CN110020598A - A kind of method and device based on foreign matter on deep learning detection electric pole - Google Patents
A kind of method and device based on foreign matter on deep learning detection electric pole Download PDFInfo
- Publication number
- CN110020598A CN110020598A CN201910150713.9A CN201910150713A CN110020598A CN 110020598 A CN110020598 A CN 110020598A CN 201910150713 A CN201910150713 A CN 201910150713A CN 110020598 A CN110020598 A CN 110020598A
- Authority
- CN
- China
- Prior art keywords
- foreign matter
- electric pole
- picture
- detection
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of method and devices based on foreign matter on deep learning detection electric pole, select the picture comprising electric pole and foreign matter of preset quantity, processing is extended to picture and obtains the picture library of presupposition multiple, indicates in each picture the location information of electric pole and foreign matter as sample database；The characteristic pattern that the 6th layer of convolutional layer of SSD object detection algorithms is extracted is as the 7th kind of grid, and then by the grid and 6 kinds of original mesh updates, as the input of last loss function, training obtains the detection model to electric pole and foreign matter；Finally the electric pole and foreign matter inputted in picture is detected simultaneously using the obtained detection model of training；And the overlapping area of electric pole and foreign matter and the first accounting of foreign matter area are calculated, the foreign matter that the first accounting is lower than preset threshold is rejected, detection serious forgiveness is promoted.Process of the present invention is more simplified, improves model to the rate of precision of wisp foreign matter.
Description
Technical field
The invention belongs to field of machine vision more particularly to a kind of methods based on foreign matter on deep learning detection electric pole
And device.
Background technique
Foreign matter (such as Bird's Nest, balloon, kite) on usual electric pole all may cause strong influence to transmission line of electricity,
Such as trigger wire short circuit, cause bulk zone power supply impaired.Therefore detection removing is carried out to the foreign matter on electric pole very must
It wants.Traditional detection method needs staff's operation on the spot, consumes a large amount of manpower and material resources, and monitoring range is also extremely limited.
The position of target object is a kind of technology hand of the supreme arrogance of a person with great power instantly in detection image in the way of deep learning
Section.It is electric in automatic identification detection image in conjunction with the ability of machine deep learning and identification by the image data of outdoor acquisition
Foreign matter in line bar can help staff to reduce workload, reduce manpower and material resources cost.
Although the prior art proposes some technical solutions for detecting track foreign matter based on deep learning, existing
Deep learning is relatively low for the object detection precision of smaller scale, and the precision of foreign bodies detection is not high.
Summary of the invention
The object of the present invention is to provide a kind of method and devices based on foreign matter on deep learning detection electric pole, improve inspection
The precision and serious forgiveness of survey.
To achieve the goals above, technical solution of the present invention is as follows:
A kind of method based on foreign matter on deep learning detection electric pole of the present invention, comprising:
The picture comprising electric pole and foreign matter for selecting preset quantity is extended processing to picture and obtains presupposition multiple
Picture library indicates in each picture the location information of electric pole and foreign matter as sample database；
The characteristic pattern that the 6th layer of convolutional layer of SSD object detection algorithms is extracted is as the 7th kind of grid, then by the grid
With 6 kinds of original mesh updates, as the input of last loss function, training obtains the detection model to electric pole and foreign matter；
Using trained obtained detection model while detecting electric pole and foreign matter in input picture；
The overlapping area of electric pole and foreign matter and the first accounting of foreign matter area are calculated, the first accounting is rejected and is lower than default threshold
The foreign matter of value promotes detection serious forgiveness.
Further, the picture is extended processing, comprising:
Left and right overturning processing is carried out to picture；
Or/and picture is carried out to spin upside down processing；
Or/and size scaling processing is carried out to picture；
Or/and random salt-pepper noise processing is carried out to picture.
Further, the training obtains the detection model to electric pole and foreign matter, further includes:
Model is trained according to preset learning rate, and when training is to preset number, to learning rate carry out according to
Secondary decaying.
The invention also provides a kind of devices based on foreign matter on deep learning detection electric pole, including processor and deposit
The step of containing the memory of several computer instructions, the above method realized when the computer instruction is executed by processor.
A kind of method and device based on foreign matter on deep learning detection electric pole proposed by the present invention, utilizes deep learning
This unsupervised segmentation mode reduces manual intervention, and compared to traditional supervised classification mode, process is more simplified.Utilize statistics
In image foreign matter is minimum wide, high accounting designs default boxes, to improve model to the accurate of wisp foreign matter
Rate.By calculating the overlap ratio of electric pole and foreign matter, to improve detection serious forgiveness.The present invention need to only be done once
Detection, the method that fine granularity is extracted again after extracting compared to universal elder generation's coarseness, efficiency are upper many fastly.
Detailed description of the invention
Fig. 1 is that the present invention is based on the method flow diagrams that deep learning detects foreign matter on electric pole；
Fig. 2 is detection model of embodiment of the present invention mesh update schematic diagram.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the application, and do not have to
In restriction the application.
In one embodiment, as shown in Figure 1, providing a kind of side based on foreign matter on deep learning detection electric pole
Method, comprising:
Step S1, select preset quantity the picture comprising electric pole and foreign matter, to picture be extended processing obtain it is pre-
If the picture library of multiple, the location information of electric pole and foreign matter is indicated in each picture as sample database.
For example, selecting about 500 or so includes the picture of electric pole and foreign matter, and the size of every picture is fixed on
(512,512) mark the XML file of a electric pole and foreign matter location information cluster then for every picture.
Every picture is all overturn respectively, is scaled, the image procossings such as random salt-pepper noise expand 20 times of picture
Amount.
The corresponding XML tag file of reverse read original image obtains the location information of electric pole and foreign matter in image, from
The location information cluster of the dynamic electric pole for calculating out expanded images and foreign matter, and XML file is saved as, it is as follows that strategy is calculated in part:
Assuming that the location information for the arbitrary objects that original image obtains are as follows: (terminating point X is sat by Sx (starting point X-coordinate), Ex
Mark), Sy (starting point Y-coordinate) and Ey (terminating point Y-coordinate), width and the height respectively W and H of image.Below TSx, TEx,
TSy and TEy then respectively represents starting point X-coordinate, terminating point X-coordinate, starting point Y-coordinate and the terminating point Y seat of image after enhancing
Mark.
Image or so overturning: TSx=W-Ex, TEx=W-Sx, TSy=Sy, TEy=Ey.
Image is spun upside down: TSx=Sx, TEx=Ex, TSy=H-Sy, TEy=H-Ey.
Picture size scaling: TSx=α * Sx, TEx=α * Ex, TSy=α * Sy, TEy=α * Ey, α are scaling.
Random salt-pepper noise: TSx=Sx, TEx=Ex, TSy=Sy, TEy=Ey.
It should be understood that the present embodiment preset quantity is about 500 or so pictures, every picture is all carried out respectively
Overturning, scaling, the image procossings such as random salt-pepper noise expand 20 times of picture amount.The number of samples pictures quantity is related to
The effect of model training, picture amount required for those skilled in the art can deduce according to experiment effect, here no longer
It repeats.
Step S2, then the characteristic pattern for extracting the 6th layer of convolutional layer of SSD object detection algorithms will as the 7th kind of grid
The grid and 6 kinds of original mesh updates, as the input of last loss function, training obtains the detection to electric pole and foreign matter
Model.
Common algorithm of target detection, such as Faster R-CNN, there is slow-footed disadvantages.SSD object detection algorithms
(Single Shot MultiBox Detector) not only increases speed, and improves accuracy.In SSD algorithm, Gu
The fixed grid (default boxes) that picture is divided into 6 kinds of different sizes, then all carries out every kind of default boxes
The recurrence of position and classification, therefore the algorithm has relatively good adaptability to multiple dimensioned object.
By the high accounting and wide accounting of foreign matter in statistical sample picture, the two minimum value is taken, it is found that the minimum value is far small
The smallest high accounting of 6 kinds of fixed default boxes and wide accounting in SSD, therefore the present embodiment calculation of adjustment SSD
Method structure increases by one layer of additional default boxes, is slightly less than the high accounting of the default boxes and wide accounting all
The smallest high accounting and wide accounting in sample.
Specifically, the present embodiment reverse read XML tag file first takes the position coordinates cluster of foreign matter.Then basis
The position of foreign matter calculates the minimum scale value of foreign matter size and dimension of picture, rounds up to obtain to the inverse of minimum scale value
The grid number of SSD object detection algorithms.
Specifically carry out following calculating:
Calculate foreign matter width: W1=Ex-Sx；
Calculate foreign matter height: H1=Ey-Sy；
It brings formula P=Min (W1/W, H1/H) into, records P value.
Repeat above-mentioned calculating, until finding out Pmin=Min (P1, P2, P3 ..., Pn), (n ∈ foreign matter total amount).
Then the inverse for seeking Pmin again, rounds up to obtain: Q=Ceil (1/Pmin).
Seek the 2 minimum power exponent N=f (N) not less than Pmin.
Original SSD algorithm classification and position frame recurrence on use 6 kinds of default boxes input, such one
Picture, which is equivalent to, has been divided into 6 parts of various sizes of grid charts, and the dimension of object that every part of grid chart can be experienced is not yet
Together.
Using the N value (calculated N=6 in implementation) calculated above come out, with 2 n times power (n ∈ 0,1,2 ..., N)
Mode design default boxes, then, it is known that final default boxes should actually have (1,1), (2,2), (4,
4), (8,8), (16,16), (32,32) and (64,64) this 7 kinds of specifications, that is, the net for the SSD object detection algorithms being calculated
Lattice number is 7, however the loss of original SSD algorithm input is fixed as 6 kinds of default boxes, has lacked (64,64) just
default boxes。
As shown in Fig. 2, the characteristic pattern (feature map) that the present embodiment extracts the 6th layer of convolutional layer is as the 7th kind
Then default boxes merges the default boxes and 6 kinds of original default boxes, as last loss
The input of function promotes the object detection precision to smaller scale with this.The SSD object detection algorithms that ought be calculated
Grid number when being greater than 6, the characteristic pattern (feature map) that the 6th layer of convolutional layer is extracted is as the 7th kind of default
Then boxes merges the default boxes and 6 kinds of original default boxes, as the defeated of last loss function
Enter.And if calculated when the grid number for the SSD object detection algorithms being calculated is less than or equal to 6 according to original SSD object detection
Method trains detection model.
The network that SSD algorithm extracts feature is made of 18 convolutional layers, and each convolutional layer includes the convolution of one (3,3)
Core.Original SSD algorithm fixation can obtain the feature vector (feature map) of image from the convolutional layer of 6 different dimensions,
The mesh shape that each feature vector can seem diagram picture divides (default boxes), and final algorithm
Classification will be carried out to the presentation object in each grid and coordinate returns.Object size to be detected is centainly always in one width figure
Not of uniform size, and by carrying out grid dividing to image, as long as the scale divided is enough, then can theoretically cover
The object of arbitrary dimension.Based on such imagination, the wide accounting of height of foreign matter in sample is then calculated, finds the smallest foreign matter shape
Shape is much smaller than the smallest original mesh shape, therefore original Gridding Method is not particularly suited for current sample, needs to do more
Small grid dividing.And in implementing, 6 kinds of original grid dividings are (1,1) respectively, (2,2), (4,4), (8,8), (16,16),
(32,32) hereafter need the form Expanding grid shape with common ratio 2, and (64,64) dimensional vector that layer 6 convolutional layer extracts is rigid
Meet current sample well, therefore the feature vector that the present embodiment extracts the 6th layer of convolutional layer is as the 7th kind of mesh shape, and will
Original 6 kinds of this feature vector sum are incorporated as the input of last loss function.
In one embodiment, when training obtains the detection model to electric pole and foreign matter, further includes:
Model is trained according to preset learning rate, and when training is to preset number, to learning rate carry out according to
Secondary decaying.
For example, learning rate is adjusted to 0.001, model is trained, iteration 120000 times, in 80000,100000 time-divisions
Learning rate is not decayed, each attenuation rate is 0.9, obtains the detection model to electric pole and foreign matter.
Step S3, using trained obtained detection model while electric pole and foreign matter in input picture being detected.
The present embodiment is after training obtains to the detection model of electric pole and foreign matter, so that it may be contained using picture pick-up device acquisition
The picture of live wire bar, to detect electric pole and foreign matter, is realized to electric pole so that the picture that will acquire is input to detection model
The detection of upper foreign matter.
Step S4, the overlapping area of electric pole and foreign matter and the first accounting of foreign matter area are calculated, it is low to reject the first accounting
In the foreign matter of preset threshold, detection serious forgiveness is promoted.
The present embodiment is after the electric pole and foreign matter in the model inspection image obtained using training, and there are false positives for discovery
The case where foreign matter, and practical most and electric pole the band of position of these false positive foreign matters does not have intersection or intersection very little.
Therefore the present embodiment by calculating the overlapping area of electric pole and foreign matter and the first accounting of foreign matter area two-by-two
Overlap ratio=(S electric pole ∩ S foreign matter)/S foreign matter, selection filter out foreign matter of the overlap ratio lower than 0.9 and come
A large amount of false positive test results are excluded, to promote serious forgiveness.
The method for calculating overlap ratio is as follows:
Assuming that the band of position of certain electric pole is (Px1, Py1, Px2, Py2), the band of position of foreign matter be (Tx1, Ty1,
Tx2, Ty2), then area St=(Tx2-Tx1) * (Ty2-Ty1) of foreign matter, overlap area So=(Min (Px2,
Tx2)-Max (Px1, Tx1)) * (Min (Py2, Ty2)-Max (Py1, Ty1)), then overlap ratio=So/St.It is comprehensive
The overlap ratio of foreign matter in all samples is counted, discovery can obtain goodr effect as overlap ratio > 0.9
Fruit.
It should be understood that although each step in the flow chart of Fig. 1 is successively shown according to the instruction of arrow, this
A little steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly state otherwise herein, these steps
It executes there is no the limitation of stringent sequence, these steps can execute in other order.Moreover, at least part in Fig. 2
Step may include that perhaps these sub-steps of multiple stages or stage are executed in synchronization to multiple sub-steps
It completes, but can execute at different times, the execution sequence in these sub-steps or stage, which is also not necessarily, successively to be carried out,
But it can be executed in turn or alternately at least part of the sub-step or stage of other steps or other steps.
In one embodiment, present invention also provides it is a kind of based on deep learning detection electric pole on foreign matter device,
Including processor and it is stored with the memories of several computer instructions, the computer instruction realizes institute when being executed by processor
State the step of method of foreign matter on electric pole is detected based on deep learning.
It may refer to about the specific restriction based on the device of foreign matter on deep learning detection electric pole above for base
In the restriction of the method for foreign matter on deep learning detection electric pole, details are not described herein.It is above-mentioned that electric wire is detected based on deep learning
Modules on bar in the device of foreign matter can be realized fully or partially through software, hardware and combinations thereof.Above-mentioned each module
It can be embedded in the form of hardware or independently of in the processor in computer equipment, computer can also be stored in a software form
In memory in equipment, the corresponding operation of the above modules is executed in order to which processor calls.
It is directly or indirectly electrically connected between memory and processor, to realize the transmission or interaction of data.For example, this
A little elements can be realized by one or more communication bus or signal wire be electrically connected from each other.Being stored in memory can
The computer program run on a processor, the computer program that the processor is stored in memory by operation, thus
Realize the network topology layout method in the embodiment of the present invention.
Wherein, the memory may be, but not limited to, random access memory (Random Access Memory,
RAM), read-only memory (Read Only Memory, ROM), programmable read only memory (Programmable Read-Only
Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM),
Electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only Memory, EEPROM) etc..
Wherein, memory is for storing program, and the processor executes described program after receiving and executing instruction.
The processor may be a kind of IC chip, the processing capacity with data.Above-mentioned processor can be with
It is general processor, including central processing unit (Central Processing Unit, CPU), network processing unit (Network
Processor, NP) etc..It may be implemented or execute each method, step disclosed in the embodiment of the present invention and logic diagram.It is logical
It can be microprocessor with processor or the processor be also possible to any conventional processor etc..
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (4)
1. a kind of method based on foreign matter on deep learning detection electric pole, which is characterized in that described to be detected based on deep learning
The method of foreign matter on electric pole, comprising:
The picture comprising electric pole and foreign matter for selecting preset quantity is extended processing to picture and obtains the picture of presupposition multiple
Library indicates in each picture the location information of electric pole and foreign matter as sample database；
The characteristic pattern that the 6th layer of convolutional layer of SSD object detection algorithms is extracted is as the 7th kind of grid, then by the grid and original
The 6 kinds of mesh updates to begin, as the input of last loss function, training obtains the detection model to electric pole and foreign matter；
Using trained obtained detection model while detecting electric pole and foreign matter in input picture；
The overlapping area of electric pole and foreign matter and the first accounting of foreign matter area are calculated, rejects the first accounting lower than preset threshold
Foreign matter promotes detection serious forgiveness.
2. the method according to claim 1 based on foreign matter on deep learning detection electric pole, which is characterized in that the figure
Piece is extended processing, comprising:
Left and right overturning processing is carried out to picture；
Or/and picture is carried out to spin upside down processing；
Or/and size scaling processing is carried out to picture；
Or/and random salt-pepper noise processing is carried out to picture.
3. the method according to claim 1 based on foreign matter on deep learning detection electric pole, which is characterized in that the instruction
Get the detection model to electric pole and foreign matter, further includes:
Model is trained according to preset learning rate, and when preset number is arrived in training, is successively declined to learning rate
Subtract.
4. a kind of device based on foreign matter on deep learning detection electric pole, including processor and it is stored with several computers and refers to
The memory of order, which is characterized in that realized in claim 1 to claim 3 when the computer instruction is executed by processor
The step of any one the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910150713.9A CN110020598B (en) | 2019-02-28 | 2019-02-28 | Method and device for detecting foreign matters on telegraph pole based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910150713.9A CN110020598B (en) | 2019-02-28 | 2019-02-28 | Method and device for detecting foreign matters on telegraph pole based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110020598A true CN110020598A (en) | 2019-07-16 |
| CN110020598B CN110020598B (en) | 2022-04-15 |
Family
ID=67189172
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910150713.9A Active CN110020598B (en) | 2019-02-28 | 2019-02-28 | Method and device for detecting foreign matters on telegraph pole based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110020598B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110826647A (en) * | 2019-12-09 | 2020-02-21 | 国网智能科技股份有限公司 | Method and system for automatically detecting foreign matter appearance of power equipment |
| CN113670929A (en) * | 2021-07-05 | 2021-11-19 | 国网宁夏电力有限公司电力科学研究院 | Power transmission line foreign matter detection method and device, storage medium and terminal equipment |
| CN116883648A (en) * | 2023-09-06 | 2023-10-13 | 南方电网数字电网研究院有限公司 | Foreign matter detection method and device, electronic equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202024734U (en) * | 2011-04-18 | 2011-11-02 | 四川电力科学研究院 | Electric wire detection device |
| CN106372662A (en) * | 2016-08-30 | 2017-02-01 | 腾讯科技（深圳）有限公司 | Helmet wearing detection method and device, camera, and server |
| CN106683075A (en) * | 2016-11-22 | 2017-05-17 | 广东工业大学 | Power transmission line tower cross arm bolt defect detection method |
| CN106778472A (en) * | 2016-11-17 | 2017-05-31 | 成都通甲优博科技有限责任公司 | The common invader object detection and recognition method in transmission of electricity corridor based on deep learning |
| CN107784634A (en) * | 2017-09-06 | 2018-03-09 | 广东工业大学 | A kind of power transmission line shaft tower Bird's Nest recognition methods based on template matches |
| CN108038846A (en) * | 2017-12-04 | 2018-05-15 | 国网山东省电力公司电力科学研究院 | Transmission line equipment image defect detection method and system based on multilayer convolutional neural networks |
| CN108197610A (en) * | 2018-02-02 | 2018-06-22 | 北京华纵科技有限公司 | A kind of track foreign matter detection system based on deep learning |
| CN108304807A (en) * | 2018-02-02 | 2018-07-20 | 北京华纵科技有限公司 | A kind of track foreign matter detecting method and system based on FPGA platform and deep learning |
-
2019
- 2019-02-28 CN CN201910150713.9A patent/CN110020598B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202024734U (en) * | 2011-04-18 | 2011-11-02 | 四川电力科学研究院 | Electric wire detection device |
| CN106372662A (en) * | 2016-08-30 | 2017-02-01 | 腾讯科技（深圳）有限公司 | Helmet wearing detection method and device, camera, and server |
| CN106778472A (en) * | 2016-11-17 | 2017-05-31 | 成都通甲优博科技有限责任公司 | The common invader object detection and recognition method in transmission of electricity corridor based on deep learning |
| CN106683075A (en) * | 2016-11-22 | 2017-05-17 | 广东工业大学 | Power transmission line tower cross arm bolt defect detection method |
| CN107784634A (en) * | 2017-09-06 | 2018-03-09 | 广东工业大学 | A kind of power transmission line shaft tower Bird's Nest recognition methods based on template matches |
| CN108038846A (en) * | 2017-12-04 | 2018-05-15 | 国网山东省电力公司电力科学研究院 | Transmission line equipment image defect detection method and system based on multilayer convolutional neural networks |
| CN108197610A (en) * | 2018-02-02 | 2018-06-22 | 北京华纵科技有限公司 | A kind of track foreign matter detection system based on deep learning |
| CN108304807A (en) * | 2018-02-02 | 2018-07-20 | 北京华纵科技有限公司 | A kind of track foreign matter detecting method and system based on FPGA platform and deep learning |
Non-Patent Citations (2)
| Title |
|---|
| WEI LIU等: ""SSD: Single Shot MultiBox Detector"", 《ARXIV》 * |
| 刘士波: ""基于数字图像的输电线故障识别与定位方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110826647A (en) * | 2019-12-09 | 2020-02-21 | 国网智能科技股份有限公司 | Method and system for automatically detecting foreign matter appearance of power equipment |
| CN113670929A (en) * | 2021-07-05 | 2021-11-19 | 国网宁夏电力有限公司电力科学研究院 | Power transmission line foreign matter detection method and device, storage medium and terminal equipment |
| CN113670929B (en) * | 2021-07-05 | 2024-06-14 | 国网宁夏电力有限公司电力科学研究院 | Power transmission line foreign matter detection method and device, storage medium and terminal equipment |
| CN116883648A (en) * | 2023-09-06 | 2023-10-13 | 南方电网数字电网研究院有限公司 | Foreign matter detection method and device, electronic equipment and storage medium |
| CN116883648B (en) * | 2023-09-06 | 2024-02-13 | 南方电网数字电网研究院股份有限公司 | Foreign matter detection method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110020598B (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112819804B (en) | Insulator defect detection method based on improved YOLOv convolutional neural network | |
| CN107358149B (en) | Human body posture detection method and device | |
| CN110188720A (en) | A kind of object detection method and system based on convolutional neural networks | |
| CN105279484B (en) | Method for checking object and object test equipment | |
| CN108710913A (en) | A kind of switchgear presentation switch state automatic identification method based on deep learning | |
| CN110020598A (en) | A kind of method and device based on foreign matter on deep learning detection electric pole | |
| CN107563412A (en) | A kind of infrared image power equipment real-time detection method based on deep learning | |
| CN110378218A (en) | A kind of image processing method, device and terminal device | |
| CN108182695B (en) | Target tracking model training method and device, electronic equipment and storage medium | |
| CN109886951A (en) | Method for processing video frequency, device and electronic equipment | |
| CN106780551B (en) | A kind of Three-Dimensional Moving Targets detection method and system | |
| CN114140683A (en) | Aerial image target detection method, equipment and medium | |
| CN109165589A (en) | Vehicle based on deep learning recognition methods and device again | |
| CN110032916A (en) | A kind of method and apparatus detecting target object | |
| CN111382808A (en) | Vehicle detection processing method and device | |
| CN110084175A (en) | A kind of object detection method, object detecting device and electronic equipment | |
| CN106407908A (en) | Training model generation method and human face detection method and device | |
| CN109671055B (en) | Pulmonary nodule detection method and device | |
| CN110148117A (en) | Power equipment defect identification method and device based on power image and storage medium | |
| CN110245555A (en) | A kind of electric system terminal box condensation determination method and system based on image recognition | |
| CN111325204A (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN111899279A (en) | Method and device for detecting motion speed of target object | |
| CN110796135A (en) | Target positioning method and device, computer equipment and computer storage medium | |
| CN108961267A (en) | Image processing method, picture processing unit and terminal device | |
| CN107704797A (en) | Real-time detection method and system and equipment based on pedestrian in security protection video and vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |