CN108446707A - Remote sensing image airplane detection method based on key point screening and DPM confirmation - Google Patents
Remote sensing image airplane detection method based on key point screening and DPM confirmation Download PDFInfo
- Publication number
- CN108446707A CN108446707A CN201810182286.8A CN201810182286A CN108446707A CN 108446707 A CN108446707 A CN 108446707A CN 201810182286 A CN201810182286 A CN 201810182286A CN 108446707 A CN108446707 A CN 108446707A
- Authority
- CN
- China
- Prior art keywords
- target
- detected
- region
- area
- candidate region
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 40
- 238000012216 screening Methods 0.000 title claims abstract description 30
- 238000012790 confirmation Methods 0.000 title abstract 2
- 238000000034 method Methods 0.000 claims abstract description 25
- 238000012876 topography Methods 0.000 claims description 12
- 230000001186 cumulative effect Effects 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 5
- 239000010410 layer Substances 0.000 description 15
- 238000012549 training Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 238000007689 inspection Methods 0.000 description 5
- 230000004044 response Effects 0.000 description 5
- 238000009432 framing Methods 0.000 description 3
- 230000003139 buffering effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 239000002344 surface layer Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides a remote sensing image airplane detection method based on key point screening and DPM confirmation, which comprises the following steps: acquiring a local image in an airport image; the local image is obtained according to the local characteristic information of the target to be detected; determining the processing result of the preprocessed local image as the primary screening result of the target to be detected; determining the area of the target to be detected according to the primary screening result and a preset model established in advance; classifying the regions, acquiring candidate regions with the same classification and overlapping parts, and determining the target to be detected according to a preset rule and the candidate regions. According to the method provided by the embodiment of the invention, the area where the target to be detected is located is determined firstly, then the areas are classified, the candidate areas with the same classification and the overlapped parts are obtained, and then the target to be detected is determined, so that the adaptability to a plurality of influence factors in an airport and the detection precision of the target to be detected can be improved.
Description
Technical field
The present embodiments relate to remote sensing images Airplane detection technical fields, and in particular to one kind based on key point screening and
The remote sensing images airplane detection method that DPM confirms.
Background technology
Currently, the method kind being detected to target to be detected (being, for example, aircraft) based on remote sensing image in airport
Class is various, since aircraft configuration size difference is larger so that the adaptability for having algorithm is limited.In addition, airport remote sensing images field
Scape is complicated, there is building interference similar with aircraft signature attribute, easily generates a large amount of false-alarms, detection difficulty is big.
Some prior arts are detected aircraft, but using the description to aircraft local feature to aircraft global feature
Declarative poor, the Aircraft Targets detection being only applicable under simple local scene, will produce under the complicated big visual field in airport big
Measure false-alarm.Some prior arts isolate target in the feature base of target surface layer by the methods of filtering, along with follow-up mirror
Other places reason realizes detection, but the scene that such detection method is not suitable for background complexity yet, building heavy dense targets mix.Also
The prior art is to carry out Airplane detection using template matching technique, but common template form is relatively fixed, and flexibility is poor, compatible
Property is poor;The inside even from weather such as remote sensing image imaging platform easily loaded, illumination so that aircraft generates in remote sensing image
There is shade in certain deformation, fuselage edge, and using common templates matching technique, adaptability is poor in actual scene;And aircraft
Often difference, the aircraft of different sizes have all needed to establish corresponding template, calculation amount and model foundation complexity size
It greatly increases.
Therefore, how to avoid drawbacks described above, improve to the adaptability of numerous influence factors in airport, target to be detected
Accuracy of detection, becoming need solve the problems, such as.
Invention content
In view of the problems of the existing technology, the embodiment of the present invention provide it is a kind of based on key point screening and DPM confirm
Remote sensing images airplane detection method.
The embodiment of the present invention provides a kind of remote sensing images airplane detection method confirmed based on key point screening and DPM, institute
The method of stating includes:
Obtain the topography in Airport Images;The topography is obtained according to the local feature information of target to be detected
It takes;
The handling result of pretreated topography is determined as to the initial screening result of the target to be detected;And according to
The initial screening result and the preset model pre-established, determine the region where the target to be detected;
Classify to the region, and obtain the candidate region for having lap of same category, and according to default rule
Then with the candidate region, the target to be detected is determined.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is passed through
It first determines the region where target to be detected, then classifies to region, and obtain the candidate for having lap of same category
Region, then determine target to be detected, the detection to the adaptability of numerous influence factors, target to be detected in airport can be improved
Precision.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Some bright embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is that the remote sensing images airplane detection method flow that the embodiment of the present invention is screened based on key point and DPM confirms is shown
It is intended to;
Fig. 2 is that aircraft DPM models of the embodiment of the present invention export result figure;
Fig. 3 is that the embodiment of the present invention determines the pervious Airport Images figure of target to be detected;
Fig. 4 is that the embodiment of the present invention determines the later Airport Images figure of target to be detected;
Fig. 5 is the entirety for the remote sensing images airplane detection method that the embodiment of the present invention is screened based on key point and DPM confirms
Flow chart.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
The every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Fig. 1 is that the remote sensing images airplane detection method flow that the embodiment of the present invention is screened based on key point and DPM confirms is shown
It is intended to, as shown in Figure 1, a kind of remote sensing images aircraft inspection confirmed based on key point screening and DPM provided in an embodiment of the present invention
Survey method, includes the following steps:
S1:Obtain the topography in Airport Images;The topography is believed according to the local feature of target to be detected
What breath obtained.
Specifically, device obtains the topography in Airport Images;The topography is the office according to target to be detected
What portion's characteristic information obtained.DPM (Deformable Part Model), just as its name suggests, deformable component model,
It is a kind of detection algorithm component-based.Key point can be understood as the local feature of target to be detected.Target to be detected can be with
It is the aircraft rested on airport, local feature information can be projecting point, i.e. angle point, marginal point, dark areas bright spot, bright area
The abundant local message such as dim spot.
S2:The handling result of pretreated topography is determined as to the initial screening result of the target to be detected;And
According to the initial screening result and the preset model pre-established, the region where the target to be detected is determined.
Specifically, the handling result of pretreated topography is determined as the initial screening of the target to be detected by device
As a result;And according to the initial screening result and the preset model pre-established, determine the region where the target to be detected.
Following method specifically may be used to realize:
1. the doubtful area screening based on key point closeness
1) key point is extracted
Since most of region is free of target to be detected in the whole figure in airport (corresponding Airport Images), to reduce whole visual field detection
Calculation amount first carries out doubtful area's screening.SIFT (Scale-invariant feature transform) algorithm pair may be used
Projecting point in Airport Images, i.e. angle point, marginal point, dark areas bright spot, bright area dim spot etc. have office existing for abundant local message
Portion's image extracts.The Local Extremum of scale space is calculated using difference of Gaussian, and these local extremums are clicked through
Row low contrast removes, skirt response removes, and obtains pinpoint characteristic point.
2) the doubtful piecemeal screening based on piecemeal key point closeness
Window size is set as 256*256 in 1m resolution ratio visible remote sensing images after being extracted to key point, and with
128 carry out the sliding window for having overlapping for step-length, and cumulative statistics is carried out to the SIFT key points number in each window.In the image point
Under resolution, key point number threshold value of the usual traffic pattern rich in information candidate area may be set to 25, and key point number is more than
The image in window of threshold value retains, the direct zero setting of remainder image, and the purpose that can reach doubtful area's screening (gets primary dcreening operation
Select result).
2. the full-scale doubtful Aircraft Targets of DPM based on dual resolution design are confined
1) the doubtful big Aircraft Targets detection of the DPM based on former image in different resolution is confined
A. HOG feature pyramids are built
Due to DPM models training when it has been determined that in for detection image various sizes of target need to carry out it is multiple dimensioned
Analysis, needs construction feature pyramid.The histograms of oriented gradients of image is calculated by HOG algorithms to build L layers of feature gold
Word tower, it is specified that
λ is sampling specification, needs the number of plies walked downwards to be in order to obtain twice of resolution ratio of a certain layer in pyramid
λ.And pyramid top layer is HOG feature of the image under former resolution ratio.
B. computation model responds score
Assuming that the model obtained in the training stage has n component, then (n+2) tuple (F can be defined as0,P1,…,Pn,
b).Wherein F0For root filter, PiFor i-th of component filter, b is amount of bias.Filter is on HOG feature pyramid l layers
Response is scored at:
pi=(xi,yi,li)
Wherein, H is the characteristic image pyramid established in previous step;piIndicate characteristic image pyramid liLayer position is
(xi,yi) point;φ indicates feature vector of the position in H.As i=0, R indicates root filter in l0Layer responds
Point;As i > 0, indicate i-th of component filter in liThe score response of layer.Since the resolution ratio of component filter is root filter
Twice of wave device, li=l0- λ, i.e. l in pyramidiLayer resolution ratio is l0Twice of layer.
According to the filter score calculated above, high subregion is expanded while considering that deformation is spent, is adjusted
Partial model is in pyramid l0Find combination in-λ layers of position
Wherein, (dx, dy) indicates displacement of the component locations relative to anchor point (component filter ideal position);diIt indicates
Offset vector;φdIndicate the cost weight of offset.
C. comprehensive score is calculated
According to given root modal position p0=(x0,y0,l0), high subregion is carried out while considering that deformation is spent
It expands, regulating member model is in pyramid l0Find partial model position grouping (p in-λ layers of position1,p2,...,pn) so that
It is maximum that all parts model responds score.
Root filter response score is added with each expansion with sub-sampling back part filter response score, is finally added
Partial model obtains the l layers of comprehensive score with respect to the offset of root model:
Wherein, viIndicate the relative position of a certain component filter anchor point and root filter;(x0,y0) 2 multiplying factors be for
Resolution ratio is unified in feature pyramidal layer where partial model;B expressions make component filter to its offset.By
This can obtain every layer of score in pyramid, choose effective score by the method for non-maxima suppression, and mapped back image, can
The rectangle frame for obtaining determining target location, stores the doubtful frame coordinate information in image, and doubtful frame is target place to be detected
Region.
2) the doubtful flivver target detection of the DPM based on interpolated resolution image is confined
Since aircraft size difference is larger in image to be detected, only by building characteristics of image pyramid in different resolution
Missing inspection can be generated to flivver by being detected on layer, therefore is detected again to image based on interpolated resolution.It will be carried in step 1
It takes the image interpolation in doubtful area to be amplified to twice of original image size, all behaviour in step 1) is repeated to image after interpolation amplification
Make, with the flivver in detection image.
Fig. 2 is that aircraft DPM models of the embodiment of the present invention export result figure, wherein (a), which is root model, exports result figure;(b)
Result figure is exported for partial model;(c) it is that the partial model expanded exports result figure.
It should be noted that:The above-mentioned preset model pre-established needs to be trained in advance before the use, method
It can be as follows:
Establish tranining database
Definition c is training objective classification, and P is positive sample collection, and N is negative sample collection, and positive and negative sample set is given target category c
Training sample.P is the positive sample database of artificial indicia framing, is the set of two tuples (I, B).Wherein I is image, and B is figure
As c classification target indicia framings in I.Since actual scene needs, c class targets are chosen to be aircraft, and N is negative sample airfield runway, machine
Shelter bridge, building corner set.All sample images are carried out manually to confine target, and by corresponding informance data storage into
In xml document, model aircraft tranining database is established.
Model aircraft is trained
Initialize root filter
Trained mixed model contains m component, and the positive sample database P of handmarking's frame is sorted and divided with length-width ratio
Class is m groups, is denoted as P1 ..., Pm.Select the value more than 80% rectangle frame area as root model area.With standard SVM algorithm
M corresponding root filters are trained, F1 ..., Fm is denoted as.
More new root filter
This m root filter is joined together, declining training algorithm by coordinate is iterated optimization.
Initialisation unit filter
It is 8 by the component settings of each component model, component is placed on to the two of root filter in the form of formed symmetrical
Side, first component are placed on root filter highest energy region, and by this region energy zero setting.This operation is repeated, until will
Component placement finishes.The resolution ratio of wherein component filter is twice of root filter.The setting of wherein 8 components numbers with it is silent
6 components recognized compare more fine, are more suitable for the target detection of complex scene, and excessively will not finely allow detection process
It complicates.
Update existing model
The positive sample indicia framing that training data is concentrated is detected with existing model, using the position of highest scoring as this
The positive sample of sample pane, is put into buffering area.Equally with existing model inspection negative sample, and the position of highest scoring is put into slow
Area is rushed, until file maximum limits, the sample training of buffering area is used in combination to go out new model.By above-mentioned Policy iteration more new model 10
It is secondary, obtain final model parameter.
S3:Classify to the region, and obtain the candidate region for having lap of same category, and according to default
The regular and described candidate region determines the target to be detected.
Specifically, device classifies to the region, and obtains the candidate region for having lap of same category, and
According to preset rules and the candidate region, the target to be detected is determined.Region may include the vertex position information of framework,
I.e. above-mentioned doubtful frame coordinate information, classification may include:
According to the vertex position information, the center position information of the framework is obtained;Traverse all center points
Confidence ceases, and calculates the Euclidean distance determined according to each two center position information;It will be greater than the Euclidean distance institute of predetermined threshold value
Corresponding two region divisions are the region of same category.
It is described as follows:Doubtful frame coordinate information can be framework top left co-ordinate position, lower right corner coordinate position, with
This calculates each framework center position coordinate (corresponding center position information).Center point coordinate with first doubtful frame is
Benchmark is compared one by one with other doubtful frame center point coordinates, calculates Euclidean distance, be will be greater than threshold value and (is corresponded to default threshold
Value, predetermined threshold value be referred to model aircraft size width carry out independently be arranged) be divided into one kind.Again to remaining unfiled
Doubtful frame repeats this operation, until all doubtful frame classification collect.
The step of determining the target area where target to be detected may include:
It adds up to the presetted pixel value in the region for having lap of each same category;Extraction is more than or equal to default
The candidate region of cumulative pixel value;The minimum enclosed rectangle of each candidate region is obtained, and according to the area of each candidate region
With the comparing result of corresponding minimum enclosed rectangle, the target area where the target to be detected is determined.
Following operation is made to the doubtful frame of each of the above class:
1 (i.e. default picture is set to the pixel value of each doubtful frame (region for having lap of corresponding each same category)
Plain value is selected as 1), having lap to do cumulative.Assuming that the doubtful frame number of the heap is n, then pixel value is more than or equal to n/2 after being superimposed
Extracting section come out, be A1,...,AiI.e. the number of candidate region is i, seeks extraction region AiMinimum enclosed rectangle Bi,
By BiWith composition AiEach of the doubtful frame area area of candidate region (corresponding each) compare, if BiArea, which is more than, constitutes Ai
Each of the half of doubtful frame area then retain Bi, A will be constitutediAll doubtful frames removals;Conversely, removing Bi, retain structure
At AiAll doubtful frames.
After carrying out this operation to every a kind of frame, the target area that the framework of reservation is determined is preserved as target to be detected
Its coordinate (location information of corresponding target area), is shown in as testing result in Airport Images.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is passed through
It first determines the region where target to be detected, then classifies to region, and obtain the candidate for having lap of same category
Region, then determine target to be detected, the detection to the adaptability of numerous influence factors, target to be detected in airport can be improved
Precision.
Fig. 3 is that the embodiment of the present invention determines the pervious Airport Images figure of target to be detected;Fig. 4 determines for the embodiment of the present invention
The later Airport Images figure of target to be detected, as shown in Figure 3, Figure 4, it can be seen that the embodiment of the present invention can be treated accurately
Detection target is detected.
Fig. 5 is the entirety for the remote sensing images airplane detection method that the embodiment of the present invention is screened based on key point and DPM confirms
Flow chart, illustrating for Fig. 5 can refer to above-described embodiment, repeat no more.
On the basis of the above embodiments, the region includes the vertex position information of framework;Correspondingly, described to described
Region is classified, including:
According to the vertex position information, the center position information of the framework is obtained.
Specifically, device obtains the center position information of the framework according to the vertex position information.It can refer to
Embodiment is stated, is repeated no more.
All center position information is traversed, the Euclidean distance determined according to each two center position information is calculated.
Specifically, device traverses all center position information, calculates and determined according to each two center position information
Euclidean distance.Above-described embodiment is can refer to, is repeated no more.
It will be greater than the region that two region divisions corresponding to the Euclidean distance of predetermined threshold value are same category.
Specifically, device will be greater than the area that two region divisions corresponding to the Euclidean distance of predetermined threshold value are same category
Domain.Above-described embodiment is can refer to, is repeated no more.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is passed through
First classify to the region where target to be detected, can ensure being normally carried out for this method.
On the basis of the above embodiments, the region for having lap that is described and obtaining same category, and according to default
Rule determines the target area where the target to be detected, including:
It adds up to the presetted pixel value in the region for having lap of each same category.
Specifically, device adds up to the presetted pixel value in the region for having lap of each same category.It can join
According to above-described embodiment, repeat no more.
Extraction is more than or equal to the candidate region of default cumulative pixel value.
Specifically, device extraction is more than or equal to the candidate region of default cumulative pixel value.Above-described embodiment is can refer to, no longer
It repeats.
The minimum enclosed rectangle of each candidate region is obtained, and outer according to the area of each candidate region and corresponding minimum
The comparing result for connecing rectangle determines the target area where the target to be detected.
Specifically, device obtains the minimum enclosed rectangle of each candidate region, and according to the area of each candidate region with
The comparing result of corresponding minimum enclosed rectangle determines the target area where the target to be detected.It can refer to above-mentioned implementation
Example, repeats no more.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is based on
Overlapping confidence level determines the target area where target to be detected, is further able to improve and be fitted to numerous influence factors in airport
It should be able to power, the accuracy of detection of target to be detected.
On the basis of the above embodiments, the candidate region for having lap that is described and obtaining same category, and according to
Preset rules and the candidate region determine the target to be detected, including:
If the area of candidate region is less than or equal to twice of the area of corresponding minimum enclosed rectangle, by minimum external square
The target area that shape determines is as the target to be detected.
Specifically, if device judges to know that the area of candidate region is less than or equal to the area of corresponding minimum enclosed rectangle
Twice, then the target area determined minimum enclosed rectangle is as the target to be detected.Above-described embodiment is can refer to, it is no longer superfluous
It states.
If the area of candidate region is more than twice of the area of corresponding minimum enclosed rectangle, and the candidate region is true
Fixed target area is as the target to be detected.
Specifically, if device judges to know that the area of candidate region is more than the two of the area of corresponding minimum enclosed rectangle
Times, then the target area determined the candidate region is as the target to be detected.Above-described embodiment is can refer to, it is no longer superfluous
It states.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is passed through
Compare the area of candidate region and the area of corresponding minimum enclosed rectangle, is further able to improve the inspection of target to be detected
Survey precision.
On the basis of the above embodiments, the method further includes:
The quantity n in the region for having lap of each same category is obtained, the presetted pixel value is 1;Correspondingly, institute
It is n/2 to state default cumulative pixel value.
Specifically, device obtains the quantity n in the region for having lap of each same category, the presetted pixel value is
1;Correspondingly, the default cumulative pixel value is n/2.Above-described embodiment is can refer to, is repeated no more.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is passed through
Reasonable value is set for presetted pixel value and default cumulative pixel value, more can control to independent and flexible the inspection of target to be detected
Survey precision.
On the basis of the above embodiments, after the step of determination target to be detected, the method further includes:
The location information of target area is obtained, and the location information is shown in the Airport Images.
Specifically, device obtains the location information of target area, and the location information is shown in the Airport Images
In.Above-described embodiment is can refer to, is repeated no more.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, by mesh
The location information in mark region is shown in Airport Images, can be convenient for checking for location information.
On the basis of the above embodiments, the target to be detected is the aircraft rested on airport.
Specifically, the target to be detected in device is the aircraft rested on airport.
The remote sensing images airplane detection method provided in an embodiment of the present invention confirmed based on key point screening and DPM, is passed through
Target to be detected is selected as resting on the aircraft on airport, is further able to improve the adaptation energy to numerous influence factors in airport
Power, the accuracy of detection of aircraft.
One of ordinary skill in the art will appreciate that:Realize that all or part of step of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can be stored in a computer read/write memory medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed;And storage medium above-mentioned includes:ROM, RAM, magnetic disc or light
The various media that can store program code such as disk.
The embodiments such as electronic equipment described above are only schematical, illustrate as separating component wherein described
Unit may or may not be physically separated, and the component shown as unit may or may not be object
Manage unit, you can be located at a place, or may be distributed over multiple network units.It can select according to the actual needs
Some or all of module therein is selected to achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying wound
In the case of the labour for the property made, you can to understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It is realized by the mode of software plus required general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be expressed in the form of software products in other words, should
Computer software product can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally it should be noted that:The above various embodiments is only to illustrate the technical solution of the embodiment of the present invention rather than right
It is limited;Although the embodiment of the present invention is described in detail with reference to foregoing embodiments, the ordinary skill of this field
Personnel should understand that:It still can be with technical scheme described in the above embodiments is modified, or to which part
Or all technical features carries out equivalent replacement;And these modifications or replacements, it does not separate the essence of the corresponding technical solution
The range of each embodiment technical solution of the embodiment of the present invention.
Claims (7)
1. a kind of remote sensing images airplane detection method confirmed based on key point screening and DPM, which is characterized in that including:
Obtain the topography in Airport Images;The topography is the local feature acquisition of information according to target to be detected
's;
The handling result of pretreated topography is determined as to the initial screening result of the target to be detected;And according to described
Initial screening result and the preset model pre-established, determine the region where the target to be detected;
Classify to the region, and obtain the candidate region for having lap of same category, and according to preset rules and
The candidate region determines the target to be detected.
2. according to the method described in claim 1, it is characterized in that, the region includes the vertex position information of framework;Accordingly
Ground, it is described to classify to the region, including:
According to the vertex position information, the center position information of the framework is obtained;
All center position information is traversed, the Euclidean distance determined according to each two center position information is calculated;
It will be greater than the region that two region divisions corresponding to the Euclidean distance of predetermined threshold value are same category.
3. according to the method described in claim 1, it is characterized in that, the area for having lap that is described and obtaining same category
Domain, and according to preset rules, determine the target area where the target to be detected, including:
It adds up to the presetted pixel value in the region for having lap of each same category;
Extraction is more than or equal to the candidate region of default cumulative pixel value;
The minimum enclosed rectangle of each candidate region is obtained, and according to the area of each candidate region and the corresponding external square of minimum
The comparing result of shape determines the target area where the target to be detected.
4. according to the method described in claim 3, it is characterized in that, the candidate for having lap that is described and obtaining same category
Region, and according to preset rules and the candidate region, determine the target to be detected, including:
If the area of candidate region is less than or equal to twice of the area of corresponding minimum enclosed rectangle, and minimum enclosed rectangle is true
Fixed target area is as the target to be detected;
If the area of candidate region is more than twice of the area of corresponding minimum enclosed rectangle, the candidate region is determined
Target area is as the target to be detected.
5. according to the method described in claim 3, it is characterized in that, the method further includes:
The quantity N in the region for having lap of each same category is obtained, the presetted pixel value is 1;Correspondingly, described pre-
If cumulative pixel value is N/2.
6. method according to any one of claims 1 to 5, which is characterized in that the step of the determination target to be detected
Later, the method further includes:
The location information of target area is obtained, and the location information is shown in the Airport Images.
7. according to the method described in claim 1, it is characterized in that, the target to be detected is the aircraft rested on airport.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810182286.8A CN108446707B (en) | 2018-03-06 | 2018-03-06 | Remote sensing image airplane detection method based on key point screening and DPM confirmation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810182286.8A CN108446707B (en) | 2018-03-06 | 2018-03-06 | Remote sensing image airplane detection method based on key point screening and DPM confirmation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108446707A true CN108446707A (en) | 2018-08-24 |
CN108446707B CN108446707B (en) | 2020-11-24 |
Family
ID=63193699
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810182286.8A Active CN108446707B (en) | 2018-03-06 | 2018-03-06 | Remote sensing image airplane detection method based on key point screening and DPM confirmation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108446707B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977965A (en) * | 2019-02-28 | 2019-07-05 | 北方工业大学 | Method and device for determining detection target in remote sensing airport image |
CN110188815A (en) * | 2019-05-24 | 2019-08-30 | 广州市百果园信息技术有限公司 | A kind of characteristic point method of sampling, device, equipment and storage medium |
CN110349184A (en) * | 2019-06-06 | 2019-10-18 | 南京工程学院 | The more pedestrian tracting methods differentiated based on iterative filtering and observation |
CN110378186A (en) * | 2019-03-22 | 2019-10-25 | 北京理工雷科电子信息技术有限公司 | The method that SAR remote sensing images Ship Target false-alarm is rejected |
CN111242088A (en) * | 2020-01-22 | 2020-06-05 | 上海商汤临港智能科技有限公司 | Target detection method and device, electronic equipment and storage medium |
CN111345815A (en) * | 2020-02-11 | 2020-06-30 | 广州视源电子科技股份有限公司 | Method, device, equipment and storage medium for detecting QRS wave in electrocardiosignal |
CN112183159A (en) * | 2019-07-03 | 2021-01-05 | 四川大学 | Construction of a skeletal model of a non-human target in an image using keypoints |
CN112541372A (en) * | 2019-09-20 | 2021-03-23 | 初速度(苏州)科技有限公司 | Difficult sample screening method and device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101783076A (en) * | 2010-02-04 | 2010-07-21 | 西安理工大学 | Method for quick vehicle type recognition under video monitoring mode |
US8306163B1 (en) * | 2007-12-13 | 2012-11-06 | Marvell International Ltd. | Method and apparatus for automatic gain control |
CN105913082A (en) * | 2016-04-08 | 2016-08-31 | 北京邦焜威讯网络技术有限公司 | Method and system for classifying objects in image |
CN106408529A (en) * | 2016-08-31 | 2017-02-15 | 浙江宇视科技有限公司 | Shadow removal method and apparatus |
CN106845496A (en) * | 2016-12-30 | 2017-06-13 | 首都师范大学 | fine target identification method and system |
-
2018
- 2018-03-06 CN CN201810182286.8A patent/CN108446707B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8306163B1 (en) * | 2007-12-13 | 2012-11-06 | Marvell International Ltd. | Method and apparatus for automatic gain control |
CN101783076A (en) * | 2010-02-04 | 2010-07-21 | 西安理工大学 | Method for quick vehicle type recognition under video monitoring mode |
CN105913082A (en) * | 2016-04-08 | 2016-08-31 | 北京邦焜威讯网络技术有限公司 | Method and system for classifying objects in image |
CN106408529A (en) * | 2016-08-31 | 2017-02-15 | 浙江宇视科技有限公司 | Shadow removal method and apparatus |
CN106845496A (en) * | 2016-12-30 | 2017-06-13 | 首都师范大学 | fine target identification method and system |
Non-Patent Citations (1)
Title |
---|
陈芝垚: ""基于DPM的行人检测和行人特征提取算法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977965B (en) * | 2019-02-28 | 2021-04-13 | 北方工业大学 | Method and device for determining detection target in remote sensing airport image |
CN109977965A (en) * | 2019-02-28 | 2019-07-05 | 北方工业大学 | Method and device for determining detection target in remote sensing airport image |
CN110378186B (en) * | 2019-03-22 | 2021-09-24 | 北京理工雷科电子信息技术有限公司 | SAR remote sensing image ship target false alarm rejection method |
CN110378186A (en) * | 2019-03-22 | 2019-10-25 | 北京理工雷科电子信息技术有限公司 | The method that SAR remote sensing images Ship Target false-alarm is rejected |
CN110188815A (en) * | 2019-05-24 | 2019-08-30 | 广州市百果园信息技术有限公司 | A kind of characteristic point method of sampling, device, equipment and storage medium |
CN110349184A (en) * | 2019-06-06 | 2019-10-18 | 南京工程学院 | The more pedestrian tracting methods differentiated based on iterative filtering and observation |
CN110349184B (en) * | 2019-06-06 | 2022-08-09 | 南京工程学院 | Multi-pedestrian tracking method based on iterative filtering and observation discrimination |
CN112183159A (en) * | 2019-07-03 | 2021-01-05 | 四川大学 | Construction of a skeletal model of a non-human target in an image using keypoints |
CN112541372A (en) * | 2019-09-20 | 2021-03-23 | 初速度(苏州)科技有限公司 | Difficult sample screening method and device |
CN112541372B (en) * | 2019-09-20 | 2023-03-28 | 魔门塔(苏州)科技有限公司 | Difficult sample screening method and device |
CN111242088A (en) * | 2020-01-22 | 2020-06-05 | 上海商汤临港智能科技有限公司 | Target detection method and device, electronic equipment and storage medium |
CN111242088B (en) * | 2020-01-22 | 2023-11-28 | 上海商汤临港智能科技有限公司 | Target detection method and device, electronic equipment and storage medium |
CN111345815A (en) * | 2020-02-11 | 2020-06-30 | 广州视源电子科技股份有限公司 | Method, device, equipment and storage medium for detecting QRS wave in electrocardiosignal |
Also Published As
Publication number | Publication date |
---|---|
CN108446707B (en) | 2020-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108446707A (en) | Remote sensing image airplane detection method based on key point screening and DPM confirmation | |
Vetrivel et al. | Identification of damage in buildings based on gaps in 3D point clouds from very high resolution oblique airborne images | |
CN113392775B (en) | Sugarcane seedling automatic identification and counting method based on deep neural network | |
CN104881865B (en) | Forest pest and disease monitoring method for early warning and its system based on unmanned plane graphical analysis | |
CN103218831B (en) | A kind of video frequency motion target classifying identification method based on profile constraint | |
Wang et al. | Photovoltaic panel extraction from very high-resolution aerial imagery using region–line primitive association analysis and template matching | |
CN103218787B (en) | Multi-source heterogeneous remote sensing image reference mark automatic acquiring method | |
CN110543837A (en) | visible light airport airplane detection method based on potential target point | |
CN103337052B (en) | Automatic geometric correcting method towards wide cut remote sensing image | |
CN103268358B (en) | Multi-source control point image database builds and update method | |
CN108009509A (en) | Vehicle target detection method | |
CN109697475A (en) | A kind of muskeg information analysis method, remote sensing monitoring component and monitoring method | |
CN104463117A (en) | Sample collection method and system used for face recognition and based on video | |
CN108710913A (en) | A kind of switchgear presentation switch state automatic identification method based on deep learning | |
CN110033516A (en) | Flat-elongated particles detection method of content based on the identification of binocular camera Image Acquisition | |
CN110163836A (en) | Based on deep learning for the excavator detection method under the inspection of high-altitude | |
CN109978882A (en) | A kind of medical imaging object detection method based on multi-modal fusion | |
CN110728252B (en) | Face detection method applied to regional personnel motion trail monitoring | |
CN104680542A (en) | Online learning based detection method for change of remote-sensing image | |
CN104751166A (en) | Spectral angle and Euclidean distance based remote-sensing image classification method | |
CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
CN108681711A (en) | A kind of natural landmark extracting method towards mobile robot | |
CN108776777A (en) | The recognition methods of spatial relationship between a kind of remote sensing image object based on Faster RCNN | |
CN111914634A (en) | Complex-scene-interference-resistant automatic manhole cover type detection method and system | |
CN112766417A (en) | Method and system for recognizing current land type of land utilization of target land block by using field photo |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |