positioning and forecasting method based on a convolutional neural network.
Background
In part of practical engineering projects, the problem that the unfavorable geological condition of a water-rich fractured zone brings great risks to tunnel construction is found, tunnel geological advanced prediction is mostly carried out on the basis of a geophysical prospecting method at present, however, in the existing advanced geological prediction, a waveform data image obtained by detecting through a geophysical prospecting sensor needs to be explained by an experienced expert, and the problems that the measured result image obtained by the geophysical prospecting sensor is low in interpretability, dependent on the experience of the expert and low in prediction accuracy exist in the prediction process, so that the method is very critical for carrying out accurate prediction and positioning on a certain specific unfavorable geology (such as the water-rich fractured zone) in the advanced prediction image.
With the rapid development of Deep Learning (Deep Learning), Convolutional Neural Networks (CNN) have been widely used in the fields of image classification, object detection, and image understanding due to the characteristics of feature sampling, weight sharing, and operation dimension reduction. The method is based on the deep convolutional neural network to automatically identify and classify the geophysical prospecting method (geological radar, TSP, transient electromagnetism and the like) detection images in a probability mode, a prediction network model system with good generalization performance is researched and developed, and the interpretability and the accuracy of poor geology of a water-rich broken zone in advance geological prediction are improved.
Disclosure of Invention
In order to solve the problems, the invention provides a bad geological positioning forecasting method based on a convolutional neural network, which mainly comprises the following steps:
the application illustrates an example of advance forecasting of unfavorable geology, namely a water-rich fractured zone, and the method can be suitable for geophysical prospecting method image data and can be used for accurately forecasting the water-rich fractured zone or other unfavorable geology.
A bad geological positioning forecasting method based on a convolutional neural network comprises the following steps:
s101: establishing a geophysical prospecting image data set;
s102: constructing a target prediction positioning neural network model based on feature extraction;
s103: training the target prediction positioning neural network model by adopting the image data set to obtain a trained target prediction positioning neural network model;
s104: and inputting image result data obtained by a certain geophysical prospecting method into the trained target prediction positioning neural network model to perform actual unfavorable geological positioning prediction.
Further, in step S101, a geophysical prospecting image dataset is established; the method specifically comprises the following steps:
s201: establishing a geophysical prospecting method detection result preliminary image data set through advanced geological forecast case collection and actual field project data acquisition; the preliminary image data set comprises a plurality of result images obtained by performing advanced geological forecast based on a geophysical prospecting method;
s202: forecasting abnormal geology on each image in the preliminary image data set according to a water-rich zone broken zone forecasting mechanism and the imaging characteristics of the water-rich zone broken zone forecasting mechanism;
s203: determining the position of a water-rich broken band on each image in the preliminary image data set again by adopting an expert experience method in combination with a forecast conclusion, and performing data annotation on each image in the preliminary image data set by taking a water-rich broken band positioning area as label content to obtain an annotated preliminary image data set;
s204: and adopting various image data augmentation methods to expand and augment the marked preliminary image data set, and detecting the negative sample data of the water-rich fractured zone by coupling a partial geophysical prospecting method to obtain a final image data set.
Further, in step S201, the geophysical prospecting method includes TSP, geological radar, transient electromagnetism, and the like; the preliminary image dataset has a sample capacity greater than 500 images.
Further, in step S102, the target prediction positioning neural network model based on feature extraction includes a feature extraction basic network and a positioning prediction network which are sequentially connected;
wherein, the characteristic extraction basic network comprises the following connected in sequence: CBR block, max pooling layer, BaseRN1 layer, 2 BaseRN0 layers, BaseRN1 layers, 3 BaseRN0 layers, BaseRN1 layers, 5 BaseRN0 layers;
the positioning prediction network comprises a maximum pooling layer, a full connection layer and a Sigmoid layer which are sequentially connected.
Further, the CBR block comprises a volume base layer, a regularization layer and an activation layer which are connected in sequence; wherein C represents a convolution layer, B represents a regularization layer Batch Norm, R represents an activation layer, and a Leaky-ReLU activation function is adopted.
Furthermore, the BaseRN1 layer and the BaseRN0 layer are designed based on a residual error network concept, and a residual error item is introduced, so that a depth network model is conveniently constructed; wherein, the trunks in the BaseRN0 layer and the BaseRN1 layer are both CBR-CBR-CB modules, and the parameters of the trunks CBR-CBR-CB modules in the BaseRN0 layer and the BaseRN1 layer are consistent, and the concrete steps are as follows: the filter size of the first CBR block is 1 × 1, 64 channels; the filter size of the second CBR block is 3 × 3, 64 channels; the filter size of the third CB block is 1 × 1, 256 channels; the filter size of the branched CB block in the BaseRN0 layer is consistent with the CB block filter size of the trunk, and the filter size of the branched CBR block in the BaseRN1 is consistent with the filter size of the first CBR block of the trunk; the CB block comprises a volume base layer and a regularization layer which are connected in sequence.
Further, in step S103, in the feature extraction base network, the output feature mapping of the feature extraction layer is down-sampled by a factor of 16 times, a tradeoff is made between spatial resolution and extracted feature strength, and the final output of the active layer leak-ReLU is 14 × 14 × 1024; the feature extraction network performs 2-time down-sampling on the input image in four stages to realize 16-time down-sampling: the method comprises the following specific steps:
the first stage is as follows: inputting a three-channel RGB image with the size of W multiplied by H multiplied by C, and inputting the three-channel RGB image into a CBR block at a first stage; w-256 is the width of the image, H-256 is the height of the image, and C-3 represents three channels of the image; the filter size in the CBR block is 1 multiplied by 1, 64 channels are sampled by the CBR block for one time, and the size of the output image is 128 multiplied by 64;
and a second stage: first pass through a maximum pooling layer of 3 × 3 and stride 2 with a filter, then pass through a BaseRN1 layer and 2 BaseRN0 layers; the size of the image output after the second stage is 64 × 64 × 256;
and a third stage: inputting the output of the second stage to 1 BaseRN1 layer and 3 BaseRN0 layers which are connected in sequence, obtaining a deeper network model structure through the combination of the BaseRN0 layer and the BaseRN1 layer, and enabling the size of an output graph to be 32 multiplied by 512;
a fourth stage: the output of the third stage is input to 1 BaseRN1 layer and 5 BaseRN0 layers connected in sequence, a further deeper network model structure is obtained by the combination of BaseRN0 and BaseRN1, and the size of the output image is 16 × 16 × 1024.
Furthermore, in the positioning prediction network, the output result of the feature extraction network is input into the positioning prediction network, the offset influence of the feature estimation mean value caused by the model parameter error is effectively reduced through the maximum pooling layer, finally, the full-connection layer and the Sigmoid layer are connected for regression positioning, and the simultaneous detection of the position and classification is realized through the design of a multi-task loss function.
Further, in step S103, during training, in order to realize accurate location prediction of a certain unfavorable geology, a loss function of multitask coupling is designed, and at the end of the location prediction network, probability confidence judgment and location area data output of the unfavorable geology are realized through a sigmoid regression method; specifically, the method comprises the following steps:
the design of the loss function aims at two tasks of judging the prediction confidence coefficient of the unfavorable geological target and positioning the target, and the target loss function is respectively and correspondingly constructed, wherein:
target prediction confidence: target prediction confidence loss function Lconf(o, p) representing the probability that the predicted target region is the unfavorable geology to be monitored, and adopting binary cross entropy loss characterization:
wherein o isiE {0,1} represents whether a target exists in the ith prediction box, 0 represents absence and 1 represents existence;
representing the probability of whether the target exists in the ith prediction box, and locating the result p output by the neural network model through predicting the targetiThe sigmoid is evaluated to obtain the result,
belongs to (0, 1); i is 1,2, …, and I is the number of predicted result frames; l isconfThe value of (o, p) is influenced by the number and value of the prediction frames, the range is not fixed, LconfThe smaller the (o, p), the smaller the loss, the more accurate the prediction result;
target positioning: target location loss function Lloc(mu, sigma), in order to improve the result positioning accuracy, Gaussian distribution modeling is introduced to the prediction bounding box result (x, y, w, h) of the positioning box to obtain the mean value and the variance of each parameter; the bounding box is a rectangular box, (x, y) represents the center position of the box, and (w, h) represents the size of the box;
the target localization loss function is characterized by negative log likelihood loss (NLL):
wherein,
w multiplied by H is the width and the height of an input image, and the input image is divided into H rows and W columns of grids; i is the number of the prediction frames, and the ith prediction frame and the base coordinate of the jth row and the jth column of the kth line are obtained by indexing I, j and k; gamma raytThe hyper-parameter is a hyper-parameter influencing the weight of the (x, y, w, h) prediction result, is a hyper-parameter of model training, gradually approaches a stable value along with the training, and can effectively ensure the accuracy of the training result; gijkRepresenting the real value { x, y, W, H } of the result of the ith prediction box at the position of the jth row pixel point on the kth line under the H multiplied by W coordinate systemtruthA positive distribution G to N (mu, sigma) satisfying a mean value of mu and a variance of sigma2)；μt,σtT ∈ { x, y, w, h } is a model prediction result, is converted into a range of (0,1) by a sigmoid function, and represents a center coordinate μ in a prediction box indexed by t ═ i, j, k at presentw,μhIs expressed in μx,μyThe length and width of a rectangular frame of the prediction region which is a central point; using the square error amount sigma simultaneouslytThe t belongs to { x, y, w, h } representation prediction result reliability, 0 represents reliable, and 1 represents unreliable;
synthesis Lconf(o, p) and Lloc(mu, sigma) to obtain a multitask loss function by a weighting factor lambda1,λ2And coupling the two loss functions to obtain a comprehensive bad geology positioning forecast loss function L (o, p, mu, sigma):
L(o,p,μ,σ)＝λ1Lconf(o,p)+λ2Lloc(μ,σ)
wherein λ is1,λ2These two weighting factors are set according to the specific task requirements.
