CN109961003A - A kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA - Google Patents
A kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA Download PDFInfo
- Publication number
- CN109961003A CN109961003A CN201811603127.7A CN201811603127A CN109961003A CN 109961003 A CN109961003 A CN 109961003A CN 201811603127 A CN201811603127 A CN 201811603127A CN 109961003 A CN109961003 A CN 109961003A
- Authority
- CN
- China
- Prior art keywords
- fpga
- transmission line
- inspection device
- electricity
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The airborne auxiliary inspection device of the embedded transmission line of electricity that the invention discloses a kind of based on FPGA, belongs to line upkeep technical field, comprising: a kind of method for designing weight shared parameter and K-Means++ cluster carries out compression optimization to neural network structure model；Compressed network model is put into the high-performance embedded FPGA system of parallel organization；The FPGA system integrated is installed on unmanned plane；Original image is acquired under unmanned flight's state；Original image is pre-processed, target image is obtained；Quality evaluation is carried out to target image, decides whether to retake；Real-time defects detection is carried out to target image, result is will test and is transmitted to ground background system.The invention also discloses system, equipment and the storage medium of realizing the above method, the exemplary technical solution of the present invention is used for the real-time inspection of power transmission line route unmanned plane, and real-time is high, result is accurate, compared with currently used manual inspection, greatly reduces cost.
Description
Technical field
The invention belongs to the maintenance of line upkeep technical field more particularly to transmission line of electricity, specifically one kind is based on
Image procossing.
Background technique
China has built up the power grid in six big areas transprovincially at present, is south, northwest, East China, Central China, North China and east respectively
This northern six bulk power grid, transmission line of electricity total length have been more than 1,150,000 kms, and the transmission line of electricity of 500kV or more has become each area's electricity
The net transmission of electricity main force.The territory in China is vast in territory, and landform is also relative complex, and hills is more, Plain is less, in addition meteorological condition
It is complicated and changeable, certain difficulty is brought to the construction of interregional grid and extra high voltage network engineering, in addition the maintenance after building up
With maintenance, relies solely on existing detection methods and routine test and be not able to satisfy efficiently quickly requirement, can not reach very
Good effect.The utilization of helicopter, unmanned air vehicle technique can be good at completing electric inspection process and construction plan task, naturally
The approval and large-scale application of electric power enterprise can be obtained.
But helicopter, unmanned plane inspection are acquiring data volume, data due to the limitation in technology and patrol mode
Validity, analysis processing timeliness etc. efficiency are lower, have the following problems:
(1) helicopter due to spatial volume it is larger, in order to avoid there is hazard exposure with transmission line of electricity, can generally guarantee
The safety patrol inspection distance of 20-30m, is easy to appear inspection blind spot, causes inspection result not accurate enough, and coverage area is limited；
(2) helicopter routing inspection detects mould due to mainly observing two ways using artificial telescope screening and infrared detection
The case where there is certain limitation, be easy to cause missing inspection in formula and identification type, cause routing inspection efficiency not high；
(3) helicopter routing inspection is due to lacking online inspection target positioning and feedback mechanism when fructufy, for single inspection
The defect of detection does not carry out service information tracking, therefore has segmental defect and can not determine newly generated or exist before
, there is the case where defect repeats screening, affects routing inspection efficiency to a certain extent；
(4) unmanned plane inspection analysis of data collected is mostly backstage processed offline, and real-time is lower, affects event to a certain degree
The precision for hindering diagnosis, the problems such as be easy to causeing failure erroneous judgement, fail to judge；
(5) it when out of focus, shake, over-exposed etc. occur during unmanned plane inspection causes image that can not identify situation, lacks
Weary information supplement acquisition methods；
With artificial intelligence, the development of embedded image processing and identification, in conjunction with helicopter at this stage, unmanned plane inspection
Business demand carries out the technologies such as the analysis of polling transmission line image intelligent, embedded high-performance computing technique, airborne integrated technology
Research and application, develop embedded airborne power transmission line auxiliary inspection device and ensured to improve polling transmission line efficiency
Inspection safety.
Summary of the invention
To solve the deficiency in the above-mentioned prior art, the purpose of the present invention is to provide a kind of based on the embedded of FPGA
The airborne auxiliary inspection device of transmission line of electricity, this method and system are using unmanned plane and helicopter as carrier, in polling transmission line
Real-time defect recognition is carried out, while can carry out in scenes such as agricultural, mining industry, traffic infrastructure, exploration and public safeties and promote
Using.
