CN110175538A - A kind of substation's Bird's Nest recognition methods and system based on machine learning - Google Patents
A kind of substation's Bird's Nest recognition methods and system based on machine learning Download PDFInfo
- Publication number
- CN110175538A CN110175538A CN201910389617.XA CN201910389617A CN110175538A CN 110175538 A CN110175538 A CN 110175538A CN 201910389617 A CN201910389617 A CN 201910389617A CN 110175538 A CN110175538 A CN 110175538A
- Authority
- CN
- China
- Prior art keywords
- bird
- substation
- nest
- image
- video image
- 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
- 244000000626 Daucus carota Species 0.000 title claims abstract description 87
- 235000005770 birds nest Nutrition 0.000 title claims abstract description 87
- 235000005765 wild carrot Nutrition 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000010801 machine learning Methods 0.000 title claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 21
- 238000013527 convolutional neural network Methods 0.000 claims description 19
- 230000003321 amplification Effects 0.000 claims description 9
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 9
- 238000007689 inspection Methods 0.000 claims description 8
- 239000000284 extract Substances 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000010835 comparative analysis Methods 0.000 claims description 5
- 238000013135 deep learning Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 238000000605 extraction Methods 0.000 abstract description 4
- 230000002411 adverse Effects 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 description 7
- 230000033001 locomotion Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 239000010902 straw Substances 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 244000025254 Cannabis sativa Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 229910052573 porcelain Inorganic materials 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The substation's Bird's Nest recognition methods and system, method that the present invention relates to a kind of based on machine learning include: image preprocessing, receive the input of transformer substation video image, and denoised and enhanced to improve the contrast of target and background to video image；Data format processing, is converted into the TFRecord file with binary storage for pretreated image；Image recognition, based on Faster R-CNN algorithm and LSTM network struction substation Bird's Nest identification model；The video image after data format analysis processing is handled based on trained substation's Bird's Nest identification model, obtains recognition result.The present invention is for the hidden features such as changeable in Bird's Nest position in substation, it is proposed that the algorithm based on Faster R-CNN algorithm and LSTM network identifies substation's Bird's Nest, with good ability in feature extraction, adverse effect of the birds activity to substation equipment can be reduced, the reliability of substation operation is improved.
Description
Technical field
The present invention relates to intelligent substation patrol technical field, in particular to a kind of substation's Bird's Nest based on machine learning
Recognition methods and system.
Background technique
Currently, substation realizes video image all standing, and using artificial or part automatic video frequency routine inspection mode into
Row real time monitoring.However, this mode causes monitoring personnel tired, inefficiency cannot find transmission line of electricity and periphery in time
The abnormal conditions of appearance, therefore be inevitable development trend by the way of intelligent patrol detection.Birds are on overhead transmission line and equipment
Problem of nesting it is very universal, Birds In Spring breeding during it is especially prominent.Statistical data shows birds to nest a little to be located at mostly
Disconnecting link joint on base interior, each phase upside-down mounting porcelain vase support seat, cross arm of tower and tower head.Bird's Nest is mainly by straw, branch
It constitutes, easily causes the dielectric strength of substation equipment to reduce in wet weather or rainy season.In addition, long straw is also extremely easy to drop and hangs
It is connected to each alternate, touches electrical body and in turn result in line short, tripping, and birds are between substation line framework and equipment
Movable same easy initiation electric fault.
It is different since Bird's Nest is not of uniform size, and Bird's Nest captured by monitoring system of electric substation is usually imperfect, part
Even human eye is all difficult to distinguish.In addition, the video image background of substation is complicated, only pass through simple video image processing algorithm
The trees being difficult to differentiate between in straw and background.Existing substation's Bird's Nest recognition methods mostly according to the special texture in Bird's Nest region,
Color, shape distinguish, but branch, thick grass etc. also have similar texture features, and grey is presented in color, interfere larger, effect
Fruit is unsatisfactory.
Summary of the invention
It is an object of the invention to overcome the deficiencies in the prior art, propose that a kind of substation's Bird's Nest based on machine learning is known
Other method and system are able to ascend existing video frequency graphic monitoring technical level, ensure that substation safety is reliably run.
