CN114187522A - Detection method for dangerous case of yellow river basin dam bank based on DETR model - Google Patents

Detection method for dangerous case of yellow river basin dam bank based on DETR model Download PDF

Info

Publication number
CN114187522A
CN114187522A CN202111541070.4A CN202111541070A CN114187522A CN 114187522 A CN114187522 A CN 114187522A CN 202111541070 A CN202111541070 A CN 202111541070A CN 114187522 A CN114187522 A CN 114187522A
Authority
CN
China
Prior art keywords
dam bank
model
dangerous case
detr
image
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
Application number
CN202111541070.4A
Other languages
Chinese (zh)
Inventor
乔保军
许冰辉
皂菲菲
左宪禹
王永图
谢浩粮
赵潇雄
张磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Information Center Of Henan Yellow River Bureau
Henan University
Original Assignee
Information Center Of Henan Yellow River Bureau
Henan University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Information Center Of Henan Yellow River Bureau, Henan University filed Critical Information Center Of Henan Yellow River Bureau
Priority to CN202111541070.4A priority Critical patent/CN114187522A/en
Publication of CN114187522A publication Critical patent/CN114187522A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a method for detecting a bank dangerous case of a yellow river basin based on a DETR model, which comprises the steps of obtaining a bank sample image set, wherein the bank sample image set comprises bank sample images when the yellow river basin is in danger and has no dangerous case; marking the dam bank sample image set to obtain an image training set and an image testing set; training the DETR network model according to the image training set and the image testing set until the training is finished, and acquiring a dam bank dangerous case detection model; and acquiring a real-time dam bank image, detecting by using a dam bank dangerous case detection model, and judging whether a dangerous case occurs. The method has the advantages that the DETR model is adopted for target detection, the DETR model has strong parallel computing characteristics, the speed is increased, the training time is shortened, the real-time performance is improved, a plurality of manually designed components in the traditional depth detection method are effectively removed to simplify the detection flow, all targets can be detected simultaneously, and the efficiency is higher.

Description

Detection method for dangerous case of yellow river basin dam bank based on DETR model
Technical Field
The invention relates to the field of yellow river basin dam bank dangerous case detection, in particular to a method for detecting yellow river basin dam bank dangerous case based on a DETR model.
Background
In recent years, strong rainfall often occurs along yellow river basin cities due to frequent natural disasters, so that the water level of a yellow river is caused to be higher than a water level warning line for many times, and the yellow danger-prevention dam bank can face collapse at any time. Therefore, the dam bank condition needs to be paid attention to all the time, emergency rescue is carried out in time at the initial stage of dangerous case occurrence, and the range of the dangerous case is prevented from being expanded. The traditional yellow river channel dam bank detection mode completely depends on manpower, and changes of the river channel and the dam bank are observed through long-time tour on the dam bank by workers. After long-term practice conclusion, the method has the great disadvantages that the underwater dangerous situation of the dam bank cannot be known in time, manpower and material resources are consumed, and the effect is poor.
Aiming at the defects, a plurality of detection technologies aiming at the bank dangerous case of the yellow river basin are proposed, and currently, the satellite remote sensing technology, the Internet of things technology and the like are mainly used. However, the satellite remote sensing technology has high use cost, unclear resolution and slow data transmission rate, and can not quickly obtain dangerous case results; in the internet of things technology, hardware devices such as sensors and control panels need to be deployed on the dam bank, but the environment on the dam bank of the yellow river basin is harsh, the hardware devices made of common materials are easily damaged to cause downtime, and once a dangerous situation occurs, the hardware devices are basically scrapped.
With the rapid development of artificial intelligence technology, many technicians combine the technology with their own fields for a great deal of research and exploration, and now have demonstrated their powerful learning capabilities. The target detection technology is also promoted rapidly by the benefit of transfer learning. Although the traditional deep learning identification method based on convolution is improved in precision, the model depends on an anchor frame, so that the calculation and storage resources are consumed, and the real-time requirement cannot be met.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a method for detecting the dangerous case of the yellow river basin dam bank based on the DETR model.
The adopted technical scheme is as follows:
a method for detecting dangerous cases of a yellow river basin dam bank based on a DETR model comprises the following steps:
acquiring a dam bank sample image set, wherein the dam bank sample image set comprises dam bank sample images when a yellow river basin is in danger and has no danger;
marking the dam bank sample image set to obtain an image training set and an image testing set;
training the DETR network model according to the image training set and the image testing set until the training is finished, and acquiring a dam bank dangerous case detection model;
and acquiring a real-time dam bank image, detecting by using the dam bank dangerous case detection model, and judging whether a dangerous case occurs.
Further, before the marking of the dam bank sample image set, the method for detecting the dangerous case of the dam bank in the yellow river basin further comprises:
and preprocessing the dam bank sample image set, and removing invalid sample images.
Further, after the preprocessing is performed on the dam bank sample image set and the invalid sample images are removed, the method for detecting the dangerous case of the dam bank in the yellow river basin further comprises the following steps:
and carrying out rotation, folding and deformation operations for a plurality of times on the dam bank sample images which are left after the invalid sample images are removed, or adding noise operations.
Further, the dam bank sample image set specifically includes: and (3) dam bank sample images or video streams when the yellow river basin is in danger and has no danger in different time periods, different angles and different weather conditions.
Further, the DETR network model comprises a convolutional neural network module, an encoder, a decoder, a feedforward neural network module and a loss function;
the training process of the DETR network model comprises the following steps:
convolving the dam bank sample image of the image training set according to the convolutional neural network module to generate an activation feature map;
reducing the channel dimension of the activation characteristic diagram by using the convolutional neural network module, spatially folding the activation characteristic diagram into a sequence expected to be input by the encoder, and then performing addition operation on the input sequence and fixed two-dimensional position embedded codes to input the sequence into an encoding layer of the encoder;
inputting the input sequence into the encoder for encoding calculation, wherein each encoding layer of the encoder has the same structure and comprises a multi-head self-attention mechanism, a feedforward network and layer normalization operation;
each decoding layer in the decoder has the same structure and comprises a multi-head self-attention mechanism, a multi-head cross-attention mechanism, a feed-forward network and layer normalization operation; taking N learnable object queries as the input of an encoder, participating in each multi-head attention layer calculation, and performing self-attention calculation and cross-attention calculation on the input object queries; in the cross attention calculation, the query element is an object query, and the key element is extracted from the features output by the encoder; self-attention calculation, object query interaction, and global reasoning is carried out by utilizing the mutual relation of the object query; finally converting the N object queries into N outputs;
decoding N outputs by using a feedforward neural network module to serve as a final result, and outputting the N outputs as class labels and box loss function predictions, wherein each feedforward prediction network consists of a 3-layer perceptron with a ReLU activation function and a linear prediction layer;
and (4) performing model optimization by using the loss function to obtain a dam bank dangerous case detection model.
Further, the performing model optimization by using the loss function to obtain a dam bank dangerous case detection model includes:
and (3) calculating the optimal bilateral matching of the prediction result set and the real set result by using a bipartite graph maximum matching algorithm, wherein the formula is as follows:
Figure BDA0003414184750000021
wherein the content of the first and second substances,
Figure BDA0003414184750000022
is true value yiAnd prediction sequences
Figure BDA0003414184750000023
The loss of the binary match between them,
Figure BDA0003414184750000024
the method comprises the steps of representing arrangement of N elements, wherein N represents a prediction set with a fixed size, and in a current data set, an artificially marked risk occurrence area is a truth value set; bipartite graph matching is obtained by the Hungarian algorithm, and the formula is as follows:
Figure BDA0003414184750000031
wherein the box loss function combines the L1 loss function and the GLOU loss function, and the formula is as follows:
Figure BDA0003414184750000032
inputting an image test set, evaluating the obtained dam bank dangerous case detection result by using the evaluation index, adjusting the parameters of the model according to the evaluation result, and repeatedly training the improved model until the value of the loss function is smaller than a set threshold value to obtain the dam bank dangerous case detection model.
Further, the detecting by using the dam bank dangerous case detection model to judge whether a dangerous case occurs includes:
preprocessing the real-time dam bank image;
and inputting the preprocessed real-time dam bank image into the dam bank dangerous case detection model, and outputting a detection result.
The embodiment of the invention at least has the following beneficial effects: obtaining a dam bank sample image set, wherein the dam bank sample image set comprises dam bank sample images when a yellow river basin is in danger and has no danger, labeling the dam bank sample image set, obtaining an image training set and an image testing set, training a DETR network model according to the image training set and the image testing set until the training is finished, obtaining a dam bank danger detection model, inputting a real-time dam bank image into the dam bank danger detection model for detection, and judging whether the danger occurs. The method has the advantages that the DETR model is adopted for target detection, the DETR model has strong parallel computing characteristics, the speed is increased, the training time is shortened, the real-time performance is improved, a plurality of manually designed components in the traditional depth detection method are effectively removed to simplify the detection flow, all targets can be detected simultaneously, and the efficiency is higher.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a method for detecting dangerous cases on the bank of a yellow river basin based on a DETR model provided by the invention;
fig. 2 is a diagram of a DETR neural network model provided by the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined invention purpose, the following detailed description is provided with reference to the accompanying drawings and preferred embodiments for the detection method of the dangerous case of the yellow river basin dam bank based on the DETR model, and the specific implementation, structure, features and effects thereof are described in detail. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The specific scheme of the method for detecting the dangerous case of the yellow river basin dam bank based on the DETR model is specifically described below with reference to the attached drawings.
Referring to fig. 1, a flowchart illustrating steps of a method for detecting a dangerous case at a bank of a yellow river basin based on a DETR model according to an embodiment of the present invention is shown, where the method includes the following steps:
step S1: obtaining a dam bank sample image set, wherein the dam bank sample image set comprises dam bank sample images when a yellow river basin is in danger and has no danger:
and acquiring a dam bank sample image set, wherein the dam bank sample image set comprises dam bank sample images when the yellow river basin is in danger and has no danger. The dam bank sample images in the dam bank sample image set can be historical images, namely the dam bank sample image set comprises dam bank images in the case of danger and no danger of a historical yellow river basin. Further, the dam bank sample image set specifically includes: and (3) dam bank sample images or video streams when the yellow river basin is in danger and has no danger in different time periods, different angles and different weather conditions. Therefore, the camera deployed on the yellow river basin dam bank is used for collecting dam bank sample images or video streams in the case of danger and no danger of the historical yellow river basin in different time periods, different angles and different weather conditions.
Before step S2, the dam bank sample image set may be preprocessed to implement basic classification and screening, eliminate invalid sample images (which may be damaged images), and unify the images into the same size, such as 512 × 512. And moreover, the residual dam bank sample images after the invalid sample images are removed can be subjected to rotation, folding and deformation operations for a plurality of times, or noise operation is added, so that the data set is expanded, and the generalization and the precision of the model are improved.
Step S2: labeling the dam bank sample image set to obtain an image training set and an image testing set:
and (4) marking the dam bank sample image set, namely marking the dangerous case occurring region in each dam bank sample image. In this embodiment, label is performed by using LabelMe software, a feature of a dam bank image in which a dangerous case (such as deformation or collapse) occurs is marked, a data set is produced according to a COCO standard format, and an image training set and an image testing set are obtained.
Step S3: training the DETR network model according to the image training set and the image testing set until the training is finished, and acquiring a dam bank dangerous case detection model:
in this embodiment, the overall training process of the DETR network model is as follows: extracting shallow layer characteristics from a dam bank sample image by using a convolutional neural network, stretching an obtained characteristic graph through a mapping layer, embedding the characteristic graph in a position, inputting the characteristic graph into a DETR network model to encode and decode the characteristics, and finally sending an encoding and decoding result to each pre-measuring head, wherein each pre-measuring head consists of three layers of perceptrons with a ReLU activation function, the hidden dimension is d, the mapping layer is one layer, model parameters are updated and modified through a gradient descent algorithm until a loss function reaches a preset threshold value or an automatic stopping method is utilized to obtain an optimal network model, and then the dam bank dangerous case detection model is obtained.
As a specific embodiment, the DETR network model includes a convolutional neural network module, an encoder, a decoder, a feed-forward neural network module, and a loss function, as shown in fig. 2. A specific training process for the DETR network model is given below:
(1) dam bank sample image of image training set according to convolutional neural network module
Figure BDA0003414184750000041
Performing convolution to generate an activation characteristic diagram which is an activation characteristic diagram f epsilon R with lower resolutionC×H×WIn this example C1024 (or 2048),
Figure BDA0003414184750000042
the convolution steps are as follows:
(i) performing convolution operation on the dam bank sample image by using 64 convolution kernels with the size of 3x3 and the stride of 1;
(ii) performing convolution operation on the dam bank sample image by using 128 convolution kernels with the size of 3x3 and stride of 1 and a residual error network;
(iii) performing convolution operation on the dam bank sample image by using 512 convolution kernels with the size of 3x3 and stride of 1 and a residual error network;
(iiii) convolving the dam bank sample image with 1024 convolution kernels of size 3x3 with stride 1 and a residual network;
(iiii) the dam bank sample image was convolved with 2048 convolution kernels of size 3x3 with stride 1 and a residual network, followed by a MaxPool pooling operation.
(2) Using a convolutional neural network module, in this embodiment, using a 1x1 convolutional neural network to reduce the channel dimension of the activation feature map f, in this embodiment, reducing the channel dimension of the activation feature map f from C to d, and then spatially folding the activation feature map f into a sequence z e R expected to be input by the encoderd×HWAnd then, performing an adding operation on the input sequence and a fixed two-dimensional position embedding code (the position embedding is transmitted to any attention layer of the DETR network model) and inputting the position embedding code into a code layer of the code module.
(3) Inputting the sequence z epsilon Rd×HWThe coding calculation is carried out by inputting the coding into an encoder, and each coding layer of the encoder has the same structure and comprises a multi-head self-attention mechanism, a feedforward network and a layer normalization operation. In this embodiment, the attention mechanism algorithm and the whole calculation process for calculating the input sequence can be described as follows:
Q＝WqX,K＝WkX,V＝WvX
Figure BDA0003414184750000051
X′＝MHA(LN(X))
X＝X+X′ (4)
X′＝MLP(LN(X))
X＝X+X′
(4) each decoding layer in the decoder has the same structure, and comprises a multi-head self-attention mechanism, a multi-head cross-attention mechanism, a feed-forward network and a layer normalization operation. Taking N learnable object queries as the input of the encoder, participating in each multi-head attention layer calculation, and performing self-attention calculation and cross-attention calculation on the input object queries. In the cross attention calculation, the query element is an object query, and the key element is extracted from the features output by the encoder; and self-attention calculation, object query interaction and global reasoning are carried out by utilizing the mutual relation of the object query. Finally, the N object queries are converted into N outputs.
(5) And decoding the N outputs as a final result by using a feed-forward neural network module, wherein the outputs are class labels and box loss function predictions, each feed-forward prediction network consists of a 3-layer perceptron with a ReLU activation function and a hidden layer with the dimension d, and a linear prediction layer.
(6) And (4) performing model optimization by using the loss function to obtain a dam bank dangerous case detection model. In this embodiment, the loss function is a bounding box loss function, and the building of the bounding box loss function includes: a bounding box loss function is constructed based on the L1 loss function and the GLOU loss function. As a specific implementation, the method for obtaining the dam bank dangerous case detection model by using the loss function to perform model optimization includes:
and (3) calculating the optimal bilateral matching of the prediction result set and the real set result by using a bipartite graph maximum matching algorithm, wherein the formula is as follows:
Figure BDA0003414184750000061
wherein the content of the first and second substances,
Figure BDA0003414184750000062
is true value yiAnd prediction sequences
Figure BDA0003414184750000063
The loss of the binary match between them,
Figure BDA0003414184750000064
the method comprises the steps of representing arrangement of N elements, wherein N represents a prediction set with a fixed size, and in a current data set, an artificially marked risk occurrence area is a truth value set; bipartite graph matching is obtained by the Hungarian algorithm (Hungarian algorithm) as follows:
Figure BDA0003414184750000065
wherein the box loss function combines the L1 loss function and the GLOU loss function, and the formula is as follows:
Figure BDA0003414184750000066
(7) inputting an image test set, evaluating the obtained dam bank dangerous case detection result by using the evaluation index, adjusting the parameters of the model according to the evaluation result, and repeatedly training the improved model until the value of the loss function is smaller than a set threshold value to obtain the dam bank dangerous case detection model.
Step S4: acquiring a real-time dam bank image, detecting by using the dam bank dangerous case detection model, and judging whether dangerous cases occur:
acquiring a real-time dam bank image, and preprocessing the real-time dam bank image in the embodiment; and inputting the preprocessed real-time dam bank image into a dam bank dangerous case detection model, and outputting a detection result. And if the dam bank is detected to have dangerous situations, such as deformation or collapse, an alarm is given.
The invention provides a method for detecting a dam bank based on a DETR model, which is characterized in that historical dam bank image data are input into the DETR neural network model for full training and learning, the trained neural network model is used for detecting a dam bank image of a yellow river basin acquired in real time, an emergence area is marked, an alarm can be sent out at the same time, a worker is informed to check on site, and assistance is provided for the worker to judge dangerous situations. Moreover, the model does not need prior knowledge, has strong parallel capability and real-time performance, and is an end-to-end model.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims (7)

1. A method for detecting the dangerous case of the yellow river basin dam bank based on a DETR model is characterized by comprising the following steps:
acquiring a dam bank sample image set, wherein the dam bank sample image set comprises dam bank sample images when a yellow river basin is in danger and has no danger;
marking the dam bank sample image set to obtain an image training set and an image testing set;
training the DETR network model according to the image training set and the image testing set until the training is finished, and acquiring a dam bank dangerous case detection model;
and acquiring a real-time dam bank image, detecting by using the dam bank dangerous case detection model, and judging whether a dangerous case occurs.
2. The method for detecting the dangerous case of the yellow river basin dam bank based on the DETR model as claimed in claim 1, wherein before the labeling of the dam bank sample image set, the method for detecting the dangerous case of the yellow river basin dam bank further comprises:
and preprocessing the dam bank sample image set, and removing invalid sample images.
3. The method for detecting the dangerous case of the yellow river basin dam bank based on the DETR model as claimed in claim 2, wherein the method for detecting the dangerous case of the yellow river basin dam bank further comprises the steps of preprocessing the sample image set of the dam bank, and after eliminating invalid sample images:
and carrying out rotation, folding and deformation operations for a plurality of times on the dam bank sample images which are left after the invalid sample images are removed, or adding noise operations.
4. The method for detecting the dangerous case of the dam bank of the yellow river basin based on the DETR model as claimed in claim 1, wherein the dam bank sample image set specifically comprises: and (3) dam bank sample images or video streams when the yellow river basin is in danger and has no danger in different time periods, different angles and different weather conditions.
5. The method for detecting the dangerous case of the yellow river basin dam bank based on the DETR model is characterized in that the DETR network model comprises a convolutional neural network module, an encoder, a decoder, a feedforward neural network module and a loss function;
the training process of the DETR network model comprises the following steps:
convolving the dam bank sample image of the image training set according to the convolutional neural network module to generate an activation feature map;
reducing the channel dimension of the activation characteristic diagram by using the convolutional neural network module, spatially folding the activation characteristic diagram into a sequence expected to be input by the encoder, and then performing addition operation on the input sequence and fixed two-dimensional position embedded codes to input the sequence into an encoding layer of the encoder;
inputting the input sequence into the encoder for encoding calculation, wherein each encoding layer of the encoder has the same structure and comprises a multi-head self-attention mechanism, a feedforward network and layer normalization operation;
each decoding layer in the decoder has the same structure and comprises a multi-head self-attention mechanism, a multi-head cross-attention mechanism, a feed-forward network and layer normalization operation; taking N learnable object queries as the input of an encoder, participating in each multi-head attention layer calculation, and performing self-attention calculation and cross-attention calculation on the input object queries; in the cross attention calculation, the query element is an object query, and the key element is extracted from the features output by the encoder; self-attention calculation, object query interaction, and global reasoning is carried out by utilizing the mutual relation of the object query; finally converting the N object queries into N outputs;
decoding N outputs by using a feedforward neural network module to serve as a final result, and outputting the N outputs as class labels and box loss function predictions, wherein each feedforward prediction network consists of a 3-layer perceptron with a ReLU activation function and a linear prediction layer;
and (4) performing model optimization by using the loss function to obtain a dam bank dangerous case detection model.
6. The method for detecting the dangerous case of the dam bank of the yellow river basin based on the DETR model as claimed in claim 5, wherein the obtaining of the dam bank dangerous case detection model by using the model optimization of the loss function comprises:
and (3) calculating the optimal bilateral matching of the prediction result set and the real set result by using a bipartite graph maximum matching algorithm, wherein the formula is as follows:
Figure FDA0003414184740000021
wherein the content of the first and second substances,
Figure FDA0003414184740000022
is true value yiAnd prediction sequences
Figure FDA0003414184740000023
The loss of the binary match between them,
Figure FDA0003414184740000024
the method comprises the steps of representing arrangement of N elements, wherein N represents a prediction set with a fixed size, and in a current data set, an artificially marked risk occurrence area is a truth value set; bipartite graph matching is obtained by the Hungarian algorithm, and the formula is as follows:
Figure FDA0003414184740000025
wherein the box loss function combines the L1 loss function and the GLOU loss function, and the formula is as follows:
Figure FDA0003414184740000026
inputting an image test set, evaluating the obtained dam bank dangerous case detection result by using the evaluation index, adjusting the parameters of the model according to the evaluation result, and repeatedly training the improved model until the value of the loss function is smaller than a set threshold value to obtain the dam bank dangerous case detection model.
7. The method for detecting the dangerous case of the dam bank of the yellow river basin based on the DETR model as claimed in claim 1, wherein the detecting by using the dam bank dangerous case detection model to determine whether the dangerous case occurs comprises:
preprocessing the real-time dam bank image;
and inputting the preprocessed real-time dam bank image into the dam bank dangerous case detection model, and outputting a detection result.
CN202111541070.4A 2021-12-16 2021-12-16 Detection method for dangerous case of yellow river basin dam bank based on DETR model Pending CN114187522A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111541070.4A CN114187522A (en) 2021-12-16 2021-12-16 Detection method for dangerous case of yellow river basin dam bank based on DETR model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111541070.4A CN114187522A (en) 2021-12-16 2021-12-16 Detection method for dangerous case of yellow river basin dam bank based on DETR model

Publications (1)

Publication Number Publication Date
CN114187522A true CN114187522A (en) 2022-03-15

Family

ID=80605293

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111541070.4A Pending CN114187522A (en) 2021-12-16 2021-12-16 Detection method for dangerous case of yellow river basin dam bank based on DETR model

Country Status (1)

Country Link
CN (1) CN114187522A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116138780A (en) * 2022-12-30 2023-05-23 北京视友科技有限责任公司 Student attention evaluation method, terminal and computer readable storage medium
CN116580328A (en) * 2023-07-12 2023-08-11 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Intelligent recognition method for leakage danger of thermal infrared image dykes and dams based on multitasking assistance
CN117034143A (en) * 2023-10-10 2023-11-10 南京邮电大学 Distributed system fault diagnosis method and device based on machine learning

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116138780A (en) * 2022-12-30 2023-05-23 北京视友科技有限责任公司 Student attention evaluation method, terminal and computer readable storage medium
CN116138780B (en) * 2022-12-30 2023-08-08 北京视友科技有限责任公司 Student attention evaluation method, terminal and computer readable storage medium
CN116580328A (en) * 2023-07-12 2023-08-11 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Intelligent recognition method for leakage danger of thermal infrared image dykes and dams based on multitasking assistance
CN116580328B (en) * 2023-07-12 2023-09-19 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Intelligent recognition method for leakage danger of thermal infrared image dykes and dams based on multitasking assistance
CN117034143A (en) * 2023-10-10 2023-11-10 南京邮电大学 Distributed system fault diagnosis method and device based on machine learning
CN117034143B (en) * 2023-10-10 2023-12-15 南京邮电大学 Distributed system fault diagnosis method and device based on machine learning

Similar Documents

Publication Publication Date Title
CN114187522A (en) Detection method for dangerous case of yellow river basin dam bank based on DETR model
Sony et al. A systematic review of convolutional neural network-based structural condition assessment techniques
US11521379B1 (en) Method for flood disaster monitoring and disaster analysis based on vision transformer
CN113191285B (en) River and lake remote sensing image segmentation method and system based on convolutional neural network and Transformer
CN114120102A (en) Boundary-optimized remote sensing image semantic segmentation method, device, equipment and medium
CN111079683A (en) Remote sensing image cloud and snow detection method based on convolutional neural network
CN111177882B (en) PIV flow field recovery method based on deep migration learning
CN112558185A (en) Bidirectional GRU typhoon track intelligent prediction and forecast system based on attention mechanism, computer equipment and storage medium
CN117253112B (en) Large-model visual language cross-modal learning method for structural health diagnosis
CN115272940B (en) Neural network-based intelligent rock debris particle metering method and system
Kamoji et al. Effective Flood prediction model based on Twitter Text and Image analysis using BMLP and SDAE-HHNN
Kaur et al. A review on natural disaster detection in social media and satellite imagery using machine learning and deep learning
CN117726954A (en) Sea-land segmentation method and system for remote sensing image
CN117746264A (en) Multitasking implementation method for unmanned aerial vehicle detection and road segmentation
CN116665114B (en) Multi-mode-based remote sensing scene identification method, system and medium
Li et al. Exposed aggregate detection of stilling basin slabs using attention U-net network
CN117056807A (en) Integrated cloud detection method based on semantic segmentation network and long-term and short-term memory network, readable storage medium and computing device
Wu et al. SlimRGBD: A geographic information photography noise reduction system for aerial remote sensing
CN111950465A (en) SAE-CNN-based SAR image flood detection and prevention method
CN113343861B (en) Remote sensing image water body region extraction method based on neural network model
CN116824352A (en) Water surface floater identification method based on semantic segmentation and image anomaly detection
CN116797941A (en) Marine oil spill risk source rapid intelligent identification and classification method for high-resolution remote sensing image
Meng et al. A modified fully convolutional network for crack damage identification compared with conventional methods
CN115984714A (en) Cloud detection method based on double-branch network model
Bahrami et al. Enhanced Flood Detection Through Precise Water Segmentation Using Advanced Deep Learning Models

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