CN114596278A - Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station - Google Patents

Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station Download PDF

Info

Publication number
CN114596278A
CN114596278A CN202210213418.5A CN202210213418A CN114596278A CN 114596278 A CN114596278 A CN 114596278A CN 202210213418 A CN202210213418 A CN 202210213418A CN 114596278 A CN114596278 A CN 114596278A
Authority
CN
China
Prior art keywords
photovoltaic panel
infrared image
hot spot
photovoltaic
low
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
CN202210213418.5A
Other languages
Chinese (zh)
Inventor
王方政
刘喜泉
李鹏
孙勇
张亚平
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.)
China Three Gorges Corp
Original Assignee
China Three Gorges Corp
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 China Three Gorges Corp filed Critical China Three Gorges Corp
Priority to CN202210213418.5A priority Critical patent/CN114596278A/en
Publication of CN114596278A publication Critical patent/CN114596278A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0004Industrial image inspection
    • G06T7/0006Industrial image inspection using a design-rule based approach
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10004Still image; Photographic image
    • YGENERAL 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
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02EREDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00Energy generation through renewable energy sources
    • Y02E10/50Photovoltaic [PV] energy

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The invention discloses a method and a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, wherein the method comprises the steps of obtaining a low-altitude infrared image of the photovoltaic panel; inputting the photovoltaic panel low-altitude infrared image into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the photovoltaic panel infrared image; reading a color mask in a binarization mode based on OpenCV, and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background; the infrared image of the segmented photovoltaic panel is input into a pre-constructed YOLOv5s photovoltaic panel hot spot defect detection model, the photovoltaic panel hot spot defect detection is carried out, and a detection result is obtained.

Description

Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station
Technical Field
The invention relates to a method and a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, and belongs to the technical field of photovoltaic panel detection.
Background
Low carbon emissions have long been a common recognition. As a clean energy, the photovoltaic power generation greatly reduces the carbon emission, and in addition, the photovoltaic power generation also has the advantages of easy equipment construction, low cost, high energy quality and the like. China started late in the photovoltaic field, but benefited from the vast territorial area in China and the abundant solar energy resources in the west, the China photovoltaic industry developed very rapidly, and the China photovoltaic packaging machine volume is the first in the world at present and is still developing rapidly.
The photovoltaic loading capacity is continuously improved, and meanwhile, some problems are brought, most of Chinese photovoltaic power stations are far away and unattended, if some photovoltaic faults are not found in time, accidents such as fire disasters can be caused, and therefore safety of the power stations is threatened. The traditional photovoltaic panel defect detection method mainly comprises a manual detection method and a photovoltaic panel electrical parameter measurement method, and the two methods mainly have the defects of low exposure efficiency and high cost.
Later, with the rapid development of image processing technology, deep learning technology and civil unmanned aerial vehicle industry, practitioners begin to use unmanned aerial vehicles to shoot infrared images of photovoltaic panels and use a deep learning image detection method to detect defects of the photovoltaic panels. However, most target detection networks have unsatisfactory detection effects, whether they are single-stage or two-stage detection networks. This is because the background in the infrared image is often complex, and more than one target of the photovoltaic panel may include various objects such as trees, weeds, rivers, animals, etc., and these heat sources may greatly affect the performance of the detection network.
Therefore, the hot spot defect detection method for the photovoltaic panel of the photovoltaic power station can eliminate the influence of an interference heat source, and has great significance for improving the defect detection precision.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a method and a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, which can eliminate the interference of a background heat source in an infrared image of the photovoltaic panel and greatly improve the hot spot defect detection precision of the photovoltaic panel.
In order to achieve the purpose, the invention is realized by adopting the following technical scheme:
in a first aspect, the present invention provides a method and an apparatus for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, comprising:
acquiring a low-altitude infrared image of the photovoltaic panel;
inputting the photovoltaic panel low-altitude infrared image into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the photovoltaic panel infrared image;
reading a color mask in a binarization mode based on OpenCV, and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background;
and inputting the divided infrared image of the photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model, and detecting the hot spot defects of the photovoltaic panel to obtain a detection result.
Further, the method for constructing the Deeplabv3+ photovoltaic panel semantic segmentation model and the YOLOv5s photovoltaic panel hot spot defect detection model comprises the following steps:
acquiring a low-altitude infrared image of the photovoltaic panel, wherein the low-altitude infrared image comprises a photovoltaic panel image and a hot spot defect image;
training a Deeplabv3+ network by using the photovoltaic panel image to obtain a Deeplabv3+ photovoltaic panel semantic segmentation model;
and training a YOLOv5s network by using the hot spot defect image to obtain a YOLOv5s photovoltaic panel hot spot defect detection model.
Further, the training of the deplabv 3+ network by using the photovoltaic panel image to obtain a deplabv 3+ photovoltaic panel semantic segmentation model includes:
carrying out pixel-level labeling on the low-altitude infrared image by using a Labelme tool to manufacture a photovoltaic panel infrared image data set;
and building a Deeplabv3+ semantic segmentation network based on a deep learning framework, and training the network by using the infrared image data set of the photovoltaic panel to obtain a Deeplabv3+ semantic segmentation model of the photovoltaic panel.
Further, the training of the YOLOv5s network by using the hot spot defect image to obtain a YOLOv5s photovoltaic panel hot spot defect detection model includes:
inputting the low-altitude infrared image into the Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the photovoltaic panel infrared image;
reading a color mask in a binarization mode based on OpenCV, and adding the color mask and an original infrared image pixel by pixel to obtain a segmented photovoltaic panel infrared image with a pure black background;
labeling the hot spot defect area in the segmented infrared image of the photovoltaic panel by using a LabelImg tool to manufacture a hot spot defect image data set of the photovoltaic panel;
and constructing a YOLOv5s target detection network based on a deep learning framework, and training the network by using the photovoltaic panel hot spot defect image data set to obtain a YOLOv5s photovoltaic panel hot spot defect detection model.
Further, the step of inputting the low-altitude infrared image of the photovoltaic panel into a pre-constructed deep pabv 3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the infrared image of the photovoltaic panel includes:
sending a photovoltaic panel infrared image into a Deeplabv3+ photovoltaic panel semantic segmentation model;
obtaining a low-level feature map by passing the infrared image of the photovoltaic panel through a feature extraction network;
further realizing multi-scale feature extraction on the low-level feature map through a void space convolution pooling pyramid layer to obtain a high-level feature map;
the high-level feature map is subjected to 4-time bilinear interpolation upsampling processing and then fused with the low-level feature map, and the high-level feature map is restored to the original image resolution after being subjected to 4-time bilinear interpolation upsampling processing;