The technical solution adopted by the present invention to solve the technical problems is:
On the one hand, a kind of substation's Bird's Nest recognition methods based on machine learning of the present invention, comprising:
Image preprocessing receives the input of transformer substation video image, and is denoised and enhanced to improve mesh to video image
The contrast of mark and background；
Data format processing, is converted into the TFRecord file with binary storage for pretreated image；
Image recognition, based on Faster R-CNN algorithm and LSTM network struction substation Bird's Nest identification model；Based on instruction
The substation's Bird's Nest identification model perfected handles the video image after data format analysis processing, obtains recognition result.
Preferably, described image, which pre-processes, includes:
By carrying out greyscale transformation to image, noise-containing gradient image is obtained, then using small probability strategy and most
Theory of error criterion is split image between major class, obtains two parts of target and background, then use the fractional order of different orders
Calculus mask handles each region, obtains self-adaptive solution and enhanced image.
Preferably, when carrying out substation's Bird's Nest identification model training, by the data of rotation to training sample figure
As carrying out amplification operation, limited training sample is expanded；It specially collects and is sorted out containing Bird's Nest from data bank
Video image, and carry out comparative analysis three times respectively, mark out the position of target manually in the form of rectangle candidate frame；So
Determine whether the position of institute's label target in every picture is correct afterwards.
Preferably, the recognition result includes output target area, target type and corresponding probability, to video image
The middle maximum target area of occupied area extracts, and obtains Bird's Nest video image.
Preferably, the training process of substation's Bird's Nest identification model includes:
(1) input marking has the substation inspection video image of Bird's Nest, and original video image is scaled 245 × 205, together
When calculate VGG-16 the last layer convolutional layer characteristic pattern size；
(2) coordinate for reading the target frame in flag data, adjusts and saves according to scaling in previous step；
(3) the bounding box size and ratio of default rectangle candidate frame are preset；
(4) with 1 × 1 window scanning feature figure, each window is an anchor point, i.e. bounding box central point；
(5) coordinate of anchor point is mapped in the picture by scaling by preset ratio amplification, obtains actual coordinate；
(6) according to the bounding box size of bounding box coordinates in flag data and update, new bounding box and corresponding is generated
Score；
(7) classify to the video image of bounding box new in step (6), and using LSTM network to efficiency frontier frame
Video image multiplexing and generate multiple sequence signatures, generate class probability update VGG-16 convolutional neural networks parameter.
On the other hand, a kind of substation's Bird's Nest identifying system based on machine learning of the present invention, comprising:
Image pre-processing module for receiving the input of transformer substation video image, and is denoised and is enhanced to video image
To improve the contrast of target and background；
Data format processing module, for pretreated image to be converted into the TFRecord text with binary storage
Part；
Deep learning identification module, for being identified based on Faster R-CNN algorithm and LSTM network struction substation's Bird's Nest
Model；The video image after data format analysis processing is handled based on trained substation's Bird's Nest identification model, is known
Other result.
Preferably, described image, which pre-processes, includes:
By carrying out greyscale transformation to image, noise-containing gradient image is obtained, then using small probability strategy and most
Theory of error criterion is split image between major class, obtains two parts of target and background, then use the fractional order of different orders
Calculus mask handles each region, obtains self-adaptive solution and enhanced image.
Preferably, when carrying out substation's Bird's Nest identification model training, by the data of rotation to training sample figure
As carrying out amplification operation, limited training sample is expanded；It specially collects and is sorted out containing Bird's Nest from data bank
Video image, and carry out comparative analysis three times respectively, mark out the position of target manually in the form of rectangle candidate frame；So
Determine whether the position of institute's label target in every picture is correct afterwards.
Preferably, the recognition result includes output target area, target type and corresponding probability, to video image
The middle maximum target area of occupied area extracts, and obtains Bird's Nest video image.
Preferably, the training process of substation's Bird's Nest identification model includes:
(1) input marking has the substation inspection video image of Bird's Nest, and original video image is scaled 245 × 205, together
When calculate VGG-16 the last layer convolutional layer characteristic pattern size；
(2) coordinate for reading the target frame in flag data, adjusts and saves according to scaling in previous step；
(3) the bounding box size and ratio of default rectangle candidate frame are preset；
(4) with 1 × 1 window scanning feature figure, each window is an anchor point, i.e. bounding box central point；
(5) coordinate of anchor point is mapped in the picture by scaling by preset ratio amplification, obtains actual coordinate；
(6) according to the bounding box size of bounding box coordinates in flag data and update, new bounding box and corresponding is generated
Score；
(7) classify to the video image of bounding box new in step (6), and using LSTM network to efficiency frontier frame
Video image multiplexing and generate multiple sequence signatures, generate class probability update VGG-16 convolutional neural networks parameter.
The invention has the following beneficial effects:
(1) a kind of substation's Bird's Nest recognition methods and system based on machine learning of the present invention, in substation
Bird's Nest position is hidden changeable and Bird's Nest textural characteristics are unobvious, proposes the improvement based on Faster R-CNN algorithm and LSTM network
Algorithm identifies have good ability in feature extraction, adapt to changeable complex environment, help to drop to substation's Bird's Nest
Adverse effect and harm of the low birds activity to substation equipment and operational safety, improve the reliability of substation operation；
(2) a kind of substation's Bird's Nest recognition methods and system based on machine learning of the present invention regards the substation of input
Frequency image is denoised and is enhanced, and noise and clutter can be reduced, and improves the contrast of target and background；
(3) a kind of substation's Bird's Nest recognition methods and system based on machine learning of the present invention, by pretreated image
It is converted into the TFRecord file with binary storage, is suitble to read high-volume data in a serial fashion, is conducive to storage, duplication
And movement.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those skilled in the art without any creative labor, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow chart of substation's Bird's Nest recognition methods based on machine learning of the embodiment of the present invention；
Fig. 2 is the Faster R-CNN detection figure of the embodiment of the present invention；
Fig. 3 is that substation's Bird's Nest identification model of the embodiment of the present invention carries out actual field substation inspection video image
Target identification result figure；
Fig. 4 is the structure chart of substation's Bird's Nest identifying system based on machine learning of the embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
It is shown in Figure 1, on the one hand, a kind of substation's Bird's Nest recognition methods based on machine learning of the present invention, comprising:
S101, image preprocessing receive the input of transformer substation video image, and are denoised and enhanced to mention to video image
The contrast of high target and background；
S102, data format processing, is converted into the TFRecord file with binary storage for pretreated image；
S103, image recognition, based on Faster R-CNN algorithm and LSTM network struction substation Bird's Nest identification model；Base
The video image after data format analysis processing is handled in trained substation's Bird's Nest identification model, obtains recognition result.
Specifically, a kind of substation's Bird's Nest recognition methods based on machine learning of the present invention, can be realized video image intelligence
It can analyze, adapt to environmental system complicated and changeable, improve existing video frequency graphic monitoring technical level, ensure that substation safety can
The operation leaned on.Described image pretreatment is denoised and is enhanced to input picture, to reduce noise and clutter, improves target and back
The contrast of scape is realized by carrying out greyscale transformation to image, obtains noise-containing gradient image, then utilize small probability
Theory of error criterion is split image between strategy and maximum kind, obtains two parts of target and background, then use different orders
Fractional calculus mask each region is handled, obtain self-adaptive solution and enhanced image.The data lattice
Video image is converted into the TFRecord format with binary storage by programming by formula processing, is suitble to read in a serial fashion
High-volume data are conducive to storage, duplication and movement；Described image identification is led to based on Faster R-CNN algorithm and LSTM network
Cross training sample and feature extraction and classification carried out to substation's Bird's Nest, such as the shape of Bird's Nest, the dendritic structure at Bird's Nest edge and
The color etc. of Bird's Nest establishes substation's Bird's Nest identification model；Finally by trained substation's Bird's Nest identification model to substation
Video image is analyzed, and recognition result is obtained.
Further, when model training, operation is expanded by the data of rotation to sample image, by limited training sample
(Bird's Nest video image collection) is expanded.Concrete operations are as follows: the video image containing Bird's Nest is collected and sorted out from data bank,
And comparative analysis three times is carried out respectively, mark out the position of target manually in the form of rectangle candidate frame；Then every is determined
Whether the position of institute's label target is correct in picture, can be significantly reduced the time of sample mark, improves working efficiency.
The feature extraction of substation's Bird's Nest uses Faster R-CNN algorithm, to the confidence level auto-sequencing of recognition result,
By comparing the overlapping area and confidence value of recognition result two-by-two, redundance is removed.This processing mode is in mesh to be identified
Target overlapping region very hour significant effect.But when identifying target in the presence of blocking, maximum probability can be removed, if adjustment faying surface
Product judges parameter, then will appear a large amount of repetition recognition result.Therefore the present invention uses LSTM on Faster R-CNN algorithm
Network makes a network convolution characteristic parameter multiplexing and generates multiple sequence signatures, to characterize the target knowledge for blocking video image
Not as a result, to improve recognition accuracy.
Faster R-CNN mainly consists of two parts, and first is to propose region motion (Region Proposal
Net-works, RPN) the full convolutional network of machine, second is detector, to RPN generate region motion carry out target detection.
As shown in Fig. 2, video image passes through the processing of CNN model, convolution characteristic pattern is obtained, then being found by RPN may include object
Region, export as the list of rectangle candidate frame and recall rate.The target of RPN is relevant to object in building needle video image
The rectangle candidate frame of different sizes, different proportion, then judge whether candidate frame includes that object (might not include by detector
Complete target), thus calculate recall rate (0-1 points).
The present invention is using trained VGG-16 convolutional neural networks model on ImageNet, only due to Faster R-CNN
Needing to identify a kind of target --- Bird's Nest, so the identifier of VGG-16 is changed to single output, i.e. RPN generates defeated in candidate region
Object is the probability of Bird's Nest out.
Further, the training process of substation's Bird's Nest identification model includes:
(1) input marking has the substation inspection video image of Bird's Nest, and original video image is scaled 245 × 205, together
When calculate VGG-16 the last layer convolutional layer characteristic pattern size；
(2) coordinate for reading the target frame in flag data, adjusts and saves according to scaling in previous step；
(3) the bounding box size and ratio of default rectangle candidate frame are preset；
(4) with 1 × 1 window scanning feature figure, each window is an anchor point, i.e. bounding box central point；
(5) coordinate of anchor point is mapped in the picture by scaling by preset ratio amplification, obtains actual coordinate；
(6) according to the bounding box size of bounding box coordinates in flag data and update, new bounding box and corresponding is generated
Score；
(7) classify to the video image of bounding box new in step (6), and using LSTM network to efficiency frontier frame
Video image multiplexing and generate multiple sequence signatures, generate class probability update VGG-16 convolutional neural networks parameter.
It is shown in Figure 3 to train the substation's Bird's Nest identification model completed to actual field substation inspection video to use
The result of image progress target identification.Model output is target area (rectangle frame in figure), target class after inputted video image
Type and corresponding probability extract the maximum target of occupied area in video image, obtain Bird's Nest video image.
Hardware components of the invention are mainly video analytics server, and using Ubuntu operating system, method passes through software
It realizes, runnable interface is built based on QT+OpenCV；Video image data is transmitted by substation's aided synthesis monitoring system host
To video analytics server, result is informed into substation's aided synthesis monitoring system host after differentiating to video image, and is realized
Automatic early-warning.
It is shown in Figure 4, on the other hand, a kind of substation's Bird's Nest identifying system based on machine learning of the present invention, comprising:
Image pre-processing module for receiving the input of transformer substation video image, and is denoised and is enhanced to video image
To improve the contrast of target and background；
Data format processing module, for pretreated image to be converted into the TFRecord text with binary storage
Part；
Deep learning identification module, for based on trained substation's Bird's Nest identification model to data format analysis processing after
Video image is handled, output target area, target type and corresponding probability, maximum to occupied area in video image
Target area extract, obtain Bird's Nest video image.
The specific implementation of each module and a kind of machine is based in a kind of substation's Bird's Nest identifying system based on machine learning
Substation's Bird's Nest recognition methods of study is identical, and the embodiment of the present invention is not repeated to illustrate.
Specific embodiment is applied in the present invention, and principle and implementation of the present invention are described, above embodiments
Explanation be merely used to help understand method and its core concept of the invention；At the same time, for those skilled in the art,
According to the thought of the present invention, there will be changes in the specific implementation manner and application range, in conclusion in this specification
Appearance should not be construed as limiting the invention.
Claims (10)
1. a kind of substation's Bird's Nest recognition methods based on machine learning characterized by comprising
Image preprocessing, receive transformer substation video image input, and video image is denoised and enhance with improve target with
The contrast of background；
Data format processing, is converted into the TFRecord file with binary storage for pretreated image；
Image recognition, based on Faster R-CNN algorithm and LSTM network struction substation Bird's Nest identification model；Based on training
Substation's Bird's Nest identification model the video image after data format analysis processing is handled, obtain recognition result.
2. substation's Bird's Nest recognition methods according to claim 1 based on machine learning, which is characterized in that described image
Pretreatment includes:
By carrying out greyscale transformation to image, noise-containing gradient image is obtained, small probability strategy and maximum kind are then utilized
Between the theory of error criterion image is split, obtain two parts of target and background, then use the micro- product of fractional order of different orders
Divide mask to handle each region, obtains self-adaptive solution and enhanced image.
3. substation's Bird's Nest recognition methods according to claim 1 based on machine learning, which is characterized in that carrying out institute
When stating the training of substation's Bird's Nest identification model, amplification operation is carried out to training sample image by the data of rotation, it will be limited
Training sample is expanded；It is specially collected from data bank and sorts out the video image containing Bird's Nest, and carry out three respectively
Secondary comparative analysis marks out the position of target manually in the form of rectangle candidate frame；Then it determines and is marked in every picture
Whether the position of target is correct.
4. substation's Bird's Nest recognition methods according to claim 1 based on machine learning, which is characterized in that the identification
It as a result include output target area, target type and corresponding probability, to the maximum target area of occupied area in video image
Domain extracts, and obtains Bird's Nest video image.
5. substation's Bird's Nest recognition methods according to claim 1 based on machine learning, which is characterized in that the power transformation
The training process for Bird's Nest identification model of standing includes:
(1) input marking has the substation inspection video image of Bird's Nest, and original video image is scaled 245 × 205, counts simultaneously
Calculate the characteristic pattern size of VGG-16 the last layer convolutional layer；
(2) coordinate for reading the target frame in flag data, adjusts and saves according to scaling in previous step；
(3) the bounding box size and ratio of default rectangle candidate frame are preset；
(4) with 1 × 1 window scanning feature figure, each window is an anchor point, i.e. bounding box central point；
(5) coordinate of anchor point is mapped in the picture by scaling by preset ratio amplification, obtains actual coordinate；
(6) according to the bounding box size of bounding box coordinates in flag data and update, new bounding box and corresponding score are generated；
(7) classify to the video image of bounding box new in step (6), and using LSTM network to the view of efficiency frontier frame
Frequency image is multiplexed and generates multiple sequence signatures, generates the parameter that class probability updates VGG-16 convolutional neural networks.
6. a kind of substation's Bird's Nest identifying system based on machine learning characterized by comprising
Image pre-processing module for receiving the input of transformer substation video image, and denoises video image and is enhanced to mention
The contrast of high target and background；
Data format processing module, for pretreated image to be converted into the TFRecord file with binary storage；
Deep learning identification module, for identifying mould based on Faster R-CNN algorithm and LSTM network struction substation's Bird's Nest
Type；The video image after data format analysis processing is handled based on trained substation's Bird's Nest identification model, is identified
As a result.
7. substation's Bird's Nest identifying system according to claim 6 based on machine learning, which is characterized in that described image
Pretreatment includes:
By carrying out greyscale transformation to image, noise-containing gradient image is obtained, small probability strategy and maximum kind are then utilized
Between the theory of error criterion image is split, obtain two parts of target and background, then use the micro- product of fractional order of different orders
Divide mask to handle each region, obtains self-adaptive solution and enhanced image.
8. substation's Bird's Nest identifying system according to claim 6 based on machine learning, which is characterized in that carrying out institute
When stating the training of substation's Bird's Nest identification model, amplification operation is carried out to training sample image by the data of rotation, it will be limited
Training sample is expanded；It is specially collected from data bank and sorts out the video image containing Bird's Nest, and carry out three respectively
Secondary comparative analysis marks out the position of target manually in the form of rectangle candidate frame；Then it determines and is marked in every picture
Whether the position of target is correct.
9. substation's Bird's Nest identifying system according to claim 6 based on machine learning, which is characterized in that the identification
It as a result include output target area, target type and corresponding probability, to the maximum target area of occupied area in video image
Domain extracts, and obtains Bird's Nest video image.
10. substation's Bird's Nest identifying system according to claim 6 based on machine learning, which is characterized in that the change
The training process of power station Bird's Nest identification model includes:
(1) input marking has the substation inspection video image of Bird's Nest, and original video image is scaled 245 × 205, counts simultaneously
Calculate the characteristic pattern size of VGG-16 the last layer convolutional layer；
(2) coordinate for reading the target frame in flag data, adjusts and saves according to scaling in previous step；
(3) the bounding box size and ratio of default rectangle candidate frame are preset；
(4) with 1 × 1 window scanning feature figure, each window is an anchor point, i.e. bounding box central point；
(5) coordinate of anchor point is mapped in the picture by scaling by preset ratio amplification, obtains actual coordinate；
(6) according to the bounding box size of bounding box coordinates in flag data and update, new bounding box and corresponding score are generated；
(7) classify to the video image of bounding box new in step (6), and using LSTM network to the view of efficiency frontier frame
Frequency image is multiplexed and generates multiple sequence signatures, generates the parameter that class probability updates VGG-16 convolutional neural networks.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910389617.XA CN110175538A (en) | 2019-05-10 | 2019-05-10 | A kind of substation's Bird's Nest recognition methods and system based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910389617.XA CN110175538A (en) | 2019-05-10 | 2019-05-10 | A kind of substation's Bird's Nest recognition methods and system based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110175538A true CN110175538A (en) | 2019-08-27 |
Family
ID=67690772
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910389617.XA Pending CN110175538A (en) | 2019-05-10 | 2019-05-10 | A kind of substation's Bird's Nest recognition methods and system based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110175538A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717490A (en) * | 2019-09-30 | 2020-01-21 | 深圳供电局有限公司 | Bird nest identification method and system for power transmission line tower and storage medium |
| CN110992257A (en) * | 2019-12-20 | 2020-04-10 | 北京航天泰坦科技股份有限公司 | Remote sensing image sensitive information automatic shielding method and device based on deep learning |
| CN111489354A (en) * | 2020-05-18 | 2020-08-04 | 国网浙江省电力有限公司检修分公司 | Method and device for detecting bird nest on power tower, server and storage medium |
| CN111510688A (en) * | 2020-05-19 | 2020-08-07 | 国网上海市电力公司 | Intelligent auxiliary control method for transformer substation with unmanned inspection function |
| CN111507249A (en) * | 2020-04-16 | 2020-08-07 | 浙江华云信息科技有限公司 | Transformer substation nest identification method based on target detection |
| CN112311092A (en) * | 2020-10-26 | 2021-02-02 | 杭州市电力设计院有限公司余杭分公司 | Method and system for identifying monitoring information of power system |
| CN112715427A (en) * | 2020-12-30 | 2021-04-30 | 青海保绿丰生态农林科技有限公司 | Automatic monitoring system for rules of hawk-leading nesting, number of spawning, brooding and predation habits |
| CN112749509A (en) * | 2020-12-30 | 2021-05-04 | 西华大学 | Intelligent substation fault diagnosis method based on LSTM neural network |
| CN112868490A (en) * | 2020-12-31 | 2021-06-01 | 重庆市工程管理有限公司 | BIM technology-based transplanting planning design method for trees with bird nests |
| CN113536989A (en) * | 2021-06-29 | 2021-10-22 | 广州博通信息技术有限公司 | Refrigerator frosting monitoring method and system based on camera video frame-by-frame analysis |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004038744A (en) * | 2002-07-05 | 2004-02-05 | Nippon Telegr & Teleph Corp <Ntt> | Image recognition algorithm generation method, image recognition algorithm generation device, image recognition algorithm generation program, and recording medium recorded with image recognition algorithm generation program |
| CN107330920A (en) * | 2017-06-28 | 2017-11-07 | 华中科技大学 | A kind of monitor video multi-target tracking method based on deep learning |
| CN109344753A (en) * | 2018-09-21 | 2019-02-15 | 福州大学 | A kind of tiny fitting recognition methods of Aerial Images transmission line of electricity based on deep learning |
| CN109559295A (en) * | 2018-06-04 | 2019-04-02 | 新影智能科技(昆山)有限公司 | Image analysis system, method, computer readable storage medium and electric terminal |
-
2019
- 2019-05-10 CN CN201910389617.XA patent/CN110175538A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004038744A (en) * | 2002-07-05 | 2004-02-05 | Nippon Telegr & Teleph Corp <Ntt> | Image recognition algorithm generation method, image recognition algorithm generation device, image recognition algorithm generation program, and recording medium recorded with image recognition algorithm generation program |
| CN107330920A (en) * | 2017-06-28 | 2017-11-07 | 华中科技大学 | A kind of monitor video multi-target tracking method based on deep learning |
| CN109559295A (en) * | 2018-06-04 | 2019-04-02 | 新影智能科技(昆山)有限公司 | Image analysis system, method, computer readable storage medium and electric terminal |
| CN109344753A (en) * | 2018-09-21 | 2019-02-15 | 福州大学 | A kind of tiny fitting recognition methods of Aerial Images transmission line of electricity based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 欧家祥等: "基于深度学习的高效电力部件识别", 《电力大数据》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717490A (en) * | 2019-09-30 | 2020-01-21 | 深圳供电局有限公司 | Bird nest identification method and system for power transmission line tower and storage medium |
| CN110992257A (en) * | 2019-12-20 | 2020-04-10 | 北京航天泰坦科技股份有限公司 | Remote sensing image sensitive information automatic shielding method and device based on deep learning |
| CN111507249A (en) * | 2020-04-16 | 2020-08-07 | 浙江华云信息科技有限公司 | Transformer substation nest identification method based on target detection |
| CN111489354A (en) * | 2020-05-18 | 2020-08-04 | 国网浙江省电力有限公司检修分公司 | Method and device for detecting bird nest on power tower, server and storage medium |
| CN111510688A (en) * | 2020-05-19 | 2020-08-07 | 国网上海市电力公司 | Intelligent auxiliary control method for transformer substation with unmanned inspection function |
| CN112311092A (en) * | 2020-10-26 | 2021-02-02 | 杭州市电力设计院有限公司余杭分公司 | Method and system for identifying monitoring information of power system |
| CN112715427A (en) * | 2020-12-30 | 2021-04-30 | 青海保绿丰生态农林科技有限公司 | Automatic monitoring system for rules of hawk-leading nesting, number of spawning, brooding and predation habits |
| CN112749509A (en) * | 2020-12-30 | 2021-05-04 | 西华大学 | Intelligent substation fault diagnosis method based on LSTM neural network |
| CN112868490A (en) * | 2020-12-31 | 2021-06-01 | 重庆市工程管理有限公司 | BIM technology-based transplanting planning design method for trees with bird nests |
| CN113536989A (en) * | 2021-06-29 | 2021-10-22 | 广州博通信息技术有限公司 | Refrigerator frosting monitoring method and system based on camera video frame-by-frame analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110175538A (en) | A kind of substation's Bird's Nest recognition methods and system based on machine learning | |
| Li et al. | Detection of fruit-bearing branches and localization of litchi clusters for vision-based harvesting robots | |
| CN109711325B (en) | Mango picking point identification method | |
| Luo et al. | A vision methodology for harvesting robot to detect cutting points on peduncles of double overlapping grape clusters in a vineyard | |
| CN106097346B (en) | A kind of video fire hazard detection method of self study | |
| CN107016405B (en) | A kind of pest image classification method based on classification prediction convolutional neural networks | |
| CN105718945B (en) | Apple picking robot night image recognition method based on watershed and neural network | |
| CN109188148B (en) | Transmission line of electricity applied to smart grid reliably monitors system | |
| CN109376605B (en) | Electric power inspection image bird-stab-prevention fault detection method | |
| CN108537154A (en) | Transmission line of electricity Bird's Nest recognition methods based on HOG features and machine learning | |
| CN107784661A (en) | Substation equipment infrared image classifying identification method based on region-growing method | |
| CN108648169A (en) | The method and device of high voltage power transmission tower defects of insulator automatic identification | |
| CN109032093A (en) | A kind of rice pest monitoring system | |
| CN109344699A (en) | Winter jujube disease recognition method based on depth of seam division convolutional neural networks | |
| CN110287882A (en) | A kind of big chrysanthemum kind image-recognizing method based on deep learning | |
| CN113553977A (en) | Improved YOLO V5-based safety helmet detection method and system | |
| CN110263794A (en) | Safety belt images steganalysis method based on data enhancing | |
| CN114627411A (en) | Crop growth period identification method based on parallel detection under computer vision | |
| CN113298077A (en) | Transformer substation foreign matter identification and positioning method and device based on deep learning | |
| CN107016360A (en) | The object detection method that electricity substation is merged based on behavioral characteristics and region | |
| CN110188624A (en) | A kind of substation's wind drift recognition methods and system based on deep learning | |
| CN110246131A (en) | Conducting wire share split defect image recognition methods based on convolutional neural networks | |
| CN109657540A (en) | Withered tree localization method and system | |
| CN117726853A (en) | Bird protection method, device, equipment and storage medium based on artificial intelligence | |
| CN113111830A (en) | Grape vine winter pruning point detection algorithm |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190827 |
|
| RJ01 | Rejection of invention patent application after publication |