The technical scheme adopted by the invention is as follows:
On the one hand, a kind of airborne auxiliary inspection device of the embedded transmission line of electricity based on FPGA is provided, comprising:
Compression processing is carried out to deep neural network structure；
The design of convolutional neural networks structure customized hardware is carried out to FPGA；
Compressed deep neural network and high-performance embedded FPGA system are integrated；
FPGA system after will be integrated is loaded on unmanned plane；
According to unmanned plane shoot power transmission line inspection picture, carry out intelligence retake, retake, identify defect in real time；
Original image is pre-processed, target image is obtained；
Further, the acquisition original image is carried out by the imaging unit being mounted in line system.
Further, described image gray processing includes: the power that will constitute the red of image, green and blue are distinguished according to it
Value is weighted and averaged the gray value as gray level image.
Further, described image, which enhances, includes:
The range of route gray value in enlarged image；
Image is handled using gray scale stretching function.
Further, described image, which filters, includes:
Determine target pixel points；
Target window is determined in the object pixel neighborhood of a point；
The gray value that the target pixel points are replaced with the intermediate value of the gray value of all pixels point in the target window, obtains
To filtered image.
Target area characteristic pattern vector in target image is extracted, and is exported；
Target area and high latitude characteristic pattern are read, is ranked up by the probability for target object occur in region is extracted, and defeated
High probability feature graph region out；
It is exported according to region screening layer, regional aim is analyzed, is classified, according to classificating requirement for pixel, area
Domain carries out classification marker；
The pictorial information and defect classification for be to label pass background system back in real time.
Further, further includes:
It is identified in real time according to the transmission line of electricity picture that unmanned plane is shot in the front end FPGA, recognition result is passed back in real time
Background system.
On the other hand, the airborne auxiliary inspection tour system of embedded transmission line of electricity that the present invention also provides a kind of based on FPGA,
Include:
Imaging unit is configured to acquisition original image；
Pretreatment unit is configured to pre-process original image, obtains target image；
Arithmetic element is configured to obtain the defect type of transmission line of electricity；
Output unit is configured to output defect type and location information；
Wherein, arithmetic element includes computing module and database module, and the database module is for storage and output line
The good location information of road actual defects type.
On the other hand, the present invention also provides a kind of equipment, the equipment includes:
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of places
It manages device and executes the exemplary airborne auxiliary inspection device of any embedded transmission line of electricity based on FPGA of the present invention.
On the other hand, the present invention also provides a kind of computer readable storage medium for being stored with computer program, the journeys
Realize that the airborne auxiliary of the exemplary any embedded transmission line of electricity based on FPGA of the present invention makes an inspection tour dress when sequence is executed by processor
It sets.
Compared with prior art, the invention has the benefit that
1, the airborne auxiliary inspection device of the exemplary embedded transmission line of electricity based on FPGA of the invention, it is opposite for unmanned plane
Target is easy to be lost under high-speed motion state, is easy to miss inspection, the problems such as diagnostic accuracy is lower, using nobody based on Faster-Rcnn
Machine inspection multi-object Recognition Model and target defect diagnostic model, in conjunction with the neural network compression optimization side of embedded type low-power consumption
Method realizes the inspection scene multi-targets recognition diagnosis under unmanned plane load limited situation, improves power transmission line unmanned machine inspection
Fault diagnosis timeliness, experiment proves that result precision is high, concept feasible is strong.
2, the airborne auxiliary inspection device of the exemplary embedded transmission line of electricity based on FPGA of the invention carries out original image
The noise in original image is eliminated in pretreatment, improves the clarity of original image, while carrying out quality evaluation to original image,
To the bad image of quality is pinpointed, positioning intelligent is retaken, retake, conducive to accurately identifying for subsequent step, the standard of result is improved
True property.
3, the airborne auxiliary inspection device of the exemplary embedded transmission line of electricity based on FPGA of the invention, uses embedded
FPGA front end recognition technology reduces unmanned plane picture amount of storage and picture transfer pressure, directly in front end by defect recognition result
It is transferred to background system, effectively increases unmanned plane routing inspection efficiency.
4, the airborne auxiliary inspection device of the exemplary embedded transmission line of electricity based on FPGA of the invention, constructs neural network
The low-power consumption of compression algorithm compression optimization, high performance embedded preposition option hardware structure model are airborne embedded platform
The DSP/GPU module with high speed image data parallel processing capability is carried, and on this basis to Embedded basic device
Low power dissipation design is carried out with application logical unit, reduces device power consumption while guaranteeing inspection precision.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is the flow chart of the embodiment of the present invention；
Fig. 2 is the pretreated flow chart of picture of the embodiment of the present invention；
Fig. 3 is grayscale bar schematic diagram；
Fig. 4 is Faster of embodiment of the present invention R-CNN model method figure；
Fig. 5 is model training of embodiment of the present invention techniqueflow chart；
Fig. 6 is the model compression flow chart of the embodiment of the present invention；
Fig. 7 is the high-performance embedded FPGA heterogeneous system architecture diagram of the embodiment of the present invention；
Fig. 8 is the hardware architecture diagram of the embodiment of the present invention；
Fig. 9 is transmission line of electricity of embodiment of the present invention defects detection result exemplary diagram.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
As shown in Figure 1, An embodiment provides a kind of, the embedded transmission line of electricity based on FPGA is airborne auxiliary
Help inspection device, comprising:
S1: model compression optimization is carried out using weight shared parameter and K-Means++ cluster to deep learning model；
S2: the design of deep learning hsrdware requirements is carried out to embedded FPGA system；
S3: compressed neural network model is integrated with FPGA system；
S4: the system after will be integrated is mounted in unmanned plane；
S5: transmission line of electricity Image Acquisition is carried out after UAV flight's embedded FPGA system；
S6: according to acquisition image imaging contexts decide whether intelligence retake, retake；
S7: front end recognition is carried out to the target image of acquisition, passes recognition result back background system in real time.
In S1, model compression optimization is carried out using weight shared parameter and K-Means++ cluster to deep learning model.Tool
The model compression of body optimizes as shown in fig. 6, reducing neural network by the technical solution for deleting network and shared parameter combines
In parameter delete the connection that weight is less than threshold value firstly, be trained to original neural network, its weight zero setting is protected
Stay the biggish connection weight of information content constant, in order to keep its network accuracy rate constant as far as possible, on the network after deleting again
It is trained fine tuning；Then, K-means++ cluster is carried out for each layer of parameter, falls within the multiple parameters of each same cluster
Sharing Center's value optimizes the value of k under the premise of keeping accuracy rate constant as far as possible as weight.
Wherein, original neural network is trained using model structure as shown in Figure 4, training process is as shown in Figure 5.
In the deep learning framework, input terminal is original power transmission line inspection video image, and output end is electric power object detection results.Its
Middle input picture is analyzed without empirical model, directly extracts high latitude characteristic pattern by depth convolutional layer.Characteristic pattern is exported to use
In three aspect functions of support:
(1) target area extract layer extracts potential target area in characteristic pattern according to characteristic pattern vector, and by target area
Domain output；
(2) region screening layer, which extracts, reads target area extraction once output and high dimensional feature figure, occurs by extracting in region
The probability of target object is ranked up, and exports high probability feature graph region；
(3) target analysis layer is exported according to region screening layer, and regional aim is analyzed, is classified, can be according to classification
It is required that carrying out classification marker for pixel, region.
Target area extract layer, region screening layer and target analysis layer are based on deep learning network in the deep learning network
Layer exploitation, is connected with existing convolutional layer, full articulamentum.The end-to-end deep learning network architecture is trained via great amount of images,
Wherein target area extract layer and screening layer are according to label results area training network layer parameter, convolutional layer and target analysis layer root
It is adjusted according to label result loss function.
In convolution domain, all convolutional layers are all: kernel_size=3, pad=1, all convolution all expand
(0) pad=1 fills a circle, causes original image (MxN) to become (M+2) x (N+2) size, then export after doing 3x3 convolution for side processing
MxN.This set causes the conv layer in Conv layers not change and outputs and inputs matrix size.Equally, all
Pooling layers of kernel_size=2, stride=2.The MxN matrix by pooling layers each in this way, can all become (M/
2) * (N/2) size.In conclusion in entire Conv layers, conv and relu layers do not change input and output size, only
Pooling layers make output length and width all become the 1/2 of input.To sum up, the matrix of a MxN size is fixed by Conv layers
Become (M/16) x (N/16), can be mapped with original image in the featuure map that such convolution domain generates, reduces target
A possibility that losing or omitting.Detection block, which is generated, generates detection using sliding window+image pyramid such as OpenCV adaboost
Frame, RCNN generate detection block using SS (Selective Search) method.And Faster R-CNN has then abandoned traditional cunning
Dynamic window and SS method, directly generate detection block using RPN.
Wherein, after model training process is as shown in figure 5, defect sample marks, by " original image+mark file " format
Labeled data carry out model training as the input of model training, Faster-RCNN defect recognition diagnostic model trained
Journey is divided into four-stage, wherein the first two stage PRN network and Fast-RCNN network parameter stand-alone training, convolution layer parameter point
From；Latter two stage deconvolution parameter is shared, and the parameter of two networks is independently iterated fine tuning, passes through survey after four-stage
Examination collection carries out model performance verifying, to obtain defect recognition diagnostic model.
In S2, as shown in figure 8, carrying out hardware structure to the airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA
Design and autonomous optimization.
Wherein, bottom uses the customization board based on FPGA, using built-in Linux operating system, support USB,
A variety of data-interfaces such as Ethernet, VUSB, DMI facilitate the various goal systems of insertion；Deep learning algorithm uses number after training
It is hardware module according to stream process mold curing, is run in bottom hardware system, provides top layer using required high disposal performance, low function
Consumption and low delay；Top layer is applied according to different target percentage regulation learning network framework, automatic to explore hardware design space (data
Width, the parallel network number of plies, network layer internal data store framework, network layer inside degree of parallelism etc.), guarantee optimal hardware
Energy.Top layer application accesses vision data in real time after optimization, supports target detection, result after detection using hardware deep learning module
It is exported by data-interface and underlying operating system to external system.The tight fit of three-tier system guarantees the target detection of top layer
Using the high-performance of custom hardware and the accuracy of deep learning can be enjoyed simultaneously, wherein the heterogeneous system of embedded FPGA is such as
Shown in Fig. 7.
In S3, the trained model of S1 and the S2 embedded system designed are integrated, being formed has identification function
Embedded equipment.
In S4, the embedded patrolling transmission line device based on FPGA integrated is loaded into unmanned plane, by image
Acquisition unit, power module and the device integrate.
In S5, transmission line of electricity is carried out by airborne device and highlights acquisition, while original image is pre-processed, is pre-processed
Process is as shown in Figure 2, comprising: image gray processing, image enhancement and image filtering.
Described image gray processing includes: that will constitute the red of image, green and blue to be added according to its weight respectively
Gray value of the weight average as gray level image.Usually our acquired images are all color images, i.e. RGB image.RGB difference
It indicates red (Red), green (Green), blue (Blue), indicates that combination forms the visible of the overwhelming majority with these three colors
Light.When the gray value of tri- kinds of colors of RGB of pixel is identical, greyscale color is just produced, wherein three kinds of equal values of primary colours claim
For gray value, intensity value or brightness value are also.Each gray scale object has the brightness from 1% (white) to 100% (black)
Value.In general, the gray value quantization means of pixel can be certain amount gray level by gray level image, such as: as shown in figure 3,
The gray level image of 8bit contains 28=256 gray levels, 256 gray levels that the gray level image in the present embodiment all uses.
The red for constituting image, green and blue are weighted and averaged according to its weight respectively, are specifically included:
Setting is red, green, blue is respectively three components, according to the importance of each component and other needs, by three
A component is weighted and averaged the gray value as gray level image according to different weights:
F (i, j)=QRR (i, j)+QGG (i, j)+QBB (i, j) (2-1)
In formula (2-1): QR, QG, QBThe respectively weight of red component, green component and blue component, often takes QR=
0.299, QG=0.587, QB=0.114 (2-2)
Image enhancement refers to and unsharp image is apparent from or is emphasized the feature of concern and inhibits the spy being not concerned with
Sign, improving image quality, abundant information amount are to reinforce image interpretation and recognition effect.Fine granularity defect image in transmission line of electricity
It mixes with complicated background and is not easy to differentiate, image grayscale at this time is relatively concentrated.Defect characteristic and background are differentiated, then
The range for needing defect gray value in enlarged image makes the brightness value of image pixel have height to have low, increases the comparison of defect image
Degree is to be easier to differentiate.
In the present embodiment, image enhancement includes: the range of route gray value in enlarged image；Using gray scale stretching function pair
Image is handled.Before image enhancement processing, the gray value integrated distribution of pixel is 100 to 160 in image, range very little,
But pass through after image enhancement processing, there are pixel distribution, the gray value of the more pixel of integrated distribution between 0 to 255
Have from 50 to 200, range expands many.Gray scale stretching function (double types of image) is used in this embodiment.
After carrying out image enhancement processing, the details of image is relatively sharp, and contrast also strengthens, but also increases simultaneously
The noise of image.In order to reduce the noise of image, need to carry out the disposal of gentle filter to image.In the present embodiment, image filtering
Comprise determining that target pixel points；Target window is determined in the object pixel neighborhood of a point；With pictures all in the target window
The intermediate value of the gray value of vegetarian refreshments replaces the gray value of the target pixel points, obtains filtered image.
Specifically, replacing this using the intermediate value of pixel g (x, y) neighborhood all pixels brightness when carrying out image filtering
Point gray scale, specific practice are to determine a window W, are then replaced with the intermediate value of the gray value of all pixels in window original
Pixel obtain filtered image h (x, y), expression formula are as follows:
H (x, y)=median { g (x-k, y-l), (k, l) ∈ W } (2-3)
In formula (2-3), the x-axis coordinate value that x refers to, the y-axis coordinate value that y refers to.By simulation comparison, the image of the present embodiment is filtered
Wave method is easy to use, and arithmetic speed is fast, and noise elimination is also fine, while also becoming mould without image caused by other filtering methods
The problem of paste.
In S6, according to acquisition image imaging contexts decide whether intelligence retake, retake, comprising:
Construct the training data sample set of picture quality, history distorted image of the sample set by handmarking, Yi Jidian
Type image quality evaluation database (LIVE, TID2008, CSIQ, IVC etc.), image quality evaluation standard are evaluated using algorithm
Value and human eye subjective assessment value (MOS and DMOS)；
Extract the spatial domain or transform domain feature of sample set；
Using down-sampled method, the real-time video frame of unmanned plane is extracted, carries out image quality estimation, when predicted value is lower than certain
When threshold value, triggering is retaken, and the threshold value of down-sampled specific sampling rate and picture quality is needed according to actual applied field
Scape is adjusted；
When image quality evaluation discovery quality is bad, the Accuracy Space domain by obtaining unmanned plane is positioned, by nobody
The combination of machine and holder controls, and realizes that live location fix is retaken.
In S7, front end recognition is carried out to the target image of acquisition, passes recognition result back background system in real time.
The experimental data of the present embodiment is as shown in table 1:
The airborne auxiliary of embedded transmission line of electricity of the table 1 based on FPGA makes an inspection tour defect recognition situation
As can be seen that the airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA is to power transmission line from test result
The preparation verification and measurement ratio of road defect is all larger than 80%, average response time within 250ms, meet unmanned plane inspection real-time and
Accuracy requirement.By the test, the basic verification exemplary airborne auxiliary of embedded transmission line of electricity based on FPGA of the invention
The feasibility of inspection device.
The airborne auxiliary inspection device of the embedded transmission line of electricity based on FPGA that the present embodiment proposes, by digital imagery mould
Block acquires transmission line of electricity image, is integrated in particularly customized FPGA system using compressed deep neural network model, so
It is intelligently retaken by the pretreatment of image, image fault afterwards, retake processing, defects detection is carried out to target image, will finally be lacked
It falls into testing result real-time Transmission and returns background system.The airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA is a kind of high
The detection method of automation is spent, this detection method eliminates the complicated arduous labor of hunting worker, while also greatling save
Inspection spending, is especially reduction of the power grid maintenance cost of component environment complex area.
On the other hand, the airborne auxiliary inspection tour system of embedded transmission line of electricity that the present invention also provides a kind of based on FPGA,
Include:
Imaging unit is configured to acquisition original image；
Pretreatment unit is configured to pre-process original image, obtains target image；
Arithmetic element is configured to obtain the defect type of transmission line of electricity；
Output unit is configured to output defect type and location information；
Wherein, arithmetic element includes computing module and database module, and the database module is for storage and output line
The good location information of road actual defects type.
On the other hand, the present invention also provides a kind of equipment, the equipment includes:
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of places
Reason device executes the airborne auxiliary inspection device of the embedded transmission line of electricity based on FPGA of the present embodiment.
Being described in the embodiment of the present application involved unit or module can be realized by way of software, can also be with
It is realized by way of hardware.Described unit or module also can be set in the processor.These units or module
Title does not constitute the restriction to the unit or module itself under certain conditions.
On the other hand, the present embodiment additionally provides a kind of computer readable storage medium for being stored with computer program, should
The airborne auxiliary inspection device of the embedded transmission line of electricity based on FPGA of the present embodiment is realized when program is executed by processor.The meter
Calculation machine readable storage medium storing program for executing can be computer readable storage medium included in system or equipment described in above-described embodiment；
It is also possible to individualism, without the computer readable storage medium in supplying equipment, such as hard disk, CD, SD card.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from the inventive concept, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Except for the technical features described in the specification, remaining technical characteristic is the known technology of those skilled in the art, is prominent
Innovative characteristics of the invention out, details are not described herein for remaining technical characteristic.
Claims (10)
1. a kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA, characterized in that include:
Compression processing is carried out to deep neural network structure；
The design of convolutional neural networks structure customized hardware is carried out to FPGA；
Compressed deep neural network and high-performance embedded FPGA system are integrated；
FPGA system after will be integrated is loaded on unmanned plane；
According to unmanned plane shoot power transmission line inspection picture, carry out intelligence retake, retake, identify defect in real time；
Background system is passed back when by unmanned plane defects detection fructufy.
2. the airborne auxiliary inspection device of the embedded transmission line of electricity according to claim 1 based on FPGA, characterized in that institute
Compression optimization is stated to carry out by neural network weight shared parameter and K-Means++ clustering algorithm.
3. the airborne auxiliary inspection device of the embedded transmission line of electricity according to claim 1 based on FPGA, characterized in that right
Convolution algorithm module, sampling computing module, activation primitive module carry out targeted design.
4. the airborne auxiliary inspection device of the embedded transmission line of electricity according to claim 3 based on FPGA, characterized in that adopt
With the mode of single data stream driving progressive scan, hop and arithmetic section are separately designed, is added in output par, c rear end
FLASH caching.
5. the airborne auxiliary inspection device of the embedded transmission line of electricity according to claim 3 based on FPGA, characterized in that adopt
With the strategy of allocation index come the input data in great-jump-forward index store, indexed according to the storage address for reading in data corresponding
Input neuron, then will input neuron be transmitted to operation comparing unit carry out subsequent processing.
6. the airborne auxiliary inspection device of the embedded transmission line of electricity according to claim 3 based on FPGA, characterized in that adopt
It uses ReLU function as the activation primitive of convolutional neural networks, is designed on FPGA using the method for sectional linear fitting.
7. the airborne auxiliary inspection device of the embedded transmission line of electricity according to claim 1 based on FPGA, characterized in that root
It is out of focus according to camera, over-exposed, shake situations such as carry out vertex retake, comprising:
Target image is subjected to optical imagery evaluation；
The bad abnormal image of quality is retaken.
8. a kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA, comprising:
Imaging unit is configured to acquisition original image；
Pretreatment unit is configured to pre-process original image, obtains target image；
Arithmetic element is configured to obtain transmission line of electricity defect information；
Output unit is configured to the result of output transmission line of electricity defect type, anchor point；
Wherein, arithmetic element includes computing module and database module, and the database module is real with outlet line for storing
Border defect type and anchor point information.
9. a kind of equipment, characterized in that the equipment includes:
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors
Execute the airborne auxiliary inspection device of the embedded transmission line of electricity based on FPGA as claimed in claim 1.