and (4) passing the processed characteristic diagram through a Softmax classification function layer to obtain a finally corresponding color mask.
Furthermore, for the photovoltaic panel with hot spot defects, the longitude and latitude information of the defective photovoltaic panel is calculated by utilizing the positioning and attitude determination data carried by the infrared image, so that positioning and alarming are realized.
Further, the step of inputting the infrared image of the segmented photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model for photovoltaic panel hot spot defect detection includes:
sending the infrared image of the divided photovoltaic panel into a YOLOv5s photovoltaic panel hot spot defect detection model;
extracting image characteristic information of the infrared image of the photovoltaic panel under different image fine granularities through a backbone network;
sending feature graphs output by three different stages of a backbone network into a neck network with a path aggregation network structure for multi-scale feature fusion;
sending the three fused feature graphs into three prediction head networks for prediction frame regression and category regression;
and screening the prediction box obtained in the last step by adopting a non-maximum inhibition algorithm, returning to the optimal prediction box, and completing the final hot spot defect detection of the photovoltaic panel.
In a second aspect, the present invention provides a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, comprising:
the acquisition unit is used for acquiring a low-altitude infrared image of the photovoltaic panel;
the color mask obtaining unit is used for inputting the low-altitude infrared image of the photovoltaic panel into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the infrared image of the photovoltaic panel;
the segmentation unit is used for reading a color mask in a binarization mode based on OpenCV and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background;
and the detection unit is used for inputting the divided infrared image of the photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model, detecting the hot spot defects of the photovoltaic panel and acquiring a detection result.
In a third aspect, the present invention provides a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to any one of the preceding claims.
In a fourth aspect, the invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method of any one of the preceding claims.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention provides a method and a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, which are used for solving the problem of hot spot defect detection of the photovoltaic panel, adopts a two-step method of firstly segmenting and then detecting, effectively reduces the influence of an environment interference heat source on the performance of a neural network, and effectively improves the detection precision;
2. the invention provides a hot spot defect detection method and a hot spot defect detection device for a photovoltaic panel of a photovoltaic power station.
Drawings
FIG. 1 is a flow chart of a method and an apparatus for detecting hot spot defects of a photovoltaic panel of a photovoltaic power plant according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a Deeplabv3+ photovoltaic panel semantic segmentation model;
FIG. 3 is a schematic diagram illustrating addition of an infrared original image of a photovoltaic panel and a corresponding pixel of a corresponding mask;
fig. 4 is a schematic diagram of a model for detecting hot spot defects of a YOLOv5s photovoltaic panel.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
Example 1
The embodiment introduces a method and a device for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station, which comprises the following steps:
acquiring a low-altitude infrared image of the photovoltaic panel;
inputting the photovoltaic panel low-altitude infrared image into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the photovoltaic panel infrared image;
reading a color mask in a binarization mode based on OpenCV, and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background;
and inputting the divided infrared image of the photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model, and detecting the hot spot defects of the photovoltaic panel to obtain a detection result.
As shown in fig. 1, the method and the apparatus for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station provided in this embodiment specifically involve the following steps:
(1) and collecting low-altitude infrared images of the photovoltaic panel. The collection mode includes that the unmanned aerial vehicle who utilizes to carry on infrared camera shoots by oneself and the web crawler mode obtains, collects photovoltaic panel low latitude infrared image 1306, wherein has 771 images to contain the hot spot defect in this application embodiment.
(2) The method comprises the following steps of training a Deeplabv3+ network by using a photovoltaic panel image to obtain a Deeplabv3+ photovoltaic panel semantic segmentation model, wherein a model network frame structure schematic diagram is shown in FIG. 2, and the specific process comprises the following steps:
(21) carrying out pixel-level labeling on the infrared image collected in the step (1) by utilizing a Labelme tool to manufacture a photovoltaic panel infrared image data set, compiling a Python script, and randomly extracting 80% of pictures from the data set to serve as a training set, and taking the rest 20% of pictures as a test set;
(22) and (3) building a Deeplabv3+ semantic segmentation network based on a Pythrch deep learning framework, and training the network by using the photovoltaic panel infrared image data set obtained in the step (21) to obtain a Deeplabv3+ photovoltaic panel semantic segmentation model. In this embodiment, the graphics processor of the server selects geoForce 1080ti of great, the operating system is Ubuntu16.04, the ResNet50 network is selected as the feature extraction network of Deeplabv3+, the model parameters are learned by adopting the SGD algorithm, the learning rate is set to 0.001, the network momentum parameter is set to 0.9, the weight attenuation is set to 0.0001, the batch size is set to 16, the training in the first stage is set to 4000 batches, and the training in the second stage is set to 6000 batches.
(3) Training a YOLOv5s network by using the hot spot defect image to obtain a YOLOv5s photovoltaic panel hot spot defect detection model, wherein the structural schematic diagram of the model network frame is shown in FIG. 4, and the specific process comprises the following steps:
(31) sending the photovoltaic panel infrared image collected in the step (1) into the Deeplabv3+ photovoltaic panel semantic segmentation model obtained in the step (2) to obtain a color mask corresponding to the photovoltaic panel infrared image;
(32) reading a color mask in a binarization mode based on OpenCV, and adding the color mask and an original infrared image pixel by pixel to obtain a photovoltaic panel infrared image with a pure black background, wherein a schematic diagram of the process is shown in FIG. 3;
(33) labeling the hot spot defect area in the segmented photovoltaic panel infrared image obtained in the step (32) by using a LabelImg tool to manufacture a photovoltaic panel hot spot defect image data set, compiling a Python script, and randomly extracting 80% of pictures from the data set to serve as a training set, and taking the rest 20% of pictures as a test set;
(34) building a YOLOv5s target detection network based on a Pythroch deep learning framework, training the network with the photovoltaic panel hot spot defect image data set obtained in the step (33) to obtain a YOLOv5s photovoltaic panel hot spot defect detection model, in the embodiment, selecting a CSPDarkenet 53 network as a feature extraction network of YOLOv5s, learning model parameters by adopting an SGD algorithm, setting the learning rate to be 0.001, setting the network momentum parameter to be 0.9, setting the weight attenuation to be 0.0001, setting the batch size to be 16, setting the first-stage training to be 8000 batches, and setting the second-stage training to be 12000 batches. (ii) a
(4) The unmanned aerial vehicle collects the low-altitude infrared images of the photovoltaic panel. The unmanned aerial vehicle used in the embodiment is a 2-step advanced version aircraft in Xinjiang imperial, 640 x 512 thermal imaging cameras are mounted, the shooting time is two points or four points in the afternoon in autumn with fine weather, and the shooting height is 28 m.
(5) The photovoltaic panel in the infrared image is divided, and the method specifically comprises the following steps:
(51) sending the photovoltaic panel low-altitude infrared image collected by the unmanned aerial vehicle in the step (4) into the Deeplabv3+ photovoltaic panel semantic segmentation model obtained in the step (2) to obtain a color mask corresponding to the photovoltaic panel infrared image, wherein the process specifically comprises the following steps:
(511) sending a photovoltaic panel infrared image into a Deeplabv3+ photovoltaic panel semantic segmentation model;
(512) obtaining a low-level feature map after the photovoltaic panel infrared image passes through a feature extraction network;
(513) the low-level feature map obtained in the last step further realizes multi-scale feature extraction through a void space convolution pooling pyramid layer to obtain a high-level feature map;
(514) the high-level feature map obtained in the last step is fused with the low-level feature map obtained in the step (512) after being subjected to 4-time bilinear interpolation upsampling processing, and then the high-level feature map is restored to the original image resolution after being subjected to 4-time bilinear interpolation upsampling processing;
(515) the feature map obtained in the last step passes through a Softmax classification function layer to obtain a final corresponding color mask;
(52) reading a color mask in a binarization mode based on OpenCV, and adding the color mask and an original infrared image pixel by pixel to obtain a segmented photovoltaic panel infrared image with a pure black background;
(6) photovoltaic panel hot spot defect detection and location specifically include:
(61) sending the infrared image of the photovoltaic panel obtained by segmentation in the step (52) into the YOLOv5s photovoltaic panel hot spot defect detection model obtained in the step (3) for photovoltaic panel hot spot defect detection, wherein the process specifically comprises the following steps:
(611) sending the infrared image of the divided photovoltaic panel into a YOLOv5s photovoltaic panel hot spot defect detection model;
(612) extracting image characteristic information of the infrared image of the photovoltaic panel under different image fine granularities through a backbone network;
(613) sending feature graphs output by three different stages of a backbone network into a neck network with a path aggregation network structure for multi-scale feature fusion;
(614) sending the three characteristic graphs obtained in the last step into three prediction head networks for prediction frame regression and category regression;
(615) screening the prediction box obtained in the last step by adopting a non-maximum inhibition algorithm, returning to the optimal prediction box, and completing the final hot spot defect detection of the photovoltaic panel;
(62) and for the photovoltaic panel with hot spot defects, calculating the longitude and latitude information of the defective photovoltaic panel by using positioning and attitude determination data carried by the infrared image, and realizing positioning and alarming.
Example 2
The embodiment provides a photovoltaic panel hot spot defect detection device for photovoltaic power plant, includes:
the acquisition unit is used for acquiring a low-altitude infrared image of the photovoltaic panel;
the color mask obtaining unit is used for inputting the low-altitude infrared image of the photovoltaic panel into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the infrared image of the photovoltaic panel;
the segmentation unit is used for reading a color mask in a binarization mode based on OpenCV and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background;
and the detection unit is used for inputting the divided infrared image of the photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model, detecting the hot spot defects of the photovoltaic panel and acquiring a detection result.
Example 3
The embodiment provides a hot spot defect detection device for a photovoltaic panel of a photovoltaic power station, which comprises a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to any of embodiment 1.
Example 4
The present embodiment provides a computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the steps of the method of any of the embodiment 1.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.

Claims (10)

1. A method for detecting hot spot defects of a photovoltaic panel of a photovoltaic power station is characterized by comprising the following steps:
acquiring a low-altitude infrared image of the photovoltaic panel;
inputting the low-altitude infrared image of the photovoltaic panel into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the infrared image of the photovoltaic panel;
reading a color mask in a binarization mode based on OpenCV, and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background;
and inputting the infrared image of the divided photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model, and detecting the hot spot defects of the photovoltaic panel to obtain a detection result.
2. The method of claim 1 for detecting hot spot defects on photovoltaic panels of photovoltaic power plants, characterized in that: the method for constructing the Deeplabv3+ photovoltaic panel semantic segmentation model and the YOLOv5s photovoltaic panel hot spot defect detection model comprises the following steps:
acquiring a low-altitude infrared image of the photovoltaic panel, wherein the low-altitude infrared image comprises a photovoltaic panel image and a hot spot defect image;
training a Deeplabv3+ network by using the photovoltaic panel image to obtain a Deeplabv3+ photovoltaic panel semantic segmentation model;
and training a YOLOv5s network by using the hot spot defect image to obtain a YOLOv5s photovoltaic panel hot spot defect detection model.
3. The method of claim 1 for detecting hot spot defects on photovoltaic panels of photovoltaic power plants, characterized in that: the method for training a Deeplabv3+ network by using the photovoltaic panel image to obtain a Deeplabv3+ photovoltaic panel semantic segmentation model comprises the following steps:
carrying out pixel-level labeling on the low-altitude infrared image by using a Labelme tool to manufacture a photovoltaic panel infrared image data set;
and building a Deeplabv3+ semantic segmentation network based on a deep learning framework, and training the network by using the infrared image data set of the photovoltaic panel to obtain a Deeplabv3+ semantic segmentation model of the photovoltaic panel.
4. The method of claim 1 for detecting hot spot defects on photovoltaic panels of photovoltaic power plants, characterized in that: the method for training a YOLOv5s network by using the hot spot defect image to obtain a YOLOv5s photovoltaic panel hot spot defect detection model comprises the following steps:
inputting the low-altitude infrared image into the Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the photovoltaic panel infrared image;
reading a color mask in a binarization mode based on OpenCV, and adding the color mask and an original infrared image pixel by pixel to obtain a segmented photovoltaic panel infrared image with a pure black background;
labeling the hot spot defect area in the segmented infrared image of the photovoltaic panel by using a LabelImg tool to manufacture a hot spot defect image data set of the photovoltaic panel;
and constructing a YOLOv5s target detection network based on a deep learning framework, and training the network by using the photovoltaic panel hot spot defect image data set to obtain a YOLOv5s photovoltaic panel hot spot defect detection model.
5. The method of claim 1 for detecting hot spot defects on photovoltaic panels of photovoltaic power plants, characterized in that: the method for inputting the photovoltaic panel low-altitude infrared image into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the photovoltaic panel infrared image comprises the following steps:
sending a photovoltaic panel infrared image into a Deeplabv3+ photovoltaic panel semantic segmentation model;
obtaining a low-level feature map by passing the infrared image of the photovoltaic panel through a feature extraction network;
further realizing multi-scale feature extraction on the low-level feature map through a void space convolution pooling pyramid layer to obtain a high-level feature map;
the high-level feature map is subjected to 4-time bilinear interpolation upsampling processing and then fused with the low-level feature map, and the high-level feature map is restored to the original image resolution after being subjected to 4-time bilinear interpolation upsampling processing;
and (4) passing the processed characteristic diagram through a Softmax classification function layer to obtain a finally corresponding color mask.
6. The method of claim 1 for detecting hot spot defects on photovoltaic panels of photovoltaic power plants, characterized in that: and for the photovoltaic panel with hot spot defects, calculating the longitude and latitude information of the defective photovoltaic panel by using positioning and attitude determination data carried by the infrared image, and realizing positioning and alarming.
7. The method of claim 1 for detecting hot spot defects on photovoltaic panels of photovoltaic power plants, characterized in that: inputting the segmented infrared image of the photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model for detecting the hot spot defects of the photovoltaic panel, wherein the detection comprises the following steps:
sending the infrared image of the divided photovoltaic panel into a YOLOv5s photovoltaic panel hot spot defect detection model;
extracting image characteristic information of the infrared image of the photovoltaic panel under different image fine granularities through a backbone network;
sending feature graphs output by three different stages of a backbone network into a neck network with a path aggregation network structure for multi-scale feature fusion;
sending the three fused feature graphs into three prediction head networks for prediction frame regression and category regression;
and screening the prediction box obtained in the last step by adopting a non-maximum inhibition algorithm, returning to the optimal prediction box, and completing the final hot spot defect detection of the photovoltaic panel.
8. A photovoltaic panel hot spot defect detection device for photovoltaic power plant, characterized in that includes:
the acquisition unit is used for acquiring a low-altitude infrared image of the photovoltaic panel;
the color mask obtaining unit is used for inputting the low-altitude infrared image of the photovoltaic panel into a pre-constructed Deeplabv3+ photovoltaic panel semantic segmentation model to obtain a color mask corresponding to the infrared image of the photovoltaic panel;
the segmentation unit is used for reading a color mask in a binarization mode based on OpenCV and adding the color mask and the low-altitude infrared image of the photovoltaic panel pixel by pixel to obtain a segmented infrared image of the photovoltaic panel with a pure black background;
and the detection unit is used for inputting the divided infrared image of the photovoltaic panel into a preset YOLOv5s photovoltaic panel hot spot defect detection model, detecting the hot spot defects of the photovoltaic panel and acquiring a detection result.
9. The utility model provides a photovoltaic panel hot spot defect detecting device for photovoltaic power plant which characterized in that: comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that: the program when executed by a processor implements the steps of the method of any one of claims 1 to 7.
CN202210213418.5A 2022-03-04 2022-03-04 Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station Pending CN114596278A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210213418.5A CN114596278A (en) 2022-03-04 2022-03-04 Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210213418.5A CN114596278A (en) 2022-03-04 2022-03-04 Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station

Publications (1)

Publication Number Publication Date
CN114596278A true CN114596278A (en) 2022-06-07

Family

ID=81815387

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210213418.5A Pending CN114596278A (en) 2022-03-04 2022-03-04 Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station

Country Status (1)

Country Link
CN (1) CN114596278A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115205246A (en) * 2022-07-14 2022-10-18 中国南方电网有限责任公司超高压输电公司广州局 Converter valve corona discharge ultraviolet image feature extraction method and device
CN116580363A (en) * 2023-07-14 2023-08-11 尚特杰电力科技有限公司 Photovoltaic panel hot spot identification method, storage medium and electronic equipment
CN116883406A (en) * 2023-09-08 2023-10-13 中交第一航务工程勘察设计院有限公司 Photovoltaic power station hot spot detection device and method based on cleaning robot

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115205246A (en) * 2022-07-14 2022-10-18 中国南方电网有限责任公司超高压输电公司广州局 Converter valve corona discharge ultraviolet image feature extraction method and device
CN115205246B (en) * 2022-07-14 2024-04-09 中国南方电网有限责任公司超高压输电公司广州局 Method and device for extracting ultraviolet image characteristics of converter valve through corona discharge
CN116580363A (en) * 2023-07-14 2023-08-11 尚特杰电力科技有限公司 Photovoltaic panel hot spot identification method, storage medium and electronic equipment
CN116580363B (en) * 2023-07-14 2023-09-26 尚特杰电力科技有限公司 Photovoltaic panel hot spot identification method, storage medium and electronic equipment
CN116883406A (en) * 2023-09-08 2023-10-13 中交第一航务工程勘察设计院有限公司 Photovoltaic power station hot spot detection device and method based on cleaning robot
CN116883406B (en) * 2023-09-08 2023-12-12 中交第一航务工程勘察设计院有限公司 Photovoltaic power station hot spot detection device and method based on cleaning robot

Similar Documents

Publication Publication Date Title
CN113674247B (en) X-ray weld defect detection method based on convolutional neural network
CN114596278A (en) Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station
CN111598860B (en) Lithium battery defect detection method based on yolov3 network embedded into self-attention door module
CN114022420B (en) Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component
CN114973032B (en) Deep convolutional neural network-based photovoltaic panel hot spot detection method and device
CN113538391A (en) Photovoltaic defect detection method based on Yolov4 and thermal infrared image
CN112184692A (en) Multi-target detection method for power transmission line
CN109741320A (en) A kind of wind electricity blade fault detection method based on Aerial Images
CN111914720B (en) Method and device for identifying insulator burst of power transmission line
CN113222896A (en) Photovoltaic panel defect detection method based on meta-learning
CN115861263A (en) Insulator defect image detection method based on improved YOLOv5 network
CN114943689A (en) Method for detecting components of steel cold-rolling annealing furnace based on semi-supervised learning
CN116385958A (en) Edge intelligent detection method for power grid inspection and monitoring
CN116503318A (en) Aerial insulator multi-defect detection method, system and equipment integrating CAT-BiFPN and attention mechanism
CN114494908A (en) Improved YOLOv5 power transmission line aerial image defect detection method
CN115908354A (en) Photovoltaic panel defect detection method based on double-scale strategy and improved YOLOV5 network
CN115424017A (en) Building internal and external contour segmentation method, device and storage medium
CN114972225A (en) Two-stage photovoltaic panel defect detection method based on deep learning
CN114627044A (en) Solar photovoltaic module hot spot detection method based on deep learning
CN114445615A (en) Rotary insulator target detection method based on scale invariant feature pyramid structure
CN116485802B (en) Insulator flashover defect detection method, device, equipment and storage medium
CN111291721B (en) Method, device, equipment and storage medium for detecting ex-warehouse of shared vehicle
CN116152685B (en) Pedestrian detection method and system based on unmanned aerial vehicle visual field
CN116664549A (en) Photovoltaic power station hot spot defect detection method based on feature perception
CN113284103B (en) Substation equipment defect online detection method based on space transformation fast R-CNN model

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