CN113989209A - Power line foreign matter detection method based on fast R-CNN - Google Patents
Power line foreign matter detection method based on fast R-CNN Download PDFInfo
- Publication number
- CN113989209A CN113989209A CN202111228906.5A CN202111228906A CN113989209A CN 113989209 A CN113989209 A CN 113989209A CN 202111228906 A CN202111228906 A CN 202111228906A CN 113989209 A CN113989209 A CN 113989209A
- Authority
- CN
- China
- Prior art keywords
- cnn
- image
- power line
- faster
- anchor
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000003745 diagnosis Methods 0.000 claims abstract description 13
- 238000007689 inspection Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000000034 method Methods 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 6
- 230000007797 corrosion Effects 0.000 claims description 5
- 238000005260 corrosion Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 238000005516 engineering process Methods 0.000 abstract description 5
- 241000490229 Eucephalus Species 0.000 abstract 1
- 238000013527 convolutional neural network Methods 0.000 description 33
- 230000005540 biological transmission Effects 0.000 description 8
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 235000005770 birds nest Nutrition 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000012212 insulator Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 235000005765 wild carrot Nutrition 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/60—Memory management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a computer vision power grid application technology, in particular to a power line foreign matter detection method based on Faster R-CNN, which comprises a training mode and a diagnosis mode; the training mode returns to the iteration loss through the pictures in the training data set to obtain a stable and usable model; and in the diagnosis mode, the diagnosis result of whether the kite foreign matter exists on the power line and the position of the kite foreign matter is obtained by inputting the image to be tested. Aerial images of the unmanned aerial vehicle can be effectively utilized, and the utilization rate of the images is improved; the cost is low, only a power line kite foreign matter detection module based on an aster R-CNN algorithm needs to be added in the existing system, and other hardware does not need to be added; the trained model has good prediction effect on the foreign matters of the power line kite under different background environments.
Description
Technical Field
The invention belongs to the technical field of computer vision power grid application, and particularly relates to a power line foreign matter detection method based on Faster R-CNN.
Background
With the development of the industrial scale of the alternating-current transmission power line, the human resources and the property resources consumed by the maintenance of the transmission line are continuously increased. Many power transmission lines are erected in places where detection personnel cannot easily reach, and the power transmission lines in natural environments are very easily damaged by 'foreign matters', such as plastic bags, kites, bird nests and the like hung on power lines, and if the 'foreign matters' cannot be found quickly, power accidents can be caused, and domestic electricity and industrial electricity of residents are affected.
At present, a patrol inspection method for a power transmission line is mainly characterized in that a worker gradually inspects the power transmission line along the direction of the power line. However, the span of the existing power transmission network is large, and the ground is rugged, which brings great difficulty and danger to the patrol work of workers. The appearance of unmanned aerial vehicle patrols and examines makes the work of patrolling and examining go on under safe, not influenced by geographical condition, weather condition. Compared with manual inspection and helicopter inspection, the intelligent unmanned aerial vehicle inspection system has the advantages that: safety: can avoid exposing the staff under the bad condition of patrolling and examining to also reduced the electric shock danger that the staff direct contact high voltage circuit patrolled and examined. High efficiency: the problem of low efficiency of manual inspection circuits is directly avoided, and the inspection time is greatly reduced. ③ cheap: the investment of manpower and financial resources is directly reduced, and the expenditure is reduced. Fourthly, the precision is high: when directly patrolling and examining by the manpower, probably because the eyes uses in a large number and leads to the attention to descend, the judgment is not enough to reduce the accuracy of judging, and intelligent unmanned aerial vehicle patrols and examines and can gather in real time and patrol and examine the image and carry out the judgement of trouble, has greatly improved accuracy nature and objectivity.
Most unmanned aerial vehicle collected images are processed by using a YOLO series algorithm at the present stage. The YOLO series is a stage of target detection algorithm, and for an input image, under the influence of a regression idea, position and kind information of a target to be detected can be obtained by using a deep convolutional neural network. The algorithm has only one stage, so the detection speed is high, and the algorithm is often applied to various environments needing real-time detection. The existing YOLO series algorithms are mainly YOLOv3 and YOLOv 4.
The background false detection rate of the YOLO algorithm is low, the universality is strong, but compared with the R-CNN series algorithm, the YOLO algorithm has low accuracy on the identification position of an object and low recall rate, but for the detection of foreign matters on a power line, firstly, the detection of the foreign matters on the power line is required, and the positions of the foreign matters on the power line can be accurately marked, but the YOLO series algorithm cannot meet the requirements; meanwhile, the YOLO series algorithm has certain requirements on the size of a training set used in deep learning, the number of pictures obtained when the unmanned aerial vehicle patrols and examines the power line is very small at present, and the situations that foreign matters or power line abnormal states such as power line breakage occur on the power line are also very small, so that the conditions of poor training effect and poor target detection effect can exist when the YOLO series algorithm with harsh requirements on the size of the training set is used.
Firstly, CNN series technology is selected, but the load of CNN on GPU is too large, so preprocessing technologies such as Sobel operator edge extraction and Hough transformation straight line detection are carried out on the image, the missing detection rate is greatly reduced, the consumption of GPU is reduced, and meanwhile, the detection efficiency is higher due to the improved Epoch and learning rate.
Disclosure of Invention
Aiming at the problems in the background technology, the invention provides a power line kite foreign matter detection method based on the Faster R-CNN algorithm, which can be used for positioning and identifying power lines in an image of a complex background and judging whether foreign matters such as kites exist or not.
In order to solve the technical problems, the invention adopts the following technical scheme: the power line foreign matter detection method based on the fast R-CNN comprises a training mode and a diagnosis mode; the method comprises the following steps:
step 1, carrying out image preprocessing on the inspection image, wherein the image preprocessing comprises image graying, image gray equalization and Gaussian filtering on the image, and Gaussian noise generated in image acquisition is removed;
step 2, performing edge extraction on the preprocessed image by using a Sobel operator, performing expansion filling processing or corrosion processing on the extracted edge binary image, and performing linear detection by using Hough transformation;
step 3, setting different learning rates, and putting the marked pictures into an Faster R-CNN network for training to obtain a power line foreign matter detection model of the Faster R-CNN network;
step 4, comparing all indexes of the training model obtained by different learning rates, and selecting a proper learning rate;
and 5, inputting the aerial image to be detected into the trained detection model, and returning the kite foreign matter position frame through calculation by the fast R-CNN network to finish diagnosis.
In the power line foreign matter detection method based on the Faster R-CNN, the step of specifically realizing the Faster R-CNN network module comprises the following steps:
1) inputting the aerial image into a Faster R-CNN network;
2) calculating the actual output of the Faster R-CNN network to obtain the information of the target prediction box;
3) calculating the numerical difference between the target prediction frame and the target real frame;
4) if the difference between the two exceeds the threshold value, adjusting the setting of the fast R-CNN network parameters;
5) and finishing training to obtain the detection information of the foreign matters on the power line.
In the above power line foreign object detection method based on the fast R-CNN, the implementation of the core anchor mechanism of the RPN network in the fast R-CNN network includes:
in the RPN network, corresponding mapping areas in the original drawing correspond to each point in the feature map, and meanwhile, in the original drawing, the center of the area is taken as an anchor point, and three sizes 2 are respectively selected7*27、28*28And 29*29And three aspect ratios 1: 1. 1: 2. 2: 1 as anchor boxes, with a total of wxhx9 in fast R-CNN, where W, H correspond to the length and width in feature map, respectively; each anchor in the feature map is 28Dimension, inputting each anchor into the RPN network to respectively obtain two groups of data of 2k and 4 k; wherein k is 9, and represents the number of anchor boxes per anchor; 2, representing two data of the anchor frame belonging to the foreground and the background, and roughly classifying the target; 4 represents the regression value of the coordinates of the four parameters, and is used for roughly positioning a Proposal frame; outputting a vector with the length of 2k +4k for each anchor, wherein the vector represents the judgment of whether k candidate regions corresponding to the anchor are in a foreground or a background and can obtain position information; and after the result obtained by the RPB network passes through the ROIPooling layer, obtaining a feature map corresponding to the Proposal frame on the feature map, and finally bringing the result into the subsequent network to perform the type judgment of the target and the detailed regression of the position of the target.
Compared with the prior art, the invention has the beneficial effects that: firstly, aerial images of the unmanned aerial vehicle can be effectively utilized, and the utilization rate of the images is improved; the cost is low, only a power line foreign matter detection module based on the Faster R-CNN algorithm needs to be added to the existing system, and other hardware does not need to be added; and the trained model has good prediction effect on power lines in different background environments. And fourthly, Gaussian filtering, Sobel operator edge extraction and Hough transformation linear detection are adopted, so that the missing rate is reduced. And fifthly, the foreign matter detection of the sample is more accurate by changing the learning rate and the Epoch size.
Drawings
FIG. 1 is a flow chart of the detection of foreign matter on a power line based on Faster R-CNN according to an embodiment of the present invention;
FIG. 2 is a structural diagram of Faster R-CNN according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the following embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The present invention is further illustrated by the following examples, which are not to be construed as limiting the invention.
The power line foreign matter detection method based on the fast R-CNN comprises a training mode and a diagnosis mode; the training mode returns to the iteration loss through the pictures in the training data set to obtain a stable and usable model; and in the diagnosis mode, the diagnosis result of whether the kite foreign matter exists on the power line and the position of the kite foreign matter is obtained by inputting the image to be tested. Firstly, removing Gaussian noise by adopting technologies such as image graying, image gray equalization, Gaussian filtering and the like. In order to reduce the load of deep learning fast RCNN on GPU, Sobel operator is adopted to carry out edge extraction, then expansion filling processing is carried out on the extracted edge binary image, partial image possibly can also be subjected to corrosion processing, and finally line detection is carried out by utilizing Hough transformation. The power lines in the general complex image of the background can be effectively extracted, and the missing rate of the power lines is reduced. The best effect of detecting the foreign matters in the power transmission line is obtained by changing the learning rate and the clock size.
The embodiment is realized by the following technical scheme, and the power line foreign matter detection method based on the Faster R-CNN algorithm comprises the following steps:
step S1, firstly, image preprocessing needs to be performed on the inspection image, including image graying, image grayscale equalization, and gaussian filtering on the image, so as to remove gaussian noise that may occur during image acquisition, and thus, edge information of the processed image is easier to extract and less in interference.
And step S2, performing edge extraction on the enhanced image by using a Sobel operator, then performing expansion filling processing on the extracted edge binary image, possibly using corrosion processing on partial images, and finally performing straight line detection by using Hough transformation. The method can effectively extract the power lines in the image with the general and complicated background, and reduce the omission ratio of the power lines.
And step S3, inputting a training data set by using a Faster R-CNN algorithm, training a model, and performing experiments by using different learning rates.
And step S4, comparing the indexes of the training models obtained by different learning rates, and selecting the most appropriate learning rate.
And step S5, inputting the aerial images to be detected into the trained model, and returning the foreign body position square of the power line kite through the network by calculation to finish diagnosis.
In specific implementation, the input of the power line foreign matter detection method based on the Faster R-CNN algorithm is an image which is aerial by the unmanned aerial vehicle and contains a power line, and the image comprises a training mode and a diagnosis mode, as shown in fig. 1.
1) Firstly, image preprocessing is needed to be carried out on an inspection image, wherein the image preprocessing comprises image graying, image gray equalization and Gaussian filtering, edge extraction is carried out on the enhanced image by utilizing a Sobel operator, then expansion filling processing is carried out on the extracted edge binary image, partial images can be also subjected to corrosion processing, and finally line detection is carried out by utilizing Hough transformation.
2) Setting different learning rates, and putting the marked pictures into the Faster R-CNN algorithm for training to obtain a power line foreign matter detection model of the Faster R-CNN algorithm.
3) And measuring each index of the model after training, and repeating the step 2) to perform five times of experiments with different learning rates.
4) And comparing all indexes of the training model obtained by different learning rates, and selecting the most appropriate learning rate.
5) And inputting the aerial image to be detected into the trained model, and performing diagnosis by a network through calculating a box for returning the positions of the foreign matters in the kite.
The concrete implementation steps of the Faster R-CNN network module comprise:
a. inputting aerial insulator string images into a Faster R-CNN network;
b. calculating the actual output of the network to obtain the information of the target prediction frame;
c. calculating the numerical difference between the target prediction frame and the target real frame;
d. if the difference between the two exceeds a threshold value, adjusting the network parameter setting;
e. and finishing training to obtain power line foreign matter detection information.
The concrete implementation of the above procedure when using the Faster R-CNN algorithm is shown in fig. 2, where important is the RPN network.
The most important core in RPN networks is the anchor mechanism.
In the RPN network, the original image has corresponding mapping area corresponding to each point in the feature map, and in the original image, the center of the area is used as an anchor point (anchor) and 2 is taken out7*27、28*28And 29*29Three sizes and 1: 1. 1: 2. 2: 1 three length to width ratios of 9 regions in total are used as anchor boxes, so the anchor boxes for the total W X H X9 are in the Faster R-CNN, where W, H correspond to the length and width in the feature map, respectively. Each anchor point in the feature map is 28And D, inputting each anchor point into the RPN network to respectively obtain two groups of data of 2k and 4 k. It is composed ofIn (3), k is 9, representing the number of anchor boxes per anchor point; 2 represents two data of the anchor frame belonging to the foreground and the background, and can be used for roughly classifying the target; 4 represents the regression of the coordinates of the four parameters, used for the coarse positioning of the Propusal box. For each anchor point, a vector with the length of 2k +4k is output, and the vector represents the judgment of the k candidate areas corresponding to the anchor point on the foreground or the background, and the approximate information of the position can be obtained. After the result obtained by the RPB network passes through the ROIPooling layer, the feature diagram corresponding to the Proposal frame can be obtained on the feature map, and finally the result is brought into the subsequent network to carry out the object type judgment and the detailed regression of the position, wherein the last process is the same as Fast R-CNN.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.
Claims (3)
1. The power line foreign matter detection method based on the fast R-CNN comprises a training mode and a diagnosis mode; the method is characterized in that: the method comprises the following steps:
step 1, carrying out image preprocessing on the inspection image, wherein the image preprocessing comprises image graying, image gray equalization and Gaussian filtering on the image, and Gaussian noise generated in image acquisition is removed;
step 2, performing edge extraction on the preprocessed image by using a Sobel operator, performing expansion filling processing or corrosion processing on the extracted edge binary image, and performing linear detection by using Hough transformation;
step 3, setting different learning rates, and putting the marked pictures into an Faster R-CNN network for training to obtain a power line foreign matter detection model of the Faster R-CNN network;
step 4, comparing all indexes of the training model obtained by different learning rates, and selecting a proper learning rate;
and 5, inputting the aerial image to be detected into the trained detection model, and returning the kite foreign matter position frame through calculation by the fast R-CNN network to finish diagnosis.
2. The power line foreign object detection method based on Faster R-CNN according to claim 1, wherein: the concrete implementation steps of the Faster R-CNN network module comprise:
1) inputting the aerial image into a Faster R-CNN network;
2) calculating the actual output of the Faster R-CNN network to obtain the information of the target prediction box;
3) calculating the numerical difference between the target prediction frame and the target real frame;
4) if the difference between the two exceeds the threshold value, adjusting the setting of the fast R-CNN network parameters;
5) and finishing training to obtain the detection information of the foreign matters on the power line.
3. The power line foreign object detection method based on Faster R-CNN according to claim 2, wherein: the implementation of the core anchor mechanism of the RPN network in the Faster R-CNN network comprises the following steps:
in the RPN network, corresponding mapping areas in the original drawing correspond to each point in the feature map, and meanwhile, in the original drawing, the center of the area is taken as an anchor point, and three sizes 2 are respectively selected7*27、28*28And 29*29And 9 regions with three length-to-width ratios of 1: 1, 1: 2, 2: 1 as anchor boxes, with a total of WXHX9 anchor boxes in fast R-CNN, wherein W, H corresponds to the length and width in feature map, respectively; each anchor in the feature map is 28Dimension, inputting each anchor into the RPN network to respectively obtain two groups of data of 2k and 4 k; wherein k is 9, and represents the number of anchor boxes per anchor; 2, representing two data of the anchor frame belonging to the foreground and the background, and roughly classifying the target; 4 represents the regression value of the coordinates of the four parameters, and is used for roughly positioning a Proposal frame; for each anchor, a vector with the length of 2k +4k is output, and the vector represents k candidate regions corresponding to the anchorJudging whether the domain is a foreground or a background and obtaining the position information; and after the result obtained by the RPB network passes through the ROIPooling layer, obtaining a feature map corresponding to the Proposal frame on the feature map, and finally bringing the result into the subsequent network to perform the type judgment of the target and the detailed regression of the position of the target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111228906.5A CN113989209A (en) | 2021-10-21 | 2021-10-21 | Power line foreign matter detection method based on fast R-CNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111228906.5A CN113989209A (en) | 2021-10-21 | 2021-10-21 | Power line foreign matter detection method based on fast R-CNN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113989209A true CN113989209A (en) | 2022-01-28 |
Family
ID=79740087
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111228906.5A Pending CN113989209A (en) | 2021-10-21 | 2021-10-21 | Power line foreign matter detection method based on fast R-CNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113989209A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114926667A (en) * | 2022-07-20 | 2022-08-19 | 安徽炬视科技有限公司 | Image identification method based on cloud edge-end cooperation |
-
2021
- 2021-10-21 CN CN202111228906.5A patent/CN113989209A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114926667A (en) * | 2022-07-20 | 2022-08-19 | 安徽炬视科技有限公司 | Image identification method based on cloud edge-end cooperation |
| CN114926667B (en) * | 2022-07-20 | 2022-11-08 | 安徽炬视科技有限公司 | Image identification method based on cloud edge cooperation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111091544B (en) | Method for detecting breakage fault of side integrated framework of railway wagon bogie | |
| CN106056619A (en) | Unmanned aerial vehicle vision wire patrol method based on gradient constraint Radon transform | |
| CN111539355A (en) | Photovoltaic panel foreign matter detection system and detection method based on deep neural network | |
| CN110910440B (en) | Power transmission line length determination method and system based on power image data | |
| CN110530876B (en) | Insulator pollution degree development prediction method based on long-term and short-term memory neural network | |
| CN108734691A (en) | A kind of transmission line of electricity defect image recognition methods | |
| CN113947555A (en) | Infrared and visible light fused visual system and method based on deep neural network | |
| CN113205039A (en) | Power equipment fault image identification and disaster investigation system and method based on multiple DCNNs | |
| Wang et al. | Insulator defect recognition based on faster R-CNN | |
| CN111667473A (en) | Insulator hydrophobicity grade judging method based on improved Canny algorithm | |
| CN113989209A (en) | Power line foreign matter detection method based on fast R-CNN | |
| CN110246131A (en) | Conducting wire share split defect image recognition methods based on convolutional neural networks | |
| CN109389152B (en) | Refined identification method for power transmission line falling object | |
| CN111415326A (en) | Method and system for detecting abnormal state of railway contact net bolt | |
| CN116309407A (en) | Method for detecting abnormal state of railway contact net bolt | |
| CN116071299A (en) | Insulator RTV spraying defect detection method and system | |
| Yang et al. | An intelligent identification algorithm for obtaining the state of power equipment in SIFT-based environments | |
| CN114037840A (en) | Power transmission line visual object extraction method and system based on multi-feature fusion | |
| CN111930982A (en) | Intelligent labeling method for power grid images | |
| Wang et al. | Research on appearance defect detection of power equipment based on improved faster-rcnn | |
| Wu et al. | Detection method based on improved faster R-CNN for pin defect in transmission lines | |
| CN117274903B (en) | Intelligent early warning device and method for electric power inspection based on intelligent AI chip | |
| Li et al. | Study on intelligent image recognition of non-linear short pointer SF6 meter readings | |
| Liu et al. | End-to-end high-speed railway dropper breakage and slack monitoring based on computer vision | |
| Jiang et al. | Optimization algorithm of transmission line defect detection and identification based on embedded AI |
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 |