CN106251332B - SAR image airport target detection method based on edge feature - Google Patents
SAR image airport target detection method based on edge feature Download PDFInfo
- Publication number
- CN106251332B CN106251332B CN201610561336.4A CN201610561336A CN106251332B CN 106251332 B CN106251332 B CN 106251332B CN 201610561336 A CN201610561336 A CN 201610561336A CN 106251332 B CN106251332 B CN 106251332B
- Authority
- CN
- China
- Prior art keywords
- edge
- feature
- runway
- gray
- subset
- 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 62
- 238000000034 method Methods 0.000 claims abstract description 32
- 230000002776 aggregation Effects 0.000 claims description 23
- 238000004220 aggregation Methods 0.000 claims description 23
- 238000003708 edge detection Methods 0.000 claims description 21
- 238000012216 screening Methods 0.000 claims description 12
- 239000007799 cork Substances 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 6
- 238000007689 inspection Methods 0.000 abstract description 18
- 238000004364 calculation method Methods 0.000 abstract description 9
- 238000000605 extraction Methods 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 description 16
- 239000011159 matrix material Substances 0.000 description 9
- 230000002093 peripheral effect Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 230000001186 cumulative effect Effects 0.000 description 2
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000003064 k means clustering Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 239000011800 void material Substances 0.000 description 2
- 210000000988 bone and bone Anatomy 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000007323 disproportionation reaction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000005094 fruit set Effects 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30256—Lane; Road marking
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
There is empty inspection, missing inspection and airfield runway when brightness of image unevenness and when noise is more when mainly solving airport target detection and positions not accurate technical problem in the SAR image airport target detection method based on edge feature that the invention discloses a kind of.Its realization process is: inputting airport target SAR image to be detected；Edge image is obtained using average ratio edge detector；Feature extraction is carried out to edge using method proposed by the present invention and cluster is screened；Candidate marginal is attached using method proposed by the present invention；The detection to airfield runway is completed by parallel lines detection technique, realizes the detection of airport target.The present invention can be accurately finished the detection of SAR image airport target, and runway positioning is more accurate, and calculation amount is small, reduce empty inspection, reduce runing time, can be used in spaceborne and on-board SAR image processing system and related objective detection system.
Description
Technical field
The invention belongs to SAR image processing technology fields, relate generally to airport target detection, specifically one kind is based on
The SAR image airport target detection method of edge feature, can be used for the detection to SAR image airport target.
Background technique
SAR image target detection is an important content in SAR image treatment research, and the purpose is to according to digital picture
The prior information of handling principle combination SAR image detects target.
Airport target identification is an important branch of SAR image auxiliary mark identifying system, and the detection of airfield runway is
The basis of entire airport target identification.Due to SAR image ground object target detection in military field using more, SAR figure
As target detection in SAR image process field in occupation of very important status, become one of most basic technology in the field.
In previous airport target detection algorithm, according to utilized airfield runway characteristic type, the thinking of runway detection
It is generally divided into two classes: first is that the threshold value of SAR image is chosen using maximum entropy criterion using gray feature, after space filtering
Target image after to segmentation, to achieve the purpose that detect airfield runway；Second is that using geometry linear feature, to binaryzation
Image afterwards is carried out Hough transform and is counted using the length information of airfield runway to Hough transform, and airfield runway is obtained
Bone Edge straight line.
Though above-mentioned airfield runway detection algorithm has certain effect, still Shortcomings.If utilizing runway
Gray feature, due to brightness of image is uneven and runway around some regions it is close with the gray value of runway, when image segmentation
Airport target position inaccurate is easily caused, these images are hardly resulted in satisfied testing result.Hough transform is that one kind makes
With the method for parameter estimation of voting principle, principle is the point-line duality using image space and Hough parameter space,
Test problems in image space are transformed into parameter space.By carrying out simple cumulative statistics in parameter space, then exist
The method that Hough parameter space finds accumulator peak value detects straight line.If using Hough transform method, although Hough transform
The straight line of extraction is longest, but it is made an uproar using the statistical property for being image since real SAR image unavoidably exists
Sound may count ideal lines in this direction, cause airfield runway if noise in one direction is more
Positioning is inaccurate, and Hough transform is computationally intensive, consumes a large amount of storage spaces and processor time, does not meet spaceborne and airborne
The requirement of radar real-time detection.
Summary of the invention
When it is an object of the invention to for existing for above-mentioned prior art due to brightness of image unevenness and noise compared with
Occur empty inspection, missing inspection and airfield runway when more and position not accurate problem, proposes a kind of empty inspection of reduction, keep target positioning more smart
Standard, while calculation amount is reduced, reduce the SAR image airport target detection method based on edge feature of runing time.
To achieve the above object, the present invention includes the following steps:
(1) airport target SAR image I to be detected is inputted；
(2) edge detection is carried out to image I using average ratio detector, average ratio edge detection is carried out to image I
(ratio of average, ROA), the gray average g of airfield runway are smaller compared near zone and in a certain range, i.e. g ∈
[g1, g2] retains gray average gmin∈[g1,g2] marginal point, obtain edge detection results image；
(3) the runway marginal point based on Edge Gradient Feature and cluster screening extracts, and tracks side using Freeman chain code
Edge detection result image obtains edge aggregation L={ l1,l2,l3,…ln, wherein lnIndicate edge aggregation L={ l1,l2,l3,…
lnIn n-th of edge subset being made of pixel, n=1,2 ..., extract edge aggregation L={ l1,l2,l3,…lnZhong Gezi
The feature of collection, the present invention choose the maximum gray scale value set G of edge probability of occurrence, the maximum gradient direction collection of edge probability of occurrence
Close T, the gray average set M in both sides of edges region, the gray variance set S in both sides of edges region and both sides of edges region
The feature vector set H of gray level co-occurrence matrixes, obtains edge feature set Feature={ G, T, M, S, H }, by edge aggregation L
={ l1,l2,l3,…lnClustered according to edge feature Feature={ G, T, M, S, H }, obtain edge cluster set LF=
{lf1,lf2,lf3,…lfk, k is cluster number, lfkSubset is clustered for kth class edge, according to prior information it is found that runway
Gray value is lower compared with peripheral region, and texture is more uniform and regular, and random noise is less, therefore can reject and not meet runway side
The edge aggregation of edge feature retains the edge cluster result set for meeting track features, i.e. Lr={ lf1,lf2,…lfp, p≤
k；
(4) track features will be met and the pixel belonged in of a sort edge aggregation of hideing is fitted, connects, according to
Runway gray feature rule is attached straightway, the linear edge set Str=for meeting track features after being connected
{s1,s2,…,sq}；
(5) detection to runway is completed by parallel lines detection technique, is extracted according to the characteristics of runway sides aligned parallel straight line
Linear edge set Str={ s1,s2,…,sqIn runway edge parallel lines pair, it is straight to obtain final runway sides aligned parallel
Line is to set Par={ P1,P2,…Pr, PrIt indicates r-th of subset, completes the airport mesh of airport target SAR image I to be detected
Target detection.
By the present invention in that obtaining edge image with average ratio detector (ratio of average, ROA), this hair is used
The Edge Gradient Feature of bright proposition and the runway edge point extracting method opposite side edge of cluster screening carry out feature extraction and cluster sieve
Choosing, is then fitted candidate marginal, is attached using runway gray feature rule proposed by the present invention, by parallel
The detection to airfield runway is completed in line detection, to realize the detection of airport target.
The present invention has the advantage that compared with prior art
1. the present invention proposes the airfield runway detection method based on edge feature, edge pixel is assigned with multiple characteristics, is made
It is more accurate to the positioning in airfield runway region to obtain, and is eliminated using cluster screening process and does not meet airfield runway edge spy largely
The edge of sign reduces a large amount of empty inspections, missing inspection target, keeps target positioning more accurate, while decreasing calculation amount, reduce
Runing time reduces memory space required when operation；
2. the present invention proposes the airfield runway detection method based on edge feature, in the fitting connection procedure at edge, according to
According to the priori knowledge of airfield runway, it is attached, is avoided due to mistake using runway gray feature rule proposed by the present invention
Occur false target, ideal lines and position inaccurate problem caused by connection.
Detailed description of the invention
Fig. 1 is implementation flow chart of the invention；
Fig. 2 is the SAR image that a width includes airfield runway；
Fig. 3 is 8 direction sliding window schematic diagram used in average ratio edge detection；
Fig. 4 is edge detection results figure of the present invention to Fig. 2；
Fig. 5 is the edge pixel classification results figure that meets track features of the present invention to Fig. 2；
Fig. 6 is airfield runway testing result figure of the present invention to Fig. 2；
Fig. 7 is a width triangle airfield runway SAR image and testing result figure, and wherein Fig. 7 (a) is that a width triangle airport is run
Road SAR image, Fig. 7 (b) are airfield runway testing result figure of the present invention to Fig. 7 (a)；
Fig. 8 is a width right angle airfield runway SAR image and testing result figure, and wherein Fig. 8 (a) is that a width right angle airport is run
Road SAR image, Fig. 8 (b) are airfield runway testing result figure of the present invention to Fig. 8 (a)；
Fig. 9 is the horizontal airfield runway SAR image of a width and testing result figure, and wherein Fig. 9 (a) is that the horizontal airport of a width is run
Road SAR image, Fig. 9 (b) are airfield runway testing result figure of the present invention to Fig. 9 (a)；
Figure 10 is the vertical airfield runway SAR image of a width and testing result figure, and wherein Figure 10 (a) is the vertical airport of a width
Runway SAR image, Figure 10 (b) are airfield runway testing result figure of the present invention to Figure 10 (a).
Specific embodiment
Referring to the drawings, technical solutions and effects of the present invention is described in further detail.
Existing airfield runway detection technique, if using the gray feature of runway, due to image in image segmentation process
The problem of brightness disproportionation and runway gray scale and some regions of surrounding are close, easily cause traffic pattern position inaccurate；If sharp
Straight line is detected with Hough transform principle, Hough transform utilizes the point-line duality of image space and Hough parameter space, leads to
It crosses and carries out simple cumulative statistics in Hough parameter space, the method for finding accumulator peak value detects straight line, due to Hough
What transform method utilized is the statistical property of image, and real SAR image unavoidably has noise, if in one direction
Noise is more, then may count ideal lines in this direction, causes airfield runway positioning inaccurate, and Hough transform
Parameter is more, computationally intensive, consumes a large amount of storage spaces and processor time, and real-time is low.
Embodiment 1
For this purpose, the present invention proposes a kind of SAR image airport target detection method based on edge feature, referring to Fig. 1, including
It has the following steps:
(1) the airport target image comprising airfield runway is obtained using airborne radar, inputs airport target SAR to be detected
Image I is the SAR image for including airfield runway as shown in Fig. 2, Fig. 7 (a), Fig. 8 (a), Fig. 9 (a), Figure 10 (a), practical
The SAR image comprising entire airport can be inputted when operation, or inputs the SAR image for needing the local airport detected.
(2) using average ratio detector to image I carry out edge detection, using size be N × N (N=2 × n-1, n=1,
2 ...) 8 direction sliding window w, as shown in figure 3, can select sliding window size according to picture size when practical operation, this example uses size
For 13 × 13 8 direction sliding windows, average ratio edge detection (ratio of average, ROA) is carried out to image I, is schemed according to SAR
As imaging characteristics, the gray average g of airfield runway is smaller compared near zone and in a certain range, i.e. g ∈ [g1, g2], wherein
Section [g1, g2] is runway intensity value ranges, and g1 is runway minimum gray value, and g2 is runway gray scale maximum value.Sliding window w trellis diagram
As marking the pixel for meeting following two condition as marginal point in operation, first condition is gray average in 8 directions
Ratio minimum and ratio are less than the pixel of given threshold, and given threshold is generally selected according to image resolution ratio and signal-to-noise ratio
It takes, belongs to routine operation；Second condition is the value of the lesser side of gray average in the smallest direction of gray average ratio
gminBelong to the pixel of airfield runway intensity value ranges, wherein gmin∈ [g1, g2], to meeting two above condition simultaneously
Pixel is marked, and obtains edge detection results image, since SAR image has multiplicative noise, average ratio edge detection more
With preferable noise immunity, other edge detection methods can also be used in practical operation.
(3) the runway marginal point based on Edge Gradient Feature and cluster screening extracts, and tracks side using Freeman chain code
Edge detection result image obtains edge aggregation L={ l1,l2,l3,…ln, wherein liFor edge aggregation L={ l1,l2,l3,…ln}
In i-th of subset, i=1,2,3 ..., n, extract edge aggregation L={ l1,l2,l3,…lnIn each subset feature, wherein
I subset liFeature be respectively the maximum gray value g of edge probability of occurrencei, the maximum gradient direction t of edge probability of occurrencei,
The gray average m in both sides of edges regioni, the gray variance s in both sides of edges regioni, and the gray level co-occurrence matrixes in both sides of edges region
Feature vector hi, the feature of each subset is integrated, i.e. the maximum gray scale value set G={ g of edge probability of occurrence1,g2,…
gn, the maximum gradient direction set T={ t of edge probability of occurrence1,t2,…,tn, the gray average set M in both sides of edges region
={ m1,m2,…,mn, the gray variance set S={ s in both sides of edges region1,s2,…,sn, the gray scale in both sides of edges region is total
The feature vector set H={ h of raw matrix1,h2,…,hn, edge feature set Feature={ G, T, M, S, H } is obtained, by side
Edge set L={ l1,l2,l3,…lnClustered according to edge feature set Feature={ G, T, M, S, H }, the present invention makes
Cluster operation is carried out with K-means algorithm, obtains edge cluster set LF={ lf1,lf2,lf3,…lfk, k is cluster number,
lfkFor kth class edge subset, according to prior information it is found that the gray value of runway is lower compared with peripheral region, texture it is more uniform and
Regular, random noise is less, therefore can reject the edge cluster set LF={ lf for not meeting runway edge feature1,lf2,
lf3,…lfkIn subset, retain and meet the subsets of track features, i.e. Lr={ lf1,lf2,…lfp, p≤k, in practical operation
Different clustering algorithms can also be chosen to be clustered to obtain edge cluster set.
The present invention is by choosing the maximum gray scale value set G of edge probability of occurrence, the maximum gradient side of edge probability of occurrence
To set T, the gray average set M in both sides of edges region, the gray variance set S in both sides of edges region and both sides of edges area
The feature vector set H of the gray level co-occurrence matrixes in domain is clustered, summarise substantially airfield runway gray scale, gradient, texture this
The feature of three aspects, has comprehensively carried out semantic description to runway.It is not met largely being eliminated using cluster screening process
The edge of airfield runway edge feature, while calculation amount is decreased, runing time is reduced, it is required when reducing operation to deposit
Store up space.
(4) track features will be met and the pixel belonged in of a sort edge aggregation of hideing is fitted, connects, this hair
It is bright that pixel is fitted using least square method, other Algorithm of fitting a straight line can be chosen in practical operation, according to the present invention
The runway gray feature rule of proposition is attached straightway, that is to say, that and the gray value according to runway edge is almost the same,
Slope, intercept are approximate, and edge to be connected these features consistent with the connection gray average size relation of both sides of edges are connected
It connects, the linear edge set Str={ s for meeting track features after being connected1,s2,…,sq}.Hideing in the present invention, it is same to belong to
Pixel in the edge aggregation of class is also referred to as candidate marginal.
(5) detection to runway is completed by parallel lines detection technique, is mentioned according to the feature that runway edge is parallel lines
Cut-off line edge aggregation Str={ s1,s2,…,sqIn runway edge parallel lines pair, obtain final runway sides aligned parallel
Straight line is to set Par={ P1,P2,…Pr},PrR-th of parallel lines is indicated to subset, final runway sides aligned parallel straight line is to collection
Closing Par is runway detection as a result, completing the target of airport target SAR image I to be detected by the detection to airfield runway
Detection.
The present invention sufficiently combines the priori knowledge of SAR image airport target, under normal circumstances, all deposits in artificial aerodrome target
In linear runway, the present invention is not suitable for the special shapes runway such as round, U-shaped only for the linear runway of ordinary straight, and runs
Road both sides of the edge straight line parallel, under the premise of this, the present invention proposes the airfield runway detection based on edge feature, assigns edge
Pixel is with feature, so that more accurate to the positioning in airfield runway region, using clustering, screening process removal is a large amount of not to meet machine
The edge of field runway edge feature, the empty inspection of reduction, while calculation amount is reduced, runing time is reduced, reduction is required when running to deposit
Storage space be attached in the fitting connection procedure at edge using runway gray feature rule proposed by the present invention, avoid by
Occur false target and position inaccurate problem caused by incorrect link.
Embodiment 2
SAR image airport target detection method based on edge feature with embodiment 1, wherein in step (3) to edge into
Row feature extraction and cluster screening, comprise the following steps that
(3.1) edge detection results figure is tracked using 8 direction Freeman chain codes, obtains edge aggregation L={ l1,l2,
l3,…ln, wherein lnIndicate n-th of edge subset being made of edge pixel point.
(3.2) edge aggregation L={ l is extracted1,l2,l3,…lnThe feature of each subset in set, the present invention uses edge picture
The maximum gray scale value set G of plain probability of occurrence, the maximum gradient direction set T of probability of occurrence of edge pixel, both sides of edges ash
Spend equal value set A, both sides of edges gray variance set S and edge pixel two sides gray level co-occurrence matrixes feature vector set H this
Several features, features described above is comprehensive, obtain edge feature set Feature={ G, T, M, S, H }.By edge aggregation L=
{l1,l2,l3,…lnClustered according to edge feature set Feature={ G, T, M, S, H }, the present invention uses K-means
Clustering algorithm obtains edge cluster set LF={ lf1,lf2,lf3,…lfk, k is cluster sum, lfkIt is poly- for the edge of kth class
Other clustering algorithms can be used in practical operation in the subset of class set.
The present invention chooses the maximum gray scale value set G of edge probability of occurrence, the maximum gradient direction collection of edge probability of occurrence
Close T, the gray average set M in both sides of edges region, the gray variance set S in both sides of edges region and both sides of edges region
The feature vector set H of gray level co-occurrence matrixes summarises the feature of the gray scale, gradient, texture of airfield runway in terms of these three and carries out
Cluster, has accurately comprehensively carried out semantic description to runway.
(3.4) by priori knowledge it is found that the gray value of runway is lower than peripheral region gray value, the texture of runway is more compared with surrounding
Adding uniform rule, random noise is less, then CON, ENT of runway zone smaller compared with runway peripheral region, runway zone
ASM, IDM, COR value are bigger compared with peripheral region, close LF={ lf to edge cluster set1,lf2,lf3,…lfkScreened, if side
Edge clusters set LF={ lf1,lf2,lf3,…lfkIn k-th of subset lfkMeet the track features in priori knowledge, i.e. the son
Collection in edge feature set Feature={ G, T, M, S, H } in corresponding k-th of character subset both sides of edges gray average to
Measure mk={ mk1,mk2Meet mk1>mk2When, then the feature vector h of gray level co-occurrence matrixesk={ ASMk,CONk,IDMk,ENTk,
CORkIn element should meet ASMk1<ASMk2、CONk1>CONk2、IDMk1<IDMk2、ENTk1>ENTk2、CORk1<CORk2；When this
Subset edge lines two sides gray scale in corresponding k-th of character subset in edge feature set Feature={ G, T, M, S, H }
Mean vector mk={ mk1,mk2Meet mk1<mk2When, then the feature vector h of gray level co-occurrence matrixesk={ ASMk,CONk,IDMk,
ENTk,CORkElement in vector should meet ASMk1>ASMk2、CONk1<CONk2、IDMk1>IDMk2、ENTk1<ENTk2、CORk1>
CORk2, other situations are then unsatisfactory for airfield runway feature, such as mk1=mk2When, runway edge can not then be differentiated, and it is poly- to retain edge
Class set LF={ lf1,lf2,lf3,…lfkIn meet the subsets of track features, obtain the edge cluster knot for meeting track features
Fruit set Lr={ lf1,lf2,…lfp, this meets the edge cluster result set Lr={ lf of track features1,lf2,…lfpIn
Pixel be meet track features and the pixel belonged in of a sort edge aggregation of hideing, alternatively referred to as candidate edge picture
Vegetarian refreshments.
The present invention proposes the airfield runway detection method based on edge feature, is carrying out feature extraction and cluster sieve to edge
It selects and sufficiently combines gray scale, gradient, the priori knowledge of texture in step, assign edge pixel with multiple characteristics, so that being run to airport
The positioning in road region is more accurate, and the edge for not meeting airfield runway edge feature largely is removed using cluster screening process, is subtracted
Few empty inspection, while calculation amount and runing time are reduced, reduce memory space required when operation.
Embodiment 3
SAR image airport target detection method based on edge feature is with embodiment 1-2, wherein to candidate in step (4)
Edge pixel point is fitted, connects, and comprises the following steps that
(4.1) according to least square method to edge cluster result set Lr={ lf1,lf2,…lfpIn each subset edge
Pixel carries out straight line fitting, the linear edge set Ls={ ls after being fitted1,ls2,…lsp, it can root in practical operation
Different Algorithm of fitting a straight line is used according to demand；
(4.2) to the linear edge set Ls={ ls after fitting1,ls2,…lspIn m-th of subset lsm={ linem1,
linem2,…linemnIn all straightways, connected two-by-two according to runway gray feature rule, that is, meet runway gray scale
Characterization rules then carry out straightway connection, are unsatisfactory for the straight line for meeting track features then without connection, after being connected
Edge set Str={ s1,s2,…,st}。
The present invention propose the airfield runway detection method based on edge feature, candidate edge pixel is fitted,
In Connection Step, rule is connected using the straightway of runway gray feature proposed by the present invention and taps into the connection of row straightway, bonding machine
The gray scale priori knowledge of field runway is avoided as generating the inspection of void caused by incorrect link and position inaccurate when noise spot is more
Problem.
Embodiment 4
SAR image airport target detection method based on edge feature combines runway with embodiment 1-3 in step (4.2)
Gray feature rule carries out straightway connection, and connection procedure includes the following steps:
(4.2.1)linemiFor linear edge set Ls={ ls1,ls2,…lspIn m-th of subset lsm={ linem1,
linem2,…linemnIn i-th of straightway, linemjFor linear edge set Ls={ ls1,ls2,…lspIn m-th of subset
lsm={ linem1,linem2,…linemnIn j-th of straightway, linemiSlope ami, intercept bmiWith linemjSlope
amj, intercept bmjApproximation meets | ami-amj|<tha,|bmi-bmj|<thb, tha、thbFor given threshold；
(4.2.2) connects line according to the similar priori knowledge of gray value at runway edgemiWith linemjStraight line
clinemijGray average gmijBe similar to the gray average g at runway edge, i.e., | gmij-g|<thg, thgFor given threshold；
(4.2.3) meets side greater than the other side, i.e. line due to the gray value of runway linear edge two side areasmiJust
Direction vector dpThe gray average in=[1,1] region is gpi, negative vector direction dnThe gray average in=[- 1, -1] region is
gniIf meeting gpi>gni, then straight line line to be connectedmjIt is equal greater than negative vector direction gray scale to meet positive vector direction gray average
Value, that is, meet gpj>gnj, i.e. (gpi>gni)∧(gpj>gnj)=1, or (gpi<gni)∧(gpj<gnj)=1.
The present invention proposes the airfield runway detection method based on edge feature, using runway gray feature rule to straightway
It is attached, makes full use of the gray scale priori knowledge and edge line set feature of runway, avoid as caused by influence of noise
The appearance and position inaccurate problem of ideal lines.
Specific attached drawing is referred again to below, and technical solutions and effects of the present invention is described in further detail.
Embodiment 5
SAR image airport target detection method based on edge feature with embodiment 1-4,
Referring to Fig.1, steps are as follows for realization of the invention:
Step 1, SAR image I to be detected is inputted, as shown in Figure 2.
Step 2, average ratio edge detection is carried out to airport target SAR image I to be detected.
Average ratio edge is carried out to image I (x, y) using the sliding window w that size is N × N (N=2 × n-1, n=1,2 ...)
It detects (ratio of average, ROA).Sliding window w and image I is subjected to convolution, i.e. w*I.Sliding window w points are 8 directions, i.e. An
This 8 directions of={ 0 °, 45 °, 90 °, 135 °, 180 °, 225 °, 270 °, 315 ° }, by taking 13 × 13 size windows as an example, such as Fig. 3 institute
Show, calculate sum of the grayscale values s1, s2 of sliding window all directions two sides, referring in Fig. 3, by taking 0 ° of direction as an example, s1 indicates window in sliding window
In the 1st row to the 6th row the sum of pixel gray value, s2 indicate window in eighth row to the 13rd row the sum of pixel gray value,
Central row i.e. the 7th row is not involved in calculating, and so on,
When An=0 ° or 180 °,
When An=90 ° or 270 °,
When An=45 ° or 225 °,
When An=135 ° or 315 °,
Then, the ratio r atio of the sum of sliding window all directions two sides gray value is calculated, it may be assumed that
The smallest direction of ratio value in 8 directions is obtained, if An=0 °, 270 °, 45 °, 135 °, then the mean value g of s1 is calculated,
That is:
G=s1/ (N* (N-1)/2),
If An=180 °, 90 °, 225 °, 315 °, then calculating the mean value g of s2, it may be assumed that
G=s2/ (N* (N-1)/2),
According to SAR image imaging characteristics it is found that the gray value of airfield runway is darker compared near runway and its gray value g
[g1, g2] in a certain range, therefore under conditions of g ∈ [g1, g2], appropriate threshold threshold is set, if in 8 directions
The smallest ratio value is less than threshold, then is marked, obtains to the current coordinate point of edge testing result figure B (x, y)
Result figure B, as shown in figure 4, the edge detection results of runway are clear, continuous in Fig. 4, and noise jamming very little, with runway periphery
The image in region can be differentiated clearly.
Step 3, the runway edge extracting based on edge feature cluster screening.
(3.1) edge detection results figure B, edge detection results figure B such as Fig. 4 institute are tracked using 8 direction Freeman chain codes
Show, obtains edge aggregation L={ l1,l2,l3,…ln, wherein lnIndicate n-th of pixel subset being made of edge pixel point.
(3.2) edge aggregation L={ l is extracted1,l2,l3,…lnIn each subset feature, using edge pixel probability of occurrence
Maximum gray scale value set G, the main gradient direction set T of edge pixel, both sides of edges gray average set M, both sides of edges
These features of gray variance set S, the feature vector set H of edge pixel two sides gray level co-occurrence matrixes assign edge with ash
Degree, gradient, texture multiple characteristics.
(3.2.1) calculates the maximum gray scale value set G of edge pixel probability of occurrence
The gray value of [0,255] is divided into K1 grades, counts each subset lnThe middle highest gray level of probability of occurrence, subset lnIn
The Probability p that the gray value of all pixels occurs in each gray leveln(i):
Wherein, Nn(i) subset l is indicatednIn the gray value of pixel fall into the number of i-th of gray level, SumnIndicate son
Collect lnIn sum of all pixels.By the intermediate value g of the maximum gray level of probability of occurrencenAs subset lnEdge pixel probability of occurrence most
Big gray value, by each subset lnCorresponding gnConstitute set, the i.e. maximum gray scale value set G of feature edge pixels probability of occurrence
={ g1,g2,…gn}。
(3.2.2) calculates the maximum gradient direction set T of edge pixel probability of occurrence
It is divided to approximate two array I to calculate x Yu y partial derivative using 2 × 2 first differencesx' (x, y) and Iy' (x, y):
Ix'(x,y)≈Gdx=[I (x+1, y)-I (x, y)+I (x+1, y+1)-I (x, y+1)]/2
Iy'(x,y)≈Gdy=[I (x, y+1)-I (x, y)+I (x+1, y+1)-I (x+1, y)]/2
θ (x, y) reflects the gradient direction of edge pixel:
θ (x, y)=arctan (Gdx(x,y)/Gdy(x,y))
Calculate subset lnThe gradient direction of upper each pixel, angle [0 °, 180 °] is classified, and is divided into K2 grades, system
Count subset lnThe probability q that the gradient direction of upper all pixels occurs in each angle graden(i):
Wherein, Dn(i) subset l is indicatednIn the gradient direction of pixel fall into the quantity of i-th of angle grade, SumnIt indicates
Subset lnIn sum of all pixels.Using the intermediate value of the maximum angle grade of probability of occurrence as subset lnEdge pixel probability of occurrence most
Big gradient direction tn, by each subset lnCorresponding tnConstitute set, the i.e. maximum gradient direction set of edge pixel probability of occurrence
T={ t1,t2,…tn}。
(3.2.3) calculates edge lines two sides gray average set M and edge lines two sides gray variance set S
It calculates and by edge pixel set L={ l1,l2,l3,…lnIn n-th of subset lnThe edge line of middle pixel composition
Item is parallel and the gray average m of the identical two side areas of distance, sizen={ mn1,mn2, wherein mn1Indicate positive vector direction dp
The gray average in=[1,1] region, mn2Indicate negative vector direction dnThe gray average in=[- 1, -1] region, obtains edge lines
Two sides gray average set M={ m1,m2,…,mn}；
It calculates and by edge pixel set L={ l1,l2,l3,…lnIn n-th of subset lnThe edge line of middle pixel composition
Item is parallel and the gray variance s of the identical two side areas of distance, sizen={ sn1,sn2, wherein sn1Indicate positive vector direction dp
The gray variance in=[1,1] region, sn2Indicate negative vector direction dnThe gray variance in=[- 1, -1] region, obtains edge lines
Two sides gray variance set S={ s1,s2,…,sn}。
(3.2.4) calculates the feature vector set H of the gray level co-occurrence matrixes in both sides of edges region
Gray level co-occurrence matrixes algorithm (Grey Level Co-occurrence Matrix, GLCM) is based on certain in image
The probability that one gray scale level structure repeats is come the method that describes image texture information.Usually ash can be characterized with some scalars
The feature for spending co-occurrence matrix enables Gr indicate that the gray level co-occurrence matrixes of SAR image I to be detected, h indicate that gray level co-occurrence matrixes are common
The member of feature vector, h is known as:
The energy of (3.2.4.1) calculating gray level co-occurrence matrixes
The quadratic sum of i.e. each matrix element.Energy is the quadratic sum of gray level co-occurrence matrixes element value, so also referred to as energy,
Image grayscale is reflected to be evenly distributed degree and texture fineness degree.If all values of co-occurrence matrix are equal, ASM value is small；
On the contrary, ASM value is big if the big and other value of some of values is small,
The contrast of (3.2.4.1) calculating gray level co-occurrence matrixes
Directly reflect the comparative situation of the brightness of some pixel value and its field pixel value.If deviateing cornerwise member
It is known as the larger value, i.e., quickly, then CON has larger value, this also complies with the definition of contrast for image brightness values variation.Reflection
The clarity of image and the degree of the texture rill depth.Texture rill is deeper, and contrast is bigger, and visual effect is more clear；
Conversely, contrast is small, then rill is shallow, and effect is fuzzy,
The inverse difference moment of (3.2.4.3) calculating gray level co-occurrence matrixes
Reflect image texture homogeney, measurement image texture localized variation number.Its value then illustrates greatly image texture
Different zones between lack variation, part is highly uniform,
The entropy of (3.2.4.4) calculating gray level co-occurrence matrixes
Entropy is the randomness metrics that image includes information content, when all values are equal in co-occurrence matrix or pixel value shows
Out when maximum randomness, entropy is maximum；Therefore entropy shows the complexity of image grayscale distribution, and entropy is bigger, and image is got over
It is complicated.
The auto-correlation of (3.2.4.5) calculating gray level co-occurrence matrixes
Wherein,
Auto-correlation has reacted the consistency of image texture.If there is horizontal direction texture in image, horizontal direction matrix
COR be greater than its complementary submatrix COR value.Its metric space gray level co-occurrence matrixes element be expert at or column direction on similarity degree,
Therefore, correlation size reflects local gray level correlation in image.When matrix element value homogeneous phase etc., correlation is with regard to big；
On the contrary, correlation is small if matrix pixel value differs greatly,
(3.2.4.2) is by edge pixel set L={ l1,l2,l3,…lnIn n-th of subset lnMiddle pixel composition
The two sides of edge lines respectively take size it is identical, apart from edge lines apart from identical one piece of region, calculate separately the energy in two regions
Measure ASMn={ ASMn1,ASMn2, wherein ASMn1Indicate positive vector direction dpThe energy value in=[1,1] region, ASMn2Indicate negative
Direction vector dnThe energy value in=[- 1, -1] region；Calculate the contrast C ON in two regionsn={ CONn1,CONn2, wherein CONn1
Indicate positive vector direction dpThe contrast value in=[1,1] region, CONn2Indicate negative vector direction dnThe comparison in=[- 1, -1] region
Angle value；Calculate the inverse difference moment IDM in two regionsn={ IDMn1,IDMn2, wherein IDMn1Indicate positive vector direction dp=[1,1] region
Inverse gap value, IDMn2Indicate negative vector direction dnThe inverse gap value in=[- 1, -1] region；Calculate the entropy ENT in two regionsn=
{ENTn1,ENTn2, wherein ENTn1Indicate positive vector direction dpThe entropy in=[1,1] region, ENTn2Indicate negative vector direction dn
The entropy in=[- 1, -1] region；Calculate the auto-correlation COR in two regionsn={ CORn1,CORn2, wherein CORn1Indicate positive vector side
To dpThe autocorrelation value in=[1,1] region, CORn2Indicate negative vector direction dnThe autocorrelation value in=[- 1, -1] region, obtains son
Collect lnThe feature vector h of the gray level co-occurrence matrixes in both sides of edges regionn={ ASMn,CONn,IDMn,ENTn,CORn, then edge collection
Close L={ l1,l2,l3,…lnEdge pixel two sides gray level co-occurrence matrixes feature vector set H={ h1,h2,…,hn,
Features described above is comprehensive, obtain edge feature set Feature={ G, T, M, S, H }.
(3.3) by edge aggregation L={ l1,l2,l3,…lnAccording to edge feature set Feature={ G, T, M, S, H }
It is clustered, the present invention uses K-means clustering algorithm, obtains edge cluster set LF={ lf1,lf2,lf3,…lfk, k is
Classification number, lfkFor kth class edge subset.
(3.4) by priori knowledge it is found that the gray value of runway is lower than surrounding gray value, the texture of runway is more compared with peripheral region
Adding uniform rule, random noise is less, then CON, ENT of runway zone smaller compared with runway peripheral region, runway zone
ASM, IDM, COR value are bigger compared with peripheral region, cluster set LF={ lf for edge1,lf2,lf3,…lfkIn k-th son
Collect lfkIf subset lfkMeet the track features in priori knowledge, i.e., when the edge feature set Feature=of the subset G, T,
M, S, H } in both sides of edges gray average vector mk={ mk1,mk2Meet mk1>mk2When, then the feature vector of gray level co-occurrence matrixes
hk={ ASMk,CONk,IDMk,ENTk,CORkIn element should meet ASMk1<ASMk2、CONk1>CONk2、IDMk1<IDMk2、
ENTk1>ENTk2、CORk1<CORk2；When edge lines two sides in the edge feature set Feature={ G, T, M, S, H } of the subset
Gray average vector mk={ mk1,mk2Meet mk1<mk2When, then the feature vector h of gray level co-occurrence matrixesk={ ASMk,CONk,
IDMk,ENTk,CORkElement in vector should meet ASMk1>ASMk2、CONk1<CONk2、IDMk1>IDMk2、ENTk1<ENTk2、
CORk1>CORk2, other situations are then unsatisfactory for airfield runway feature, such as mk1=mk2When, runway edge can not then be differentiated, and retain
Edge clusters set LF={ lf1,lf2,lf3,…lfkIn meet the subsets of track features, reject the son for not meeting track features
Collection, obtains the edge cluster result set Lr={ lf for meeting track features1,lf2,…lfp, therefore calculation amount of the present invention reduces,
As shown in Figure 5, edge cluster result set Lr={ lf1,lf2,…lfpIn belong to the edge same color table of same subset
Show, it can be seen that runway edge cluster is accurate, and the edge pixel with same characteristic features property is clustered same class, and in figure
Only it is left to meet the edge pixel point of track features after cluster screening.
Step 4, candidate marginal is fitted, connected.
(4.1) according to least square method to edge cluster result set Lr={ lf1,lf2,…lfpIn each subset edge
Pixel carries out the linear edge set Ls={ ls after straight line fitting is fitted1,ls2,…lst}。
(4.2) to set Ls={ ls1,ls2,…lstIn straightway in each subset advised according to runway gray feature
Then it is attached, every two straightway is attached if meeting runway gray feature rule in subset, otherwise without connecting,
Wherein k-th of subset is lsk={ linek1,linek2,…linekm, k=1,2 ..., t, wherein linekiFor subset lsk=
{linek1,linek2,…linekmIn i-th of straightway, linekjFor subset lsk={ linek1,linek2,…linekmIn
J-th of straightway, the linear edge set Str={ s after being connected1,s2,…,st, runway gray feature rule of the invention
It is then as follows:
(4.2.1)linekiSlope aki, intercept bkiWith linekjSlope akj, intercept bkjApproximation meets | aki-akj
|<tha,|bki-bkj|<thb；
(4.2.2) connects line according to the similar priori knowledge of gray value at runway edgekiWith linekjStraight line
clinekijGray average gmijBe similar to the gray average g at runway edge, i.e., | gmij-g|<thg, thgFor given threshold；
(4.2.3) is necessarily greater than the other side, i.e. line since the gray value of runway linear edge two sides meets sidekiJust
Direction vector dpThe gray average in=[1,1] region is gpi, negative vector direction dnThe gray average in=[- 1, -1] region is
gniIf meeting gpi>gni, then straight line line to be connectedkjIt is equal greater than negative vector direction gray scale to meet positive vector direction gray average
Value, that is, meet gpj>gnj, i.e. (gpi>gni)∧(gpj>gnj)=1, or (gpi<gni)∧(gpj<gnj)=1；
Step 5, the detection to airport target is completed by parallel lines detection technique.
(5.1) feature parallel according to runway edge line extracts edge line set Str={ s1,s2,…,stIn institute
There are the parallel lines pair of straight line, wherein t-th of subset st={ sline1,sline2,…slinen, slinenIndicate subset st=
{sline1,sline2,…slinenIn n-th of straightway, slineiFor Str={ s1,s2,…,stIn it is straight in arbitrary collection
Line segment, equally, slinejFor Str={ s1,s2,…,stIn in arbitrary collection straightway, extracting rule it is as follows:
(5.1.1) straight line slineiSlope ai, intercept biWith slinejSlope aj, intercept bjMeet | ai-aj|<tha1,
|bi-bj|∈[thb1,thb2], wherein section [thb1,thb2] be the corresponding number of pixels of runway width range；
(5.1.2)slineiPositive vector direction dpThe gray average of=[1,1] is gpi, negative vector direction dn=[- 1 ,-
1] gray average is gni, similarly, straight line slinejGray average in positive negative direction is respectively gpj、gnj, then slineiWith to
Match straight line slinejPositive negative direction on gray value relationship meet (gpi>gni)∧(gpj<gnj)=1, or (gpi<gni)∧
(gpj>gnj)=1；
(5.1.3) aperture closes the straight line pair of above-mentioned condition, obtains final parallel lines to set Par={ P1,P2,…
Pr, subset Pi(i=1,2 ... r) is i-th of parallel lines to subset.
For airfield runway testing result figure as shown in fig. 6, two runway edges are accurately detected, accurate positioning, no void examines mesh
Mark.
The present invention can be seen by completing the detection to runway so as to complete the detection to airport target by Fig. 2, Fig. 6
Out, by assigning gray scale, gradient, texture multiple characteristics to edge, so that priori knowledge is fully utilized, edge is clustered and is screened
Process improves the positioning accuracy to target, reduces calculation amount, is connected using runway gray feature rule to edge straightway
It connects, avoids the generation of the falseness straight line as caused by influence of noise, reduce empty inspection.
Effect of the present invention can be verified by following emulation experiment:
Embodiment 6
SAR image airport target detection method based on edge feature with embodiment 1-5,
1. experiment condition and method
Hardware platform are as follows: processor is Intel (R) Core (TM) i5-4200U CPU 1.60GHz, inside saves as 4.0G, firmly
Disk 500G, operating system are Microsoft windows sever 2007；
Software platform: MATLAB2011b；
Experimental method: the method for the present invention.
2. emulation content and result
Under these experimental conditions, the 10 different SAR pictures of width size content are chosen to be tested, is chosen here wherein
4 width are shown.Fig. 7 (a) is the typical triangle airfield runway image of a width, and Fig. 8 (a) is the typical right angle airfield runway of a width
Image, Fig. 9 (a) are a width typically horizontal airfield runway images, and Figure 10 (a) is a width typically vertical airfield runway image.
Here, detection accuracy=target correctly detects number/real goal number, and empty inspection rate=false target detects number/(mesh
The correct detection number+target error of mark detects number).By Fig. 7 (b), Fig. 8 (b), Fig. 9 (b), Figure 10 (b) testing result figure
See, verification and measurement ratio reaches 100%, and empty inspection rate is 0%, reaches expected detection effect.
Table 1 illustrates the airport target test result for the 10 width images that experiment is chosen, it is seen that and verification and measurement ratio reaches 100%,
Empty inspection rate is 0%, illustrates the robustness and validity of inventive algorithm.
Table 1SAR image airport target test result
In conclusion the invention discloses a kind of SAR image airport target detection method based on edge feature, main to solve
It is not smart to occur empty inspection, missing inspection and airfield runway positioning when certainly airport target detects when brightness of image unevenness and when noise is more
Quasi- technical problem.Its realization process is: inputting airport target SAR image to be detected；Use average ratio edge detector
(ratio of average, ROA) obtains edge image；Feature extraction is carried out to edge using method proposed by the present invention and is gathered
Class screening；Candidate marginal is attached using method proposed by the present invention；It is completed by parallel lines detection technique to airport
The detection of airport target is realized in the detection of runway.The present invention can be accurately finished the detection of SAR image airport target, runway
It is more accurate to position, and calculation amount is small, reduces empty inspection, reduces runing time, can be used for spaceborne and on-board SAR image processing system with
And in related objective detection system.
Claims (2)
1. a kind of SAR image airport target detection method based on edge feature, which is characterized in that comprise the following steps that
(1) airport target SAR image I to be detected is inputted；
(2) edge detection is carried out to image I using average ratio detector: average ratio edge detection, airfield runway is carried out to image I
Gray average g it is smaller compared near zone and in a certain range, i.e. g ∈ [g1, g2], g1 are runway minimum gray value, and g2 is
Runway gray scale maximum value retains gray average gminThe marginal point of ∈ [g1, g2], obtains edge detection results image；
(3) the runway marginal point based on Edge Gradient Feature and cluster screening extracts, and comprises the following steps that
(3.1) edge detection results figure is tracked using Freeman chain code, obtains edge aggregation L={ l1,l2,l3,K ln, wherein
lnIndicate n-th of pixel collection being made of edge pixel point, i.e. edge subset；
(3.2) edge aggregation L={ l is extracted1,l2,l3,K lnIn each subset feature, it is maximum to extract edge pixel probability of occurrence
Gray scale value set G, edge pixel the maximum gradient direction set T of probability of occurrence, both sides of edges gray average set M, side
The feature vector set H of edge two sides gray variance set S and edge pixel two sides gray level co-occurrence matrixes, obtain edge feature set
Feature={ G, T, M, S, H }, the maximum gray value of edge probability of occurrence refer to the maximum gray level of probability of occurrence
Intermediate value as the maximum gray value of edge probability of occurrence, the maximum gradient value of edge probability of occurrence refer to will occur it is general
The intermediate value of the maximum angle grade of rate is as the maximum gradient value of edge probability of occurrence；
(3.3) by edge aggregation L={ l1,l2,l3,K lnGathered according to edge feature set Feature={ G, T, M, S, H }
Class obtains edge cluster set LF={ lf1,lf2,lf3,K lfk, k is cluster sum, lfkSon is clustered for the edge of kth class
Collection；
(3.4) all subsets closed to edge cluster set are screened, if edge clusters set LF={ lf1,lf2,lf3,K lfk}
In k-th of subset lfkMeet the track features in airfield runway priori knowledge, i.e., as subset lfkIn edge feature set
Both sides of edges gray average vector m in corresponding k-th of character subset in Feature={ G, T, M, S, H }k={ mk1,mk2Full
Sufficient mk1> mk2When, then the feature vector h of gray level co-occurrence matrixesk={ ASMk,CONk,IDMk,ENTk,CORkIn element,
In, ASMk={ ASMk1,ASMk2},CONk={ CONk1,CONk2},IDMk={ IDMk1,IDMk2},ENTk={ ENTk1,ENTk2},
CORk={ CORk1,CORk2, meet ASMk1< ASMk2、CONk1> CONk2、IDMk1< IDMk2、ENTk1> ENTk2、CORk1<
CORk2；As subset lfkThe edge in corresponding k-th of character subset in edge feature set Feature={ G, T, M, S, H }
Two sides gray average vector mk={ mk1,mk2Meet mk1< mk2When, then the feature vector h of gray level co-occurrence matrixesk={ ASMk,
CONk,IDMk,ENTk,CORkElement in vector meets ASMk1> ASMk2、CONk1< CONk2、IDMk1> IDMk2、ENTk1<
ENTk2、CORk1> CORk2, other situations are then unsatisfactory for airfield runway feature, retain edge and cluster set LF={ lf1,lf2,
lf3,K lfkIn meet the subset of runway feature, reject the subset for not meeting runway feature, obtain the edge for meeting track features
Cluster result set Lr={ lf1,lf2,K lfp},p≤k；
(4) track features will be met and be under the jurisdiction of of a sort edge cluster result set Lr={ lf1,lf2,K lfpIn picture
Vegetarian refreshments is fitted, connects: being attached according to runway gray feature rule to straightway, meets runway spy after being connected
The linear edge set Str={ s of sign1,s2,K,sq, candidate edge pixel is fitted, is connected, includes following step
It is rapid:
(4.1) according to least square method to edge cluster result set Lr={ lf1,lf2,K lfpIn each subset edge pixel
Point carries out straight line fitting, the linear edge set Ls={ ls after being fitted1,ls2,K lsp}；
(4.2) to the linear edge set Ls={ ls after fitting1,ls2,K lspEach subset lsm={ linem1,linem2,K
linemn, m=1,2,3, K p, in all straightways, connected two-by-two according to runway gray feature rule, do not met race
Road gray feature rule is then without connection, the linear edge set Str={ s for meeting track features after being connected1,s2,
K,st}；
(5) detection to runway is completed by parallel lines detection technique: straight line is extracted according to the characteristics of runway sides aligned parallel straight line
Edge aggregation Str={ s1,s2,K,sqIn runway edge parallel lines pair, obtain final runway sides aligned parallel straight line to collection
Close Par={ P1,P2,K Pr},PrIndicate that r-th of parallel lines to subset, completes the airport of airport target SAR image I to be detected
The detection of target.
2. the SAR image airport target detection method according to claim 1 based on edge feature, which is characterized in that step
Suddenly the straightway connection of combination runway gray feature rule described in (4.2), connection procedure include the following steps:
(4.2.1) straightway linemiSlope ami, intercept bmiWith straightway linemjSlope amj, intercept bmjApproximation, i.e., it is full
Foot | ami-amj| < tha,|bmi-bmj| < thb, tha、thbFor given threshold；
(4.2.2) connects straightway linemiWith straightway linemjStraightway clinemijGray average gmijIt is similar to airport
The gray average g of runway, i.e., | gmij- g | < thg, thgFor given threshold；
(4.2.3) straightway linemiPositive vector direction dpThe gray average in=[1,1] region is gpi, negative vector direction dn
The gray average in=[- 1, -1] region is gniIf meeting gpi> gni, then straight line line to be connectedmjMeet positive vector direction gray scale
Mean value is greater than negative vector direction gray average, that is, meets gpj> gnj, i.e. (gpi> gni)∧(gpj> gnj)=1, or (gpi< gni)
∧(gpj< gnj)=1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610561336.4A CN106251332B (en) | 2016-07-17 | 2016-07-17 | SAR image airport target detection method based on edge feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610561336.4A CN106251332B (en) | 2016-07-17 | 2016-07-17 | SAR image airport target detection method based on edge feature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106251332A CN106251332A (en) | 2016-12-21 |
| CN106251332B true CN106251332B (en) | 2019-05-21 |
Family
ID=57613842
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610561336.4A Active CN106251332B (en) | 2016-07-17 | 2016-07-17 | SAR image airport target detection method based on edge feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106251332B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107564027B (en) * | 2017-10-25 | 2020-07-14 | 中国电子科技集团公司第五十四研究所 | Edge vector calculation method of synthetic aperture radar image |
| CN108510467B (en) * | 2018-03-28 | 2022-04-08 | 西安电子科技大学 | SAR image target identification method based on depth deformable convolution neural network |
| CN111141208B (en) * | 2019-01-09 | 2021-08-27 | 银河水滴科技（北京）有限公司 | Parallel line detection method and device |
| CN110210418B (en) * | 2019-06-05 | 2021-07-23 | 西安电子科技大学 | SAR image airplane target detection method based on information interaction and transfer learning |
| CN111862128B (en) * | 2020-06-12 | 2024-04-16 | 广州市申迪计算机系统有限公司 | Image segmentation method and device |
| CN111539429B (en) * | 2020-06-19 | 2020-11-03 | 天津施格机器人科技有限公司 | Automatic circulation box positioning method based on image geometric features |
| CN113313052B (en) * | 2021-06-15 | 2024-05-03 | 杭州萤石软件有限公司 | Cliff area detection and mobile robot control method and device and mobile robot |
| CN113828892B (en) * | 2021-10-29 | 2022-05-10 | 广东福维德焊接股份有限公司 | HDR image-based molten pool center identification system and weld joint tracking method |
| CN116758065B (en) * | 2023-08-14 | 2023-10-20 | 山东嘉利诺新材料科技有限公司 | Rapid detection method for surface defects of fireproof plate |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101944233A (en) * | 2010-09-25 | 2011-01-12 | 西北工业大学 | Method for quickly extracting airport target in high-resolution remote sensing image |
| CN102254162A (en) * | 2011-07-17 | 2011-11-23 | 西安电子科技大学 | Method for detecting airport runway in synthetic aperture radar (SAR) image based on minimum linear ratio |
| CN104156722A (en) * | 2014-08-14 | 2014-11-19 | 西北工业大学 | Airport target detection method based on high-resolution remote sensing image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2584683A1 (en) * | 2006-04-20 | 2007-10-20 | Optosecurity Inc. | Apparatus, method and system for screening receptacles and persons |
-
2016
- 2016-07-17 CN CN201610561336.4A patent/CN106251332B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101944233A (en) * | 2010-09-25 | 2011-01-12 | 西北工业大学 | Method for quickly extracting airport target in high-resolution remote sensing image |
| CN102254162A (en) * | 2011-07-17 | 2011-11-23 | 西安电子科技大学 | Method for detecting airport runway in synthetic aperture radar (SAR) image based on minimum linear ratio |
| CN104156722A (en) * | 2014-08-14 | 2014-11-19 | 西北工业大学 | Airport target detection method based on high-resolution remote sensing image |
Non-Patent Citations (4)
| Title |
|---|
| "基于复杂背景的红外机场目标自动识别算法研究";姚克明等;《红外与激光工程》;20070630;第36卷(第3期);第400页第3.2节 |
| "基于改进分水岭及区域合并的图像分割方法";杨海峰;《微计算机应用》;20071130;第28卷(第11期);第1134-1135页第3.2.1节 |
| "数字图像的分割技术";俞学刚等;《常州工学院学报》;20031031;第16卷(第4期);第50页第4.4.2节 |
| "高分辨率遥感图像中机场跑道快速检测方法";应龙等;《小型微型计算机系统》;20060228;第27卷(第2期);第283页第3节，图1 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106251332A (en) | 2016-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106251332B (en) | SAR image airport target detection method based on edge feature | |
| CN108038476B (en) | A kind of facial expression recognition feature extracting method based on edge detection and SIFT | |
| CN108510467B (en) | SAR image target identification method based on depth deformable convolution neural network | |
| CN106778595B (en) | Method for detecting abnormal behaviors in crowd based on Gaussian mixture model | |
| CN105518709B (en) | The method, system and computer program product of face for identification | |
| CN103035013B (en) | A kind of precise motion shadow detection method based on multi-feature fusion | |
| CN104063706B (en) | Video fingerprint extraction method based on SURF algorithm | |
| CN107895139B (en) | SAR image target identification method based on multi-feature fusion | |
| CN102496016B (en) | Infrared target detection method based on space-time cooperation framework | |
| CN108960404B (en) | Image-based crowd counting method and device | |
| CN101930547A (en) | Method for automatically classifying remote sensing image based on object-oriented unsupervised classification | |
| CN104408482A (en) | Detecting method for high-resolution SAR (Synthetic Aperture Radar) image object | |
| CN106778680B (en) | A kind of hyperspectral image band selection method and device based on critical bands extraction | |
| CN110414538A (en) | Defect classification method, defect classification based training method and device thereof | |
| CN109726649B (en) | Remote sensing image cloud detection method and system and electronic equipment | |
| CN102542293A (en) | Class-I extraction and classification method aiming at high-resolution SAR (Synthetic Aperture Radar) image scene interpretation | |
| CN101364263A (en) | Method and system for detecting skin texture to image | |
| CN102928822A (en) | Radar target length calculation method based on high-resolution one-dimensional range profiles | |
| CN103093243B (en) | The panchromatic remote sensing image clouds of high-resolution sentences method | |
| Pamplona Segundo et al. | Pore-based ridge reconstruction for fingerprint recognition | |
| Zaharieva et al. | Image based recognition of ancient coins | |
| CN102592280A (en) | Hyperspectral image anomaly detection method using multi-window feature analysis | |
| CN103366373B (en) | Multi-time-phase remote-sensing image change detection method based on fuzzy compatible chart | |
| CN110334760A (en) | A kind of optical component damage detecting method and system based on resUnet | |
| CN104537384B (en) | A kind of SAR target discrimination methods of combination likelihood ratio judgement |
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 | ||
| OL01 | Intention to license declared |