CN114782679A - Hardware defect detection method and device in power transmission line based on cascade network - Google Patents

Hardware defect detection method and device in power transmission line based on cascade network Download PDF

Info

Publication number
CN114782679A
CN114782679A CN202210480944.8A CN202210480944A CN114782679A CN 114782679 A CN114782679 A CN 114782679A CN 202210480944 A CN202210480944 A CN 202210480944A CN 114782679 A CN114782679 A CN 114782679A
Authority
CN
China
Prior art keywords
detection
network
hardware
image
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210480944.8A
Other languages
Chinese (zh)
Inventor
方登洲
严波
许义
陈龙庆
朱炳翔
周豪
张承习
孙建明
李方耀
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
State Grid Anhui Electric Power Co Ltd
Hefei Power Supply Co of State Grid Anhui Electric Power Co Ltd
Taizhou Power Supply Co of State Grid Jiangsu Electric Power Co Ltd
Original Assignee
State Grid Corp of China SGCC
State Grid Anhui Electric Power Co Ltd
Hefei Power Supply Co of State Grid Anhui Electric Power Co Ltd
Taizhou Power Supply Co of State Grid Jiangsu 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, State Grid Anhui Electric Power Co Ltd, Hefei Power Supply Co of State Grid Anhui Electric Power Co Ltd, Taizhou Power Supply Co of State Grid Jiangsu Electric Power Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN202210480944.8A priority Critical patent/CN114782679A/en
Publication of CN114782679A publication Critical patent/CN114782679A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

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

Abstract

The invention discloses a method and a device for detecting hardware defects in a power transmission line based on a cascade network, which are implemented by acquiring a target image of hardware; preprocessing the target image to obtain a test sample image; inputting the test sample image into a first detection model, and detecting the link fitting to obtain a detection result graph of the link fitting; cutting a mark detection part area of the detection result image to obtain a first cut image, and zooming the first cut image to obtain a first result image; and inputting the first result image into a second detection model to obtain a small hardware defect detection result. According to the invention, the cascade network is arranged, the improved YOLOv4 network model is used for detecting the link fitting, and the PVANet network model is used for detecting the small-size fitting defect of the link fitting, so that the phenomena of missing detection and wrong detection caused by the loss of the information of a small target are avoided, and the detection accuracy is greatly improved.

Description

Hardware defect detection method and device in power transmission line based on cascade network
Technical Field
The invention belongs to the field of target detection, and relates to a hardware defect detection method and device in a power transmission line based on a cascade network.
Background
In recent years, along with the fact that the construction speed of transmission lines in China is faster and faster, the lean management requirements of the transmission lines are higher and higher, and in order to ensure the stable operation of a power system, inspection personnel need to regularly inspect overhead transmission lines. Due to the factors of wide distribution of overhead transmission lines and complex and changeable landforms, the traditional manual inspection mode is greatly limited. Under the novel digital infrastructure construction strategy background of national grid company, intelligent inspection technology based on unmanned aerial vehicle is being extensively promoted the pilot point and is being used, and unmanned aerial vehicle carries imaging device and shoots transmission line shaft tower part, utilizes the thought of deep learning to patrol and examine the magnanimity picture that brings to unmanned aerial vehicle and carries out the defect analysis.
The induction and the arrangement of the defect samples shot by the unmanned aerial vehicle by the Chinese institute of Electrical science and technology are as follows: towers, ground wires, insulators, large-size fittings, small-size fittings, foundations, channel environments, grounding devices and accessory facilities 9. The defect type proportion of the small-size hardware reaches 62%, and the correct identification of the defects is particularly important. The small-size hardware defect mainly comprises split pin defect, split pin non-opening, split pin installation not in place, bolt looseness, split pin detachment, split pin defect, nut detachment, nut defect, bolt and nut corrosion and the like.
In the prior art, a defect analysis method based on deep learning has a good recognition rate for defects of large-volume power components such as insulators and towers, and for small-size hardware, when a deep neural network recognition model is used for processing the defects, information of small targets is lost due to convolution and down sampling, so that detection omission and false detection are caused. Therefore, the invention provides a method and a device for detecting hardware defects in a power transmission line based on a cascade network, and aims to solve the technical defects.
Disclosure of Invention
The invention mainly aims to provide a hardware defect detection method and device in a power transmission line based on a cascade network, and aims to solve the technical problems of missing detection and error detection caused by the fact that a detection method in the prior art easily loses information of a small target, so that the defect of the power transmission line is accurately detected.
In order to achieve the purpose, the method for detecting the hardware defect in the power transmission line based on the cascade network comprises the following steps:
s1, collecting a target image of the hardware; preprocessing the target image to obtain a test sample image;
s2, inputting the test sample image into a first detection model, detecting the link fitting, and obtaining a detection result diagram of the link fitting;
s3, cutting the mark detection part area of the detection result image to obtain a first cut image, and zooming the first cut image to obtain a first result image;
and S4, inputting the first result image to a second detection model to obtain a small hardware defect detection result.
Preferably, the first detection model is a modified YOLOv4 target detection network model; the second detection model is a PVANet network detection model;
the method further comprises the following steps: pre-training the YOLOv4 target detection network model and the PVANet network detection model;
the improved YOLOv4 target detection network model by using k-meansThe size of the labeling box is clustered by a + clustering algorithm, and the size of the prior box is obtained again; applying the obtained size of the prior frame to YOLOv4 detecting in the network.
Preferably, said pre-training said YOLOv4 target detection network model and PVANethe t-network detection model comprises the following steps:
s11, reading related images acquired by the unmanned aerial vehicle in advance, and carrying out preprocessing operation on the related images to obtain a training data set;
s12, randomly selecting a half of data set, and labeling the connection fitting position of the images in the data set to form a connection fitting part data set; randomly selecting a half of data sets, and labeling small hardware defect positions of the images to form small hardware defect part data sets;
s13, connecting goldWith partial data sets as YOLOv4, detecting the input of the network model; taking a small hardware fitting defect part data set as PVANet detecting the input of the network model;
s14, according to YOLOv4 and PVANethe t-network structure is used for inputting the selected training data into a network to perform one-time forward propagation calculation to obtain a predicted value;
s15, calculating the parameter gradient by using a back propagation algorithm through the predicted value, the real value and the network loss function, and updating the network parameters;
the network loss function includes YLOLv4 network loss function and PVANet network loss function;
the YLOLv4 network loss function is: l ═ Lcoord+Liou+Lclass(ii) a PVANet network loss function is L ═ Lcoord+Lclass；
Wherein L is the total loss function, LcoordFor the confidence coefficient loss function, whether object information exists is regressed; l isiouReturning the position information as a frame loss function; l isclassTo classify the loss function, class information is regressed.
S16, stopping training if the training target is reached; if not, judging whether the preset training times are reached, if so, stopping training, otherwise, repeating the steps S13 to S16.
Preferably, the preprocessing operation comprises: rotation, translation, sharpening, and/or gamut transformation.
Preferably, in S12, the labeled information includes position coordinates of the link fitting and a fitting type to which the link fitting belongs.
Preferably, after the step S16, the method further includes:
detecting a half data set without connection part labeling by the trained YOLOv4 network, outputting position coordinates and the class of the hardware fitting, and adding the position coordinates and the class of the hardware fitting into a partial data set of the connection hardware fitting after manual examination and adjustment to form a data set of the connection hardware fitting;
and detecting a half of data set without the small hardware defect labeling by the trained PVANet network, outputting position coordinates and the belonging defect types, and adding the position coordinates and the belonging defect types into the small hardware defect part data set after manual examination and adjustment to form a small hardware defect data set.
In addition, still provide a gold utensil defect detecting device in transmission line based on cascade network, include:
the acquisition module is used for acquiring a target image of the hardware; preprocessing the target image to obtain a test sample image;
the first detection module is used for inputting the test sample image into a first detection model, detecting the link fitting and obtaining a detection result graph of the link fitting;
the processing module is used for cutting the mark detection part area of the detection result image to obtain a first cutting image, and zooming the first cutting image to obtain a first result image;
and the second detection module is used for inputting the first result image to a second detection model to obtain a small hardware defect detection result.
Preferably, the first detection model is a modified YOLOv4 target detection network model; the second detection model is a PVANet network detection model;
the method further comprises the following steps: pre-training the YOLOv4 target detection network model and the PVANet network detection model;
clustering the size of the labeling box by the improved YOLOv4 target detection network model by using a k-means + + clustering algorithm to obtain the size of a prior box again; and applying the obtained size of the prior frame to a YOLOv4 detection network.
In addition, an electronic device is further provided, and includes a memory and a processor, where the memory stores a program, and the processor executes the program to implement the operation of the hardware defect detection method in the power transmission line based on the cascaded network.
In addition, a computer storage medium is also provided, and the computer storage medium stores a program, and the program is loaded and executed by a processor to implement the hardware defect detection method operation in the power transmission line based on the cascade network.
The method and the device for detecting the hardware defects in the power transmission line based on the cascade network, provided by the invention, are implemented by acquiring a target image of hardware; preprocessing the target image to obtain a test sample image; inputting the test sample image into a first detection model, and detecting the link fitting to obtain a detection result graph of the link fitting; cutting a mark detection part area of the detection result image to obtain a first cut image, and zooming the first cut image to obtain a first result image; and inputting the first result image into a second detection model to obtain a small hardware defect detection result. According to the invention, the cascade network is arranged, the improved YOLOv4 network model is used for detecting the link fitting, and the PVANet network model is used for detecting the small-size fitting defect of the link fitting, so that the phenomena of missing detection and wrong detection caused by losing the information of a small target can be avoided, and the detection accuracy is greatly improved.
Drawings
Fig. 1 is a flowchart of a hardware defect detection method in a power transmission line based on a cascade network according to an embodiment of the present invention;
FIG. 2 is a flow chart of a k-means clustering algorithm adopted for clustering a pair of link fitting labeling boxes in the embodiment of the invention;
fig. 3 is a flowchart of an embodiment of a method for detecting a small-size hardware defect target of a power transmission line based on a cascade network according to an embodiment of the present invention;
fig. 4 is a sample diagram of defect detection performed on a connection hardware according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a device for detecting a metal defect in a power transmission line based on a cascade network according to a second embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, it being understood that the preferred embodiments described herein are for purposes of illustration and explanation only and are not intended to be limiting of the present invention, and that the embodiments and features of the embodiments may be combined with each other without conflict.
Example one
Referring to fig. 1, fig. 1 is a flowchart of a hardware defect detection method in a power transmission line based on a cascade network according to the present invention, and in this embodiment, the hardware defect detection method in the power transmission line based on the cascade network includes the following steps:
s1, collecting a target image of the hardware; preprocessing the target image to obtain a test sample image;
preferably, the preprocessing operation comprises: rotation, translation, sharpening, and/or gamut transformation.
S2, inputting the test sample image into a first detection model, detecting the link fitting, and obtaining a detection result diagram of the link fitting;
s3, cutting the mark detection part area of the detection result image to obtain a first cutting image, and zooming the first cutting image to obtain a first result image;
specifically, the first detection model is a modified YOLOv4 target detection network model; inputting a test sample of an actual application scene into a trained and improved Yolov4 target detection network model to obtain a detection result diagram of the link fitting; cutting a detection result graph of the connecting hardware fitting, only reserving a network model marking part, removing a marking frame, and scaling the size of the marking frame, such as: the adjustment is 256 × 256.
The improved YOLOv4 target detection network model clusters the size of the labeling box by using a k-means + + clustering algorithm to obtain the size of the prior box again; and applying the obtained size of the prior frame to a YOLOv4 detection network.
And S4, inputting the first result image into a second detection model to obtain a small hardware defect detection result.
Preferably, the second detection model is a PVANet network detection model; the method further comprises the following steps: the YOLOv4 target detection network model and the PVANet network detection model are trained in advance.
Specifically, the first detection model and the second detection model are established cascade network target detectionAn algorithmic model, the object detection model comprising using modified YOLOv4 network model detection connection hardware fitting using PVANethe method comprises the following steps that a t network model detects small-size hardware defects of a connecting hardware; in the embodiment, by establishing a cascade network target detection algorithm model, as shown in fig. 2, k-m adopted for clustering link fitting labeling boxes in the embodiment is showneansClustering algorithm flow chart, this example, the improved YOLOv4 target detection network model by using k-meansThe size of the labeling box is clustered by a + clustering algorithm, and the size of the prior box is obtained again; applying the obtained prior frame size to YOLOv4 detecting in the network.
Preferably, said pre-training said YOLOv4 target detection network model and PVANethe t-network detection model comprises the following steps:
s11, reading related images acquired by the unmanned aerial vehicle in advance, and carrying out preprocessing operation on the related images to obtain a training data set;
s12, randomly selecting a half of data set, and labeling the connection fitting position of the images in the data set to form a connection fitting part data set; randomly selecting a half of data sets, and labeling the small hardware defect part of the image to form a small hardware defect part data set;
s13, using the link fitting part data set as YOLOv4, detecting the input of the network model; taking a small hardware fitting defect part data set as PVANet detecting the input of the network model;
s14, according to YOLOv4 and PVANethe t-network structure is used for inputting the selected training data into a network to perform one-time forward propagation calculation to obtain a predicted value;
s15, calculating the parameter gradient by using a back propagation algorithm through the predicted value, the real value and the network loss function, and updating the network parameters;
wherein, the YLOLv4 network loss function is: l ═ Lcoord+Liou+Lclass(ii) a PVANet network loss function is L ═ Lcoord+Lclass；
Wherein L is the total loss function, LcoordFor the confidence coefficient loss function, whether object information exists in the regression is judged; l isiouReturning the position information for the frame loss function; l is a radical of an alcoholclassRegressing category information for a classification loss function;
s16, if the training target is reached, stopping training; if not, judging whether the preset training times are reached, if so, stopping training, otherwise, repeating the steps S13 to S16.
Specifically, the achievement of the training target means that the loss value does not have a descending trend in the process of continuous rounds of training, such as: in the five continuous training processes, the loss value corresponding to the loss function has no descending trend.
Fig. 3 is a flowchart of an embodiment of the method for detecting the defect target of the small-size hardware of the power transmission line based on the cascade network according to the present embodiment.
Preferably, in S12, the labeled information includes position coordinates of the link fitting and a fitting type to which the link fitting belongs.
Preferably, after the step S16, the method further includes:
will train good YOLOv4, detecting a half data set without connection part labeling by the network, outputting position coordinates and the class of the hardware fitting, and adding the position coordinates and the class of the hardware fitting into a partial data set of the connection hardware fitting after manual examination and adjustment to form a data set of the connection hardware fitting;
will train the PVANeAnd the t network detects a half data set without small hardware defect labeling, outputs position coordinates and the defect types, and adds the position coordinates and the defect types into the data set of the small hardware defects after manual auditing and adjustment to form the small hardware defect data set.
In addition, still provide a gold utensil defect detecting device in transmission line based on cascade network, include:
the acquisition module is used for acquiring a target image of the hardware; preprocessing the target image to obtain a test sample image;
the first detection module is used for inputting the test sample image into a first detection model, detecting the link fitting and obtaining a detection result graph of the link fitting;
the processing module is used for cutting the mark detection part area of the detection result image to obtain a first cutting image, and zooming the first cutting image to obtain a first result image;
and the second detection module is used for inputting the first result image to a second detection model to obtain a small hardware defect detection result.
Preferably, the preprocessing operation comprises: rotation, translation, sharpening, and/or gamut transformation.
Preferably, the first detection model is a modified YOLOv4 target detection network model; the second detection model is a PVANet network detection model;
the method further comprises the following steps: and pre-training the YOLOv4 target detection network model and the PVANet network detection model.
Clustering the size of the labeling box by the improved YOLOv4 target detection network model by using a k-means + + clustering algorithm to obtain the size of a prior box again; and applying the obtained size of the prior frame to a YOLOv4 detection network.
As shown in fig. 4, it is a sample diagram of the defect detection performed on the connection hardware in this embodiment; wherein, fig. 4(a) is a detection sample image; fig. 4(b) is a detection result image of the link, in which the label name LJJU is the link; fig. 4(c) is an image obtained by cutting and resizing the detection result of one of the link fittings to 256 × 256; fig. 4(d) is a defect result image obtained by identifying the cut link fitting image; wherein, the label name LSQXZ is the bolt missing pin.
Example two
The invention also provides a device for detecting the hardware defects in the power transmission line based on the cascade network, which comprises the following components:
the acquisition module is used for acquiring a target image of the hardware; preprocessing the target image to obtain a test sample image;
the first detection module inputs the test sample image into a first detection model, and detects the connecting hardware to obtain a detection result graph of the connecting hardware;
the processing module is used for cutting the mark detection part area of the detection result image to obtain a first cutting image, and zooming the first cutting image to obtain a first result image;
specifically, the first detection model is a modified YOLOv4 target detection network model; inputting a test sample of an actual application scene into a trained and improved Yolov4 target detection network model to obtain a detection result diagram of the link fitting; cutting a detection result graph of the connecting hardware fitting, only reserving a network model marking part, removing a marking frame, and scaling the size of the marking frame, such as: adjusted to 256 x 256. The improved YOLOv4 target detection network model clusters the size of the labeling box by using a k-means + + clustering algorithm to obtain the size of the prior box again; and applying the obtained size of the prior frame to a YOLOv4 detection network.
And the second detection module is used for inputting the first result image to a second detection model to obtain a small hardware defect detection result.
Preferably, the preprocessing operation comprises: rotation, translation, sharpening, and/or color gamut transformation.
Preferably, the second detection model is a PVANet network detection model; the method further comprises the following steps: the YOLOv4 target detection network model and the PVANet network detection model are trained in advance.
Specifically, the first detection model and the second detection model are established cascade network target detection algorithm models, the target detection models comprise hardware fittings detected by using an improved YOLOv4 network model, and small-size hardware fitting defects of the hardware fittings are detected by using a PVANet network model; the established cascade network target detection algorithm model comprises the following steps of 1: clustering the size of the labeling frame by using a k-means + + clustering algorithm to obtain the size of the prior frame again; and 2, step: and applying the obtained size of the prior frame to a YOLOv4 detection network.
In the scheme of the embodiment of the invention, the target image of the hardware is acquired; preprocessing the target image to obtain a test sample image; inputting the test sample image into a first detection model, and detecting the link fitting to obtain a detection result graph of the link fitting; cutting a mark detection part area of the detection result image to obtain a first cut image, and zooming the first cut image to obtain a first result image; and inputting the first result image into a second detection model to obtain a small hardware defect detection result. According to the invention, the cascade network is arranged, the improved YOLOv4 network model is used for detecting the link fitting, and the PVANet network model is used for detecting the small-size fitting defect of the link fitting, so that the phenomena of missing detection and wrong detection caused by losing the information of a small target can be avoided, and the detection accuracy is greatly improved.
It should be noted that the division of each module of the above apparatus is only a logical division, and all or part of the actual implementation may be integrated into one physical entity or may be physically separated. And these modules may all be implemented in the form of software calls by the processing element. Or may be implemented entirely in hardware. And part of the modules can be realized in the form of calling software by the processing element, and part of the modules can be realized in the form of hardware. In addition, an electronic device is also provided, comprising a memory and a processor; the memory stores a program, and the processor executes the program to realize the operation of the hardware defect detection method in the power transmission line based on the cascade network.
In addition, a computer storage medium is also provided, and the computer storage medium stores a program, and the program is loaded and executed by a processor to implement the hardware defect detection method operation in the power transmission line based on the cascade network.
Having thus described the basic concept, it will be apparent to those skilled in the art that the foregoing detailed disclosure is to be regarded as illustrative only and not as limiting the present specification. Various modifications, improvements, and offset processing may occur to those skilled in the art, though not expressly stated herein. Such modifications, improvements, and offset processing are suggested in this specification and still fall within the spirit and scope of the exemplary embodiments of this specification.
Moreover, those skilled in the art will appreciate that aspects of the present description may be illustrated and described in terms of several patentable species or situations, including any new and useful combination of processes, machines, manufacture, or materials, or any new and useful modification thereof. Accordingly, aspects of this description may be performed entirely by hardware, entirely by software (including firmware, resident software, micro-code, etc.) or by a combination of hardware and software. The above hardware or software may be referred to as "data block," module, "" engine, "" unit, "" component, "or" system. Furthermore, aspects of the present description may be represented as a computer product, including computer readable program code, embodied in one or more computer readable media.
It is to be understood that the descriptions, definitions and/or uses of terms in the accompanying materials of this specification shall control if they are inconsistent or contrary to the descriptions and/or uses of terms in this specification.
Finally, it should be understood that the embodiments described herein are merely illustrative of the principles of the embodiments of the present disclosure. Other variations are also possible within the scope of the present description. Thus, by way of example, and not limitation, alternative configurations of the embodiments of the specification can be considered consistent with the teachings of the specification. Accordingly, the embodiments of the present description are not limited to only those embodiments explicitly described and depicted herein.

Claims (11)

1. A hardware defect detection method in a power transmission line based on a cascade network is characterized by comprising the following steps:
s1, collecting a target image of the hardware; preprocessing the target image to obtain a test sample image;
s2, inputting the test sample image into a first detection model, and detecting the link fitting to obtain a detection result diagram of the link fitting;
s3, cutting the mark detection part area of the detection result image to obtain a first cut image, and zooming the first cut image to obtain a first result image;
and S4, inputting the first result image into a second detection model to obtain a small hardware defect detection result.
2. The method for detecting the hardware defects in the power transmission line based on the cascade network as claimed in claim 1, wherein the first detection model is an improved YOLOv4 target detection network model; the second detection model is a PVANet network detection model;
the method further comprises the following steps: pre-training the YOLOv4 target detection network model and the PVANet network detection model;
clustering the size of the labeling box by the improved YOLOv4 target detection network model by using a k-means + + clustering algorithm to obtain the size of a prior box again; and applying the obtained size of the prior frame to a YOLOv4 detection network.
3. The method for detecting the hardware defects in the power transmission line based on the cascade network as claimed in claim 2, wherein the pre-training of the YOLOv4 target detection network model and the PVANet network detection model comprises the following steps:
s11, reading related images acquired by the unmanned aerial vehicle in advance, and carrying out preprocessing operation on the related images to obtain a training data set;
s12, randomly selecting a half of data set, and labeling the connection fitting position of the images in the data set to form a connection fitting part data set; randomly selecting a half of data sets, and labeling small hardware defect positions of the images to form small hardware defect part data sets;
s13, detecting the input of the network model from the link fitting part data set as YOLOv 4; taking the small hardware defect part data set as input of a PVANet detection network model;
s14, according to the YOLOv4 and the PVANet network structure, inputting the selected training data into the network to perform one-time forward propagation calculation to obtain a predicted value;
s15, calculating the parameter gradient by using a back propagation algorithm through the predicted value, the real value and the network loss function, and updating the network parameters;
s16, stopping training if the training target is reached; if not, judging whether the preset training times are reached, if so, stopping training, otherwise, repeating the steps S13 to S16.
4. The method for detecting the hardware defects in the power transmission line based on the cascade network as claimed in claim 3, wherein the network loss function includes a YLOLv4 network loss function and a PVANet network loss function;
the YLOLv4 network loss function is: l ═ Lcoord+Liou+Lclass(ii) a PVANet network loss function is L ═ Lcoord+Lclass；
Wherein L is the total loss function, LcoordFor the confidence coefficient loss function, whether object information exists is regressed; l isiouReturning the position information for the frame loss function; l isclassTo classify the loss function, the class information is regressed.
5. The method for detecting the hardware defects in the power transmission line based on the cascade network as claimed in claim 1, wherein the preprocessing operation includes: rotation, translation, sharpening, and/or color gamut transformation.
6. The method according to claim 3 or 4, wherein in S12, the labeled information includes position coordinates of the link hardware and the hardware category to which the link hardware belongs.
7. The method for detecting the hardware defect in the power transmission line based on the cascade network according to claim 3 or 4, wherein after the step S16, the method further comprises:
detecting a half data set without connection part labeling by the trained YOLOv4 network, outputting position coordinates and the class of the hardware fitting, and adding the position coordinates and the class of the hardware fitting into a partial data set of the connection hardware fitting after manual examination and adjustment to form a data set of the connection hardware fitting;
and detecting a half of data set without the small hardware defect labeling by the trained PVANet network, outputting position coordinates and the belonging defect types, and adding the position coordinates and the belonging defect types into the small hardware defect part data set after manual examination and adjustment to form a small hardware defect data set.
8. The utility model provides a gold utensil defect detecting device in transmission line based on cascade network which characterized in that includes:
the acquisition module is used for acquiring a target image of the hardware; preprocessing the target image to obtain a test sample image;
the first detection module is used for inputting the test sample image into a first detection model, detecting the link fitting and obtaining a detection result graph of the link fitting;
the processing module is used for cutting the mark detection part area of the detection result image to obtain a first cutting image, and zooming the first cutting image to obtain a first result image;
and the second detection module inputs the first result image into a second detection model to obtain a small hardware defect detection result.
9. The device for detecting the hardware defects in the power transmission line based on the cascade network as claimed in claim 8, wherein the first detection model is an improved YOLOv4 target detection network model; the second detection model is a PVANet network detection model.
The method further comprises the following steps: pre-training the YOLOv4 target detection network model and the PVANet network detection model;
the improved YOLOv4 target detection network model clusters the size of the labeling box by using a k-means + + clustering algorithm to obtain the size of the prior box again; and applying the obtained size of the prior frame to a YOLOv4 detection network.
10. An electronic device comprising a memory and a processor, the memory having a program stored thereon; the processor executes the program to realize the operation of the hardware defect detection method in the power transmission line based on the cascade network according to any one of claims 1 to 7.
11. A computer storage medium storing a program; the program is loaded and executed by a processor to realize the operation of the hardware defect detection method in the power transmission line based on the cascade network according to any one of the claims 1-7.
CN202210480944.8A 2022-05-05 2022-05-05 Hardware defect detection method and device in power transmission line based on cascade network Pending CN114782679A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210480944.8A CN114782679A (en) 2022-05-05 2022-05-05 Hardware defect detection method and device in power transmission line based on cascade network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210480944.8A CN114782679A (en) 2022-05-05 2022-05-05 Hardware defect detection method and device in power transmission line based on cascade network

Publications (1)

Publication Number Publication Date
CN114782679A true CN114782679A (en) 2022-07-22

Family

ID=82435915

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210480944.8A Pending CN114782679A (en) 2022-05-05 2022-05-05 Hardware defect detection method and device in power transmission line based on cascade network

Country Status (1)

Country Link
CN (1) CN114782679A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117422689A (en) * 2023-10-31 2024-01-19 南京邮电大学 Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7
CN117422689B (en) * 2023-10-31 2024-05-31 南京邮电大学 Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117422689A (en) * 2023-10-31 2024-01-19 南京邮电大学 Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7
CN117422689B (en) * 2023-10-31 2024-05-31 南京邮电大学 Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7

Similar Documents

Publication Publication Date Title
CN109948469B (en) Automatic inspection robot instrument detection and identification method based on deep learning
CN114581855B (en) Information collection method and system based on big data
CN111754498A (en) Conveyor belt carrier roller detection method based on YOLOv3
CN108830332A (en) A kind of vision vehicle checking method and system
CN110175519B (en) Method and device for identifying separation and combination identification instrument of transformer substation and storage medium
CN107833211B (en) Infrared image-based zero value insulator automatic detection method and device
CN110533654A (en) The method for detecting abnormality and device of components
CN112541372B (en) Difficult sample screening method and device
CN114463296B (en) Light-weight part defect detection method based on single sample learning
CN109003271A (en) A kind of Wiring harness connector winding displacement quality determining method based on deep learning YOLO algorithm
CN111598913B (en) Image segmentation method and system based on robot vision
Wang et al. Pointer meter recognition in UAV inspection of overhead transmission lines
CN115019294A (en) Pointer instrument reading identification method and system
CN110728269A (en) High-speed rail contact net support pole number plate identification method
CN112884795A (en) Power transmission line inspection foreground and background segmentation method based on multi-feature significance fusion
CN111931721B (en) Method and device for detecting color and number of annual inspection label and electronic equipment
CN113408630A (en) Transformer substation indicator lamp state identification method
CN114782679A (en) Hardware defect detection method and device in power transmission line based on cascade network
CN109359637B (en) Pointer instrument value reading method based on machine vision
CN112101107B (en) Intelligent recognition method for intelligent network connection model vehicle on-loop simulation traffic signal lamp
CN113591548B (en) Target ring identification method and system
CN115147591A (en) Transformer equipment infrared image voltage heating type defect diagnosis method and system
CN112132088B (en) Inspection point missing inspection identification method
CN111626323A (en) Electric meter state detection and evaluation method based on deep learning
CN113822148B (en) Intelligent identification method for trace tiny carryover based on convolutional neural network

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