CN110929618A - Potential safety hazard detection and evaluation method for power distribution network crossing type building - Google Patents

Potential safety hazard detection and evaluation method for power distribution network crossing type building Download PDF

Info

Publication number
CN110929618A
CN110929618A CN201911118585.6A CN201911118585A CN110929618A CN 110929618 A CN110929618 A CN 110929618A CN 201911118585 A CN201911118585 A CN 201911118585A CN 110929618 A CN110929618 A CN 110929618A
Authority
CN
China
Prior art keywords
potential safety
safety hazard
power distribution
distribution network
detection
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201911118585.6A
Other languages
Chinese (zh)
Other versions
CN110929618B (en
Inventor
刘蓓
安义
杜敏
刘珣
陈世金
尚银辉
戚沁雅
周求宽
欧阳文华
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.)
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Jiangxi Electric Power Co Ltd
Original Assignee
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Jiangxi Electric Power Co Ltd
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 State Grid Corp of China SGCC, Electric Power Research Institute of State Grid Jiangxi Electric Power Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN201911118585.6A priority Critical patent/CN110929618B/en
Publication of CN110929618A publication Critical patent/CN110929618A/en
Application granted granted Critical
Publication of CN110929618B publication Critical patent/CN110929618B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06V20/176—Urban or other man-made structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Tourism & Hospitality (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Game Theory and Decision Science (AREA)
  • Educational Administration (AREA)
  • Development Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Operations Research (AREA)
  • Evolutionary Biology (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a potential safety hazard detection and evaluation method for a power distribution network crossing type building, which comprises the following steps of: obtaining a remote sensing image data set by establishing a channel environment remote sensing image block set based on fast-RCNN; extracting characteristic data from channel environment remote sensing image information by establishing a safety hazard detection model based on a fast-RCNN power distribution network tower pole construction, and determining a possible candidate target region; training a potential safety hazard detection model constructed on the basis of the tower pole of the fast-RCNN power distribution network to obtain a learning model, inputting the characteristic data into the learning model, and generating a model output result; the method comprises the steps of detecting and evaluating the output result by establishing a power distribution tower pole construction potential safety hazard intelligent detection and evaluation method based on fast-RCNN, and outputting the evaluation result.

Description

Potential safety hazard detection and evaluation method for power distribution network crossing type building
Technical Field
The invention relates to the field of artificial intelligence and ubiquitous power internet of things, in particular to a potential safety hazard detection and evaluation method for a power distribution network crossing type building.
Background
The power distribution network is the last kilometer of electric energy transmitted to users by the power system, and potential safety hazards caused by the fact that the power distribution network is distributed in a crowd-dense area cannot be ignored. In recent years, national grid companies issue notifications for potential safety hazards such as 'three-span' (crossing roads, railways and bridges), crossing fishponds and the like for troubleshooting and governing power distribution networks for many times, but the troubleshooting modes adopted in the industry at present are all manual on-site investigation modes, and the problems of large workload, incomplete troubleshooting, inaccurate positions and the like exist.
With the rapid development of artificial intelligence technology, remote sensing image recognition technology has been widely applied in the fields of satellite transmission, address survey, city planning and construction, earthquake relief and the like. Potential safety hazards such as 'three spans' of the power distribution network, fishpond crossing and the like are closely related to geographic information, so that if the remote sensing image recognition technology is applied to the investigation of the potential safety hazards such as 'three spans' of the power distribution network, fishpond crossing and the like, the manual workload can be effectively reduced, and the comprehensiveness and accuracy of the investigation of the potential safety hazards can be improved.
Disclosure of Invention
In order to solve the existing problems, the invention discloses a potential safety hazard detection and evaluation method for a power distribution network spanning type building, which is applied to the detection and evaluation of potential safety hazards of a power distribution network 'three spans' and a spanning fishpond, and comprises the following steps:
s1: obtaining a remote sensing image data set by establishing a channel environment remote sensing image block set based on a fast-RCNN network detection model;
s2: extracting characteristic data from the remote sensing image data set obtained in the step S1 by establishing a power distribution network tower pole construction potential safety hazard detection model based on a fast-RCNN network detection model, and determining a possible candidate target area;
s3: training a potential safety hazard detection model constructed on the basis of a power distribution network tower pole of a fast-RCNN network detection model to obtain a learning model, inputting the characteristic data in the step S2 into the learning model, and generating a model output result;
s4: and detecting and evaluating the output result by establishing an intelligent detection and evaluation method for potential safety hazard in the construction of the power distribution tower pole based on the fast-RCNN network detection model, and outputting the evaluation result.
Further, the step S1 further includes the following steps:
carrying out classification labeling on the remote sensing image data set, generating a corresponding class label by the remote sensing image data set, and dividing the class label into a training set and a verification set according to a certain proportion or step length;
the remote sensing image data set contains potential safety hazards of spanning construction of tower poles of the power distribution network to be detected;
the potential safety hazards include one or more of railways, highways, and fish ponds.
Further, step S2 further includes the following steps:
s21: sequentially constructing a tower pole three-span feature extraction submodel, an RPN region suggestion framework submodel and a classification detection position correction submodel;
s22: and performing effect simulation on the tower pole three-span feature extraction submodel, the RPN region proposed framework submodel and the classification detection position correction submodel by adopting a training set and a verification set.
Furthermore, the simulation effect of the tower three-span feature extraction submodel is that the tower three-span feature extraction submodel is used for extracting features of a training set and a verification set;
the tower pole three-span feature extraction submodel adopts at least 13 convolution layers, at least 13 Relu layers and at least 4 pooling layers;
and setting parameters of the convolution layer, and reducing the size of the picture after the features are extracted.
Further, the RPN region suggestion framework sub-model adopts a full convolution network to realize a recommendation possibility candidate target region;
the RPN region suggested framework sub-model comprises a CNN model frame and a convolution layer and a two-layer structure which are connected behind the CNN model frame; the two-layer structure comprises a classification structure and a target detection positioning structure, wherein the classification structure is used for classifying targets, and the target detection positioning structure is used for accurately positioning a possible candidate target area;
and the ROI-Pooling layer in the fast-RCNN network detection model is used for collecting the extracted feature data and calculating a feature map.
Further, the simulation effect of the classification detection position correction submodel is to calculate the specific attribution category of each abstracted feature by adopting a full connection layer and Softmax and realize regression adjustment of a target detection frame.
Further, the safety hazard detection model for building the tower pole of the fast-RCNN power distribution network is built in the step S2 and is calculated by adopting a loss function of the safety hazard detection model, wherein the formula is as follows:
Figure BDA0002274775760000021
where i denotes the candidate frame index, piRepresenting the probability of the candidate box being predicted as the target,
Figure BDA0002274775760000022
a calibration value representing a candidate box is shown,
Figure BDA0002274775760000031
representing two classes of logarithmic loss, NregRepresenting the number of anchor positions, NclsDenotes the minimum batch size, lambda denotes the equilibrium parameter, LregRepresenting the smoothing L1 penalty function, t representing the offset of the prediction candidate block, t*Representing the actual offset of the calibration frame corresponding to the candidate frame;
Figure BDA0002274775760000032
Figure BDA0002274775760000033
Figure BDA0002274775760000034
further, step S3 further includes the following steps:
s31: selecting VGG16 weight to initialize a fast-RCNN network detection model;
performing feature extraction by adopting a VGG16 convolutional neural network trained on ImageNet, and abandoning a full connection layer for classification; carrying out training initialization on a fast-RCNN network detection model by adopting the trained VGG16 weight, and continuously training the fast-RCNN network detection model by utilizing a distribution network channel environment remote sensing image set;
s32: and setting parameters of the fast-RCNN network detection model.
Further, the loss function of the potential safety hazard detection model is optimized by adopting a random gradient descent method.
Further, step S4 further includes the following steps:
s41: selecting a target area: carrying out target area detection on the power distribution network channel environment based on a fast-RCNN trained fast-RCNN network detection model, wherein the target area is divided into a starting point and an end point for road, railway, fishpond and tower pole construction; if the starting point and the ending point of the tower pole construction are not on the same side of the target area, the tower pole is judged to be in a crossing construction, and potential safety hazards exist;
s42: detection and evaluation: if one side of the target area simultaneously contains a starting point and an end point of tower construction, judging the potential safety hazard to be a low potential safety hazard;
if the initial point and the end point of the tower pole construction are respectively distributed on two sides of the target area, the potential safety hazard is judged to be a moderate potential safety hazard;
different detection targets contain the initial point and the end point of specific tower pole construction, and then the potential safety hazard that the tower pole construction exists in the power distribution network is judged to be the high-risk potential hazard.
The invention has the beneficial effects that: the invention is based on the multi-potential safety hazard detection model for the intelligent power distribution network tower-rod crossing construction of remote sensing images of the fast-RCNN, and the recall ratio and the precision are effectively improved; and the multi-potential-hazard rapid detection network constructed in a tower-pole crossing manner under the power distribution network channel environment is realized, and the potential-hazard detection network is trained by using the self-constructed intelligent power distribution network channel environment remote sensing image on the premise of maximum safety, robustness and accuracy. Therefore, the method utilizes the multi-scale remote sensing image set and the batch regularization method to train and improve the generalization capability of the fast-RCNN network detection model. The multi-potential-hazard detection model based on the fast-RCNN network detection model is provided for potential hazards existing in crossing construction of the system tower poles of the intelligent power distribution network, multi-potential-hazard detection of construction of the system tower poles of the power distribution network is achieved, and effective guarantee is provided for construction safety of the system tower poles of the intelligent power distribution network.
Drawings
FIG. 1 is a block diagram of a technical route in a preferred embodiment of the present invention;
FIG. 2 is a block diagram of a power distribution network tower construction three-span potential safety hazard detection model in the preferred embodiment of the invention;
fig. 3 is a block diagram of potential safety hazard detection and evaluation in the spanning construction of tower poles of a power distribution network in the preferred embodiment of the present invention;
FIG. 4 is a feature extraction network of a detection model in a preferred embodiment of the present invention;
fig. 5 is a diagram of an RPN network architecture in a preferred embodiment of the present invention;
FIG. 6 is a classification and detection model in a preferred embodiment of the invention;
FIG. 7 is a fast-RCNN network detection model in accordance with a preferred embodiment of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The following detailed description of specific embodiments of the invention refers to the accompanying drawings.
Referring to fig. 1 to 7, the method for detecting and evaluating potential safety hazards of a power distribution network spanning type building, provided by the invention, is applied to detection and evaluation of potential safety hazards of a power distribution network in a 'three-span' manner and across a fishpond, and specifically applied as shown in fig. 1 to 3: the method is applied to a power distribution tower pole construction potential safety hazard assessment system, the system comprises a power distribution channel remote sensing image block set and a power distribution tower pole construction potential safety hazard detection model, and the method comprises the following steps:
s1: obtaining a remote sensing image data set by establishing a channel environment remote sensing image block set based on a fast-RCNN network detection model;
s2: extracting characteristic data from the remote sensing image set obtained in the step S1 by establishing a power distribution network tower pole construction potential safety hazard detection model based on a fast-RCNN network detection model, and determining a possible candidate target region;
s3: training a potential safety hazard detection model constructed on the basis of a power distribution network tower pole of a fast-RCNN network detection model to obtain a learning model, inputting characteristic data into the learning model, and generating a model output result;
s4: and detecting and evaluating the output result by establishing an intelligent detection and evaluation method for potential safety hazard in the construction of the power distribution tower pole based on the fast-RCNN network detection model, and outputting the evaluation result.
Specifically, step S1 further includes the steps of:
carrying out classification labeling on the remote sensing image data set, generating corresponding class labels for the remote sensing image data set, and dividing the class labels into a training set and a verification set according to step length; the training set is train, and the verification set is valid;
train＝[I1,I2,...,In]
valid＝[In+1,In+2,...,IN]
the remote sensing image data set contains potential safety hazards of spanning construction of tower poles of the power distribution network to be detected;
the potential safety hazards include railways, highways and fish ponds, and other potential safety hazards can be increased in specific application.
Specifically, as shown in fig. 5 to 7, step S2 further includes the steps of:
s21: sequentially constructing a tower pole three-span feature extraction submodel, an RPN region suggestion framework submodel and a classification detection position correction submodel;
s22: and performing effect simulation on the tower pole three-span feature extraction submodel, the RPN region proposed framework submodel and the classification detection position correction submodel by adopting a training set and a verification set.
Specifically, the simulation effect of the tower three-span feature extraction submodel is as follows:
a tower-pole three-span feature extraction sub-model is adopted for feature extraction of the training set and the verification set;
the tower pole three-span feature extraction submodel adopts 13 convolutional layers, 13 Relu layers and 4 pooling layers;
and setting parameters of the convolution layer, and reducing the size of the picture after the features are extracted.
In specific application, the parameters are set as follows:
the setup of the convolutional layer in the feature extraction network is as follows: keenenl _ size ═ 3, pad ═ 1, stride ═ 1; the arrangement of the pooling layer is as follows: keenenl _ size ═ 2, stride ═ 2; the picture size after passing through the feature extraction network is reduced to 16 x 16 times of the original size
Specifically, the RPN region suggestion framework sub-model adopts a full convolution network to realize a recommendation possibility candidate target region;
the RPN region suggested framework sub-model comprises a CNN model frame and a convolution layer and a two-layer structure which are connected behind the CNN model frame; the two-layer structure comprises a classification structure and a target detection positioning structure, wherein the classification structure is used for classifying targets, and the target detection positioning structure is used for accurately positioning a possible candidate target area;
and the ROI-Pooling layer in the fast-RCNN network detection model is used for collecting the extracted feature data and calculating a feature map.
Specifically, the simulation effect of the classification detection position correction submodel is to calculate the specific attribution category of each abstracted feature by adopting a full connection layer and Softmax and realize regression adjustment of a target detection frame.
Specifically, the hidden danger detection model for building the tower pole of the fast-RCNN power distribution network in the step S2 is calculated by using a loss function of the hidden danger detection model, and the formula is as follows:
Figure BDA0002274775760000061
where i denotes the candidate frame index, piRepresenting the probability of the candidate box being predicted as the target,
Figure BDA0002274775760000062
a calibration value representing a candidate box is shown,
Figure BDA0002274775760000063
representing two classes of logarithmic loss, NregRepresenting the number of anchor positions, NclsDenotes the minimum batch size, lambda denotes the equilibrium parameter, LregRepresenting the smoothing L1 penalty function, t representing the offset of the prediction candidate block, t*Representing the actual offset of the calibration frame corresponding to the candidate frame;
Figure BDA0002274775760000064
Figure BDA0002274775760000065
Figure BDA0002274775760000066
specifically, step S3 further includes the steps of:
s31: selecting VGG16 weight to initialize a fast-RCNN network detection model;
performing feature extraction by adopting a VGG16 convolutional neural network trained on ImageNet, and abandoning a full connection layer for classification; carrying out training initialization on a fast-RCNN network detection model by adopting the trained VGG16 weight, and continuously training the fast-RCNN network detection model by utilizing a distribution network channel environment remote sensing image set;
s32: and setting parameters of the fast-RCNN network detection model.
Specifically, a loss function of the potential safety hazard detection model is optimized by adopting a random gradient descent method, and specific parameters are set as follows: max _ iters 10000; leaving _ rate ═ 0.001; batch _ size 64; step _ size ═ 4000; momentum is 0.9, and weight _ decay is 0.0005.
Specifically, step S4 further includes the steps of:
s41: selecting a target area: carrying out target area detection on the power distribution network channel environment based on a fast-RCNN trained fast-RCNN network detection model, wherein the target area comprises a road, a railway or a fishpond; if the starting point and the ending point of the tower pole construction are not on the same side of the target area, the tower pole is judged to be in a crossing construction, and potential safety hazards exist;
s42: detection and evaluation: if one side of the target area simultaneously contains a starting point and an end point of tower construction, judging the potential safety hazard to be a low potential safety hazard;
if the initial point and the end point of the tower pole construction are respectively distributed on two sides of the target area, the potential safety hazard is judged to be a moderate potential safety hazard;
different detection targets contain the initial point and the end point of specific tower pole construction, and then the potential safety hazard that the tower pole construction exists in the power distribution network is judged to be the high-risk potential hazard.
The above embodiments only describe the best mode of use of the existing device, and similar common means are used to replace the elements in the present embodiments, which fall into the protection scope.

Claims (10)

1. A potential safety hazard detection and evaluation method for a power distribution network crossing type building is characterized by comprising the following steps: the method comprises the following steps:
s1: obtaining a remote sensing image data set by establishing a channel environment remote sensing image block set based on a fast-RCNN network detection model;
s2: extracting characteristic data from the remote sensing image data set obtained in the step S1 by establishing a power distribution network tower pole construction potential safety hazard detection model based on a fast-RCNN network detection model, and determining a possible candidate target area;
s3: training a potential safety hazard detection model constructed on the basis of a power distribution network tower pole of a fast-RCNN network detection model to obtain a learning model, inputting the characteristic data in the step S2 into the learning model, and generating a model output result;
s4: and detecting and evaluating the output result by establishing an intelligent detection and evaluation method for potential safety hazard in the construction of the power distribution tower pole based on the fast-RCNN network detection model, and outputting the evaluation result.
2. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 1, wherein the method comprises the following steps: the step S1 further includes the following steps:
carrying out classification labeling on the remote sensing image data set, generating a corresponding class label by the remote sensing image data set, and dividing the class label into a training set and a verification set according to a certain proportion or step length;
the remote sensing image data set contains potential safety hazards of spanning construction of tower poles of the power distribution network to be detected;
the potential safety hazard comprises one or more of railways, roads and fish ponds.
3. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 1, wherein the method comprises the following steps: step S2 further includes the steps of:
s21: sequentially constructing a tower pole three-span feature extraction submodel, an RPN region suggestion framework submodel and a classification detection position correction submodel;
s22: and performing effect simulation on the tower pole three-span feature extraction submodel, the RPN region proposed framework submodel and the classification detection position correction submodel by adopting a training set and a verification set.
4. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 3, wherein the method comprises the following steps: the simulation effect of the tower pole three-span feature extraction submodel is to extract features of a training set and a verification set by adopting the tower pole three-span feature extraction submodel;
the tower pole three-span feature extraction submodel adopts at least 13 convolution layers, at least 13 Relu layers and at least 4 pooling layers;
and setting parameters of the convolution layer, and reducing the size of the picture after the features are extracted.
5. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 3, wherein the method comprises the following steps: the RPN region suggested framework sub-model adopts a full convolution network;
the RPN region suggested framework sub-model comprises a CNN model frame and a convolution layer and a two-layer structure which are connected behind the CNN model frame; the two-layer structure comprises a classification structure and a target detection positioning structure, wherein the classification structure is used for classifying targets, and the target detection positioning structure is used for accurately positioning a possible candidate target area;
and the model of the ROI-Pooling layer in the fast-RCNN network detection model is used for collecting the extracted feature data and calculating a feature map.
6. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 5, wherein the method comprises the following steps: the simulation effect of the classification detection position correction submodel is to calculate the specific attribution category of each abstracted feature by adopting a full connection layer and a Softmax classification function and realize regression adjustment of a target detection frame.
7. The method for detecting and evaluating the potential safety hazard of the power distribution network spanning type building according to any one of claims 1 to 6, wherein the method comprises the following steps:
the method comprises the following steps of S2, constructing a Faster-RCNN distribution network tower pole construction potential safety hazard detection model, and calculating by adopting a potential safety hazard detection model loss function, wherein the formula is as follows:
Figure FDA0002274775750000021
where i denotes the candidate frame index, piRepresenting the probability of the candidate box being predicted as the target,
Figure FDA0002274775750000022
a calibration value representing a candidate box is shown,
Figure FDA0002274775750000023
representing two classes of logarithmic loss, NregRepresenting the number of anchor positions, NclsDenotes the minimum batch size, lambda denotes the equilibrium parameter, LregRepresenting the smoothing L1 penalty function, t representing the offset of the prediction candidate block, t*Representing the actual offset of the calibration frame corresponding to the candidate frame;
Figure FDA0002274775750000024
Figure FDA0002274775750000025
Figure FDA0002274775750000026
Figure FDA0002274775750000031
8. the potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 7, wherein the method comprises the following steps:
the step S3 further includes the following steps:
s31: selecting a VGG16 weight to initialize a fast-RCNN network detection model, adopting a VGG16 convolutional neural network trained on ImageNet to extract characteristics, and abandoning a full connection layer for classification; carrying out training initialization on a fast-RCNN network detection model by adopting the trained VGG16 weight, and continuously training the fast-RCNN network detection model by utilizing a distribution network channel environment remote sensing image set;
s32: and setting parameters of the fast-RCNN network detection model.
9. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 7, wherein the method comprises the following steps: and the loss function of the potential safety hazard detection model is optimized by adopting a random gradient descent method.
10. The potential safety hazard detection and evaluation method for the power distribution network spanning type building according to claim 9, wherein the method comprises the following steps:
step S4 further includes the steps of:
s41: selecting a target area: carrying out target area detection on the power distribution network channel environment by the trained fast-RCNN network detection model, wherein the target area comprises a road, a railway or a fishpond; if the starting point and the ending point of the tower pole construction are not on the same side of the target area, the tower pole is judged to be in a crossing construction, and potential safety hazards exist;
s42: detection and evaluation: if one side of the target area simultaneously contains a starting point and an end point of tower construction, judging the potential safety hazard to be a low potential safety hazard;
if the initial point and the end point of the tower pole construction are respectively distributed on two sides of the target area, the potential safety hazard is judged to be a moderate potential safety hazard;
different detection targets contain the initial point and the end point of specific tower pole construction, and then the potential safety hazard that the tower pole construction exists in the power distribution network is judged to be the high-risk potential hazard.
CN201911118585.6A 2019-11-15 2019-11-15 Potential safety hazard detection and assessment method for power distribution network crossing type building Active CN110929618B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911118585.6A CN110929618B (en) 2019-11-15 2019-11-15 Potential safety hazard detection and assessment method for power distribution network crossing type building

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911118585.6A CN110929618B (en) 2019-11-15 2019-11-15 Potential safety hazard detection and assessment method for power distribution network crossing type building

Publications (2)

Publication Number Publication Date
CN110929618A true CN110929618A (en) 2020-03-27
CN110929618B CN110929618B (en) 2023-06-20

Family

ID=69854044

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911118585.6A Active CN110929618B (en) 2019-11-15 2019-11-15 Potential safety hazard detection and assessment method for power distribution network crossing type building

Country Status (1)

Country Link
CN (1) CN110929618B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111950488A (en) * 2020-08-18 2020-11-17 山西大学 Improved fast-RCNN remote sensing image target detection method

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009129013A1 (en) * 2008-04-18 2009-10-22 The Boeing Company Assessing conditions of aircraft wiring
WO2014101636A1 (en) * 2012-12-31 2014-07-03 北京邮电大学 Method for evaluating risk in electric power communications network
TW201434004A (en) * 2013-02-19 2014-09-01 Univ Nat Kaohsiung Marine Distribution network enhanced decision-making system with increased distributed generation grid connection safety
CN105787501A (en) * 2015-12-17 2016-07-20 武汉大学 Vegetation classification method capable of automatically selecting features in power transmission line corridor area
CN108564109A (en) * 2018-03-21 2018-09-21 天津大学 A kind of Remote Sensing Target detection method based on deep learning
CN109765462A (en) * 2019-03-05 2019-05-17 国家电网有限公司 Fault detection method, device and the terminal device of transmission line of electricity
CN109977921A (en) * 2019-04-11 2019-07-05 广东电网有限责任公司 A kind of transmission line of electricity perils detecting method
CN110222641A (en) * 2019-06-06 2019-09-10 北京百度网讯科技有限公司 The method and apparatus of image for identification

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009129013A1 (en) * 2008-04-18 2009-10-22 The Boeing Company Assessing conditions of aircraft wiring
WO2014101636A1 (en) * 2012-12-31 2014-07-03 北京邮电大学 Method for evaluating risk in electric power communications network
TW201434004A (en) * 2013-02-19 2014-09-01 Univ Nat Kaohsiung Marine Distribution network enhanced decision-making system with increased distributed generation grid connection safety
CN105787501A (en) * 2015-12-17 2016-07-20 武汉大学 Vegetation classification method capable of automatically selecting features in power transmission line corridor area
CN108564109A (en) * 2018-03-21 2018-09-21 天津大学 A kind of Remote Sensing Target detection method based on deep learning
CN109765462A (en) * 2019-03-05 2019-05-17 国家电网有限公司 Fault detection method, device and the terminal device of transmission line of electricity
CN109977921A (en) * 2019-04-11 2019-07-05 广东电网有限责任公司 A kind of transmission line of electricity perils detecting method
CN110222641A (en) * 2019-06-06 2019-09-10 北京百度网讯科技有限公司 The method and apparatus of image for identification

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘宁: "高压输电线路智能检测技术研究与应用", 《中国优秀硕士学位论文全文数据库（电子期刊）工程科技Ⅱ辑》 *
孙梓超 等: "基于深度卷积神经网络的遥感影像目标检测", 《上海航天》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111950488A (en) * 2020-08-18 2020-11-17 山西大学 Improved fast-RCNN remote sensing image target detection method
CN111950488B (en) * 2020-08-18 2022-07-19 山西大学 Improved Faster-RCNN remote sensing image target detection method

Also Published As

Publication number Publication date
CN110929618B (en) 2023-06-20

Similar Documents

Publication Publication Date Title
US20210209939A1 (en) Large-scale real-time traffic flow prediction method based on fuzzy logic and deep LSTM
CN108805348B (en) Method and device for controlling and optimizing intersection signal timing
CN108629978A (en) A kind of traffic trajectory predictions method based on higher-dimension road network and Recognition with Recurrent Neural Network
CN106651036A (en) Air quality forecasting system
CN109359166B (en) Space growth dynamic simulation and driving force factor contribution degree synchronous calculation method
CN109948547A (en) Urban green space landscape evaluation method, device, storage medium and terminal device
CN111667101A (en) Personalized electric power field operation path planning method and system integrating high-resolution remote sensing image and terrain
CN112668375B (en) Tourist distribution analysis system and method in scenic spot
CN112801399B (en) Path generation method and device, terminal equipment and storage medium
CN112712138A (en) Image processing method, device, equipment and storage medium
CN116046008A (en) Situation awareness-based route planning method, system and efficiency evaluation device
CN115985104A (en) Traffic flow prediction device, prediction method and prediction model construction method
CN115564174A (en) Method, system, computer device and medium for measuring living street space quality
Zheng et al. A deep learning–based approach for moving vehicle counting and short-term traffic prediction from video images
CN110929618B (en) Potential safety hazard detection and assessment method for power distribution network crossing type building
CN116884222B (en) Short-time traffic flow prediction method for bayonet nodes
CN114550016B (en) Unmanned aerial vehicle positioning method and system based on context information perception
Huang et al. Estimating urban noise along road network from street view imagery
CN112785083B (en) Arrival time estimation method and device, electronic equipment and storage medium
CN115511280A (en) Urban flood toughness evaluation method based on multi-mode data fusion
CN113128559A (en) Remote sensing image target detection method based on cross-scale feature fusion pyramid network
Luo et al. Recognition and Extraction of Blue-roofed Houses in Remote Sensing Images based on Improved Mask-RCNN
CN117591674B (en) Automatic classification method for bridge inspection text based on text classification model
Ou et al. Floodet: A Lightweight Edge AI Model for Intelligent Flood Detection
EP4273502A1 (en) Method, apparatus, and computer program product for map geometry generation based on data aggregation and conflation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant