CN107507198B - Aircraft brake disc detection and method for tracing - Google Patents
Aircraft brake disc detection and method for tracing Download PDFInfo
- Publication number
- CN107507198B CN107507198B CN201710724698.5A CN201710724698A CN107507198B CN 107507198 B CN107507198 B CN 107507198B CN 201710724698 A CN201710724698 A CN 201710724698A CN 107507198 B CN107507198 B CN 107507198B
- Authority
- CN
- China
- Prior art keywords
- aircraft
- region
- stage
- brake disc
- detector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to target tracking technical field, a kind of aircraft brake disc detection and method for tracing are provided.This method includes that original image is inputted convolutional neural networks, generate convolution characteristic pattern, original image is divided into using image partition method by different regions, to the convolution characteristic pattern in each extracted region region being partitioned into, and be standardized, the corresponding provincial characteristics figure in each region is obtained, the provincial characteristics figure in each region is examined, according to the corresponding region of characteristic pattern for being determined as aircraft, calculates and export aircraft region.Aircraft brake disc detection of the present invention and method for tracing, can be improved the reliability and real-time of target acquisition and tracking during takeoff and landing, avoid similar target jamming, improve algorithm execution efficiency.
Description
Technical field
The present invention relates to target tracking technical fields, and in particular to a kind of detection of aircraft brake disc and method for tracing.
Background technique
The landing of aircraft with to take off be that accident rate highest two stages are compared in flight course.To flying during landing
Machine carries out video tracking shooting automatically, is a direction of current airport security surveillance technology development.
It in the prior art, is mostly by automatic control turntable, autozoom telephoto lens and image enhancement technique, with biography
The working method of system is compared, and the vision signal of captured in real-time being capable of the apparent state for more stably observing aircraft.This is also certainly
The data basis for the accident potential such as dynamic survey mission posture is abnormal, undercarriage is abnormal.Entire landing process view under automatically recording
Frequency information is also the Primary Reference foundation of subsequent problem investigation.
When takeoff and landing process automatic tracing monitoring system is run, end aircraft region to be taken off is gone in alignment first, or
Person's landing preceding low latitude navigation channel region, starts to monitor.If there is object enters monitor area, after judgement is Aircraft Targets, locking
Target starts tracking shooting.In tracing process, keep aircraft in picture center, root according to the position adjust automatically turntable of aircraft
Keep aircraft imaging size suitable apart from automatic focus adjustable according to aircraft.Descent is tracked after aircraft clears the runway and is tied
Beam, take-off process, which tracks after aircraft leaves monitoring range, to be terminated.
From system operating mode as can be seen that takeoff and landing tracking and common target tracking problem make a big difference.
Firstly, capture apparatus keeps rotation in whole process, cause the background in video in high-speed motion.Secondly as shooting process
Target lock-on, aircraft is in picture central area erratic motion, and size is imaged, and there is also change to a certain degree.Finally, due to see
The lasting variation of angle and distance is examined, there is also very big variations for aircraft imaging.Other than these differences, takeoff and landing tracking is also
It faces some technological challenges, for example aircraft background at low clearance area is more complicated, there is other fly when activity in runway zone
Machine target jamming.These differences and challenge cause to be unsuitable for based on conventional methods such as Kalman filter motion tracking, light stream trackings
Takeoff and landing tracking.
The reliability and real-time for how improving target acquisition and tracking during takeoff and landing, avoid similar target dry
The problem of disturbing, being those skilled in the art's urgent need to resolve.
Summary of the invention
For the defects in the prior art, it the present invention provides a kind of detection of aircraft brake disc and method for tracing, can be improved
The reliability and real-time of target acquisition and tracking, avoid similar target jamming during takeoff and landing.
In a first aspect, the present invention provides a kind of aircraft brake disc detection method, this method comprises:
Original image is inputted into convolutional neural networks, generates convolution characteristic pattern, convolution characteristic pattern includes the N-dimensional of each pixel
Feature vector;
Original image is divided into using image partition method by different regions;
It to the convolution characteristic pattern in each extracted region region being partitioned into, and is standardized, obtains each area
The corresponding provincial characteristics figure in domain;
The provincial characteristics figure for examining each region is calculated and is exported according to the corresponding region of characteristic pattern for being determined as aircraft
Aircraft region.
Further, original image is inputted into convolutional neural networks, generates convolution characteristic pattern, specifically includes:
Convolution operation is carried out to original image using the convolution kernel of predefined parameter, generates convolved image;
Convolution operation is carried out to convolved image using the convolution kernel of predefined parameter, generates convolution characteristic pattern, convolutional Neural net
Network includes the convolution kernel of predefined parameter.
Further, original image is divided into using image partition method by different regions, specifically included:
Calculate the color difference in adjacent subarea domain in original image;
By position is adjacent and color difference is no more than the subregion of preset color threshold and merges, the field color after merging is two
Sub-regions press pixel quantity weighted average, traverse all subregions, until all adjacent subarea domains color difference be all larger than it is pre-
If color threshold, wherein color difference be three channel strength difference absolute values of RGB sum;
Statistics each subregion is distributed in most value horizontal, on ordinate, generates rectangular shaped rim, as current segmentation result,
It is exported;
Preset color threshold is adjusted, repetition is compared, until the region quantity in image is one, and is merged identical
Output is as a result, then obtain the final result of image segmentation.
Further, the provincial characteristics figure for examining each region, specifically includes:
Using PCA algorithm, dimensionality reduction is carried out to the provincial characteristics figure in each region, obtains feature vector, each feature vector
The position of middle characteristic value is determined according to the separating capacity power of this feature value;
A characteristic value in feature vector is successively taken from front to back, and is tested according to the classification thresholds of the dimension: if
The test fails, then determines the region without aircraft;
If all characteristic values pass through inspection in feature vector, determine that the region is aircraft.
As shown from the above technical solution, aircraft brake disc detection method provided in this embodiment, it is raw by convolutional neural networks
At convolution characteristic pattern, same classification is excluded so that this method has better separating capacity relative to traditional feature describing mode
Mark interference, helps to improve the accuracy of target acquisition.Meanwhile this method divides original image using image partition method
It cuts, relative to traditional sliding window way of search, search time is greatly shortened.Also, this method to each provincial characteristics figure into
Performing check, and then judge the region with the presence or absence of aircraft, effectively prevents that background complexity, movement velocity, object variations are big etc. to be done
Disturb influence of the factor to target acquisition.
Therefore, the present embodiment aircraft brake disc detection method can be improved the reliability of target acquisition during takeoff and landing
And real-time, similar target jamming is avoided, algorithm execution efficiency is improved.
Second aspect, the present invention provide a kind of aircraft brake disc method for tracing, this method comprises:
Obtain the empirical weight of mission phase;
According to the empirical weight of mission phase, the detection result of the detector in a preceding adjacent flight stage is calculated respectively
Likelihood probability, the three-dimensional space position in detection result are the positions according to the region conversion for being determined as aircraft in advance;
According to the calculated result of likelihood probability, mission phase locating for current aircraft is judged, and switch the work of corresponding detector
Make state, each detector and each mission phase correspond.
Further, according to the empirical weight of mission phase, respectively to the spy of the detector in a preceding adjacent flight stage
It surveys result and calculates likelihood probability, specifically include:
According to the empirical weight of mission phase, using likelihood probability formula, to a preceding SiThe detection of the detector in stage
As a result it is calculated:
b(Oi(α))=ti_iexp(-|A'-α|)
Wherein, OiIndicate SiThe detection result of the detector in stage, α indicate the center of search coverage, ti_iExpression experience
S is kept in weightiThe probability of stage condition, A' indicate the anticipation position of current aircraft;
According to the empirical weight of mission phase, using likelihood probability formula, to a preceding Si+1The detection of the detector in stage
As a result it is calculated:
b(Oi+1(α))=ti_i+1exp(-|A'-α|)
Wherein, Oi+1Indicate Si+1The detection result of the detector in stage, α indicate the center of search coverage, ti_i+1It indicates
From S in empirical weightiStage condition transfer is Si+1The probability of stage condition, A' indicate the anticipation position of current aircraft;
According to the calculated result of likelihood probability, judges mission phase locating for current aircraft, specifically includes:
Determine the maximum probing result of likelihood probability;
By the stage locating for the maximum detection result of likelihood probability, as the stage locating for current aircraft.
Further, the empirical weight for obtaining mission phase, specifically includes:
Acquire the sample data of takeoff and landing process;
Statistics calculating is carried out to sample data, obtains the empirical weight of mission phase.
As shown from the above technical solution, aircraft brake disc method for tracing provided in this embodiment, can be according to the warp of pre-acquiring
Weight is tested, the likelihood probability of the detection result of a preceding adjacent phases is calculated, mission phase locating for aircraft is determined, in order to cut
The working condition of Current detector is changed, all detector concurrent workings are not used, only uses a detector as far as possible, improves algorithm
Operational efficiency.
Therefore, the present embodiment aircraft brake disc method for tracing can be improved the reliability of target tracking during takeoff and landing
And real-time, improve algorithm execution efficiency.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art are briefly described.In all the appended drawings, similar element
Or part is generally identified by similar appended drawing reference.In attached drawing, each element or part might not be drawn according to actual ratio.
Fig. 1 shows a kind of method flow diagram of aircraft brake disc detection method provided by the present invention;
Fig. 2 shows a kind of method flow diagrams of aircraft brake disc method for tracing provided by the present invention;
Fig. 3 shows the relation schematic diagram of three dimensional space coordinate system provided by the present invention;
Fig. 4 shows the state transition diagram in stage locating for a kind of aircraft provided by the present invention.
Specific embodiment
It is described in detail below in conjunction with embodiment of the attached drawing to technical solution of the present invention.Following embodiment is only used for
Clearly illustrate technical solution of the present invention, therefore be intended only as example, and cannot be used as a limitation and limit protection of the invention
Range.
It should be noted that unless otherwise indicated, technical term or scientific term used in this application should be this hair
The ordinary meaning that bright one of ordinary skill in the art are understood.
In a first aspect, a kind of aircraft brake disc detection method provided by the embodiment of the present invention, in conjunction with Fig. 1, this method comprises:
Original image is inputted convolutional neural networks by step S11, generates convolution characteristic pattern, and convolution characteristic pattern includes each
The N-dimensional feature vector of pixel such as generates the feature vector of one 64 dimension in each pixel.Wherein, the parameter of convolutional neural networks
Be as mass data training obtained by, it is consistent to all mission phases.
Original image is divided into different areas using image partition method according to the continuity of picture material by step S12
Domain.
Step S13 to the convolution characteristic pattern in each extracted region region being partitioned into, and is standardized, obtains
Take the corresponding provincial characteristics figure in each region.For example, feature extraction is carried out to each region after segmentation, the characteristic pattern of extraction
Having a size of (W, H, 64), this feature figure is then zoomed into standard size (30,10,64).
Step S14 examines the provincial characteristics figure in each region, according to the corresponding region of characteristic pattern for being determined as aircraft, meter
It calculates and exports aircraft region.
As shown from the above technical solution, aircraft brake disc detection method provided in this embodiment, it is raw by convolutional neural networks
At convolution characteristic pattern, same classification is excluded so that this method has better separating capacity relative to traditional feature describing mode
Mark interference, helps to improve the accuracy of target acquisition.Meanwhile this method divides original image using image partition method
It cuts, relative to traditional sliding window way of search, search time is greatly shortened.Also, this method to each provincial characteristics figure into
Performing check, and then judge the region with the presence or absence of aircraft, effectively prevents that background complexity, movement velocity, object variations are big etc. to be done
Disturb influence of the factor to target acquisition.
Therefore, the present embodiment aircraft brake disc detection method can be improved the reliability of target acquisition during takeoff and landing
And real-time, similar target jamming is avoided, algorithm execution efficiency is improved.
In order to further increase the accuracy of the present embodiment aircraft brake disc detection method, specifically, in terms of convolutional calculation,
Original image is inputted into convolutional neural networks, when generating convolution characteristic pattern, this method the specific implementation process is as follows:
Convolution operation is carried out to original image using the convolution kernel of predefined parameter, generates convolved image.
Convolution operation is carried out to convolved image using the convolution kernel of predefined parameter, generates convolution characteristic pattern, convolutional Neural net
Network includes the convolution kernel of predefined parameter.
For example, the present embodiment aircraft brake disc detection method is handled using convolutional neural networks model, convolutional Neural net
Network haves three layers altogether, and each layer parameter and calculating process are as follows:
First layer carries out convolution operation to original image by 47 × 7 convolution kernels, generates 4 convolved images.
The second layer carries out convolution operation to 4 convolved images of first layer output respectively by 4 13 × 13 convolution kernels,
Generate 16 convolved images.
Third layer carries out convolution behaviour to 16 convolved images of second layer output respectively by 4 25 × 25 convolution kernels
Make, 64 convolved images is generated, as convolution characteristic pattern.
By above-mentioned convolution operation, the feature vector of 64 dimensions can be generated to each pixel.Also, in convolutional neural networks
The parameter of 12 convolution kernels be as mass data training obtained by, it is consistent to all mission phases.
Here, the present embodiment aircraft brake disc detection method uses convolutional Neural net relative to traditional character description method
Network handles original image, can be realized better separating capacity, reduces the interference of noise or complex background.
Specifically, in terms of image dividing processing, original image is divided into using image partition method by different regions
When, this method the specific implementation process is as follows:
Calculate the color difference in adjacent subarea domain in original image.Here, each pixel in original image is considered as a son
The color in region, each subregion is the color of the pixel.
When initial, color threshold T=12 can be set.
By the subregion merging that position is adjacent and color difference is no more than preset color threshold T, the field color after merging is
Two sub-regions press the weighted average of pixel quantity, traverse all subregions, until the color difference in all adjacent subarea domains is all larger than
Preset color threshold, wherein color difference is the sum of three channel strength difference absolute values of RGB.
Statistics each subregion is distributed in most value horizontal, on ordinate, the horizontal seat of minimum of the subregion after merging such as statistics
Mark, minimum ordinate, maximum abscissa and maximum ordinate, generate rectangular shaped rim and are exported as current segmentation result.
Preset color threshold is adjusted, such as by color threshold multiplied by 2, repetition is compared, until the region quantity in image is
One, and merge identical output as a result, can then obtain the final result of image segmentation.
Here, the present embodiment aircraft brake disc detection method is according between adjacent pixel relative to traditional sliding window method
Color difference merges processing, determines cut zone, increases substantially image segmentation efficiency, shortens the image segmentation time.Also, this
Embodiment aircraft brake disc detection method can be split original image by the way of successive ignition, and will be identical
Region help to improve the accuracy of image segmentation as the region finally divided, to judge that region locating for aircraft is provided with
The Informational support of effect.
Meanwhile calculate original image in adjacent subarea domain color difference when, this method the specific implementation process is as follows: count respectively
Calculate intensity difference of the adjacent subarea domain on three channels in original image.
It takes absolute value, and sums up to the intensity difference on three channels.Such as to the intensity on three channels of red, yellow, and green
Difference takes absolute value.
Using the numerical value after adduction as the color difference in adjacent subarea domain.
Here, the present embodiment aircraft brake disc detection method can determine color difference according to the intensity difference on different channels, improve
The accuracy of Colorimetry.
Specifically, in terms of flight range judgement, when examining the provincial characteristics figure in each region, the specific implementation of this method
Process is as follows:
Using PCA algorithm, dimensionality reduction is carried out to the provincial characteristics figure in each region, 10 dimensions is such as reduced to, obtains feature
Vector, the position of characteristic value is determined according to the separating capacity power of this feature value in each feature vector.
A characteristic value in feature vector is successively taken from front to back, and is tested according to the classification thresholds of the dimension: if
The test fails, then determines the region without aircraft.
If all characteristic values pass through inspection in feature vector, determine that the region is aircraft.
Here, plane prevention algorithm is cascaded by several weak typing algorithms.In i-stage classification, dimensionality reduction is extracted
I-th of dimension of characteristic pattern is excluded non-aircraft target by comparing with the classification thresholds of this grade step by step.
The affiliated image-region of the characteristic pattern that every level testing is passed through, is determined as the region of aircraft.Pass through 10 cascade classification
It after exclusion, is still retained, then it is assumed that the region is aircraft.
Wherein, every grade of classification thresholds in the dimensionality reduction transition matrix and tandem type classifier of PCA algorithm, are by each
What the positive and negative aircraft sample training of mission phase obtained.Also, the better characteristic of division of elimination ability, that is, separating capacity is more
The feature that good subregion generates, more preferably uses in cascade process.In this way, most of non-aircraft target can be by comparing
What few classification is excluded.
Here, the present embodiment aircraft brake disc detection method, replaces neural-network classification method using tandem type classifier, mentions
The high execution speed of algorithm, improves the accuracy for judging region locating for aircraft.
Second aspect, a kind of aircraft brake disc method for tracing provided by the embodiment of the present invention, in conjunction with Fig. 2, this method comprises:
Step S21 obtains the empirical weight of mission phase.
Step S22, according to the empirical weight of mission phase, respectively to the detection of the detector in a preceding adjacent flight stage
As a result likelihood probability is calculated, the three-dimensional space position in detection result is the position according to the region conversion for being determined as aircraft in advance
It sets.Here, can be converted to region locating for the aircraft determined in above-mentioned aircraft brake disc detection method, to determine detection knot
Coordinate in fruit.
Step S23 judges mission phase locating for current aircraft according to the calculated result of likelihood probability, and switches corresponding visit
The working condition of device is surveyed, each detector and each mission phase correspond.
Here, the takeoff and landing of aircraft has totally been divided into five stages in advance, it is respectively as follows: downslide, puts down and float, slide
It runs, is liftoff, rising, the performance for supporting total algorithm can be reached.One aircraft detector of each stage-training.Each detection
The course of work of device is aircraft brake disc method for tracing.Five aircraft detectors are only in the area parameter Shang You of tandem type classifier
Not, other parts are consistent.
In landing tracing process, five aircraft detectors are used alternatingly, and specific alternately rule is as follows:
First, descent starts since first stage detector;
Second, take-off process starts since phase III detector;
Third can only be kept using current detector, or be switched to the detector of next stage, and cannot jump rank
Section, can not switch forward.Here, this method can calculate the likelihood probability of the detection result of adjacent phases detector, determine
Stage locating for current aircraft, and switch the detector of respective stage.
Wherein, the three-dimensional space position in detection result is according to the position for the region conversion for being determined as aircraft in advance, tool
Body realizes that process is as follows:
Using runway centerline as X-axis, vertical direction is Y-axis, and vertical runway centerline direction is Z axis in runway plane, is run
Road starting point is origin, establishes three-dimensional coordinate system, as shown in Figure 3.Surveillance camera is mounted on runway side, is denoted as P (pX,
pY,pZ) point, aircraft location is denoted as A point, and the unit vector for being directed toward A point from P point is denoted as V (vX,vY,vZ)。
After video camera installation is fixed, measurement obtains P point coordinate.Video camera obtains current be horizontally directed in real time in operation
Angle and pitching orientation angle.After detecting aircraft in the picture, the pixel at aircraft region center and picture centre is calculated
Deviation.Misalignment angle is calculated in conjunction with CCD component pixel size and focal length value, using misalignment angle to being horizontally directed to angle and bow
It faces upward orientation angle to be modified, the direction of V is calculated by modified goniometer.
During normal takeoff and landing, flight path is in X/Y plane.Therefore, in the direction of known P point coordinate and V
In the case where, by following formula, calculate the intersection point of V extended line and X/Y plane, it is believed that be the space that aircraft is currently located
Position A.
Whether aircraft is moved in X/Y plane, can be determined roughly by the size that aircraft is imaged.Positive reason
Under condition, distance of the P point along the direction V to X/Y plane can be calculated directly.In conjunction with CCD component pixel size and focal length value, Yi Jiyi
The aircraft size known can calculate expected imaging size.If imaging area is significantly greater than or is significantly less than desired value,
Then think that aircraft deviates X/Y plane.
As shown from the above technical solution, aircraft brake disc method for tracing provided in this embodiment, can be according to the warp of pre-acquiring
Weight is tested, the likelihood probability of the detection result of a preceding adjacent phases is calculated, mission phase locating for aircraft is determined, in order to cut
The working condition of Current detector is changed, all detector concurrent workings are not used, only uses a detector as far as possible, improves algorithm
Operational efficiency.
Therefore, the present embodiment aircraft brake disc method for tracing can be improved the reliability of target tracking during takeoff and landing
And real-time, improve algorithm execution efficiency.
Also, according to the empirical weight of mission phase, respectively to the detection knot of the detector in a preceding adjacent flight stage
When fruit calculates likelihood probability, the specific implementation process of this method are as follows:
According to the empirical weight of mission phase, using likelihood probability formula, to a preceding SiThe detection of the detector in stage
As a result it is calculated:
b(Oi(α))=ti_iexp(-|A'-α|)
Wherein, OiIndicate SiThe detection result of the detector in stage, α indicate the center of search coverage, ti_iExpression experience
S is kept in weightiThe probability of stage condition, A' indicate the anticipation position of current aircraft.
According to the empirical weight of mission phase, using likelihood probability formula, to a preceding Si+1The detection of the detector in stage
As a result it is calculated:
b(Oi+1(α))=ti_i+1exp(-|A'-α|)
Wherein, Oi+1Indicate Si+1The detection result of the detector in stage, α indicate the center of search coverage, ti_i+1It indicates
From S in empirical weightiStage condition transfer is Si+1The probability of stage condition, A' indicate the anticipation position of current aircraft.
According to the calculated result of likelihood probability, when judging mission phase locating for current aircraft, the specific implementation of this method
Journey are as follows: determine the maximum probing result of likelihood probability.
By the stage locating for the maximum detection result of likelihood probability, as the stage locating for current aircraft.
Here, the present embodiment aircraft brake disc method for tracing, can count the state transition probability in stage, adjacent phases are calculated
The likelihood probability of detection result, accurately to calculate mission phase locating for current aircraft according to previous detection result, so as to
In the working condition of switching detector.
When obtaining the empirical weight of mission phase, the specific implementation process of this method are as follows:
Acquire the sample data of takeoff and landing process.
Statistics calculating is carried out to sample data, obtains the empirical weight of mission phase, such as keeps SiThe probability of mission phase
ti_i, or from SiMission phase transfer is Si+1The probability t of mission phasei_i+1。
In actual application, aircraft position and state in statistical sample data, determine the empirical weight of mission phase.
For example, time of day when once detecting before enabling aircraft is S (A, L), wherein A is the three-dimensional coordinate position of aircraft,
L is the three-dimensional velocity vector of aircraft;
The anticipation state for enabling current aircraft is S'(A', L'), wherein A'=A+L, L'=L;
The time of day for enabling current aircraft is S " (A ", L ").
With stage Si、Si+1For, transition probability parameters between adjacent phases as shown in figure 4,
Wherein, ti_i+ti_i+1=1, ti_i=fi(A), ti+1_i=0, ti_iIt indicates to keep SiThe probability in stage, ti_i+1Table
Show from SiPhase transition is Si+1The probability in stage, ti+1_i+1It indicates to keep Si+1The probability in stage, ti+1_iIt indicates from the
Si+1Phase transition is SiThe probability in stage, bi_iIt indicates to keep SiThe likelihood probability in stage, bi_i+1It indicates from SiStage turns
It is changed to Si+1The likelihood probability in stage, bi+1_i+1It indicates to keep Si+1The likelihood probability in stage, bi+1_iIt indicates from Si+1Stage
Be converted to SiThe likelihood probability in stage, function f () indicate table lookup operation, can be obtained herein according to aircraft position and state
Corresponding transition probability, later state conversion and so on.
Here, the present embodiment aircraft brake disc method for tracing, acquires the sample data of takeoff and landing process, to sample number in advance
According to statistics calculating is carried out, according to aircraft position and state, and then determines the state transition probability of aircraft, determine the warp of mission phase
Weight is tested, provides effective Informational support for the stage locating for subsequent judgement aircraft.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
It should be noted that the flow chart and block diagram in the drawings show the services of multiple embodiments according to the present invention
The architecture, function and operation in the cards of device, method and computer program product.In this regard, flowchart or block diagram
In each box can represent a part of a module, section or code, one of the module, section or code
Subpackage is containing one or more executable instructions for implementing the specified logical function.It should also be noted that at some as replacement
Realization in, function marked in the box can also occur in a different order than that indicated in the drawings.For example, two continuous
Box can actually be basically executed in parallel, they can also be executed in the opposite order sometimes, this is according to related function
Depending on energy.It is also noted that each box in block diagram and or flow chart and the box in block diagram and or flow chart
Combination, can the dedicated hardware based server of as defined in executing function or movement realize, or can be with dedicated
The combination of hardware and computer instruction is realized.
Configuration device provided by the embodiment of the present invention can be computer program product, including storing program code
Computer readable storage medium, the instruction that said program code includes can be used for executing previous methods side as described in the examples
Method, specific implementation can be found in embodiment of the method, and details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the service of foregoing description
The specific work process of device, device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided herein, it should be understood that disclosed server, device and method, it can
To realize by another way.The apparatus embodiments described above are merely exemplary, for example, the unit is drawn
Point, only a kind of logical function partition, there may be another division manner in actual implementation, in another example, multiple units or group
Part may be combined or can be integrated into another server, or some features can be ignored or not executed.Another point is shown
The mutual coupling, direct-coupling or communication connection shown or discussed can be through some communication interfaces, device or unit
Indirect coupling or communication connection, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or can also be published to multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic or disk.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations;To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement;And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme should all cover within the scope of the claims and the description of the invention.
Claims (6)
1. a kind of aircraft brake disc detection method characterized by comprising
Original image is inputted into convolutional neural networks, generates convolution characteristic pattern, the convolution characteristic pattern includes the N-dimensional of each pixel
Feature vector;
The original image is divided into using image partition method by different regions;
It to the convolution characteristic pattern in each extracted region region being partitioned into, and is standardized, obtains each region pair
The provincial characteristics figure answered;
The provincial characteristics figure for examining each region calculates according to the corresponding region of characteristic pattern for being determined as aircraft and exports aircraft
Region;
The original image is divided into using image partition method by different regions, is specifically included:
Calculate the color difference in adjacent subarea domain in the original image;
By position is adjacent and color difference is no more than the subregion of preset color threshold and merges, the field color after merging is two sons
Region press pixel quantity weighted average, traverse all subregions, until all adjacent subarea domains color difference be all larger than it is preset
Color threshold, wherein color difference is the sum of three channel strength difference absolute values of RGB;
Statistics each subregion is distributed in most value horizontal, on ordinate, generates rectangular shaped rim, as current segmentation result, carries out
Output;
Preset color threshold is adjusted, repetition is compared, until the region quantity in image is one, and merges identical output
As a result, then obtaining the final result of image segmentation.
2. aircraft brake disc detection method according to claim 1, which is characterized in that
Original image is inputted into convolutional neural networks, convolution characteristic pattern is generated, specifically includes:
Convolution operation is carried out to the original image using the convolution kernel of predefined parameter, generates convolved image;
Convolution operation is carried out to the convolved image using the convolution kernel of predefined parameter, generates the convolution characteristic pattern, the volume
Product neural network includes the convolution kernel of the predefined parameter.
3. aircraft brake disc detection method according to claim 1, which is characterized in that
The provincial characteristics figure for examining each region, specifically includes:
Using PCA algorithm, dimensionality reduction is carried out to the provincial characteristics figure in each region, obtains feature vector, spy in each feature vector
The position of value indicative is determined according to the separating capacity power of this feature value;
A characteristic value in feature vector is successively taken from front to back, and is examined according to the classification thresholds of the dimension of this feature vector
It tests: if the test fails, determining the region without aircraft;
If all characteristic values pass through inspection in feature vector, determine that the region is aircraft.
4. a kind of aircraft brake disc method for tracing, which is characterized in that
Obtain the empirical weight of mission phase;
According to the empirical weight of mission phase, likelihood is calculated to the detection result of the detector in a preceding adjacent flight stage respectively
Probability, the three-dimensional space position in the detection result are the positions according to the region conversion for being determined as aircraft in advance;
According to the calculated result of likelihood probability, mission phase locating for current aircraft is judged, and switch the work shape of corresponding detector
State, each detector and each mission phase correspond.
5. aircraft brake disc method for tracing according to claim 4, which is characterized in that
According to the empirical weight of mission phase, likelihood is calculated to the detection result of the detector in a preceding adjacent flight stage respectively
Probability specifically includes:
According to the empirical weight of mission phase, using likelihood probability formula, to a preceding SiThe detection result of the detector in stage into
Row calculates:
b(Oi(α))=ti_iexp(-|A′-α|)
Wherein, OiIndicate SiThe detection result of the detector in stage, α indicate the center of search coverage, ti_iIndicate empirical weight
Middle holding SiThe probability of stage condition, A' indicate the anticipation position of current aircraft;
According to the empirical weight of mission phase, using likelihood probability formula, to a preceding Si+1The detection result of the detector in stage
It is calculated:
b(Oi+1(α))=ti_i+1exp(-|A′-α|)
Wherein, Oi+1Indicate Si+1The detection result of the detector in stage, α indicate the center of search coverage, ti_i+1Expression experience
From S in weightiStage condition transfer is Si+1The probability of stage condition, A' indicate the anticipation position of current aircraft;
According to the calculated result of likelihood probability, judges mission phase locating for current aircraft, specifically includes:
Determine the maximum probing result of likelihood probability;
By the stage locating for the maximum detection result of likelihood probability, as the stage locating for current aircraft.
6. aircraft brake disc method for tracing according to claim 4 or 5, which is characterized in that
The empirical weight for obtaining mission phase, specifically includes:
Acquire the sample data of takeoff and landing process;
Statistics calculating is carried out to the sample data, obtains the empirical weight of mission phase.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710724698.5A CN107507198B (en) | 2017-08-22 | 2017-08-22 | Aircraft brake disc detection and method for tracing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710724698.5A CN107507198B (en) | 2017-08-22 | 2017-08-22 | Aircraft brake disc detection and method for tracing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107507198A CN107507198A (en) | 2017-12-22 |
CN107507198B true CN107507198B (en) | 2019-11-19 |
Family
ID=60691419
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710724698.5A Active CN107507198B (en) | 2017-08-22 | 2017-08-22 | Aircraft brake disc detection and method for tracing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107507198B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108520527A (en) * | 2018-03-28 | 2018-09-11 | 东北大学 | A kind of space-time context fast track method based on color attribute and PCA |
CN110211159A (en) * | 2019-06-06 | 2019-09-06 | 中国民航科学技术研究院 | A kind of aircraft position detection system and method based on image/video processing technique |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102142078A (en) * | 2010-02-03 | 2011-08-03 | 中国科学院自动化研究所 | Method for detecting and identifying targets based on component structure model |
CN105512684A (en) * | 2015-12-09 | 2016-04-20 | 江苏大为科技股份有限公司 | Vehicle logo automatic identification method based on principal component analysis convolutional neural network |
CN105989336A (en) * | 2015-02-13 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Scene recognition method based on deconvolution deep network learning with weight |
CN106295695A (en) * | 2016-08-08 | 2017-01-04 | 中国民用航空总局第二研究所 | A kind of takeoff and landing process automatic tracing image pickup method and device |
-
2017
- 2017-08-22 CN CN201710724698.5A patent/CN107507198B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102142078A (en) * | 2010-02-03 | 2011-08-03 | 中国科学院自动化研究所 | Method for detecting and identifying targets based on component structure model |
CN105989336A (en) * | 2015-02-13 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Scene recognition method based on deconvolution deep network learning with weight |
CN105512684A (en) * | 2015-12-09 | 2016-04-20 | 江苏大为科技股份有限公司 | Vehicle logo automatic identification method based on principal component analysis convolutional neural network |
CN106295695A (en) * | 2016-08-08 | 2017-01-04 | 中国民用航空总局第二研究所 | A kind of takeoff and landing process automatic tracing image pickup method and device |
Non-Patent Citations (6)
Title |
---|
Adaptive Visual Tracking with Minimum Uncertainty Gap Estimation;Junseok Kwon 等;《IEEE Transactions on Pattern Analysis and Machine Intelligence》;20160302;第39卷(第1期);18-31 * |
Cascade Convolutional Neural Network Based on Transfer-Learning for Aircraft Detection on High-Resolution Remote Sensing Images;Bin Pan 等;《Journal of Sensors》;20170727;第2017卷;1-14 * |
Rich feature hierarchies for accurate object detection and semantic segmentation;Ross Girshick 等;《Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition》;20141231;摘要,第2.1节,附图1 * |
Weakly Supervised Learning Based on Coupled Convolutional Neural Networks for Aircraft Detection;Fan Zhang 等;《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》;20160930;第54卷(第9期);5553-5563 * |
基于卷积神经网络的高分辨率SAR图像飞机目标检测方法;王思雨 等;《雷达学报》;20170430;第6卷(第2期);195-203 * |
极大似然法在飞机起飞性能参数辨识中的应用;于雪梅 等;《飞行力学》;20120630;第30卷(第3期);201-205 * |
Also Published As
Publication number | Publication date |
---|---|
CN107507198A (en) | 2017-12-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108986064B (en) | People flow statistical method, equipment and system | |
US10275709B2 (en) | Method and system for tracking an object | |
JP6943338B2 (en) | Image processing equipment, systems, methods and programs | |
CN105678811B (en) | A kind of human body anomaly detection method based on motion detection | |
CN106845364B (en) | Rapid automatic target detection method | |
WO2019096952A1 (en) | A system and method for single image object density estimation | |
CN104378582A (en) | Intelligent video analysis system and method based on PTZ video camera cruising | |
CN108957453A (en) | A kind of high-precision pre-filter method and recognition methods based on multiple target tracking | |
CN109918971A (en) | Number detection method and device in monitor video | |
CN113538347B (en) | Image detection method and system based on efficient bidirectional path aggregation attention network | |
KR102199252B1 (en) | Method and Apparatus for Analyzing Traffic Situation | |
CN107507198B (en) | Aircraft brake disc detection and method for tracing | |
Varior et al. | Multi-scale attention network for crowd counting | |
GB2585933A (en) | System and method for processing images | |
KR20100093797A (en) | Method and system for vision-based people counting in cctv | |
Manninen et al. | Multi-stage deep learning networks for automated assessment of electricity transmission infrastructure using fly-by images | |
CN110532853A (en) | The classification method and device of remote sensing time-out phase data | |
Yao et al. | Substation object detection based on enhance RCNN model | |
Bolten et al. | Evaluation of Deep Learning based 3D-Point-Cloud Processing Techniques for Semantic Segmentation of Neuromorphic Vision Sensor Event-streams. | |
Tienin et al. | Comparative ship classification in heterogeneous dataset with pre-trained models | |
RU2698157C1 (en) | System for searching for violations in order of location of objects | |
CN107730535A (en) | A kind of cascaded infrared video tracing method of visible ray | |
Mei et al. | Supervised learning for semantic segmentation of 3D LiDAR data | |
Qiu et al. | Yolo-based detection technology for aerial infrared targets | |
Cani et al. | Illicit object detection in X-ray images using Vision Transformers |
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 |