CN106022217A - Civil airport runway area detection method free from supervision multistage classification - Google Patents
Civil airport runway area detection method free from supervision multistage classification Download PDFInfo
- Publication number
- CN106022217A CN106022217A CN201610298469.7A CN201610298469A CN106022217A CN 106022217 A CN106022217 A CN 106022217A CN 201610298469 A CN201610298469 A CN 201610298469A CN 106022217 A CN106022217 A CN 106022217A
- Authority
- CN
- China
- Prior art keywords
- pixel
- runway
- class
- scattering
- entropy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a civil airport runway area detection method free from supervision multistage classification. The method comprises the following steps: polarized SAR images are classified through combining object scattering randomness with a Freeman decomposition method; classification results are classified by use of polarization difference degrees between two types of objects, and class labels of each pixel is updated; suspected runway areas are extracted from the classification results and a binary graph is obtained; isolated small areas are removed by performing morphological processing on ROIs in the binary graph; results are classified by use of a Wishart classifier, and an internal structure of an airport runway area is distinguished from a surrounding natural object; suspected runway areas are extracted from classification results, and the morphological processing is performed on extracted ROIs again; and the suspected runway areas in the binary graph are identified, and finally the airport runway area is determined. According to the invention, the civic airport runway area in the polarized SAR images in a complex scene can be effectively detected, used prior information is small, the false alarm rate is quite low, and the promotion is high.
Description
Technical field
The invention belongs to Polarimetric SAR Image target detection technique field, particularly relate to a kind of nothing supervision multi-level images and divide
The civil airport runway zone detection method of class.
Technical background
Runway as the core facility of traffic pattern, it is detected automatically all have on dual-use the heaviest
The using value wanted, thus paid close attention to by the most widely.Relative to optical remote sensing, synthetic aperture radar (SAR) is because having
Round-the-clock, round-the-clock feature and be widely used in military target identification and location, battlefield surroundings monitoring, the condition of a disaster detection etc. and lead
Territory.Compared to single polarization SAR, multipolarization SAR data contains more target polarization information, is therefore applied to civil airport
Runway zone detection tool is of great significance.
At present, civil airport runway detection technique study based on SAR image is primarily directed to single polarization SAR image, and
Mainly use based on edge extracting and the method for straight-line detection.This method is primarily present problems with: 1. mainly for SAR merit
Rate image, utilizes the strength information of target to carry out feature extraction and target detection, does not utilize the more polarization information of target, because of
This correct verification and measurement ratio is affected；2. there is strong coherent speckle noise in SAR image, thus increases the difficulty of edge extracting；
3. in the SAR image of scene complexity, the method using single linear feature to carry out runway detection, it is easily subject to road, bridge
The interference of the similar track object such as beam, waters, robustness is poor.Thus this kind of method applies in general to little scene and picture quality
Preferably in airfield runway detection.And polarization SAR data can preferably characterize the scattering properties of atural object, currently for polarization
The civil airport runway zone detection research of SAR image has begun to come into one's own, and Xu Jiansa et al. proposes target classification
The detection method combined with runway mechanism characteristics, achieves good Detection results.But, this method has used prison
The method superintending and directing classification, and employ too much prior information, therefore there is certain limitation.
Summary of the invention
In order to solve the problems referred to above, it is an object of the invention to provide a kind of civil airport runway without supervision multiclass classification
Method for detecting area.
In order to achieve the above object, the civil airport runway zone detection method without supervision multiclass classification that the present invention provides
The following step including carrying out in order:
1) target scattering randomness and Freeman decomposition method are combined Polarimetric SAR Image is classified, calculate every
The average coherence matrix of one class pixel as initial classes center；
2) utilize the polarization difference degree between two class targets that above-mentioned classification results carries out second time to classify, and update each picture
The category label of vegetarian refreshments；
3) utilize the weak scattering echoing characteristics of airfield runway, from step 2) classification results extract doubtful runway zone,
Obtain comprising the binary map of doubtful runway zone；
4) according to dimensional standard and the image resolution ratio of civil airport runway, runway zone doubtful in above-mentioned binary map is entered
Row Morphological scale-space is to remove isolated zonule；
5) application Wishart grader to step 4) process after result carry out third time classify, by airfield runway region
Internal structure distinguish with surrounding atural object；
6) utilize the weak scattering echoing characteristics of airfield runway, again from above-mentioned 3rd class classification results, extract doubtful runway
Region, and the area-of-interest extracted is carried out again Morphological scale-space, obtain comprising the binary map of doubtful runway zone；
7) utilize civil airport track elements feature, runway zone doubtful in above-mentioned binary map is carried out identification, the most really
Determine airfield runway region.
In step 1) in, described combines target scattering randomness and Freeman decomposition method to Polarimetric SAR Image
Classify, calculate the average coherence matrix of each class pixel and the method as initial classes center is: first with pixel
Coherence matrix calculate Polarization scattering entropy H, according to the size of Polarization scattering entropy H the pixel in image is divided into low entropy class, in
Entropy class and high entropy class, the Polarization scattering entropy H < 0.5 of the lowest entropy class, the Polarization scattering entropy of middle entropy class is 0.5 < H < 0.9, high entropy class
Polarization scattering entropy H > 0.9；The most again view picture Polarimetric SAR Image is carried out Freeman decomposition, try to achieve the surface of each pixel
Scattered power Ps, even scattered power PdAnd volume scattering power Pv；According to surface scattering power Ps, even scattered power Pd, body
Scattered power PvThe size of these three parameter, is divided three classes low entropy class: low entropy surface scattering (Ps>Pd,Pv), low entropy even dissipate
Penetrate (Pd>Ps,Pv) and low entropy volume scattering (Pv>Ps,Pd)；Middle entropy class is subdivided into six classes: middle entropy surface-even scatters (Ps>Pd>
Pv), middle entropy surface-volume scattering (Ps>Pv>Pd), middle entropy even-surface scattering (Pd>Ps>Pv), middle entropy even-volume scattering (Pd>Pv>
Ps), middle entropy body-surface scattering (Pv>Ps>Pd) and middle entropy body-even scattering (Pv>Pd>Ps)；High entropy class is not segmented, and all belongs to
The pixel of high entropy is a class；There are 10 classes, respectively with label 1,2 ... the class that 10 labellings are corresponding；Finally according to category label
Calculate the average coherence matrix of every class pixel and as its class center.
In step 2) in, described utilize the polarization difference degree between two class targets that classification results carries out second time to classify,
And the method updating the category label of each pixel is: calculate each pixel and step 1 in Polarimetric SAR Image) in 10
Polarization difference degree between class center, is divided into each pixel the apoplexy due to endogenous wind minimum with its diversity factor, and updates the class of pixel
Alias；After the classification of all pixels adjusts, recalculate the average coherence matrix of every class pixel and as next
Subseries Shi Lei center；Repeat this process, until the pixel number that in Polarimetric SAR Image, classification changes is less than given
Threshold value or reach iterations.
In step 3) in, the described weak scattering echoing characteristics utilizing airfield runway, from step 2) classification results carry
Taking doubtful runway zone, the method for the binary map obtaining comprising doubtful runway zone is: calculation procedure 2) classification results for the second time
In all kinds of average power content；This feature of weak scattering echoing characteristics according to civil airport runway, extracts second time classification knot
The class pixel that in Guo, mean power is minimum, the region constituted by these pixels is as doubtful runway zone；By view picture pole
Changing the pixel of doubtful runway zone in SAR image to be labeled as " 1 ", the pixel of non-runway zone is labeled as " 0 ", obtains a width
Binary map.
In step 4) in, the described dimensional standard according to civil airport runway and image resolution ratio, to above-mentioned binary map
In doubtful runway zone carry out Morphological scale-space to remove the method for isolated zonule and be: first according to civil airport runway
Size characteristic and the resolution of image, calculate the area threshold Th0 of runway zone in Polarimetric SAR Image；Binary map is carried out
After connected region labelling processes, calculate the pixel number in each connected region；When in certain region, pixel number is less than face
During long-pending threshold value Th0, pixel in this region is all labeled as " 0 ", thus removes zonule isolated in binary map.
In step 5) in, described application Wishart grader is to step 4) process after result carry out third time classify,
By the method that the internal structure in airfield runway region and surrounding atural object distinguish it is: by step 2) the class center that obtains and step
4) in the doubtful runway zone after processing, non-" 0 " pixel is as the initial input of Wishart grader；Calculate non-" 0 " pixel
Pixel, to the Wishart distance at each class center, is categorized into from its closest apoplexy due to endogenous wind according to minimum distance criterion by point,
And update the category label of pixel；According to new classification results, recalculate all kinds of average coherence matrixes and as next
Secondary iterative process Zhong Lei center；Repeat Wishart Iterative classification process until the pixel that in image, classification changes is individual
Number is less than given threshold value or reaches iterations.
In step 6) in, the described weak scattering echoing characteristics utilizing airfield runway, again from above-mentioned third time classification knot
Extract doubtful runway zone in Guo, and the area-of-interest extracted is carried out again Morphological scale-space, obtain comprising doubtful race
The method of the binary map in region, road is: according to this feature of weak scattering echoing characteristics of civil airport runway, extraction step 5) point
The class that in class result, mean power is minimum is as doubtful runway zone；By runway zone doubtful in view picture Polarimetric SAR Image
Pixel is labeled as " 1 ", and the pixel of non-runway zone is labeled as " 0 ", obtains a width binary map；Binary map is carried out connected region
After labelling processes, calculate the pixel number in each connected region；If in certain connected region, pixel number is less than step 4)
The area threshold Th0 tried to achieve, then be all labeled as pixel in this region " 0 ", thus removes community isolated in binary map
Territory.
In step 7) in, described utilizes civil airport track elements feature, to runway zone doubtful in above-mentioned binary map
Carry out identification, finally determine that the method in airfield runway region is: utilize the architectural feature of runway, i.e. topological characteristic, parallel lines spy
Levy, to step 6) the doubtful runway zone of gained screens further, finally determines civil airport runway zone.
The present invention has the advantage that compared with existing detection method (1) present invention utilizes without supervision multi-level images classification
Method carries out ROI (Region of Interest, area-of-interest) and extracts, it is not necessary to use training sample and power threshold,
Applied range and be applicable to Classification of Polarimetric SAR Image and other target detection, therefore this method has more generalization.(2) profit
With the polarization difference degree between target, eliminate the pixel of a large amount of nontarget areas in image, greatly reduce Wishart and divide
The operand of class device.(3) feature for runway identification is less.(4) it is capable of detecting when under complex scene in Polarimetric SAR Image
Civil airport runway zone, and structural integrity, false alarm rate is low.
Accompanying drawing explanation
The civil airport runway zone detection method flow chart based on nothing supervision multiclass classification that Fig. 1 provides for the present invention.
Fig. 2 is Freeman decomposition process figure.
Fig. 3 is runway zone identification flow chart.
Fig. 4 (a) is the optical picture that airport, HALF MOON BAY is corresponding.
Fig. 4 (b) is the Pauli figure on airport, HALF MOON BAY.
Fig. 4 (c) is second time classification results based on target diversity factor iteration.
Fig. 4 (d) is the thick ROI extracted.
Fig. 4 (e) is ROI based on Wishart grader third time classification results.
Fig. 4 (f) is that ROI extracts result.
Fig. 4 (g) is the inventive method testing result.
Detailed description of the invention
The civil airport runway district without supervision multiclass classification with specific embodiment, the present invention provided below in conjunction with the accompanying drawings
Area detecting method is described in detail.
As it is shown in figure 1, the present invention provide without supervision multiclass classification civil airport runway zone detection method include by
The following step that order is carried out:
(1) target scattering randomness and Freeman decomposition method are combined Polarimetric SAR Image is classified, calculate
The average coherence matrix of each class pixel as initial classes center:
Ground object target has uncertainty to the scattering of radar illumination ripple, referred to as target scattering randomness.Generally use pole
Change scattering entropy metric objective scattering randomness.Cloude and Pottier definition Polarization scattering entropy H is:
In formula (1),λi(i=1,2,3) it is the ith feature value of coherence matrix T.
Under low scattering stochastic condition, the value of Polarization scattering entropy H is less, now there is certain typical speckle in target
Back scattering occupies main advantage；Under height scattering randomness situation, the value of Polarization scattering entropy H is relatively big, now target bag
The scattering contained is almost equal to backward scattered contribution；When scattering randomness is between above-mentioned two-value, multiple typical speckle
The biggest to backward scattered contribution.
Freeman decomposes based on the physical model of radar scattering echo, the reflection of different ground object contrast ejected waves is divided
Solve as surface scattering, even scattering and the weighted sum of three kinds of scattering mechanisms of volume scattering:
T=PsTsurface+PdTdouble+PvTvolume (2)
In formula (2), T is the coherence matrix of target；Tsurface、Tdouble、TvolumeBe respectively surface scattering, even scattering and
The normalization creep function of volume scattering；Ps、Pd、PvRepresent surface scattering, even scattering and the watt level of volume scattering respectively.It decomposes
Process refers to Fig. 2.
All kinds of class centers is represented with the average coherence matrix of every class pixel:
N in formula (3)iRepresent class ωiIn pixel number, < Tj> represent class ωiThe coherence matrix of middle jth pixel.
Coherence matrix first with pixel calculates Polarization scattering entropy H, according to the size of Polarization scattering entropy H by image
Pixel is divided into low entropy class, middle entropy class and high entropy class, the Polarization scattering entropy H < 0.5 of the lowest entropy class, the Polarization scattering of middle entropy class
Entropy is 0.5<H<0.9, the Polarization scattering entropy H>0.9 of high entropy class；The most again view picture Polarimetric SAR Image is carried out Freeman decomposition,
Try to achieve the surface scattering power P of each pixels, even scattered power PdAnd volume scattering power Pv；According to surface scattering power
Ps, even scattered power Pd, volume scattering power PvThe size of these three parameter, is divided three classes low entropy class: low entropy surface scattering
(Ps>Pd,Pv), low entropy even scattering (Pd>Ps,Pv) and low entropy volume scattering (Pv>Ps,Pd)；Middle entropy class is subdivided into six classes: middle entropy
Surface-even scattering (Ps>Pd>Pv), middle entropy surface-volume scattering (Ps>Pv>Pd), middle entropy even-surface scattering (Pd>Ps>Pv), in
Entropy even-volume scattering (Pd>Pv>Ps), middle entropy body-surface scattering (Pv>Ps>Pd) and middle entropy body-even scattering (Pv>Pd>Ps)；High
Entropy class is not segmented, and all pixels belonging to high entropy are classes；There are 10 classes, respectively with label 1,2 ... 10 labellings are corresponding
Class；The average coherence matrix of every class pixel is calculated and as its class center finally according to category label.
(2) utilize the polarization difference degree between two class targets that above-mentioned classification results carries out second time to classify, and update each
The category label of pixel, thus obtains second time classification results；
Based on K mean algorithm thought, Euclidean distance is utilized to measure the polarization difference degree between two targets.If two mesh
Target coherence matrix is respectively TiAnd Tj, then its polarization difference degree is:
dij=| | Ti-Tj|| (4)
Wherein | | | | it is 2 norms calculating matrix.dijValue the least, then target TiAnd TjBetween difference the least,
The two is the most similar.So, atural object classification based on target polarization difference degree adjustment criterion is: be provided with M class atural object, wherein m class
It is designated as ωmIf to all of class j (1≤j≤M and j ≠ m), having dj,T≥dm,T, then target and class ωmDiversity factor minimum, will
The coherence matrix T of target puts class ω undermIn.
It is as follows that atural object classification adjusts process:
1) using 10 Ge Lei centers in step (1) classification results as initial classes center.
2) each pixel and the polarization difference degree at 10 Ge Lei centers in Polarimetric SAR Image is calculated according to formula (4).According to ground
Species does not adjust criterion, each pixel is divided into the apoplexy due to endogenous wind minimum with its diversity factor, and updates the classification number of pixel.
3) according to new classification results, all kinds of average coherence matrix V is recalculated by formula (3)i(i=1,2 ... 10), makees
For the class center during next iteration.
4) check whether and meet iteration convergence condition.If being unsatisfactory for, repeat step 2), 3) until meet iteration convergence bar
Part.The condition of convergence is that the pixel number that in image, classification changes less than given threshold value or reaches iterations.
(3) utilize the weak scattering echoing characteristics of airfield runway, from step 2) classification results extract doubtful runway zone,
Obtain comprising the binary map of doubtful runway zone；
Average power contents all kinds of in calculation procedure (2) classification results for the second time；Weak scattering according to civil airport runway
This feature of echoing characteristics, extracts the class pixel that in second time classification results, mean power is minimum, by these pixel structures
The region become is as ROI；The pixel of runway zone doubtful in view picture Polarimetric SAR Image is labeled as " 1 ", non-runway zone
Pixel is labeled as " 0 ", obtains a width binary map.
(4) according to dimensional standard and the image resolution ratio of civil airport runway, to runway zone doubtful in above-mentioned binary map
Carry out Morphological scale-space to remove isolated zonule；
By civil airport runway dimensional standard (landing airdrome length scope: 900m-4200m；Runway width range: 30m-100m)
And image resolution ratio, the area threshold Th0 of minimum runway zone in Polarimetric SAR Image can be calculated.Area threshold Th0 is utilized to go
Shown in the judgment condition such as formula (5) of isolated zonule:
Num<Th0 (5)
The number of pixel during wherein Num is connected region.Detailed process is as follows:
1) binary map obtained in step (3) is carried out connected component labeling process, and calculate in each connected region
Pixel number Num comprised；
2) pixel number Num of each connected region is substituted into formula (5), meet formula (5) and then this connected region is owned
Pixel is set to " 0 ", does not the most make any process.
The further screening to doubtful runway zone can be realized by above-mentioned process.
(5) application Wishart grader to step 4) process after result carry out third time classify, by airfield runway region
Internal structure distinguish with surrounding atural object；
Coherence matrix obeys Wishart distribution W (n, T) that degree of freedom is n, the various classification being distributed based on multiple Wishart
Device is widely used in Classification of Polarimetric SAR Image.The Wishart grader that the present invention uses is a kind of based on Wishart distance
The sorting technique of measure.Distance metric factor d (T, the ω of Wishart graderm) calculate shown in process such as formula (6).
ω in formulamRepresent m class, VmIt is the average coherence matrix of m class, | | and the row of Tr () representing matrix respectively
Column and mark.