Further, the image dataset constructed in S101 belongs to a small sample dataset; in step S103, the target prediction positioning neural network model is trained by a transfer learning method to increase the training speed.
The technical scheme provided by the invention has the beneficial effects that: the technical scheme provided by the application can accurately forecast the position scale and the property state of the bad geological body in the passing range in the underground engineering construction process such as a tunnel and the like, provides decision basis for the engineering design and construction management part, reduces the problems of low interpretability, dependence on expert experience and low prediction accuracy of the existing geological forecast geophysical prospecting method, and improves the safety of engineering construction.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flow chart of a method for forecasting unfavorable geological positioning based on a convolutional neural network in an embodiment of the present invention;
fig. 2 is a block diagram of an object prediction positioning neural network in an embodiment of the present invention.
Detailed Description
For a more clear understanding of the technical features, objects and effects of the present invention, embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
The embodiment of the invention provides a method for forecasting unfavorable geological positioning based on a convolutional neural network.
Referring to fig. 1, fig. 1 is a flowchart of a method for forecasting unfavorable geology based on a convolutional neural network in an embodiment of the present invention, which specifically includes the following steps:
s101: establishing a geophysical prospecting image data set;
s102: constructing a target prediction positioning neural network model based on feature extraction;
s103: training the target prediction positioning neural network model by adopting the image data set to obtain a trained target prediction positioning neural network model;
s104: and inputting image result data obtained by a certain geophysical prospecting method into the trained target prediction positioning neural network model to perform actual unfavorable geological positioning prediction.
In order to solve the problems of over-training fitting and low prediction precision caused by small sample amount of detection data in advance geological prediction by a geophysical prospecting method, the capacity expansion of a data set sample is realized based on an image data augmentation method.
In the step S101, a geophysical prospecting image data set is established; the method specifically comprises the following steps:
s201: establishing a geophysical prospecting method detection result preliminary image data set through advanced geological forecast case collection and actual field project data acquisition; the preliminary image data set comprises a plurality of result images obtained by advanced geological forecast based on a geophysical prospecting method (TSP, geological radar, transient electromagnetism and the like), and a sample set is obtained after collection and sorting and is original data to be processed; the sample capacity in the preliminary image dataset is greater than 500 images;
s202: forecasting abnormal geology on each image in the preliminary image data set according to a water-rich fractured zone forecasting mechanism and the imaging characteristics of the water-rich fractured zone forecasting mechanism; and guiding the network structure design of the feature extraction model;
s203: determining the position of a water-rich broken band on each image in the preliminary image data set again by adopting an expert experience method in combination with a forecast conclusion, and performing data annotation on each image in the preliminary image data set by taking a water-rich broken band positioning area as label content to obtain an annotated preliminary image data set;
s204: adopting a plurality of image data augmentation methods to expand and augment the marked preliminary image data set, and coupling a partial geophysical prospecting method to detect the negative sample data of the water-rich fractured zone (namely the detection result indicates that the data do not contain the water-rich fractured zone) to obtain a final image data set; the image data augmentation method includes: image mirroring, translation, scaling, rotation, clipping, gaussian noise, and the like;
wherein the image data in the final image dataset is greater than 1000; according to the following steps of 8: 2, dividing the image data into a training data set and a testing data set to complete the construction of the image data set.
Aiming at the problem of high difficulty in identifying and classifying adverse geological types in a tunnel engineering complex geological environment, a neural network model with a target prediction positioning function is designed based on a deep learning feature extraction method, and whether a water-rich fractured zone positioning area exists in an image or not is judged.
According to the graphic and image mechanism of various unfavorable geologies on the detection data, a proper target prediction positioning network model structure is selected and designed, based on a general principle, a deeper network can learn a more complex function and an input representation form, so that the model performance is better, and the method selectively introduces a ResNet residual error network concept design basic module and designs an ultra-deep convolution network model structure.
Referring to fig. 2, fig. 2 is a structural diagram of a target prediction positioning neural network according to an embodiment of the present invention; in step S102, the target prediction positioning neural network model based on feature extraction includes a feature extraction basic network and a positioning prediction network that are connected in sequence;
wherein, the characteristic extraction basic network comprises the following connected in sequence: CBR block, max pooling layer, BaseRN1 layer, 2 BaseRN0 layers, BaseRN1 layers, 3 BaseRN0 layers, BaseRN1 layers, 5 BaseRN0 layers;
the CBR block comprises a volume base layer, a regularization layer and an activation layer which are connected in sequence; wherein C represents a convolution layer, B represents a regularization layer Batch Norm, and R represents an activation layer; in the embodiment of the invention, a Leaky-ReLU activation function is adopted; the problem that neurons in a negative region are silenced in a traditional ReLU activation function is solved;
the BaseRN1 layer and the BaseRN0 layer are designed based on a residual error network concept, and a residual error item is introduced, so that a depth network model is conveniently constructed; wherein, the trunks in the BaseRN0 layer and the BaseRN1 layer are both CBR-CBR-CB modules, and the parameters of the trunks CBR-CBR-CB modules in the BaseRN0 layer and the BaseRN1 layer are consistent, and the concrete steps are as follows: the filter size of the first CBR block is 1 × 1, 64 channels; the filter size of the second CBR block is 3 × 3, 64 channels; the filter size of the third CB block is 1 × 1, 256 channels; the filter size of the branched CB block in the BaseRN0 layer is consistent with the CB block filter size of the trunk, and the filter size of the branched CBR block in the BaseRN1 is consistent with the filter size of the first CBR block of the trunk; the CB block comprises a volume base layer and a regularization layer which are connected in sequence;
the positioning prediction network comprises a maximum pooling layer, a full connection layer and a Sigmoid layer which are sequentially connected.
In step S103, in the feature extraction basic network, the output feature mapping of the feature extraction layer is subjected to four-level down-sampling by a factor of 16 times, a tradeoff is made between spatial resolution and extracted feature strength, and the final output of the active layer leak-ReLU is 14 × 14 × 1024; the feature extraction network performs 2-time down-sampling on the input image in four stages to realize 16-time down-sampling: the method comprises the following specific steps:
the first stage is as follows: inputting a three-channel RGB image with dimension of W × H × C, inputting into a CBR block at a first stage, as shown in FIG. 2; w-256 is the width of the image, H-256 is the height of the image, and C-3 represents three channels of the image; the filter size in the CBR block is 1 multiplied by 1, 64 channels are sampled by the CBR block for one time, and the size of the output image is 128 multiplied by 64;
and a second stage: first pass through a maximum pooling layer of 3 × 3 and stride 2 with a filter, then pass through a BaseRN1 layer and 2 BaseRN0 layers; the size of the image output after the second stage is 64 × 64 × 256;
the BaseRN0 layer and the BaseRN1 layer are designed based on the idea of a residual error network and are improved and upgraded versions of a ResNet residual error module; the layers of the BaseRN0 and the BaseRN1 are connected with a leave-ReLU activation function, and the method is a general processing mode;
and a third stage: inputting the output of the second stage to 1 BaseRN1 layer and 3 BaseRN0 layers which are connected in sequence, obtaining a deeper network model structure through the combination of the BaseRN0 layer and the BaseRN1 layer, and enabling the size of an output graph to be 32 multiplied by 512;
a fourth stage: inputting the output of the third stage to 1 BaseRN1 layer and 5 BaseRN0 layers which are connected in sequence, further obtaining a deeper network model structure through different combinations of BaseRN0 and BaseRN1, and enabling the size of an output image to be 16 multiplied by 1024;
in the third stage and the fourth stage, the model depth may also be increased by adding or reducing the BaseRNO layer and the BaseRN1 layer, as required.
In the positioning prediction network, the output result of the feature extraction network is input into the positioning prediction network, the offset influence of a feature estimation mean value caused by model parameter errors is effectively reduced through a maximum pooling layer, finally, a full-connection layer and a Sigmoid layer are connected for regression positioning, and the simultaneous detection of prediction and positioning is realized through the design of a multi-task loss function; predicting whether unfavorable geology to be detected exists or not; positioning is to position the area range of the unfavorable geology if the unfavorable geology exists;
in the step S103, during training, in order to realize accurate positioning forecast of a certain unfavorable geology, a multitask coupling loss function is designed, and finally probability confidence judgment and positioning area data output of the unfavorable geology are realized through a sigmoid regression method in a positioning prediction network; specifically, the method comprises the following steps:
the design of the loss function aims at two tasks of judging the prediction confidence coefficient of the unfavorable geological target and positioning the target, and the target loss function is respectively and correspondingly constructed, wherein:
target prediction confidence: target prediction confidence loss function Lconf(o, p) representing the probability that the predicted target region is the unfavorable geology to be monitored, and adopting binary cross entropy loss characterization:
wherein o isiE {0,1} represents whether a target exists in the ith prediction box, 0 represents absence and 1 represents existence;
representing the probability of whether the target exists in the ith prediction box, and locating the result p output by the neural network model through predicting the targetiThe sigmoid is evaluated to obtain the result,
belongs to (0, 1); i is 1,2, …, and I is the number of prediction frames; l isconfThe result of the predicted value of (o, p) is boxedInfluence of the sum of the numbers, extent of uncertainty, LconfThe smaller the (o, p), the smaller the loss, the more accurate the prediction result;
target positioning: target location loss function Lloc(mu, sigma), in order to improve the result positioning accuracy, Gaussian distribution modeling is introduced to the prediction bounding box result (x, y, w, h) of the positioning box to obtain the mean value of each parameter
And variance
The prediction precision is improved by training the mean value and the variance to obtain the result with the minimum loss, and the result is used as the representation of a positioning area in the output feature tensor; the bounding box is a rectangular box, (x, y) represents the center position of the box, and (w, h) represents the size of the box;
the target localization loss function is characterized by negative log likelihood loss (NLL):
wherein,
w multiplied by H is the width and the height of an input image, and the input image is divided into H rows and W columns of grids; i is the number of prediction blocks,
because the result of each prediction is a plurality of target prediction frames, a column is represented by j e [1, W ], k e [1, H ] represents a row index to a corresponding grid, I e [1, I ] represents an ith prediction frame region predicted by the grid, the ith prediction frame and the base coordinate of the jth column of the kth row are obtained by I, j and k indexes, and a plurality of scale prediction frames are arranged on each grid to carry out result approximation;