10. a kind of computer readable storage medium for being stored with computer program, characterized in that when the program is executed by processor
Realize the airborne auxiliary inspection device of the embedded transmission line of electricity based on FPGA as claimed in claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811603127.7A CN109961003A (en) | 2018-12-26 | 2018-12-26 | A kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811603127.7A CN109961003A (en) | 2018-12-26 | 2018-12-26 | A kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109961003A true CN109961003A (en) | 2019-07-02 |
Family
ID=67023417
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811603127.7A Pending CN109961003A (en) | 2018-12-26 | 2018-12-26 | A kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109961003A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110362907A (en) * | 2019-07-03 | 2019-10-22 | 安徽继远软件有限公司 | Based on the identification of SSD neural network transmission line of electricity target defect and diagnostic method |
| CN110909667A (en) * | 2019-11-20 | 2020-03-24 | 北京化工大学 | Lightweight design method for multi-angle SAR target recognition network |
| CN110990658A (en) * | 2019-11-15 | 2020-04-10 | 西安交通大学 | Method for realizing image processing algorithm of power transmission line on embedded system |
| CN111651649A (en) * | 2020-04-10 | 2020-09-11 | 安徽继远软件有限公司 | Virtual fence construction method and system for power transmission line and tower |
| CN112288711A (en) * | 2020-10-28 | 2021-01-29 | 浙江华云清洁能源有限公司 | Unmanned aerial vehicle inspection image defect image identification method, device, equipment and medium |
| CN112506214A (en) * | 2020-11-16 | 2021-03-16 | 天津航天中为数据***科技有限公司 | Operation flow of autonomous fan inspection system of unmanned aerial vehicle |
| CN112511800A (en) * | 2020-11-19 | 2021-03-16 | 杭州新纽科技有限公司 | Unmanned aerial vehicle power line patrol data transmission equipment and method |
| CN112668696A (en) * | 2020-12-25 | 2021-04-16 | 杭州中科先进技术研究院有限公司 | Unmanned aerial vehicle power grid inspection method and system based on embedded deep learning |
| CN112668416A (en) * | 2020-12-17 | 2021-04-16 | 国网山西省电力公司运城供电公司 | Real-time analysis and identification method for distribution network inspection video of unmanned aerial vehicle and embedded inspection device |
| CN113419140A (en) * | 2021-08-06 | 2021-09-21 | 广东电网有限责任公司 | Distribution network line fault diagnosis method and device considering human-computer-object cooperation |
| CN113780535A (en) * | 2021-09-27 | 2021-12-10 | 华中科技大学 | Model training method and system applied to edge equipment |
| CN114663404A (en) * | 2022-03-25 | 2022-06-24 | 广州中科云图智能科技有限公司 | Line defect identification method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015043363A1 (en) * | 2013-09-30 | 2015-04-02 | 华中科技大学 | Infrared image recognition device for ground moving object of aircraft |
| WO2015131462A1 (en) * | 2014-03-07 | 2015-09-11 | 国家电网公司 | Centralized monitoring system and monitoring method for unmanned aerial vehicle to patrol power transmission line |
| CN106326932A (en) * | 2016-08-25 | 2017-01-11 | 北京每刻风物科技有限公司 | Power line inspection image automatic identification method based on neural network and power line inspection image automatic identification device thereof |
| CN106932688A (en) * | 2017-04-24 | 2017-07-07 | 国家电网公司 | Power transmission line detector and the power transmission line detecting system based on unmanned plane |
| CN107729808A (en) * | 2017-09-08 | 2018-02-23 | 国网山东省电力公司电力科学研究院 | A kind of image intelligent acquisition system and method for power transmission line unmanned machine inspection |
-
2018
- 2018-12-26 CN CN201811603127.7A patent/CN109961003A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015043363A1 (en) * | 2013-09-30 | 2015-04-02 | 华中科技大学 | Infrared image recognition device for ground moving object of aircraft |
| WO2015131462A1 (en) * | 2014-03-07 | 2015-09-11 | 国家电网公司 | Centralized monitoring system and monitoring method for unmanned aerial vehicle to patrol power transmission line |
| CN106326932A (en) * | 2016-08-25 | 2017-01-11 | 北京每刻风物科技有限公司 | Power line inspection image automatic identification method based on neural network and power line inspection image automatic identification device thereof |
| CN106932688A (en) * | 2017-04-24 | 2017-07-07 | 国家电网公司 | Power transmission line detector and the power transmission line detecting system based on unmanned plane |
| CN107729808A (en) * | 2017-09-08 | 2018-02-23 | 国网山东省电力公司电力科学研究院 | A kind of image intelligent acquisition system and method for power transmission line unmanned machine inspection |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110362907B (en) * | 2019-07-03 | 2023-10-20 | 安徽继远软件有限公司 | SSD neural network-based power transmission line target defect identification and diagnosis method |
| CN110362907A (en) * | 2019-07-03 | 2019-10-22 | 安徽继远软件有限公司 | Based on the identification of SSD neural network transmission line of electricity target defect and diagnostic method |
| CN110990658A (en) * | 2019-11-15 | 2020-04-10 | 西安交通大学 | Method for realizing image processing algorithm of power transmission line on embedded system |
| CN110909667A (en) * | 2019-11-20 | 2020-03-24 | 北京化工大学 | Lightweight design method for multi-angle SAR target recognition network |
| CN111651649A (en) * | 2020-04-10 | 2020-09-11 | 安徽继远软件有限公司 | Virtual fence construction method and system for power transmission line and tower |
| CN112288711A (en) * | 2020-10-28 | 2021-01-29 | 浙江华云清洁能源有限公司 | Unmanned aerial vehicle inspection image defect image identification method, device, equipment and medium |
| CN112506214A (en) * | 2020-11-16 | 2021-03-16 | 天津航天中为数据***科技有限公司 | Operation flow of autonomous fan inspection system of unmanned aerial vehicle |
| CN112511800A (en) * | 2020-11-19 | 2021-03-16 | 杭州新纽科技有限公司 | Unmanned aerial vehicle power line patrol data transmission equipment and method |
| CN112668416A (en) * | 2020-12-17 | 2021-04-16 | 国网山西省电力公司运城供电公司 | Real-time analysis and identification method for distribution network inspection video of unmanned aerial vehicle and embedded inspection device |
| CN112668696A (en) * | 2020-12-25 | 2021-04-16 | 杭州中科先进技术研究院有限公司 | Unmanned aerial vehicle power grid inspection method and system based on embedded deep learning |
| CN113419140A (en) * | 2021-08-06 | 2021-09-21 | 广东电网有限责任公司 | Distribution network line fault diagnosis method and device considering human-computer-object cooperation |
| CN113780535A (en) * | 2021-09-27 | 2021-12-10 | 华中科技大学 | Model training method and system applied to edge equipment |
| CN113780535B (en) * | 2021-09-27 | 2024-06-04 | 华中科技大学 | Model training method and system applied to edge equipment |
| CN114663404A (en) * | 2022-03-25 | 2022-06-24 | 广州中科云图智能科技有限公司 | Line defect identification method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109961003A (en) | A kind of airborne auxiliary inspection device of embedded transmission line of electricity based on FPGA | |
| CN112380952B (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| Eisenbach et al. | How to get pavement distress detection ready for deep learning? A systematic approach | |
| Chakma et al. | Image-based air quality analysis using deep convolutional neural network | |
| CN112734692B (en) | Defect identification method and device for power transformation equipment | |
| CN105404886B (en) | Characteristic model generation method and characteristic model generating means | |
| CN109961460A (en) | A kind of multiple target method for inspecting based on improvement YOLOv3 model | |
| CN111462128B (en) | Pixel-level image segmentation system and method based on multi-mode spectrum image | |
| CN106709903B (en) | PM2.5 concentration prediction method based on image quality | |
| CN109753878A (en) | Imaging recognition methods and system under a kind of bad weather | |
| CN109297978B (en) | Binocular imaging-based power line unmanned aerial vehicle inspection and defect intelligent diagnosis system | |
| CN116543241B (en) | Detection method and device for leakage gas cloud, storage medium and electronic equipment | |
| CN113033385A (en) | Deep learning-based violation building remote sensing identification method and system | |
| CN116704273A (en) | Self-adaptive infrared and visible light dual-mode fusion detection method | |
| CN115620090A (en) | Model training method, low-illumination target re-recognition method and device and terminal equipment | |
| Luo et al. | RBD-Net: robust breakage detection algorithm for industrial leather | |
| Tang et al. | Foreign object detection for transmission lines based on Swin Transformer V2 and YOLOX | |
| Zheng et al. | Overwater image dehazing via cycle-consistent generative adversarial network | |
| Wu et al. | Real-time vehicle color recognition based on yolo9000 | |
| CN117333807A (en) | Cloud edge cooperative target locking system for air-ground integrated unmanned aerial vehicle group | |
| CN111898427A (en) | Multispectral pedestrian detection method based on feature fusion deep neural network | |
| CN116580324A (en) | Yolov 5-based unmanned aerial vehicle ground target detection method | |
| Li et al. | Infrared and visible image fusion method based on principal component analysis network and multi-scale morphological gradient | |
| CN114565597A (en) | Nighttime road pedestrian detection method based on YOLOv3-tiny-DB and transfer learning | |
| CN114926456A (en) | Rail foreign matter detection method based on semi-automatic labeling and improved deep learning |
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 |