In this step, categorizing process based on Wishart grader is as follows:
1) in the ROI after the class center that obtains in step (2) being processed with step (4), non-" 0 " pixel is as Wishart
The initial input of grader.
2) by average coherence matrix Vi(i=1,2 ... 10) ViIt substitution formula (6) is not the pixel of " 0 " to labelling in binary map
Point is classified.Sorting criterion: calculate each pixel to after the Wishart distance at each class center, be grouped into distance
Little apoplexy due to endogenous wind, and update the classification number of pixel.
3) according to new classification results, recalculate all kinds of average coherence matrixes and as next iteration during
Class center.
4) checking whether and meet stopping criterion for iteration, if being unsatisfactory for, repeating step 2 by formula (3)), 3) until meeting iteration
The condition of convergence.The condition of convergence is that the pixel number that in image, classification changes less than given threshold value or reaches iteration time
Number.
6) utilize the weak scattering echoing characteristics of airfield runway, again from above-mentioned 3rd class classification results, extract doubtful runway
Region, and the area-of-interest extracted is carried out again Morphological scale-space, obtain comprising the binary map of doubtful runway zone；
This feature of weak scattering echoing characteristics according to civil airport runway, average merit in extraction step (5) classification results
One class of rate minimum is as ROI；The pixel of runway zone doubtful in view picture Polarimetric SAR Image is labeled as " 1 ", non-runway district
The pixel in territory is labeled as " 0 ", obtains a width binary map；After binary map is carried out connected region labelling process, calculate each connection
Number of pixels in region；If area threshold Th0, the Ze Jianggai district that in certain connected region, number of pixels is tried to achieve less than step (4)
In territory, pixel is all labeled as " 0 ", thus removes zonule isolated in binary map.
7) utilize civil airport track elements feature, runway zone doubtful in above-mentioned binary map is carried out identification, the most really
Determine airfield runway region.
In the binary map obtained in step (6) in addition to comprising civil airport runway zone, also have its similar with runway
Its atural object；Need to utilize the architectural feature of runway, i.e. topological characteristic, parallel lines feature that each connected region in binary map is entered one
Step screening, finally determines that Fig. 3 is seen in civil airport runway zone, identification flow chart.
1) size characteristic: civil airport runway zone is usually the stripe region of a similar rectangle, its length typically exists
Between 900～4200 meters, width is generally at 30～100 meters.
2) topological characteristic: general Euler's numbers characterize the topological characteristic of airfield runway.Euler's numbers are equal to connected component
Number and the difference of numbers of hole.From prior information, if runway neighboring area is contained within hayfield region.Connected component number is
The connected region number that in ROI binary map, white pixel is constituted, numbers of hole is the black picture element structure that white connected region is surrounded
The connected region number become.
3) parallel lines feature: straight line is the typical characteristic of runway.Doubtful ROI is carried out Hough transform ask for meeting length
After the long straight line required.The slope differences of any two long straight lines meets certain error, then be considered as parallel lines pair.Putting down of runway both sides
Row rectilineal interval should meet the requirement of runway width.
The present invention provide without supervision multiclass classification civil airport runway zone detection method effect can by with
Lower experimental result further illustrates.
Experimental data describes: these data are that the L-band that U.S.'s USVSAR system gathers in overhead, area, HALF MOON BAY of the U.S. is complete
Polarization data, regards through 4 and processes.Image size is 501 × 701, as shown in Fig. 4 (a), Fig. 4 (b).Except including first quarter moon in scene
Outside airport, gulf, also building, ocean, farmland, forest, pond, meadow, bare area etc..Image resolution ratio is 7.2m (distance to),
5m (orientation to).
Experiment parameter describes: in experiment, resolution according to image can calculate Morphological scale-space threshold value is Th0=450；
Euler's numbers E, the threshold value of parallel lines logarithm Dis are respectively 0,10.
Fig. 4 (c) classification results based on diversity factor iteration.Comparison diagram 4 (a), Fig. 4 (b) and Fig. 4 (c), we can be clear
The blue region in Fig. 4 (c) is seen on ground, and it contains the whole region of civil airport runway.Poor according to the polarization between target
Image is classified by different degree, can be made a distinction with non-runway zone civil airport runway zone well.The most non-race
Region, road accounts for the 71% of image size, after removing the pixel of these nontarget areas, can greatly reduce Wishart and divide
The operand of class device.
In Fig. 4 (d) Fig. 4 (c) that has been the weak scattering clawback feature extraction according to civil airport runway, mean power is minimum
The pixel of one class atural object, and obtain ROI through morphologic filtering.This ROI cannot embody the internal structure of runway, it is impossible to is used for
ROI identification.Comparison diagram 4 (d) and Fig. 4 (e) are it can be seen that Wishart grader can run the civil airport in the thick ROI of extraction
The internal structure in region, road distinguishes with other atural object.
Target area is extracted, as shown in Fig. 4 (f) also according to weak scattering echoing characteristics.In Fig. 4 (f), except runway district
Territory, also similar to runway polarization characteristic at 5 atural object.Feature further according to runway carries out identification to ROI, the final district retained
Territory is i.e. civil airport runway, as shown in Fig. 4 (g).Comparison diagram 4 (f) and Fig. 4 (g) are it can be seen that the structure chosen of the present invention is special
Levying and can correctly detect civil airport runway zone, without false-alarm, and the track elements detected is complete, and edge details keeps relatively
Good.
Claims (8)
1. the civil airport runway zone detection method without supervision multiclass classification, it is characterised in that described method includes
The following step carried out in order:
1) target scattering randomness and Freeman decomposition method are combined Polarimetric SAR Image is classified, calculate each class
The average coherence matrix of pixel as initial classes center；
2) utilize the polarization difference degree between two class targets that above-mentioned classification results carries out second time to classify, and update each pixel
Category label；
3) utilize the weak scattering echoing characteristics of airfield runway, from step 2) classification results extract doubtful runway zone, obtain
Comprise the binary map of doubtful runway zone；
4) according to dimensional standard and the image resolution ratio of civil airport runway, runway zone doubtful in above-mentioned binary map is carried out shape
State processes to remove isolated zonule；
5) application Wishart grader to step 4) process after result carry out third time classify, by airfield runway region
Portion's structure distinguishes with surrounding atural object；
6) utilize the weak scattering echoing characteristics of airfield runway, again from above-mentioned 3rd class classification results, extract doubtful runway district
Territory, and the area-of-interest extracted is carried out again Morphological scale-space, obtain comprising the binary map of doubtful runway zone；
7) utilize civil airport track elements feature, runway zone doubtful in above-mentioned binary map is carried out identification, finally determines machine
Runway zone, field.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 1) in, Polarimetric SAR Image is carried out point by described target scattering randomness and Freeman decomposition method being combined
Class, calculates the average coherence matrix of each class pixel and the method as initial classes center is: being concerned with first with pixel
Matrix calculus Polarization scattering entropy H, according to the size of Polarization scattering entropy H the pixel in image is divided into low entropy class, middle entropy class and
High entropy class, the Polarization scattering entropy H < 0.5 of the lowest entropy class, the Polarization scattering entropy of middle entropy class is 0.5 < H < 0.9, the polarization of high entropy class
Scattering entropy H > 0.9；The most again view picture Polarimetric SAR Image is carried out Freeman decomposition, try to achieve the surface scattering merit of each pixel
Rate Ps, even scattered power PdAnd volume scattering power Pv；According to surface scattering power Ps, even scattered power Pd, volume scattering merit
Rate PvThe size of these three parameter, is divided three classes low entropy class: low entropy surface scattering (Ps>Pd,Pv), low entropy even scattering (Pd>
Ps,Pv) and low entropy volume scattering (Pv>Ps,Pd)；Middle entropy class is subdivided into six classes: middle entropy surface-even scatters (Ps>Pd>Pv), middle entropy
Surface-volume scattering (Ps>Pv>Pd), middle entropy even-surface scattering (Pd>Ps>Pv), middle entropy even-volume scattering (Pd>Pv>Ps), middle entropy
Body-surface scattering (Pv>Ps>Pd) and middle entropy body-even scattering (Pv>Pd>Ps)；High entropy class is not segmented, all pictures belonging to high entropy
Vegetarian refreshments is a class；There are 10 classes, respectively with label 1,2 ... the class that 10 labellings are corresponding；Every class is calculated finally according to category label
The average coherence matrix of pixel as its class center.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 2) in, described utilize the polarization difference degree between two class targets that classification results carries out second time to classify, and update every
The method of the category label of individual pixel is: calculate each pixel and step 1 in Polarimetric SAR Image) in 10 Ge Lei centers it
Between polarization difference degree, each pixel is divided into and apoplexy due to endogenous wind that its diversity factor is minimum, and updates the classification number of pixel；Work as institute
Have pixel classification adjust after, recalculate every class pixel average coherence matrix and as next time classify time
Class center；Repeat this process, until the pixel number that in Polarimetric SAR Image, classification changes less than given threshold value or reaches
To iterations.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 3) in, the described weak scattering echoing characteristics utilizing airfield runway, from step 2) classification results extract doubtful race
Region, road, the method for the binary map obtaining comprising doubtful runway zone is: calculation procedure 2) all kinds of in classification results for the second time
Average power content；This feature of weak scattering echoing characteristics according to civil airport runway, extracts in second time classification results average
The class pixel that power is minimum, the region constituted by these pixels is as doubtful runway zone；By view picture Polarimetric SAR Image
In the pixel of doubtful runway zone be labeled as " 1 ", the pixel of non-runway zone is labeled as " 0 ", obtains a width binary map.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 4) in, the described dimensional standard according to civil airport runway and image resolution ratio, to race doubtful in above-mentioned binary map
Region, road carries out Morphological scale-space to remove the method for isolated zonule: first according to the size characteristic of civil airport runway
With the resolution of image, calculate the area threshold Th0 of runway zone in Polarimetric SAR Image；Binary map is carried out connected region mark
After note processes, calculate the pixel number in each connected region；When in certain region, pixel number is less than area threshold Th0
Time, pixel in this region is all labeled as " 0 ", thus removes zonule isolated in binary map.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 5) in, described application Wishart grader is to step 4) process after result carry out third time classify, by airport run
The method that the internal structure in region, road distinguishes with surrounding atural object is: by step 2) the class center that obtains and step 4) process after
Doubtful runway zone in non-" 0 " pixel as the initial input of Wishart grader；Calculate non-" 0 " pixel and arrive each
The Wishart distance at class center, is categorized into pixel from its closest apoplexy due to endogenous wind according to minimum distance criterion, and updates picture
The category label of vegetarian refreshments；According to new classification results, recalculate all kinds of average coherence matrixes and as next iteration mistake
Cheng Zhonglei center；Repeat Wishart Iterative classification process until in image the pixel number that changes of classification less than to
Determine threshold value or reach iterations.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 6) in, the described weak scattering echoing characteristics utilizing airfield runway, again extract from above-mentioned third time classification results
Doubtful runway zone, and the area-of-interest extracted is carried out again Morphological scale-space, obtain comprising doubtful runway zone
The method of binary map is: according to this feature of weak scattering echoing characteristics of civil airport runway, extraction step 5) in classification results
One class of mean power minimum is as doubtful runway zone；Pixel mark by runway zone doubtful in view picture Polarimetric SAR Image
Being designated as " 1 ", the pixel of non-runway zone is labeled as " 0 ", obtains a width binary map；Binary map is carried out connected region labelling process
After, calculate the pixel number in each connected region；If pixel number is less than step 4 in certain connected region) face tried to achieve
Long-pending threshold value Th0, then be all labeled as pixel in this region " 0 ", thus removes zonule isolated in binary map.
Civil airport runway zone detection method without supervision multiclass classification the most according to claim 1, it is characterised in that:
In step 7) in, described utilizes civil airport track elements feature, and runway zone doubtful in above-mentioned binary map is carried out identification,
The method finally determining airfield runway region is: utilize the architectural feature of runway, i.e. topological characteristic, parallel lines feature, to step
6) the doubtful runway zone of gained is screened further, finally determines civil airport runway zone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610298469.7A CN106022217A (en) | 2016-05-09 | 2016-05-09 | Civil airport runway area detection method free from supervision multistage classification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610298469.7A CN106022217A (en) | 2016-05-09 | 2016-05-09 | Civil airport runway area detection method free from supervision multistage classification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106022217A true CN106022217A (en) | 2016-10-12 |
Family
ID=57081869
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610298469.7A Pending CN106022217A (en) | 2016-05-09 | 2016-05-09 | Civil airport runway area detection method free from supervision multistage classification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106022217A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305274A (en) * | 2018-03-08 | 2018-07-20 | 中国民航大学 | The Aircraft Targets detection method of PolSAR image multiple features fusions |
| CN108364011A (en) * | 2018-02-05 | 2018-08-03 | 中国民航大学 | PolSAR image multi-stage characteristics extract and unsupervised segmentation method |
| CN110031816A (en) * | 2019-03-22 | 2019-07-19 | 中国民航科学技术研究院 | Based on the Flying Area in Airport noncooperative target classifying identification method for visiting bird radar |
| CN112487977A (en) * | 2020-11-30 | 2021-03-12 | 东南大学 | Airport runway extraction method based on domestic GF2 remote sensing image |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102999762A (en) * | 2012-10-25 | 2013-03-27 | 西安电子科技大学 | Method for classifying polarimetric SAR (synthetic aperture radar) images on the basis of Freeman decomposition and spectral clustering |
| CN105095914A (en) * | 2015-08-13 | 2015-11-25 | 中国民航大学 | Airport runway detection method based on combination of h/q decomposition and Bayesian iterative classification |
-
2016
- 2016-05-09 CN CN201610298469.7A patent/CN106022217A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102999762A (en) * | 2012-10-25 | 2013-03-27 | 西安电子科技大学 | Method for classifying polarimetric SAR (synthetic aperture radar) images on the basis of Freeman decomposition and spectral clustering |
| CN105095914A (en) * | 2015-08-13 | 2015-11-25 | 中国民航大学 | Airport runway detection method based on combination of h/q decomposition and Bayesian iterative classification |
Non-Patent Citations (5)
| Title |
|---|
| PING HAN ET AL.: "Automatic Runway Detection Based On Unsupervised Classification in Polsar Image", 《2016 INTEGRATED COMMUNICATIONS NAVIGATION AND SURVEILLANCE》 * |
| 王文光 等: "一种基于差异度的极化SAR图像迭代分类方法", 《电子与信息学报》 * |
| 韩萍 等: "基于多级分类的PolSAR图像机场跑道检测", 《系统工程与电子技术》 * |
| 韩萍 等: "复杂场景下的极化SAR图像机场跑道检测", 《信号处理》 * |
| 韩萍 等: "散射模型与 Wishart 分类相结合的极化SAR图像分类", 《信号处理》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108364011A (en) * | 2018-02-05 | 2018-08-03 | 中国民航大学 | PolSAR image multi-stage characteristics extract and unsupervised segmentation method |
| CN108305274A (en) * | 2018-03-08 | 2018-07-20 | 中国民航大学 | The Aircraft Targets detection method of PolSAR image multiple features fusions |
| CN108305274B (en) * | 2018-03-08 | 2021-11-23 | 中国民航大学 | PolSAR image multi-feature fusion aircraft target detection method |
| CN110031816A (en) * | 2019-03-22 | 2019-07-19 | 中国民航科学技术研究院 | Based on the Flying Area in Airport noncooperative target classifying identification method for visiting bird radar |
| CN110031816B (en) * | 2019-03-22 | 2021-04-27 | 中国民航科学技术研究院 | Airport flight area non-cooperative target classification and identification method based on bird detection radar |
| CN112487977A (en) * | 2020-11-30 | 2021-03-12 | 东南大学 | Airport runway extraction method based on domestic GF2 remote sensing image |
| CN112487977B (en) * | 2020-11-30 | 2024-03-26 | 东南大学 | Airport runway extraction method based on domestic GF2 remote sensing image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Adaptively constrained dynamic time warping for time series classification and clustering | |
| Lei et al. | Landslide inventory mapping from bitemporal images using deep convolutional neural networks | |
| CN110287932B (en) | Road blocking information extraction method based on deep learning image semantic segmentation | |
| Tupin et al. | Detection of linear features in SAR images: Application to road network extraction | |
| CN106778595B (en) | Method for detecting abnormal behaviors in crowd based on Gaussian mixture model | |
| CN103456018B (en) | Remote sensing image change detection method based on fusion and PCA kernel fuzzy clustering | |
| Gao et al. | A novel target detection method for SAR images based on shadow proposal and saliency analysis | |
| CN109033950A (en) | Vehicle based on multiple features fusion cascade deep model, which is disobeyed, stops detection method | |
| Rogerson | Change detection thresholds for remotely sensed images | |
| CN105160355B (en) | A kind of method for detecting change of remote sensing image based on region correlation and vision word | |
| CN105389799B (en) | SAR image object detection method based on sketch map and low-rank decomposition | |
| CN103955926A (en) | Method for remote sensing image change detection based on Semi-NMF | |
| CN106295498A (en) | Remote sensing image target area detection apparatus and method | |
| CN106022217A (en) | Civil airport runway area detection method free from supervision multistage classification | |
| CN109635789B (en) | High-resolution SAR image classification method based on intensity ratio and spatial structure feature extraction | |
| CN105512622B (en) | A kind of visible remote sensing image sea land dividing method based on figure segmentation and supervised learning | |
| CN103093243B (en) | The panchromatic remote sensing image clouds of high-resolution sentences method | |
| Bai et al. | Object-based building damage assessment methodology using only post event ALOS-2/PALSAR-2 dual polarimetric SAR intensity images | |
| Lam | Methodologies for mapping land cover/land use and its change | |
| Mahrooghy et al. | A machine learning framework for detecting landslides on earthen levees using spaceborne SAR imagery | |
| CN105205807B (en) | Method for detecting change of remote sensing image based on sparse automatic coding machine | |
| CN109684986A (en) | A kind of vehicle analysis method and system based on automobile detecting following | |
| CN109961028A (en) | SAR detection method based on three-dimensional Block- matching and full condition of contact random field | |
| Huang et al. | Superpixel-based change detection in high resolution sar images using region covariance features | |
| CN105095914A (en) | Airport runway detection method based on combination of h/q decomposition and Bayesian iterative classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20161012 |