γtis pair (x)Y, w, h) the hyperparameter of the weight influence of the prediction result is a hyperparameter of model training, and gradually approaches to a stable value along with the training, so that the accuracy of the training result can be effectively guaranteed; gijkRepresenting the real value { x, y, W, H } of the result of the ith prediction box at the position of the jth row pixel point on the kth line under the H multiplied by W coordinate systemtruthA positive distribution G to N (mu, sigma) satisfying a mean value of mu and a variance of sigma2)；μt、σtAnd t ∈ { x, y, w, h } is the model prediction result, μx,μyConverted into the range of (0,1) by sigmoid function, representing the center coordinate, μ, within the prediction box indexed by current t-kjiw,μhIs expressed in μx,μyThe length and width of the prediction frame as the central point;
using the square error amount sigma simultaneouslytAnd t belongs to { x, y, w, h } representation prediction result reliability, 0 represents reliable, 1 represents unreliable, and the solving process is as follows:
comprehensively obtaining a multi-task loss function of the probability of the confidence degree of target positioning and target prediction, and obtaining the probability of the confidence degree of the target positioning and the target prediction through a weight factor lambda1,λ2And coupling the two loss functions to obtain a comprehensive bad geology positioning forecast loss function L (o, p, mu, sigma):
L(o,p,μ,σ)＝λ1Lconf(o,p)+λ2Lloc(μ,σ)
wherein λ is1,λ2The two weight factors are set according to specific task requirements, and if the target is in an abnormal geological condition and the probability accuracy of the target is determined preferentially in some scenes, then lambda is1The weight is lifted; some task scenarios focus on the location prediction accuracy of the anomalous geology, then lambda2The weight is lifted.
The image dataset constructed in S101 belongs to a small sample dataset; in step S103, the training speed is increased by a transfer learning method; the method comprises the following specific steps:
s301: firstly, training an original target prediction positioning neural network model in advance based on a large sample data set A; obtaining a preliminarily trained target prediction positioning neural network model A; the large sample dataset may use ImageNet or the like;
s302: freezing the first t-layer parameters of the model A, and performing migration training on the model A by taking a training data set in the image data set as a small sample data set B; obtaining a trained target prediction positioning neural network model B; wherein t is (1, L) which is a value preset empirically;
s303: testing the trained target prediction positioning neural network model by adopting a test data set in the image data set; and judging whether the test meets the requirements; if so, taking the current target prediction positioning neural network model as a trained target prediction positioning neural network model; otherwise, retraining the parameters from the last t layer to the L layer of the model B by adopting the training data set, and performing feedback training according to the training data set labels to obtain all the parameters of the model B; meanwhile, model parameters can be finely adjusted to optimize results until the trained target prediction positioning neural network model B can meet task requirements.
Fine-tune is that all parameters of the original target prediction positioning neural network model are trained on a large sample ImageNet data set, and then the first t layer parameters of A are frozen, and the first t layer parameters do not need to be trained again; then, training the t-L layer parameters of the A by using the constructed training data set of the small sample to obtain the t-L layer parameters; and (3) judging whether the parameters are well trained and whether the model can meet the task requirements or not according to the result of test evaluation (the evaluation threshold value can be specifically set according to the precision of the task requirements), repeating the steps for training if the parameters do not meet the requirements, and simultaneously adjusting parameters such as the frozen layer number and the like to optimize the result.
A training data set and a test data set are needed in the training process, the parameters are continuously updated by repeating the epoch for N times in the training process until the result is optimal, in each training process, model parameters are obtained by training through the training data set, then the group of parameters are evaluated through the test data set, and then the next epoch training is carried out until the set N times of epoch training are finished or the evaluation result meets the preset requirement
In practical application, image result data obtained by a certain geophysical prospecting method is input into the trained target prediction positioning neural network model, and whether the probability confidence of unfavorable geology to be predicted and the area where the probability confidence exists or not are automatically output, so that the dependence on the experience of professionals and experts is reduced, and the working efficiency is improved.
The beneficial effects of the invention include:
1) a geological advanced prediction data set is constructed, and the data set is the basis of deep learning and network model training;
2) based on a residual error network thought, a deep layer feature extraction network model with ultrahigh depth is designed, and the hidden and abstract information in the geological advanced prediction image can be activated and detected through the depth of the model, so that the extraction of corresponding unfavorable geological features is completed;
3) after the feature extraction is finished, the probability confidence coefficient judgment and the region positioning of a certain unfavorable geological type are realized by designing a multi-task loss function;
4) the method can be continuously upgraded and evolved along with continuous collection and expansion of the image data set, and performance, detection precision, detection range and the like are improved.
The technical scheme provided by the application can accurately forecast the position scale and the property state of the bad geological body in the passing range in the underground engineering construction process such as a tunnel and the like, provides decision basis for the engineering design and construction management part, reduces the problems of low interpretability, dependence on expert experience and low prediction accuracy of the existing geological forecast geophysical prospecting method, and improves the safety of engineering construction.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A bad geological positioning forecasting method based on a convolutional neural network is characterized by comprising the following steps: the method comprises the following steps:
s101: establishing a geophysical prospecting image data set;
s102: constructing a target prediction positioning neural network model based on feature extraction;
s103: training the target prediction positioning neural network model by adopting the image data set to obtain a trained target prediction positioning neural network model;
s104: and inputting image result data obtained by a geophysical prospecting method into the trained target prediction positioning neural network model to perform actual unfavorable geological positioning prediction.
2. The method for forecasting unfavorable geologic positioning based on convolutional neural network as defined in claim 1, wherein: in the step S101, a geophysical prospecting image data set is established; the method specifically comprises the following steps:
s201: establishing a geophysical prospecting method detection result preliminary image data set through advanced geological forecast case collection or actual field project data acquisition; the preliminary image data set comprises a plurality of result images obtained by performing advanced geological forecast based on a geophysical prospecting method;
s202: forecasting abnormal geology on each image in the preliminary image data set according to a water-rich fractured zone forecasting mechanism and the imaging characteristics of the water-rich fractured zone forecasting mechanism;
s203: determining the position of a water-rich broken band on each image in the preliminary image data set again by adopting an expert experience method in combination with a forecast conclusion, and performing data annotation on each image in the preliminary image data set by taking a water-rich broken band positioning area as label content to obtain an annotated preliminary image data set;
s204: and adopting an image data augmentation method to perform capacity expansion augmentation on the marked preliminary image data set, and coupling a partial geophysical prospecting method to detect the sample data of the water-rich broken belt to obtain a final image data set.
3. The method for forecasting the unfavorable geologic localization of a convolutional neural network as claimed in claim 2, wherein: in step S201, the geophysical prospecting method includes TSP, geological radar, and transient electromagnetism; the preliminary image dataset has a sample capacity greater than 500 images.
4. The method for forecasting unfavorable geologic positioning based on convolutional neural network as defined in claim 1, wherein: in step S102, the target prediction positioning neural network model based on feature extraction includes a feature extraction basic network and a positioning prediction network that are connected in sequence;
wherein, the characteristic extraction basic network comprises the following connected in sequence: CBR block, max pooling layer, BaseRN1 layer, 2 BaseRN0 layers, BaseRN1 layers, 3 BaseRN0 layers, BaseRN1 layers, 5 BaseRN0 layers;
the positioning prediction network comprises a maximum pooling layer, a full connection layer and a Sigmoid layer which are sequentially connected.
5. The method for forecasting unfavorable geologic positioning based on convolutional neural network as defined in claim 4, wherein: the CBR block comprises a volume base layer, a regularization layer and an activation layer which are sequentially connected; wherein C represents a convolution layer, B represents a regularization layer Batch Norm, R represents an activation layer, and a Leaky-ReLU activation function is adopted.
6. The method for forecasting unfavorable geologic positioning based on a convolutional neural network as claimed in claim 5, wherein: the BaseRN1 layer and the BaseRN0 layer are designed based on a residual error network concept, and a residual error item is introduced, so that a depth network model is conveniently constructed; wherein, the trunks in the BaseRN0 layer and the BaseRN1 layer are both CBR-CBR-CB modules, and the parameters of the trunks CBR-CBR-CB modules in the BaseRN0 layer and the BaseRN1 layer are consistent, and the concrete steps are as follows: the filter size of the first CBR block is 1 × 1, 64 channels; the filter size of the second CBR block is 3 × 3, 64 channels; the filter size of the third CB block is 1 × 1, 256 channels; the filter size of the branched CB block in the BaseRN0 layer is consistent with the CB block filter size of the trunk, and the filter size of the branched CBR block in the BaseRN1 is consistent with the filter size of the first CBR block of the trunk; the CB block comprises a volume base layer and a regularization layer which are connected in sequence.
7. The method for forecasting unfavorable geologic positioning based on convolutional neural network as defined in claim 4, wherein: in step S103, in the feature extraction basic network, the output feature mapping of the feature extraction layer is subjected to four-level down-sampling by a factor of 16 times, a tradeoff is made between spatial resolution and extracted feature strength, and the final output of the active layer leak-ReLU is 14 × 14 × 1024; the feature extraction network performs 2-time down-sampling on the input image in four stages to realize 16-time down-sampling: the method comprises the following specific steps:
the first stage is as follows: inputting a three-channel RGB image with the size of W multiplied by H multiplied by C, and inputting the three-channel RGB image into a CBR block at a first stage; w-256 is the width of the image, H-256 is the height of the image, and C-3 represents three channels of the image; the filter size in the CBR block is 1 multiplied by 1, 64 channels are sampled by the CBR block for one time, and the size of the output image is 128 multiplied by 64;
and a second stage: first pass through a maximum pooling layer of 3 × 3 and stride 2 with a filter, then pass through a BaseRN1 layer and 2 BaseRN0 layers; the size of the image output after the second stage is 64 × 64 × 256;
and a third stage: inputting the output of the second stage to 1 BaseRN1 layer and 3 BaseRN0 layers which are connected in sequence, obtaining a deeper network model structure through the combination of the BaseRN0 layer and the BaseRN1 layer, and enabling the size of an output graph to be 32 multiplied by 512;
a fourth stage: the output of the third stage is input to 1 BaseRN1 layer and 5 BaseRN0 layers connected in sequence, a further deeper network model structure is obtained by the combination of BaseRN0 and BaseRN1, and the size of the output image is 16 × 16 × 1024.
8. The method for forecasting unfavorable geologic positioning based on convolutional neural network as defined in claim 4, wherein: in the positioning prediction network, the output result of the feature extraction network is input into the positioning prediction network, the offset influence of the feature estimation mean value caused by the parameter error of the model is effectively reduced through the maximum pooling layer, finally, the full-connection layer and the Sigmoid layer are connected for regression positioning, and the simultaneous detection of prediction and positioning is realized through the design of a multi-task loss function.
9. The method for forecasting unfavorable geologic positioning based on a convolutional neural network as claimed in claim 8, wherein: in the step S103, during training, in order to realize accurate positioning forecast of a certain unfavorable geology, a multitask coupling loss function is designed, and finally probability confidence judgment and positioning area data output of the unfavorable geology are realized through a sigmoid regression method in a positioning prediction network; specifically, the method comprises the following steps:
the design of the loss function aims at two tasks of judging the prediction confidence coefficient of the unfavorable geological target and positioning the target, and the target loss function is respectively and correspondingly constructed, wherein:
target prediction confidence: target prediction confidence loss function Lconf(o, p) representing the probability that the predicted target region is the unfavorable geology to be monitored, and adopting binary cross entropy loss characterization:
wherein o isiE {0,1} represents whether a target exists in the ith prediction box, 0 represents absence and 1 represents existence;
representing the probability of whether the target exists in the ith prediction box, and locating the result p output by the neural network model through predicting the targetiThe sigmoid is evaluated to obtain the result,
belongs to (0, 1); i is 1,2, …, and I is the number of prediction frames; l isconfThe value of (o, p) is influenced by the number and value of the prediction frames, the range is not fixed, LconfThe smaller the (o, p), the smaller the loss, the more accurate the prediction result;
target positioning: target location loss function Lloc(mu, sigma), in order to improve the result positioning accuracy, Gaussian distribution modeling is introduced to the prediction bounding box result (x, y, w, h) of the positioning box to obtain the mean value and the variance of each parameter; the bounding box is a rectangular box, (x, y) represents the center position of the box, and (w, h) represents the size of the box;
the target positioning loss function is characterized by adopting negative log likelihood loss:
wherein,
w multiplied by H is the width and the height of an input image, and the input image is divided into H rows and W columns of grids; i is the number of the prediction frames, and the ith prediction frame and the base coordinate of the jth row and the jth column of the kth line are obtained by indexing I, j and k; gamma raytIs a hyper-parameter affecting the (x, y, w, h) prediction result weight; gijkRepresenting the real value { x, y, W, H } of the result of the ith prediction box at the position of the jth row pixel point on the kth line under the H multiplied by W coordinate systemtruthA positive distribution G to N (mu, sigma) satisfying a mean value of mu and a variance of sigma2)；μt,σtAnd t ∈ { x, y, w, h } is the model prediction result, passing sigmThe oid function translates to a (0,1) range, representing the center coordinate, μ, within the current t ═ i, j, k indexed prediction boxw,μhIs expressed in μx,μyThe length and width of a rectangular frame of the prediction region which is a central point; using the square error amount sigma simultaneouslytThe t belongs to { x, y, w, h } representation prediction result reliability, 0 represents reliable, and 1 represents unreliable;
synthesis Lconf(o, p) and Lloc(mu, sigma) to obtain a multitask loss function by a weighting factor lambda1,λ2And coupling the two loss functions to obtain a comprehensive bad geology positioning forecast loss function L (o, p, mu, sigma):
L(o,p,μ,σ)＝λ1Lconf(o,p)+λ2Lloc(μ,σ)
wherein λ is1,λ2These two weighting factors are set according to the specific task requirements.
10. The method for forecasting unfavorable geologic positioning based on convolutional neural network as defined in claim 1, wherein: the image dataset constructed in S101 belongs to a small sample dataset; in step S103, the target prediction positioning neural network model is trained by a transfer learning method to increase the training speed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010838249.5A CN112149502A (en) | 2020-08-19 | 2020-08-19 | Unfavorable geology positioning forecasting method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010838249.5A CN112149502A (en) | 2020-08-19 | 2020-08-19 | Unfavorable geology positioning forecasting method based on convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112149502A true CN112149502A (en) | 2020-12-29 |
Family
ID=73887532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010838249.5A Pending CN112149502A (en) | 2020-08-19 | 2020-08-19 | Unfavorable geology positioning forecasting method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112149502A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113447536A (en) * | 2021-06-24 | 2021-09-28 | 山东大学 | Concrete dielectric constant inversion and disease identification method and system |
| CN113836999A (en) * | 2021-08-16 | 2021-12-24 | 山东大学 | Tunnel construction risk intelligent identification method and system based on ground penetrating radar |
| CN115346114A (en) * | 2022-07-21 | 2022-11-15 | 中铁二院工程集团有限责任公司 | Method and equipment for identifying and positioning bad geologic body by railway tunnel aviation electromagnetic method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109709603A (en) * | 2018-11-23 | 2019-05-03 | 中国石油天然气股份有限公司 | Seismic horizon identification and method for tracing, system |
| CN111310861A (en) * | 2020-03-27 | 2020-06-19 | 西安电子科技大学 | License plate recognition and positioning method based on deep neural network |
| CN111428625A (en) * | 2020-03-23 | 2020-07-17 | 武汉理工大学 | Traffic scene target detection method and system based on deep learning |
-
2020
- 2020-08-19 CN CN202010838249.5A patent/CN112149502A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109709603A (en) * | 2018-11-23 | 2019-05-03 | 中国石油天然气股份有限公司 | Seismic horizon identification and method for tracing, system |
| CN111428625A (en) * | 2020-03-23 | 2020-07-17 | 武汉理工大学 | Traffic scene target detection method and system based on deep learning |
| CN111310861A (en) * | 2020-03-27 | 2020-06-19 | 西安电子科技大学 | License plate recognition and positioning method based on deep neural network |
Non-Patent Citations (2)
| Title |
|---|
| 周忠义 等: "基于多路特征融合和深度学习的露霜图像分类", 《计算机应用与软件》 * |
| 戴舒等: "基于YOLO算法的行人检测方法", 《无线电通信技术》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113447536A (en) * | 2021-06-24 | 2021-09-28 | 山东大学 | Concrete dielectric constant inversion and disease identification method and system |
| CN113447536B (en) * | 2021-06-24 | 2022-09-30 | 山东大学 | Concrete dielectric constant inversion and disease identification method and system |
| CN113836999A (en) * | 2021-08-16 | 2021-12-24 | 山东大学 | Tunnel construction risk intelligent identification method and system based on ground penetrating radar |
| CN115346114A (en) * | 2022-07-21 | 2022-11-15 | 中铁二院工程集团有限责任公司 | Method and equipment for identifying and positioning bad geologic body by railway tunnel aviation electromagnetic method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111259930B (en) | General target detection method of self-adaptive attention guidance mechanism | |
| CN112130216B (en) | Geological advanced fine forecasting method based on convolutional neural network multi-geophysical prospecting method coupling | |
| CN112149502A (en) | Unfavorable geology positioning forecasting method based on convolutional neural network | |
| CN108447057B (en) | SAR image change detection method based on significance and depth convolution network | |
| CN111562612B (en) | Deep learning microseismic event identification method and system based on attention mechanism | |
| CN108171119B (en) | SAR image change detection method based on residual error network | |
| CN115758212A (en) | Mechanical equipment fault diagnosis method based on parallel network and transfer learning | |
| CN112800934A (en) | Behavior identification method and device for multi-class engineering vehicle | |
| Lechgar et al. | Detection of cities vehicle fleet using YOLO V2 and aerial images | |
| CN115393631A (en) | Hyperspectral image classification method based on Bayesian layer graph convolution neural network | |
| CN114997501A (en) | Deep learning mineral resource classification prediction method and system based on sample unbalance | |
| CN114973019A (en) | Deep learning-based geospatial information change detection classification method and system | |
| CN114357372A (en) | Aircraft fault diagnosis model generation method based on multi-sensor data driving | |
| CN111860591A (en) | Cervical cell image classification method based on interval adaptive feature selection fusion | |
| CN116664826A (en) | Small sample point cloud semantic segmentation method | |
| Xu et al. | Compressed YOLOv5 for oriented object detection with integrated network slimming and knowledge distillation | |
| CN115834161A (en) | Power grid false data injection attack detection method of artificial intelligence four-layer architecture | |
| Wan et al. | RSSM-Net: Remote sensing image scene classification based on multi-objective neural architecture search | |
| CN114358058A (en) | Wireless communication signal open set identification method and system based on deep neural network | |
| Yu et al. | A lightweight ship detection method in optical remote sensing image under cloud interference | |
| CN114140524A (en) | Closed loop detection system and method for multi-scale feature fusion | |
| CN113341391B (en) | Radar target multi-frame joint detection method in unknown environment based on deep learning | |
| CN118112663B (en) | Aviation transient electromagnetic rapid imaging method and system | |
| CN114842264B (en) | Hyperspectral image classification method based on multi-scale spatial spectrum feature joint learning | |
| CN115546096A (en) | Teleoperation mechanical arm target detection method based on scaled-FCOS |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201229 |