CN105825207B - The high-voltage line detection method and device of fragmentation - Google Patents
The high-voltage line detection method and device of fragmentation Download PDFInfo
- Publication number
- CN105825207B CN105825207B CN201610249136.5A CN201610249136A CN105825207B CN 105825207 B CN105825207 B CN 105825207B CN 201610249136 A CN201610249136 A CN 201610249136A CN 105825207 B CN105825207 B CN 105825207B
- Authority
- CN
- China
- Prior art keywords
- image block
- voltage line
- line
- edge feature
- line segment
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 44
- 238000013467 fragmentation Methods 0.000 title claims abstract description 20
- 238000006062 fragmentation reaction Methods 0.000 title claims abstract description 20
- 239000012634 fragment Substances 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims abstract description 25
- 238000002788 crimping Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 30
- 238000000034 method Methods 0.000 description 20
- 238000013527 convolutional neural network Methods 0.000 description 18
- 230000008569 process Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 238000009435 building construction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000004126 nerve fiber Anatomy 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/60—Static or dynamic means for assisting the user to position a body part for biometric acquisition
- G06V40/63—Static or dynamic means for assisting the user to position a body part for biometric acquisition by static guides
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention provides the high-voltage line detection method and device of a kind of fragmentation, wherein detection method includes: acquisition image to be detected, and described image to be detected is divided at least two image blocks；Classify to described image block, obtains the first kind image block comprising high-voltage line and the second class image block not comprising high-voltage line, and give up the second class image block；High pressure line feature extraction is carried out to the first kind image block, obtains the edge feature figure of high-voltage line；Line segment detection is carried out to the edge feature figure of the high-voltage line, obtains the fragment figure of high-voltage line；Line segment in the fragment figure of the high-voltage line is spliced, high-voltage line instruction figure is obtained.The present invention can complete to restore to obtain complete high-voltage line instruction figure to the panorama of high-voltage line, to effectively avoid the flight hazard occurred due to the false judgment to high pressure line position.
Description
Technical field
The present invention relates to a kind of a kind of spies of the high-voltage line of deterrent detection method more particularly to fragmentation when low-latitude flying
Method and apparatus are surveyed, aviation safety technical field is belonged to.
Background technique
With the continuous opening that uses low latitude field of country in recent years, unmanned plane be usually utilized to the rescue of auxiliary mountain area,
The tasks such as goods and materials conveying, sample collection.And during low-latitude flying, on ground, there are many barriers to pacify to its flight
It threatens entirely, such as mountain peak, house, high-voltage line etc..In these deterrents, high-voltage line be most dangerous deterrent it
One, if unmanned plane cannot detect high-voltage line, the cable of high pressure is awing knocked, is not only caused damages, can also drawn
Ignition calamity, the accidents such as explosion.So high-voltage line detection is extremely important safely to unmanned plane during flying, unmanned plane must be in flight course
In it detected and implement necessary avoidance strategy.
It compares with other target detections, high-voltage line detection has the characteristics that itself and difficult point.High-voltage line is being taken photo by plane
Appearance features in video image have intermittence, and the part occurred in the picture is flickering, seldom in the form of complete line
Occur；And the visual saliency of high-voltage line performance is lower, the information content that can be used for detecting is few, is difficult complete high by extracting
Crimping feature is detected.In addition, the background for video of taking photo by plane is more complicated, there are some to have line feature as high-voltage line
Suspected target, such as branch, the edge of mountain peak, river, building construction etc. all have line feature.So in detection process
In, high-voltage line is easy mutually to obscure with the suspected target in background, it is difficult to distinguish in area.This be also many high pressure line detecting methods not
It can be suitable for space base scene well.Therefore, it in order to ensure the low-latitude flying of aircraft, needs to propose a kind of effective high-voltage line
Detection method.
Summary of the invention
The present invention provides the high-voltage line detection method and device of a kind of fragmentation, flies in the prior art in low latitude for overcoming
When detecting when row to high-voltage line, it is easy the technological deficiency for mutually obscuring high-voltage line with the suspected target in background.
The present invention provides a kind of high-voltage line detection method of fragmentation, comprising:
Image to be detected is obtained, described image to be detected is divided at least two image blocks；
Classify to described image block, obtains the first kind image block comprising high-voltage line and second not comprising high-voltage line
Class image block, and give up the second class image block；
High pressure line feature extraction is carried out to the first kind image block, obtains the edge feature figure of high-voltage line；
Line segment detection is carried out to the edge feature figure of the high-voltage line, obtains the fragment figure of high-voltage line；
Line segment in the fragment figure of the high-voltage line is spliced, high-voltage line instruction figure is obtained.
Further, described to classify to described image block, it obtains the first kind image block comprising high-voltage line and does not wrap
The second class image block containing high-voltage line, and give up the second class image block, comprising:
Attributive analysis is carried out at least two image block, when the attribute of described image block is the first attribute, by institute
It states image block and is denoted as first kind image block；It is note second by described image block when non-first attribute of the attribute of described image block
Class image block.
Further, described to carry out high pressure line feature extraction to the first kind image block, the edge for obtaining high-voltage line is special
Sign figure, comprising:
High pressure line feature extraction is carried out to each described first kind image block according to following formula 1, formula 2, is obtained described every
The final edge characteristic pattern of one first kind image block carries out the final edge characteristic pattern of each first kind image block
Combination, obtains the edge feature figure of the high-voltage line,
Wherein, I is first kind image block, IiIt is the characteristic pattern on each described i-th of direction of first kind image block, Ei
It is edge feature figure final on each described i-th direction of first kind image block, 0 Ω of < i≤360/, Ω≤360 °, Fi=Ka
(θi)*Fa+Kb(θi)*Fb+Kc(θi)*Fc+Kd(θi)*FdIt is θiFilter function on direction, θi=i Ω, wherein Fa,Fb,Fc,Fd
It is to be existed using three rank Gaussian derivativesThe filter function generated on four direction, Ka(θi),Kb(θi),Kc(θi),Kd
(θi) it is to calculate θiFilter function F on directioniWhen interpolating function.
Further, described to carry out high pressure line feature extraction to the first kind image block, the edge for obtaining high-voltage line is special
Before sign figure further include: according to the autonomous selection parameter σ of determining each first kind image block of formula 3, described in the σ decision
The weight of each first kind image block filter function,
Wherein, C is the contrast of each first kind image block；
Filter function FiFor the filter function for meeting default weight.
Further, the edge feature figure to the high-voltage line carries out Line segment detection, obtains the fragment figure of high-voltage line,
Include:
The first line segment in the edge feature figure of the high-voltage line is excluded, wherein the length of the First Line section is less than first
Threshold value；
Relating dot in the edge feature figure of the high-voltage line and associated point are connected, wherein the relating dot and associated
The distance of point is less than or equal to the second threshold, and first threshold is greater than second threshold.
Further, the line segment in the fragment figure of the high-voltage line is spliced, obtains high-voltage line instruction figure, packet
It includes:
Line segment aggregate in the fragment figure of the high-voltage line is denoted as S={ l1,l2,l3,...,li-1,li,li+1,...,ln,
liIndicate the segment of the high-voltage line in set；
As the liWith the ljAll meet following three conditions, connects the liWith the lj, obtain the high-voltage line
Instruction figure, in which:
(1)ljStarting point abscissa ratio liTerminal abscissa it is big；
(2)liTerminal and ljThe ordinate of starting point differs minimum；
(3)liAnd ljSlope differ minimum.
The present invention also provides a kind of high pressure line detectors of fragmentation, comprising:
Image segmentating device, described image divider are used to described image to be detected being divided at least two image blocks；
Classifier, the classifier obtain the First Kind Graph picture comprising high-voltage line for classifying to described image block
Block and the second class image block not comprising high-voltage line, and give up the second class image block；
Filter, the filter are used to carry out high pressure line feature extraction to the first kind image block, obtain high-voltage line
Edge feature figure；
Line segment detector, the line segment detector are used to carry out Line segment detection to the edge feature figure of the high-voltage line, obtain
Take the fragment figure of high-voltage line；
Line segment splicer, the line segment splicer are used to splice the line segment in the fragment figure of the high-voltage line,
Obtain high-voltage line instruction figure.
Further, the parameter for defining described image block's attribute is stored in advance in the classifier,
The classifier carries out attributive analysis to described image block, when the parameter of described image block meets the ginseng of the first attribute
When number, described image block is denoted as first kind image block by the classifier；
When the parameter of described image block is unsatisfactory for the parameter of the first attribute, described image block is denoted as by the classifier
Two class image blocks.
Further, the filter carries out high-voltage line to each described first kind image block according to following formula 1, formula 2
Feature extraction obtains the final edge characteristic pattern of each first kind image block, will each described first kind image block
Final edge characteristic pattern be combined, obtain the edge feature figure of the high-voltage line,
Wherein, I is first kind image block, IiIt is the characteristic pattern on each described i-th of direction of first kind image block, Ei
It is edge feature figure final on each described i-th direction of first kind image block, 0 Ω of < i≤360/, Ω≤360 °, Fi=Ka
(θi)*Fa+Kb(θi)*Fb+Kc(θi)*Fc+Kd(θi)*FdIt is θiFilter function on direction, θi=i Ω, wherein Fa,Fb,Fc,Fd
It is to be existed using three rank Gaussian derivativesThe filter function of the filter generated on four direction, Ka(θi),Kb
(θi),Kc(θi),Kd(θi) it is to calculate θiFilter function F on directioniWhen interpolating function.
Further, before the filter carries out high pressure line feature extraction to each described first kind image block, also
Autonomous selection parameter σ including determining each first kind image block according to formula 3, determines the filter according to the σ
To the filtering weighting of each first kind image block,
Wherein, C is the contrast of each first kind image block；
The filter function FiFor the filter function for meeting default weight.
The high pressure line detecting method and device of fragmentation provided by the invention can by being divided to image to be detected,
Classification, high pressure line feature extraction, line segment and line segment splicing are finally completed the panorama recovery to high-voltage line, obtain complete
High-voltage line instruction figure is low so as to effectively avoid the flight hazard occurred due to the false judgment to high pressure line position
The high-voltage line threat of empty flying condition provides the solution with obvious application value.
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 this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow chart of the high pressure line detecting method embodiment of fragmentation of the present invention；
Fig. 2 is the schematic diagram of the high pressure line detecting method Edge Gradient Feature process of fragmentation of the present invention；
Fig. 3 is the effect picture of the high-voltage line detection method of fragmentation of the present invention；
Fig. 4 is the structural schematic diagram of the high-voltage line detection device of fragmentation of the invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Fig. 1 is the flow chart of the high pressure line detecting method embodiment of fragmentation of the present invention.As shown in Figure 1, the present embodiment mentions
The high pressure line detecting method of the fragmentation of confession includes:
S101: image to be detected is obtained, image to be detected is divided at least two image blocks.
Wherein, the object that image to be detected is implemented as high-voltage line detection method of the present invention, from the camera shooting system of aircraft
It unites and the scene in front of aircraft flight is shot.
After getting image to be detected, it is necessary first to be divided to image to be detected, each frame image to be detected is drawn
It is divided into many image blocks (patch), which can be divided into image to be detected more patch, thus in subsequent behaviour
Each patch can be tested and analyzed in work, provide effective detection basis for the clear discrimination of high-voltage line.Generally into
The subject of implementation of row division operation is an image segmentating device, is stored with preset division size in the image segmentating device, is carrying out
When division operation, image segmentating device can be divided according to the image to be detected of pre-set dimension to each frame, obtain size uniformity
Consistent multiple patch.Specifically, there is no specific size requirement to pre-set dimension in the present invention, as long as the pre-set dimension energy
Enough meet subsequent sort operation.
S102: classifying to image block, obtains the first kind image block comprising high-voltage line and the not comprising high-voltage line
Two class image blocks, and give up the second class image block.
Due to dividing in obtained multiple patch, includes high-voltage line segment in some patch, do not include in some patch
High-voltage line segment, therefore point can will be opened comprising high-voltage line segment with the patch for not including high-voltage line segment by classifying, when stroke
After point, the patch for not including high-voltage line can be given up to fall, emphasis analyzes the patch for including high-voltage line.
S103: high pressure line feature extraction is carried out to first kind image block, obtains the edge feature figure of high-voltage line.
It is not the institute in first kind patch although wherein containing high-voltage line segment for first kind patch
Content is all high-voltage line, therefore the Edge Gradient Feature of high-voltage line is carried out to first kind patch, only obtains and wherein contains high pressure
The part image data amount of line filters out the part image data without containing high-voltage line, therefore the high-voltage line finally obtained
Edge feature figure be most important structure about high-tension line graph picture.
S104: Line segment detection is carried out to the edge feature figure of high-voltage line, obtains the fragment figure of high-voltage line.
Line segment detection is carried out to the edge feature figure of high-voltage line obtained above, is contained in the fragment figure of obtained high-voltage line
There is the segment of a plurality of high-voltage line, the segment of these high-voltage lines is the segment of all high-voltage lines included in image to be detected.
S105: splicing the line segment in the fragment figure of high-voltage line, obtains high-voltage line instruction figure.
Due to all segments obtained in the fragment figure of high-voltage line be it is discontinuous, can not indicate complete in image to be detected
Whole high pressure line position, therefore can use the space correlation relationship between high-voltage line segment, by trifling multiple high-voltage line pieces
Duan Jinhang rational joint, the final high-voltage line instruction figure obtained containing complete high-voltage line, thus in high-voltage line instruction figure just
It can clearly identify the specific location where high-voltage line.
The high pressure line detecting method of fragmentation provided by the invention can by being divided, being classified to image to be detected,
High pressure line feature extraction, line segment and line segment splicing are finally completed the panorama recovery to high-voltage line, obtain complete high-voltage line
Instruction figure, so as to clearly identify the specific location where high-voltage line, effectively avoids due to high pressure line position
False judgment and the flight hazard occurred provide the solution with obvious application value for the high-voltage line threat of low-latitude flying condition
Certainly scheme.
Further, S102 includes: to carry out attributive analysis at least two image blocks, when the attribute of image block is the first category
When property, image block is denoted as first kind image block；It is the second class figure of note by image block when non-first attribute of the attribute of image block
As block.
The present invention is not intended to limit the specific executing subject for executing sort operation, generally, can be somebody's turn to do using classifier
Item sort operation, it is preferable that can be specifically using based on convolutional neural networks (Convolutional in present embodiment
Neural Network, referred to as: CNN) classifier carry out.
CNN classification layer in the embodiment includes 7 layers of nerve fiber, and it is implicit to be divided into 1 input layer, 1 output layer and 5
Layer, 5 hidden layers include two convolutional layers, two pond layers and a full articulamentum.Wherein input layer and output layer are in two
End position outputs and inputs layer and is used to output and input data, and hidden layer is used to analyze the data of input.
Before CNN classifier carries out sort operation, usually require to be trained CNN classifier.In the present invention,
It trains and refers to will include that the image of high-voltage line and the image not comprising high-voltage line are input in CNN classifier, wherein can incite somebody to action
Include high-voltage line image tagged be 10, by do not include high-voltage line image tagged be 01, CNN classifier analytic learning this
It can carry out the Parameter analysis of attribute when a little images to each image by hidden layer, wherein must have one group of identical parameter meeting
Appear in it is each include high-voltage line image in (the identical parameter of this group is high-voltage line parameter), therefore CNN classifier
Can obtain analyzing result as follows: when containing this group of identical parameters in an image, which is the first attribute and is labeled as 10.
Likewise, being free from high-voltage line parameter in the image not comprising high-voltage line, therefore CNN classifier can will join without containing such
Several image definitions is non-first attribute and is labeled as 01, by carrying out above-mentioned training operation until CNN to CNN classifier repeatedly
Classifier can correctly classify to the image of input, and just training finishes the CNN classifier.
Therefore, the CNN classifier that training is completed classifies to each patch, CNN classifier can be to the category of patch
Property carry out probability analysis, when patch belong to the first attribute probability be greater than patch belong to the probability of non-first attribute when, just will
The patch is denoted as first kind patch, when the probability that CNN detection of classifier to patch belongs to the first attribute belongs to less than patch
When the probability of non-first attribute, the patch is just denoted as the second class patch, and the second class patch is given up, so as to complete
To the sort operation of patch.It is noted here that when the size of the patch of division should be trained with CNN classifier
The size of used image is consistent.
Further, S103 includes: and carries out high-voltage line feature to each first kind image block according to following formula 1, formula 2 to mention
It takes, obtains the final edge characteristic pattern of each first kind image block, by the final edge feature of each first kind image block
Figure is combined, and obtains the edge feature figure of high-voltage line,
Wherein, I is first kind image block, IiIt is the characteristic pattern on each i-th of direction of first kind image block, EiIt is every
Final edge feature figure on one the i-th direction of first kind image block, 0 Ω of < i≤360/, Ω≤360 °, Fi=Ka(θi)*Fa+
Kb(θi)*Fb+Kc(θi)*Fc+Kd(θi)*FdIt is θiFilter function on direction, θi=i Ω, wherein Fa,Fb,Fc,FdIt is using three
Rank Gaussian derivative existsThe filter function generated on four direction, Ka(θi),Kb(θi),Kc(θi),Kd(θi) it is meter
Calculate θiFilter function F on directioniWhen interpolating function.
Specifically, the present invention devises a kind of based on three rank Gaussian derivatives when carrying out feature extraction to first kind patch
Edge Gradient Feature algorithm, wherein using three rank Gaussian derivatives be kernel function filter function as basic function, pass through formula 1, formula
2 couples of each first kind patch carry out feature extraction.
To each first kind patch carry out feature extraction when, using formula 1, formula 2 from i direction to each first
Class patch is calculated, then the calculated result on this i direction is iterated weighting and obtains each first kind patch's
Final edge feature figure obtains the final edge feature figure of each first kind patch, then according to every in this way
Position of one first kind patch in image to be detected carries out the final edge feature figure of each first kind patch
Combination, finally obtains the edge feature figure of high-voltage line.
As soon as example, when carrying out Edge Gradient Feature to first kind patch, we every Ω=10 ° to this first
Class patch is once calculated, and is carried out i=36 in total and is calculated simultaneously iteration update.Fig. 2 is that the high-voltage line of fragmentation of the present invention is examined
The schematic diagram of survey method Edge Gradient Feature process, as shown in Fig. 2, first in θ when iteration starts1(θ1=10 °) on direction to this
A kind of patch carries out that I is calculated1, because of E0=Ο is null matrix, so E1=I1, then in θ2(θ1=20 °) it is right on direction
First kind patch carries out that I is calculated2, according to formula 1, formula 2 by E1andI2It is iterated to obtain E2, loop back and forth like this into
Row goes down, and until the calculated result in 36 directions has all been carried out iteration, the result of last wheel iteration is the first kind
The final edge characteristic pattern of patch, is denoted as E=EM。
According to above-mentioned steps until completing the iteration on 36 directions of all first kind patch, obtain all
The final edge characteristic pattern of each first kind image block is combined, obtains by the final edge feature figure of first kind patch
To the edge feature figure of high-voltage line.
By being calculated on 36 directions, we obtain informative edge feature figure, are in next step
Detection high-voltage line segment provides sufficient information, reduces omission factor when system detection high-voltage line.
Further, before S103 further include: determine the autonomous selection parameter of each first kind image block according to formula 3
σ, σ determine the weight of each first kind image block filter function,
Wherein, C is the contrast of each first kind image block；
Filter function FiFor the filter function for meeting weight.
Since each first kind patch is again distinct, in order to be directed to when carrying out Edge Gradient Feature
Property obtains the best final edge characteristic pattern of each first kind patch, before carrying out S103, it is also necessary to determine each
The autonomous selection parameter σ of first kind patch is determined according to the autonomous selection parameter σ and is carried out edge to each first kind patch
The weight of filter function when feature extraction, to guarantee that omission factor is lower when carrying out the detection of subsequent line segmentization, improves whole
The detection performance of body.
Wherein, the C in formula 3 is the contrast of each first kind patch, σ value according to the variation of contrast C, σ's
Value range is between 0 to 1.
Further, S104 includes: the first line segment in the edge feature figure for exclude high-voltage line, wherein the length of the first line segment
Degree is less than first threshold；
The relating dot and associated point in the edge feature figure of high-voltage line are connected, wherein relating dot is at a distance from associated point
Less than or equal to second threshold, first threshold is greater than second threshold.
In the present invention, it specifically can use Hough transformation algorithm and Line segment detection carried out to the edge feature figure of high-voltage line,
The largest interval threshold value connected by the first threshold MinL of setting detection line segment length with two o'clock, i.e. second threshold MaxG,
In, line segment of the line segment length less than first threshold MinL will be excluded, and two points of the interval less than second threshold MaxG will be connected
It is connected together, ultimately forms the fragment figure of high-voltage line.It is noted here that different first threshold and second threshold take
Value can obtain different Line segment detection effects, therefore in order to reduce omission factor, make to obtain more line segments during line segments extraction, can
To set 20 for first threshold MinL, 7-8 is set by second threshold MaxG.
Further, S105 includes: that the line segment aggregate in the fragment figure of high-voltage line is denoted as S={ l1,l2,l3,...,li-1,
li,li+1,...,ln, liIndicate the segment of the high-voltage line in set；
Work as liAnd ljAll meet following three conditions, connects liAnd lj, obtain high-voltage line instruction figure, in which:
(1)ljStarting point abscissa ratio liTerminal abscissa it is big；
(2)liTerminal and ljThe ordinate of starting point differs minimum；
(3)liAnd ljSlope differ minimum.
After the line segment for generating multiple high-voltage lines in S104, these line segments are not continuous, complete high-voltage line, cannot
Indicate the position of complete high-voltage line in image to be detected.Therefore, after generating the fragment figure of high-voltage line, utilization can be passed through
Trifling high-voltage line fragment assembly is become complete high pressure line segment, will belonged to by the space correlation relationship between high-voltage line segment
It is connected in the segment of same high-voltage line, it is eventually by high-voltage line instruction figure is generated that the high-voltage line in front of flying is really clear
Chu restores.
Specifically, several line segments in the fragment figure of high-voltage line can be indicated with set S, S={ l1,l2,l3,...,
li-1,li,li+1,...,ln, every time using the smallest line segment of abscissa in set S as initial segment, according to above-mentioned three in S set
A feature is found optimal line segment and is attached with it, moves in circles and goes on, so that it may find and belong to same high-voltage line
Line segment and complete splicing to the high-voltage line, accurate location locating for high-voltage line can significantly be indicated by having finally obtained
High-voltage line instruction figure, effectively keeps away to can be made by high-voltage line instruction figure to front high-voltage line when aircraft is in flight
Allow movement.
High pressure line detecting method provided by the invention carries out high for the front video that carry-on video camera takes
Crimping detects, and implements the processing of S101 to S105 in detection process to each frame image of video.For clearer description detection
Process, we are described by Fig. 3, and Fig. 3 is the effect picture of the high-voltage line detection method of fragmentation of the present invention, please refer to Fig. 3.
Wherein, (a) of Fig. 3 is a certain frame image (image to be detected) in video, if do not handled it, aircraft would not
High-voltage line is avoided, will cause greatly dangerous and loss in this way.The height in (a) of Fig. 3 is detected using our method
Crimping needs to divide the image in (a) of Fig. 3 more patch and obtains (b) of Fig. 3, the size of each patch in (b) of Fig. 3
It is uniform.It all include high-voltage line in not all patch after division obtains more patch, in order to exclude not comprising high-voltage line
Patch, the present embodiment classifies to these patch using CNN classifier, as can be seen that passing through CNN from (c) of Fig. 3
Only retain those after classifier classification and be classified device be judged as include high-voltage line first kind patch, eliminate a large amount of
Not comprising the second class patch for having high-voltage line.It is the side carried out for first kind patch in multiple directions shown in (d) of Fig. 3
The edge feature figure obtained after edge feature extraction, by being weighted the calculated result on different directions to have obtained high pressure
The abundant information at line edge, thus the line segmentization detection being unlikely to after influencing.(e) of Fig. 3 is detected using Hough transformation
The fragment figure of high-voltage line out.Finally, it is associated to obtain the instruction figure of high-voltage line using the spatial relationship between line segment, i.e.,
The specific location of high-voltage line is marked in original image in (f) of (f) of Fig. 3, Fig. 3, so that high-voltage line is in video image
Position is clearer.
Fig. 4 is the structural schematic diagram of the high-voltage line detection device of fragmentation of the invention, as shown in figure 4, piece of the invention
The high-voltage line detection device of sectionization includes:
Image segmentating device 1, image segmentating device 1 are used to image to be detected being divided at least two image blocks；Classifier 2,
Classifier 2 obtains the first kind image block comprising high-voltage line and second not comprising high-voltage line for classifying to image block
Class image block, and give up the second class image block；Filter 3, filter 3 are used to carry out high-voltage line feature to first kind image block to mention
It takes, obtains the edge feature figure of high-voltage line；Line segment detector 4, line segment detector 4 are used to carry out the edge feature figure of high-voltage line
Line segment detection obtains the fragment figure of high-voltage line；Line segment splicer 5, line segment splicer 5 are used for the line in the fragment figure of high-voltage line
Segment is spliced, and high-voltage line instruction figure is obtained.
Further, be stored in advance the parameter for defining image block's attribute in classifier 2, classifier 2 to image block into
Row attributive analysis, when the parameter of image block meets the parameter of the first attribute, image block is denoted as First Kind Graph picture by classifier 2
Block；When the parameter of image block is unsatisfactory for the parameter of the first attribute, image block is denoted as the second class image block by classifier 2.
Further, filter 3 proposes each first kind image block progress high-voltage line feature according to following formula 1, formula 2
It takes, obtains the final edge characteristic pattern of each first kind image block, by the final edge feature of each first kind image block
Figure is combined, and obtains the edge feature figure of high-voltage line,
Wherein, I is first kind image block, IiIt is the characteristic pattern on each i-th of direction of first kind image block, EiIt is every
Final edge feature figure on one the i-th direction of first kind image block, 0 Ω of < i≤360/, Ω≤360 °, Fi=Ka(θi)*Fa+
Kb(θi)*Fb+Kc(θi)*Fc+Kd(θi)*FdIt is θiFilter function on direction, θi=i Ω, wherein Fa,Fb,Fc,FdIt is using three
Rank Gaussian derivative existsThe filter function of the filter generated on four direction, Ka(θi),Kb(θi),Kc(θi),Kd
(θi) it is to calculate θiFilter function F on directioniWhen interpolating function.
It further, further include basis before filter 3 carries out high pressure line feature extraction to each first kind image block
Formula 3 determines the autonomous selection parameter σ of each first kind image block, determines filter 3 to each first kind image block according to σ
Filtering weighting,
Wherein, C is the contrast of each first kind image block；
Filter function FiFor the filter function for meeting default weight.
Further, first threshold and second threshold is stored in advance in line segment detector 4；Line segment detector 4 excludes high pressure
The first line segment in the edge feature figure of line, wherein the length of the first line segment is less than first threshold；Line segment detector 4 connects high pressure
Relating dot and associated point in the edge feature figure of line, wherein relating dot is less than or equal to the second threshold at a distance from associated point
Value, first threshold are greater than second threshold.
Further, the line segment aggregate in the fragment figure of high-voltage line is denoted as S={ l1,l2,l3,...,li-1,li,
li+1,...,ln, liIndicate the segment of the high-voltage line in set；Work as liAnd ljAll meet following three conditions, line segment splicer
5 connection liAnd lj, in which:
(1)ljStarting point abscissa ratio liTerminal abscissa it is big；
(2)liTerminal and ljThe ordinate of starting point differs minimum；
(3)liAnd ljSlope differ minimum.
The device of the present embodiment can be used for executing the technical solution of embodiment of the method shown in Fig. 1 to Fig. 3, realize former
Manage similar, details are not described herein again.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above-mentioned each method embodiment can lead to
The relevant hardware of program instruction is crossed to complete.Program above-mentioned can be stored in a computer readable storage medium.The journey
When being executed, execution includes the steps that above-mentioned each method embodiment to sequence；And storage medium above-mentioned include: ROM, RAM, magnetic disk or
The various media that can store program code such as person's CD.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (4)
1. a kind of high-voltage line detection method of fragmentation characterized by comprising
Image to be detected is obtained, described image to be detected is divided at least two image blocks, described image to be detected is flight
The scene in front shoots gained；
Classify to described image block, obtains the first kind image block comprising high-voltage line and the second class figure not comprising high-voltage line
As block, and give up the second class image block；
High pressure line feature extraction is carried out to the first kind image block, obtains the edge feature figure of high-voltage line；
Line segment detection is carried out to the edge feature figure of the high-voltage line, obtains the fragment figure of high-voltage line；
Line segment in the fragment figure of the high-voltage line is spliced, high-voltage line instruction figure is obtained；
It is described that high pressure line feature extraction is carried out to the first kind image block, obtain the edge feature figure of high-voltage line, comprising:
High pressure line feature extraction is carried out each described first kind image block according to following formula 1, formula 2, is obtained described in each
The final edge characteristic pattern of each first kind image block is carried out group by the final edge characteristic pattern of first kind image block
It closes, obtains the edge feature figure of the high-voltage line,
Wherein, I is first kind image block,It is the characteristic pattern on each described i-th of direction of first kind image block,It is every
Final edge feature figure on one i-th direction of first kind image block, 0 Ω of < i≤360/, Ω≤360 °,It isFilter function on direction, θi=i
Ω, whereinIt is to be existed using three rank Gaussian derivativesThe filter function generated on four direction,It is to calculateFilter function on directionWhen interpolating function；
It is described that high pressure line feature extraction is carried out to the first kind image block, it is also wrapped before obtaining the edge feature figure of high-voltage line
It includes: determining that the autonomous selection parameter σ of each first kind image block, the σ determine each described first kind according to formula 3
The weight of image block filter function,
Wherein, C is the contrast of each first kind image block；
Filter functionFor the filter function for meeting the weight；
The edge feature figure to the high-voltage line carries out Line segment detection, obtains the fragment figure of high-voltage line, comprising:
The first line segment in the edge feature figure of the high-voltage line is excluded, wherein the length of the First Line section is less than the first threshold
Value；
The relating dot and associated point in the edge feature figure of the high-voltage line are connected, wherein the relating dot and associated point
Distance is less than or equal to second threshold, and the first threshold is greater than the second threshold；
Line segment in the fragment figure of the high-voltage line is spliced, high-voltage line instruction figure is obtained, comprising:
Line segment aggregate in the fragment figure of the high-voltage line is denoted as,Indicate collection
The segment of high-voltage line in conjunction；
As the liAnd ljAll meet following three conditions, connects the liWith the lj, the high-voltage line instruction figure is obtained,
In:
(1) ljStarting point abscissa ratio liTerminal abscissa it is big；
(2) liTerminal and ljThe ordinate of starting point differs minimum；
(3) liAnd ljSlope differ minimum.
2. detection method according to claim 1, which is characterized in that it is described to classify to described image block, it is wrapped
First kind image block containing high-voltage line and the second class image block not comprising high-voltage line, and give up the second class image block, it wraps
It includes:
Attributive analysis is carried out at least two image block, when the attribute of described image block is the first attribute, by the figure
As block is denoted as first kind image block；It is the second class figure of note by described image block when non-first attribute of the attribute of described image block
As block.
3. a kind of high-voltage line detection device of fragmentation characterized by comprising
Image segmentating device, described image divider is used to image to be detected being divided at least two image blocks, described to be detected
Image is the scene shooting gained in flight front；
Classifier, the classifier for classifying to described image block, obtain the first kind image block comprising high-voltage line and
The second class image block not comprising high-voltage line, and give up the second class image block；
Filter, the filter are used to carry out high pressure line feature extraction to the first kind image block, obtain the side of high-voltage line
Edge characteristic pattern；
Line segment detector, the line segment detector are used to carry out Line segment detection to the edge feature figure of the high-voltage line, obtain high
The fragment figure of crimping；
Line segment splicer, the line segment splicer are obtained for splicing to the line segment in the fragment figure of the high-voltage line
High-voltage line instruction figure；
The filter carries out high pressure line feature extraction each described first kind image block according to following formula 1, formula 2, obtains
The final edge characteristic pattern of each first kind image block, by the final edge feature of each first kind image block
Figure is combined, and obtains the edge feature figure of the high-voltage line,
Wherein, I is first kind image block,It is the characteristic pattern on each described i-th of direction of first kind image block,It is every
Final edge feature figure on one i-th direction of first kind image block, 0 Ω of < i≤360/, Ω≤360 °,It isFilter function on direction, θi=i
Ω, whereinIt is to be existed using three rank Gaussian derivativesThe filter generated on four direction
Filter function,It is to calculateFilter function on directionWhen interpolating function；
It further include true according to formula 3 before the filter carries out high pressure line feature extraction each described first kind image block
The autonomous selection parameter σ of each fixed first kind image block, according to the σ determine the filter each described the
The filtering weighting of a kind of image block,
Wherein, C is the contrast of each first kind image block；
The filter functionFor the filter function for meeting default weight.
4. detection device according to claim 3, which is characterized in that be stored in advance in the classifier described for defining
The parameter of image block's attribute,
The classifier carries out attributive analysis to described image block, when the parameter of described image block meets the parameter of the first attribute
When, described image block is denoted as first kind image block by the classifier；
When the parameter of described image block is unsatisfactory for the parameter of the first attribute, described image block is denoted as the second class by the classifier
Image block.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610249136.5A CN105825207B (en) | 2016-04-20 | 2016-04-20 | The high-voltage line detection method and device of fragmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610249136.5A CN105825207B (en) | 2016-04-20 | 2016-04-20 | The high-voltage line detection method and device of fragmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105825207A CN105825207A (en) | 2016-08-03 |
| CN105825207B true CN105825207B (en) | 2019-05-14 |
Family
ID=56526296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610249136.5A Active CN105825207B (en) | 2016-04-20 | 2016-04-20 | The high-voltage line detection method and device of fragmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105825207B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934455B (en) * | 2017-02-14 | 2019-09-06 | 华中科技大学 | Remote sensing image optics adapter structure choosing method and system based on CNN |
| CN109559306B (en) * | 2018-11-27 | 2021-03-12 | 广东电网有限责任公司广州供电局 | Crosslinked polyethylene insulating layer surface smoothness detection method based on edge detection |
| CN111866484B (en) * | 2019-04-30 | 2023-06-20 | 华为技术有限公司 | Point cloud encoding method, point cloud decoding method, device and storage medium |
| CN110428424B (en) * | 2019-07-02 | 2023-04-07 | 中国航空工业集团公司雷华电子技术研究所 | Radar echo image high-voltage line segmentation method based on deep learning |
| CN110349104A (en) * | 2019-07-08 | 2019-10-18 | 创新奇智(南京)科技有限公司 | A kind of method of the COMPUTER DETECTION with groove bolt looseness, computer-readable medium and system |
| CN112040117A (en) * | 2020-07-22 | 2020-12-04 | 广东顺德电力设计院有限公司 | Method and device for polling and shooting high-voltage wire |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930280A (en) * | 2012-10-05 | 2013-02-13 | 中国电子科技集团公司第十研究所 | Method for identifying overhead high-voltage wire automatically from infrared image |
| CN103761722A (en) * | 2014-01-16 | 2014-04-30 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Fixed wing unmanned aerial vehicle touring image accurately-splicing method for power transmission line |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8306327B2 (en) * | 2008-12-30 | 2012-11-06 | International Business Machines Corporation | Adaptive partial character recognition |
-
2016
- 2016-04-20 CN CN201610249136.5A patent/CN105825207B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930280A (en) * | 2012-10-05 | 2013-02-13 | 中国电子科技集团公司第十研究所 | Method for identifying overhead high-voltage wire automatically from infrared image |
| CN103761722A (en) * | 2014-01-16 | 2014-04-30 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Fixed wing unmanned aerial vehicle touring image accurately-splicing method for power transmission line |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105825207A (en) | 2016-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105825207B (en) | The high-voltage line detection method and device of fragmentation | |
| Jain et al. | Weapon detection using artificial intelligence and deep learning for security applications | |
| CN109034044B (en) | Pedestrian re-identification method based on fusion convolutional neural network | |
| US8503792B2 (en) | Patch description and modeling for image subscene recognition | |
| CN104573111B (en) | Pedestrian's data structured in a kind of monitor video stores and preindexing method | |
| KR102328734B1 (en) | Method for automatically evaluating labeling reliability of training images for use in deep learning network to analyze images, and reliability-evaluating device using the same | |
| CN108985192A (en) | A kind of video smoke recognition methods based on multitask depth convolutional neural networks | |
| CN106915462A (en) | Forestry pests ＆ diseases intelligent identifying system based on multi-source image information | |
| CN109472193A (en) | Method for detecting human face and device | |
| CN112287983B (en) | Remote sensing image target extraction system and method based on deep learning | |
| CN105139429A (en) | Fire detecting method based on flame salient picture and spatial pyramid histogram | |
| CN111985374A (en) | Face positioning method and device, electronic equipment and storage medium | |
| CN111191531A (en) | Rapid pedestrian detection method and system | |
| CN111062950A (en) | Method, storage medium and equipment for multi-class forest scene image segmentation | |
| Petso et al. | Individual animal and herd identification using custom YOLO v3 and v4 with images taken from a uav camera at different altitudes | |
| CN115272882A (en) | Discrete building detection method and system based on remote sensing image | |
| CN115527098A (en) | Infrared small target detection method based on global mean contrast space attention | |
| Zheng et al. | A lightweight algorithm capable of accurately identifying forest fires from UAV remote sensing imagery | |
| CN116503715B (en) | Forest fire detection method based on cascade network | |
| Kang et al. | Robust adversarial attack against explainable deep classification models based on adversarial images with different patch sizes and perturbation ratios | |
| CN112598032A (en) | Multi-task defense model construction method for anti-attack of infrared image | |
| He et al. | Real-time pedestrian warning system on highway using deep learning methods | |
| CN116188442A (en) | High-precision forest smoke and fire detection method suitable for any scene | |
| Tiwari et al. | Detection of Camouflaged Drones using Computer Vision and Deep Learning Techniques | |
| Zhou et al. | An anchor-free vehicle detection algorithm in aerial image based on context information and transformer |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |