CN115171045A - YOLO-based power grid operation field violation identification method and terminal - Google Patents
YOLO-based power grid operation field violation identification method and terminal Download PDFInfo
- Publication number
- CN115171045A CN115171045A CN202210830306.4A CN202210830306A CN115171045A CN 115171045 A CN115171045 A CN 115171045A CN 202210830306 A CN202210830306 A CN 202210830306A CN 115171045 A CN115171045 A CN 115171045A
- Authority
- CN
- China
- Prior art keywords
- source image
- yolo
- image
- power grid
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
Abstract
The invention discloses a YOLO-based power grid operation field violation identification method and a YOLO-based power grid operation field violation identification terminal, which comprise the steps of obtaining an operation field source image, and dividing the source image into a training set and a test set; extracting the histogram of directional gradient and image texture characteristics of a source image of the training set before a first convolution layer of a Darknet-53 target recognition model, and fusing the histogram of directional gradient and the image texture characteristics to obtain a characteristic diagram; fusing the feature graph and a corresponding convolution layer in the Darknet-53 target recognition model to obtain a target feature image, inputting the target feature image into a YOLO neural network for training to obtain a detection model based on the YOLO neural network; inputting the test set into the Darknet-53 target recognition model and the detection model based on the YOLO neural network to obtain a test result, and determining an optimal detection model based on the YOLO neural network according to the test result; the safety monitoring of the violation behaviors of the intelligent identification operation site is realized, and the safety construction of the power grid operation site is ensured.
Description
Technical Field
The invention relates to the technical field of power construction monitoring, in particular to a method and a terminal for identifying violation of regulations on a power grid operation site based on YOLO.
Background
At present, with the rapid development of new energy industry, the demand of electric power construction is continuously expanded, the field workload of electric power construction is continuously increased, and safety supervision and detection are important means for supervising field safety construction, and generally require safety management personnel to patrol, but the manpower of an inspection department is insufficient and is influenced by field environment factors, so that the field safety monitoring by manpower has a leak and the efficiency is low; in addition, the installation position of the camera in the prior art is fixed and the monitoring mode is single due to the limitation of the environment of the operation field, so that the monitoring system cannot perform accurate control and cannot form timely and effective safety monitoring.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a YOLO-based power grid operation site violation identification method and a YOLO-based terminal, so that accurate and effective safety monitoring is automatically realized, and the safety construction of a power grid operation site is ensured.
In order to solve the technical problems, the invention adopts the technical scheme that:
a power grid operation field violation identification method based on YOLO comprises the following steps:
s1, obtaining a source image of an operation site, and dividing the source image into a training set and a test set;
s2, extracting a direction gradient histogram and image texture features of a source image of the training set before a first convolution layer of a Darknet-53 target recognition model, and fusing the direction gradient histogram and the image texture features to obtain a feature map;
s3, fusing the feature graph with a corresponding convolution layer in the Darknet-53 target recognition model to obtain a target feature image, and inputting the target feature image into a YOLO neural network for training to obtain a detection model based on the YOLO neural network;
s4, inputting the test set into the Darknet-53 target identification model and the detection model based on the YOLO neural network to obtain a test result, determining the optimal detection model based on the YOLO neural network according to the test result, and carrying out violation identification on the power grid operation field.
Further, the S2 specifically is:
s21, before a first convolution layer of a Darknet-5 target recognition model, performing brightness enhancement on a source image of the training set;
s22, carrying out gray scale normalization processing on the source image with enhanced brightness to obtain a normalized gray scale image;
s23, after calculating the direction gradient histogram and the image texture features of the normalized gray level image, combining the direction gradient histogram and the image texture feature weight value to obtain an optimal weight value, and obtaining a feature matrix according to the optimal weight value;
and S24, convolving the feature matrix by using the convolution core with the corresponding step length and scale to obtain feature maps with different scales.
Further, the S3 specifically is:
s31, fusing the feature maps with different scales with the convolution layers with corresponding scales in the Darknet-53 target identification model respectively to obtain target feature images with different scales;
s32, obtaining a plurality of clustering result K values by the target characteristic images with different scales through a dimension clustering algorithm;
and S33, determining an optimal K value by adopting a GIOU loss function to obtain the detection model.
Further, S1 specifically is:
acquiring a source image of the operation field from a source address, wherein the source image comprises a first source image and a second source image; the data sources of the first source image and the second source image are different.
Further, after the S4, the method further includes:
s5, collecting the first source image of an operation site, inputting the first source image into the optimal detection model to obtain a first monitoring result, judging whether the first source image has violation behaviors or not according to the first monitoring result, if so, marking the operation site of the first source image as a target area, and executing S6;
s6, collecting the second source image of the target area, inputting the second source image into the optimal detection model to obtain a second monitoring result, judging whether the second source image has violation behaviors or not according to the second monitoring result, if yes, uploading the second source image to a cloud system, and sending warning information.
Further, the S5 further includes: and if the first source image has no violation behaviors, storing the first source image into a preset address.
The S6 further includes: and if the second source image does not have the violation behaviors, storing the second source image into the preset address.
Further, after the S6, the method further includes:
and S7, after the storage space of the preset address reaches a preset threshold value, marking the preset address as a source address, and returning to execute the S1.
A YOLO-based power grid operation site violation identification terminal is used for completing YOLO-based power grid operation site violation identification, and comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the processor executes the computer program to realize the following steps:
s1, obtaining a source image of an operation site, and dividing the source image into a training set and a test set;
s2, extracting a direction gradient histogram and image texture features of a source image of the training set before a first convolution layer of a Darknet-53 target recognition model, and fusing the direction gradient histogram and the image texture features to obtain a feature map;
s3, fusing the feature graph with a corresponding convolution layer in the Darknet-53 target recognition model to obtain a target feature image, and inputting the target feature image into a YOLO neural network for training to obtain a detection model based on the YOLO neural network;
s4, inputting the test set into the Darknet-53 target recognition model and the detection model based on the YOLO neural network to obtain a test result, and determining the optimal detection model based on the YOLO neural network according to the test result.
The invention has the beneficial effects that:
the method comprises the steps of monitoring acquired field source images through a power grid operation field video, extracting global characteristics of a target and local detail texture characteristics of the target by respectively adopting a direction gradient histogram and an image texture characteristic method in combination with a target recognition technology of a deep learning model YOLO, preprocessing the image, highlighting contour characteristic information of the target in the image, and reducing wrong characteristics learned by a YOLO neural network in a training process, so that the precision of detection and classification is improved, the calculated amount is reduced, the violation behaviors of the intelligent recognition operation field are realized, manual inspection and monitoring are not needed, accurate and effective safety monitoring can be automatically realized, and the safety construction of the power grid operation field is ensured.
Drawings
Fig. 1 is a flow chart illustrating steps of a method for identifying violations of regulations on a power grid operation site based on YOLO according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a power grid operation field violation identification terminal based on YOLO according to an embodiment of the present invention;
fig. 3 is a network structure diagram of a Darknet-53 target identification model of a method for identifying violations of the power grid operation site based on YOLO according to the embodiment of the present invention;
description of the reference symbols:
1. a power grid operation field violation identification terminal based on YOLO; 2. a memory; 3. a processor.
Detailed Description
In order to explain the technical contents, the objects and the effects of the present invention in detail, the following description is made with reference to the accompanying drawings in combination with the embodiments.
Referring to fig. 1, a method for identifying violations of regulations on a power grid operation site based on YOLO includes the steps:
s1, obtaining a source image of an operation field, and dividing the source image into a training set and a test set;
s2, extracting a direction gradient histogram and image texture features of a source image of the training set before a first convolution layer of a Darknet-53 target recognition model, and fusing the direction gradient histogram and the image texture features to obtain a feature map;
s3, fusing the feature graph with a corresponding convolution layer in the Darknet-53 target recognition model to obtain a target feature image, and inputting the target feature image into a YOLO neural network for training to obtain a detection model based on the YOLO neural network;
s4, inputting the test set into the Darknet-53 target identification model and the detection model based on the YOLO neural network to obtain a test result, determining the optimal detection model based on the YOLO neural network according to the test result, and carrying out violation identification on the power grid operation field.
Further, S2 specifically is:
s21, before a first convolution layer of a Darknet-5 target recognition model, performing brightness enhancement on a source image of the training set;
s22, carrying out gray scale normalization processing on the source image with enhanced brightness to obtain a normalized gray scale image;
s23, after calculating the direction gradient histogram and the image texture features of the normalized gray level image, combining the direction gradient histogram and the image texture feature weight value to obtain an optimal weight value, and obtaining a feature matrix according to the optimal weight value;
and S24, convolving the feature matrix by using the convolution core with the corresponding step length and scale to obtain feature maps with different scales.
The description shows that the image texture features have gray scale invariance and rotation invariance but are sensitive to direction information, and the direction gradient histogram can better describe the appearance and the shape of an object but neglects the overall features, so that the two feature extraction methods are fused, the respective defects are eliminated, and the respective advantages are exerted; for depth feature fusion in the target recognition model, the extracted features are converted into feature maps of corresponding convolutional layers by convolution.
Further, the S3 specifically is:
and S31, fusing the feature maps with different scales with the convolution layers with corresponding scales in the Darknet-53 target identification model respectively to obtain target feature images with different scales.
S32, obtaining a plurality of clustering result K values by the target characteristic images with different scales through a means algorithm;
and S33, determining an optimal K value by adopting a GIOU loss function to obtain the detection model.
According to the description, the dimension clustering algorithm is to count or cluster a plurality of frames with different sizes from the real frames of the training set, so that blind searching of the model during training is avoided, rapid convergence of the model is facilitated, and parameter values of the candidate frames are determined by a means algorithm; the GIOU loss function can change along with the different intersecting shapes of the two frames, the problem that the IOU loss algorithm adopted by the YOLO algorithm cannot truly reflect the intersection of the real frame and the prediction frame is solved, the K value is accurately determined, and errors caused by the dimension problem can be solved.
Further, S4 specifically is:
inputting the test set into the Darknet-53 target recognition model and the detection model to obtain a test result, and determining an optimal detection model based on a YOLO neural network according to the test result.
From the above description, it can be known that the problem that the single test result is too single and the training data is insufficient is solved by performing different sets of training and verification on the model through different training sets and test sets.
Further, S1 specifically is:
acquiring the source image of the operation field from a source address, wherein the source image comprises a first source image and a second source image;
the data source of the first source image and the data source of the second source image are different.
According to the description, the source images are not operation site images acquired at the same time, the first source image is a panoramic image acquired for the first time, the second source image is a detailed image acquired for the second time, misjudgment of single monitoring violation behaviors can be avoided through double capturing monitoring of the panorama and the details, and the accuracy of safety monitoring is improved.
Further, after S4, the method further includes:
s5, collecting the first source image of an operation site, inputting the first source image into the optimal detection model to obtain a first monitoring result, judging whether the first source image has violation behaviors or not according to the first monitoring result, if yes, marking the operation site of the first source image as a target area, and executing S6;
s6, collecting the second source image of the target area, inputting the second source image into the optimal detection model to obtain a second monitoring result, judging whether the second source image has violation behaviors or not according to the second monitoring result, if yes, uploading the second source image to a cloud system, and sending warning information.
According to the description, the first source image and the second source image monitor the target with the violation behaviors through the optimal detection model of the YOLO neural network; the first source image is monitored and detected for the first time, the monitoring range is wide, misjudgment is easy to occur, the operation construction site condition is complex, an accident exists, the time difference exists between the detail image of the second source image and the first source image, misjudgment alarm under the accident condition can be avoided, the monitoring range is accurately locked, the detection target is more clear, and the correctness of the monitoring result is ensured.
Further, the S5 further includes:
and if the first source image does not have the violation behaviors, storing the first source image into a preset address.
According to the description, if the collected first source image has no violation behaviors, the collected first source image is stored in the preset address to be used as a training material for continuous deep learning of the detection model based on the YOLO neural network.
Further, the S6 further includes:
and if the second source image does not have the violation behaviors, storing the second source image into the preset address.
According to the description, if the collected second source image has no violation behaviors, the second source image is stored in the preset address to serve as a training material for the detection model based on the YOLO neural network to continuously and deeply learn.
Further, after S6, the method further includes:
and S7, after the storage space of the preset address reaches a preset threshold value, marking the preset address as a source address, and returning to execute the S1.
According to the above description, when the storage space of the preset address reaches the threshold value, that is, the number of the source images reaches a certain threshold value, the deep learning and training of the YOLO neural network are performed again, and the optimal detection model is updated.
Referring to fig. 2, the YOLO-based power grid operation field violation identification terminal comprises a memory and a processor, and is characterized in that: and the processor identifies the violation of the power grid operation field through an optimal detection model based on the YOLO neural network according to the power grid operation field violation identification method based on the YOLO. The method comprises the following specific steps:
a YOLO-based violation identification terminal 1 for a power grid operation site comprises a memory 2, a processor 3 and a computer program which is stored on the memory and can run on the processor, wherein the processor realizes any step of the YOLO-based violation identification method for the power grid operation site when executing the computer program.
The YOLO-based power grid operation field violation identification method provided by the invention can automatically monitor whether the operation field has violation behaviors or not, and ensures the safety construction of the operation field, and the following description is provided by specific embodiments:
referring to fig. 1, a first embodiment of the present invention is:
a power grid operation field violation identification method based on YOLO comprises the following steps:
s1, obtaining a source image of an operation site, and dividing the source image into a training set and a test set;
the S1 specifically comprises the following steps:
acquiring the source image of the operation field from a source address, wherein the source image comprises a first source image and a second source image;
the data sources of the first source image and the second source image are different;
in an alternative embodiment, the same group of source images may be grouped multiple times to obtain training sets and test sets of different combinations; in multiple grouping, the source images are respectively used as a training set and a test set to train a detection model of the YOLO neural network;
in an alternative embodiment, the first source image is a 360 ° panoramic image acquired by using 6 wide-angle lenses, and the second source image is a partial image acquired by using a binocular vision camera; the first source image and the second source image are images obtained aiming at the same target area;
s2, extracting a direction gradient histogram and image texture features of a source image of the training set before a first convolution layer of a Darknet-53 target recognition model, and fusing the direction gradient histogram and the image texture features to obtain a feature map;
the S2 specifically comprises the following steps:
s21, before a first convolution layer of a Darknet-5 target recognition model, performing brightness enhancement on a source image of the training set;
s22, carrying out gray scale normalization processing on the source image with enhanced brightness to obtain a normalized gray scale image;
in an optional implementation manner, the source image after brightness enhancement is subjected to gray scale conversion to obtain a gray scale image, then the gray scale image is subjected to denoising, and finally normalization processing of the gray scale image is performed;
s23, after calculating the direction gradient histogram and the image texture features of the normalized gray level image, combining the direction gradient histogram and the image texture feature weight value to obtain an optimal weight value, and obtaining a feature matrix according to the optimal weight value;
in an optional implementation manner, combining the histogram of directional gradients and the feature vectors of the image texture features according to columns, and combining weights to obtain an optimal weight value;
in an optional implementation manner, obtaining the feature matrix according to the optimal weight value specifically includes: multiplying the column characteristic vectors under the direction gradient histogram and the optimal weight value of the image texture characteristic by the line characteristic vectors with dimensions fitting and numerical values all being 1 to obtain a source matrix, and performing multiple convolution to obtain the characteristic matrix;
s24, convolving the feature matrix by using the convolution core with the corresponding step length and scale to obtain feature maps with different scales;
in an alternative embodiment, the feature maps of different scales are respectively a 13 × 13 × 1 feature map, a 26 × 26 × 1 feature map, and a 52 × 52 × 1 feature map;
s3, fusing the feature graph with a corresponding convolution layer in the Darknet-53 target recognition model to obtain a target feature image, and inputting the target feature image into a YOLO neural network for training to obtain a detection model based on the YOLO neural network;
the S3 specifically comprises the following steps:
s31, fusing the feature maps with different scales with the convolution layers with corresponding scales in the Darknet-53 target identification model respectively to obtain target feature images with different scales;
in an alternative embodiment, referring to fig. 3, the 13 × 13 × 1 feature map, the 26 × 26 × 1 feature map, and the 52 × 52 × 1 feature map are fused with the previous convolutional layer of the YOLO layers with the scales of 13 × 13, 26 × 26, and 52 × 52 in the Darknet-53 object recognition model, respectively;
s32, obtaining a plurality of clustering result K values by the target characteristic images with different scales through a dimension clustering algorithm;
s33, determining an optimal K value by adopting a GIOU loss function to obtain the detection model;
s4, inputting the test set into the Darknet-53 target recognition model and the detection model based on the YOLO neural network to obtain a test result, and determining an optimal detection model based on the YOLO neural network according to the test result;
the S4 specifically comprises the following steps:
inputting the test set into the Darknet-53 target recognition model and the detection model to obtain a test result, and determining an optimal detection model based on a YOLO neural network according to the test result;
s5, collecting the first source image of an operation site, inputting the first source image into the optimal detection model to obtain a first monitoring result, judging whether the first source image has violation behaviors or not according to the first monitoring result, if so, marking the operation site of the first source image as a target area, and executing S6;
if the first source image does not have the violation behaviors, storing the first source image into a preset address;
s6, collecting the second source image of the target area, inputting the second source image into the optimal detection model to obtain a second monitoring result, judging whether the second source image has violation behaviors or not according to the second monitoring result, if yes, uploading the second source image to a cloud system, and sending warning information;
if the second source image does not have the violation behaviors, storing the second source image into the preset address;
s7, after the storage space of the preset address reaches a preset threshold value, marking the preset address as a source address, and returning to execute the S1;
in an optional implementation manner, the preset address and the source address are the same address, and the preset address is a source address for storing an untrained source image;
in an optional implementation manner, after S1 is returned, the source image in the source address is deleted, the source address space is released, and the source address is marked as a preset address.
In conclusion, according to the YOLO-based power grid operation site violation identification method and the YOLO-based power grid operation site violation identification terminal, the monitoring range is accurately locked through batch acquisition of panoramic images and local images, the detection target is more definite, and the accuracy of the source image for acquiring information is ensured; meanwhile, a target recognition technology of a deep learning model YOLO is combined, a global feature of a target and a local detail texture feature of the target are extracted by adopting a direction gradient histogram and an image texture feature method respectively, the image is preprocessed, outline feature information of the target in the image is highlighted, wrong features learned in a training process of a YOLO neural network are reduced, and accuracy of a violation judgment result is guaranteed; in addition, the obtained source image continuously performs learning training on the YOLO model, and the YOLO model is adjusted in time; the accuracy is ensured for a plurality of times from the source of the acquired information to the process of monitoring and distinguishing, so that the precision of the monitoring effect is improved, accurate and effective safety monitoring is automatically realized, and the safety construction of a power grid operation field is ensured.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent changes made by using the contents of the present specification and the drawings, or applied directly or indirectly to the related technical fields, are included in the scope of the present invention.
Claims (9)
1. A power grid operation field violation identification method based on YOLO is characterized by comprising the following steps:
s1, obtaining a source image of an operation site, and dividing the source image into a training set and a test set;
s2, extracting a direction gradient histogram and image texture features of a source image of the training set before a first convolution layer of a Darknet-53 target recognition model, and fusing the direction gradient histogram and the image texture features to obtain a feature map;
s3, fusing the feature graph with a corresponding convolutional layer in the Darknet-53 target recognition model to obtain a target feature image, and inputting the target feature image into a YOLO neural network for training to obtain a detection model based on the YOLO neural network;
s4, inputting the test set into the Darknet-53 target identification model and the detection model based on the YOLO neural network to obtain a test result, determining the optimal detection model based on the YOLO neural network according to the test result, and carrying out violation identification on the power grid operation field.
2. The method for identifying violations of the YOLO-based power grid operation site as claimed in claim 1, wherein S2 specifically is:
s21, before a first convolution layer of a Darknet-5 target recognition model, performing brightness enhancement on a source image of the training set;
s22, carrying out gray scale normalization processing on the source image with enhanced brightness to obtain a normalized gray scale image;
s23, after calculating the direction gradient histogram and the image texture features of the normalized gray level image, combining the direction gradient histogram and the image texture feature weight value to obtain an optimal weight value, and obtaining a feature matrix according to the optimal weight value;
and S24, convolving the characteristic matrix by using the convolution kernel of the corresponding step length and scale to obtain characteristic diagrams of different scales.
3. The method for identifying violations of the YOLO-based power grid operation site as claimed in claim 1, wherein S3 specifically is:
s31, fusing the different-scale feature maps with the convolution layers with the corresponding scales in the Darknet-53 target identification model respectively to obtain target feature images with different scales;
s32, obtaining a plurality of clustering result K values by the target characteristic images with different scales through a dimension clustering algorithm;
and S33, determining an optimal K value by adopting a GIOU loss function to obtain the detection model.
4. The method for identifying violations of the YOLO-based power grid operation site as claimed in claim 1, wherein S1 specifically is:
acquiring a source image of the operation field from a source address, wherein the source image comprises a first source image and a second source image; the data sources of the first source image and the second source image are different.
5. The YOLO-based power grid job site violation identification method according to claim 1, further comprising after S4:
s5, collecting the first source image of an operation site, inputting the first source image into the optimal detection model to obtain a first monitoring result, judging whether the first source image has violation behaviors or not according to the first monitoring result, if so, marking the operation site of the first source image as a target area, and executing S6;
s6, collecting the second source image of the target area, inputting the second source image into the optimal detection model to obtain a second monitoring result, judging whether the second source image has violation behaviors or not according to the second monitoring result, if yes, uploading the second source image to a cloud system, and sending warning information.
6. The method for identifying violations on the YOLO-based power grid operating site as recited in claim 5, wherein S5 further comprises:
and if the first source image does not have the violation behaviors, storing the first source image into a preset address.
7. The method for identifying violations of YOLO-based power grid operation site as claimed in claim 5, wherein S6 further comprises:
and if the second source image does not have the violation behaviors, storing the second source image into the preset address.
8. The method for YOLO-based grid job site violation identification according to claim 5, further comprising after S6:
and S7, after the storage space of the preset address reaches a preset threshold value, marking the preset address as a source address, and returning to execute the S1.
9. A power grid operation field violation identification terminal based on YOLO comprises a memory and a processor, and is characterized in that: and the processor identifies the violation of the power grid operation field through an optimal detection model based on the YOLO neural network according to the power grid operation field violation identification method based on the YOLO.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210830306.4A CN115171045A (en) | 2022-07-15 | 2022-07-15 | YOLO-based power grid operation field violation identification method and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210830306.4A CN115171045A (en) | 2022-07-15 | 2022-07-15 | YOLO-based power grid operation field violation identification method and terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115171045A true CN115171045A (en) | 2022-10-11 |
Family
ID=83494302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210830306.4A Pending CN115171045A (en) | 2022-07-15 | 2022-07-15 | YOLO-based power grid operation field violation identification method and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115171045A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116108397A (en) * | 2022-12-22 | 2023-05-12 | 福建亿榕信息技术有限公司 | Electric power field operation violation identification method integrating multi-mode data analysis |
| CN116740654A (en) * | 2023-08-14 | 2023-09-12 | 安徽博诺思信息科技有限公司 | Substation operation prevention and control method based on image recognition technology |
-
2022
- 2022-07-15 CN CN202210830306.4A patent/CN115171045A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116108397A (en) * | 2022-12-22 | 2023-05-12 | 福建亿榕信息技术有限公司 | Electric power field operation violation identification method integrating multi-mode data analysis |
| CN116108397B (en) * | 2022-12-22 | 2024-01-09 | 福建亿榕信息技术有限公司 | Electric power field operation violation identification method integrating multi-mode data analysis |
| CN116740654A (en) * | 2023-08-14 | 2023-09-12 | 安徽博诺思信息科技有限公司 | Substation operation prevention and control method based on image recognition technology |
| CN116740654B (en) * | 2023-08-14 | 2023-11-07 | 安徽博诺思信息科技有限公司 | Substation operation prevention and control method based on image recognition technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ukhwah et al. | Asphalt pavement pothole detection using deep learning method based on YOLO neural network | |
| CN108734143A (en) | A kind of transmission line of electricity online test method based on binocular vision of crusing robot | |
| CN108711148B (en) | Tire defect intelligent detection method based on deep learning | |
| CN115171045A (en) | YOLO-based power grid operation field violation identification method and terminal | |
| CN111797890A (en) | Method and system for detecting defects of power transmission line equipment | |
| CN109858367B (en) | Visual automatic detection method and system for worker through supporting unsafe behaviors | |
| CN110264444B (en) | Damage detection method and device based on weak segmentation | |
| CN111178206A (en) | Building embedded part detection method and system based on improved YOLO | |
| CN113379712B (en) | Steel bridge bolt disease detection method and system based on computer vision | |
| CN112581445A (en) | Detection method and device for bolt of power transmission line, storage medium and electronic equipment | |
| CN114638784A (en) | Method and device for detecting surface defects of copper pipe based on FE-YOLO | |
| CN110992307A (en) | Insulator positioning and identifying method and device based on YOLO | |
| CN115993096A (en) | High-rise building deformation measuring method | |
| CN110910440A (en) | Power transmission line length determination method and system based on power image data | |
| CN116258980A (en) | Unmanned aerial vehicle distributed photovoltaic power station inspection method based on vision | |
| CN116843999A (en) | Gas cylinder detection method in fire operation based on deep learning | |
| CN113850799A (en) | YOLOv 5-based trace DNA extraction workstation workpiece detection method | |
| CN113420619A (en) | Remote sensing image building extraction method | |
| CN113393442A (en) | Method and system for detecting abnormality of train parts, electronic device and storage medium | |
| CN110765900A (en) | DSSD-based automatic illegal building detection method and system | |
| CN115984759A (en) | Substation switch state identification method and device, computer equipment and storage medium | |
| CN115908843A (en) | Superheat degree recognition model training method, recognition method, equipment and storage medium | |
| CN117677969A (en) | Defect detection method and device | |
| CN114120097A (en) | Distribution network engineering on-site transformer detection method and system based on machine vision | |
| CN113837178A (en) | Deep learning-based automatic positioning and unified segmentation method for meter of transformer substation |
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 |