CN104331910B - A kind of track obstacle detecting system based on machine vision - Google Patents
A kind of track obstacle detecting system based on machine vision Download PDFInfo
- Publication number
- CN104331910B CN104331910B CN201410681599.XA CN201410681599A CN104331910B CN 104331910 B CN104331910 B CN 104331910B CN 201410681599 A CN201410681599 A CN 201410681599A CN 104331910 B CN104331910 B CN 104331910B
- Authority
- CN
- China
- Prior art keywords
- image
- track
- value
- pixel
- barrier
- 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.)
- Expired - Fee Related
Links
- 238000001514 detection method Methods 0.000 claims abstract description 54
- 238000000034 method Methods 0.000 claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 22
- 230000003068 static effect Effects 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 230000008569 process Effects 0.000 claims abstract description 12
- 239000000284 extract Substances 0.000 claims abstract description 11
- 238000006243 chemical reaction Methods 0.000 claims abstract description 5
- 230000004888 barrier function Effects 0.000 claims description 39
- 230000033001 locomotion Effects 0.000 claims description 19
- 230000009545 invasion Effects 0.000 claims description 9
- 230000009021 linear effect Effects 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 8
- 238000004088 simulation Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 6
- 238000005314 correlation function Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 5
- 230000003287 optical effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 229910003460 diamond Inorganic materials 0.000 claims description 3
- 239000010432 diamond Substances 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims 1
- 238000012372 quality testing Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 239000000523 sample Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 239000012491 analyte Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 208000010877 cognitive disease Diseases 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 230000003137 locomotive effect Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
- B61L23/00—Control, warning or like safety means along the route or between vehicles or trains
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
A kind of track obstacle detecting system based on machine vision, belongs to railway security field.Black and white video image is obtained by the vehicle-mounted vidicon installed in train head, whether the excess of stroke judges that track is class straight way or bend according to video camera.Bend receives wireless image and realizes manual detection bend situation using fixed high-definition camera and wireless launcher.The real-time image sequences that class straight way will get are analyzed, using the rail flanges extraction algorithm based on top cap conversion and Otsu threshold values, the frame differential method improved based on mathematical morphology is combined and Corner Feature Matching pursuitalgorithm with background subtraction, and preceding object thing is divided into without dangerous (static the small-scale obstacle thing, the quick moving obstacle across track) and dangerous (moving obstacle that static large obstacle, influence train pass through) two classes in real time.The present invention can quickly process orbital image, more effectively can more accurately extract rail flanges, also higher to track obstacle accuracy in detection.
Description
Technical field
The invention belongs to railway security field, and in particular to a kind of track obstacle detecting system based on machine vision.
Background technology
In recent years, with train raising speed by a large scale, the change of train operation pattern, passenger-cargo freight volume is significantly improved, to iron
The safety and reliability of road transport proposes requirement higher.Although foreign countries have had in terms of track obstacle research at present
The product of comparative maturity, but the design concept of most products is all to send some form of signal by interception to be checked
(mainly including laser, radar, magnetic induction, ultrasonic wave etc.), and the signal for detecting and reflecting by sensor is analyzed, as
Judge the foundation of cognitive disorders thing.In these detection methods, such as railway track detection is carried out using ultrasonic wave, can be than calibrated
Really detect the position for identifying target obstacle, but still exist larger for volume and with the target disorders of certain altitude
Analyte detection effect is preferable, it may appear that the problem of the flat small barrier of missing inspection;Laser and detections of radar have spatial coverage limited with
And resolution ratio shortcoming not high.Meanwhile, such method belongs to the detection of infringement formula, inevitably increases ambient noise, and
Interference can be also produced between sensor.These shortcoming and defect, none does not interfere with barrier and accurately and effectively detects and recognize.
China's patent of invention, " locomotive ground signal and the barrier automatic identification system of Publication No. CN201825066U
System ", installs electronic image identifying system, the figure on analysis operation forward box on proposition car body.Judged by system identification
Colour-light signals, switch location and track switch open and-shut mode on figure, blind siding soil shelves and standing car remind driver or control machine
Car Braking mode occurs preventing accident.But said system also has good effect just for the detection of indivedual predetermined substances
Really, in the field of the pedestrian, falling rocks and vehicle of automatic detection identification train operation ahead, its effect cannot still be played.
The content of the invention
The present invention is in view of the shortcomings of the prior art and shortcoming, there is provided a kind of track obstacle detection system based on machine vision
System, it can quickly process orbital image, and rail flanges extract more accurate, higher to track obstacle accuracy in detection.
Technical scheme proposed by the present invention is a kind of track obstacle detecting system based on machine vision, by being arranged on
The vehicle-mounted vidicon of train head obtains black and white video image, and according to car, in video camera, whether the excess of stroke judges that track is class straight way or curved
Road.Bend receives wireless image and realizes manual detection bend situation using fixed high-definition camera and wireless launcher.Class is straight
The real-time image sequences that road will get are analyzed, and are extracted using the rail flanges based on top cap conversion and Otsu threshold values and calculated
Method, the frame differential method improved based on mathematical morphology is combined and Corner Feature Matching pursuitalgorithm with background subtraction,
Preceding object thing is divided into without dangerous (static the small-scale obstacle thing, the quick moving obstacle across track) and dangerous (quiet in real time
The moving obstacle that only large obstacle, influence train pass through) two classes.The rear result for testing and analyzing identification is notified to department and multiplies people
Member, reaches the purpose for being prevented effectively from that erroneous judgement or traffic accident occur.Comprise the following steps that:
1st, realtime graphic is gathered
Using monocular-camera mode, optical anti-vibration, super remote focal length B/W camera, in train traveling process in real time
Collection image sequence.
2nd, classification of track is judged
When the vehicle-mounted vidicon excess of stroke, it is judged as bend (wide-angle turning track, right angle track);Then using fixed high definition
Video camera and wireless launcher receive wireless image and realize manual detection bend track condition;
If vehicle-mounted vidicon does not have the excess of stroke, it is judged to class straight way (rectilinear orbit, small radian turning track), continues to adopt
Realtime graphic is obtained with vehicle-mounted vidicon, image is processed into next step.
3rd, image preprocessing
First, enhancing treatment is done to image, top cap computing, image X is done using a diameter of 4 " diamond " structural element
Top cap computing on structural element B is designated as X ° of B, is defined asDual-rail part is highlighted, suppresses background high
Bright part;
Then, Otsu threshold calculations optimal threshold T are used to enhanced image, by image binaryzation, obtains binary map
Picture;Multiple discontinuity zones are included by the image after Threshold segmentation;
Finally, the connectedness of each pixel pixel adjacent thereto in discontinuity zone is checked, mark extracts region, by connection
Field mark retains double track characteristic straight line.
4th, detection window is set up
Double track characteristic straight line is retained according to connected component labeling, using fitting a straight line, is realized to track using least square method
Edge positioning is extracted, by two rail linears, linear equation is obtained;The position of detection window window is determined with the particular location of rail
Put, size and dimension, detection window rail portion is confined in the range of it.
5th, barrier is determined whether
The characteristics of having excluded background based on image in class straight way track rule consistency and detection window, can ignore track
Motion, when there is suspicious barrier intrusion detection window, will certainly cause characteristics of image to change, and this method chooses histogram feature.
Threshold value T is defined based on experience value1And T2For differentiating histogram variances variation delta σk 2With histogram burr number
The value changes situation of BurrNum, by two-value monochrome pixels ratio r ati o, Δ σk 2With tri- monitorings of parameter of BurrNum,
Determining whether barrier invasion may.Algorithm is as follows:
Histogram average:
(wherein rkIt is histogram feature data, pkProbability corresponding to it, L is characterized the number of data)
Histogram variances:
Two-value monochrome pixels ratio:
(nbIt is the quantity of black pixel, nkIt is the quantity of white pixel)
From the point of view of significance level and reliability, ratio nb/nkIt is more directly perceived, actual conditions can be more showed, therefore enumerate
Basis for estimation:
(1)ratio<When 4.5, no matter Δ σk 2With the value of BurrNum how, all judge there is invasion barrier;
(2)ratio>When 4.5, then Δ σ is observedk 2How value with BurrNum changes, if both exceed threshold value, judges
There is invasion barrier;
(3)ratio>When 4.5, and Δ σk 2With the value of BurrNum all without departing from threshold value or one of them exceed threshold value, then
Judge no barrier, it may be possible to the influence of the external interference such as illumination condition.
If without barrier, return starts anew;If there is barrier, next step is carried out.
6th, detection of obstacles
Stationary obstruction also has relative motion to train, so it is static going back to be differentiated by movement locus and motion feature
It is moving obstacle.
Using the frame differential method improved by mathematical morphology, calculus of differences is carried out to the image in detection window, extracted
Segmentation moving target;Frame differential method is exactly in sequence of video images, to extract the adjacent frame of two frame three or multiple image, and right
The two field picture of extraction carries out the image operation that pixel is subtracted each other.If the difference of adjacent pixel is less than the threshold value of setting, then it is assumed that should
Pixel is static background, conversely, then extracting the pixel as moving target, according to this principle, motion mesh is met by all
The pixel for marking threshold value joins together that the moving target in scene just can be extracted, and removes jamming pattern, and extraction is reached with this
Split the purpose of moving target.
Frame differential method principle is as follows:
Dk(x, y)=| Fk(x,y)-Fk-1(x,y)|
Fk(x, y) represents continuous image, D in videok(x, y) represents the differential chart obtained by two continuous frames image subtraction
Picture, then by Dk(x, y) does following treatment:
(P is given threshold, and by experiment, simulation is obtained repeatedly)
Rk(x, y) is to judge the foundation whether target moves:If 1, then target discrimination is motion;Conversely, being static.
Meanwhile, using background subtraction, will detect that the detection image in window carries out difference with background image, then use threshold value
To detect moving region.If b (x, y) is background image, image sequence f (x, y, i) is defined, wherein (x, y) sits for picture position
Mark, i is number of image frames.The gray scale that the gray value of each two field picture subtracts background is worth to error image:
Id (x, y, i)=f (x, y, i)-b (x, y)
A binaryzation difference image is can obtain by setting threshold value T:
(threshold value T is obtained by experimental simulation)
Finally, pixel is carried out to the image obtained by two kinds of algorithms and takes common factor, more accurately target is extracted, obtained
The size of target, the shape facility of target.
7th, the identification and classification of barrier
Barrier size, speed and direction discernment are carried out using Corner Feature Matching pursuitalgorithm, feature extraction is adopted
Harris Corner Detection Algorithms are used, is that a kind of point feature based on signal extracts operator, in this method, will be with adjoint point brightness contrast
(intensity contrast of pixel neighborhoods point) sufficiently large point is defined as angle point.
Pixel correlation function used is as follows:
Wherein, w is the smoothing windows for carrying out noise reduction process, and (u, v) is offset coordinates, and I is image pixel matrix, I (x, y) value
The pixel value at image midpoint (x, y).IxAnd IyRepresentative image pixel single order partial differential in the horizontal direction, in vertical direction respectively,
Ix 2And Iy 2Then it is respectively the second order Grad in both direction.Then figure can just be detected by calculating angle point receptance function
Angle point as in:
R=det (M)-k*tr2(M)
Wherein, tr (M) and det (M) represent the mark and determinant of matrix M respectively, and Harris Corner Detection Algorithms recommend k
Take 0.04.
The characteristic matching Feature Points Matching algorithm related using gray scale is based on, by taking the extraction of three frames as an example:
Firstly, for each characteristic point pi∈E1, pj∈E3, respectively centered on characteristic point, construct N × N size
Window, W is designated as respectivelyi,Wj, the correlation function of each window is then calculated respectively
Can be according to CcorThe value of (i, j) is the correlation that can determine that individual features point, and the value is bigger, and corresponding characteristic point is adjacent
Domain gray scale closer to.During match point is looked for, matched by the way of mutually matching, i.e., as the characteristic point pi in E1
Optimal match point be the p in E3i, while pjOptimal match point also be piWhen, it is considered as (pipj) it is a pair of best match spy
Levy a little.Final corners Matching result is obtained accordingly.
The type of preceding object thing is judged with this, is without dangerous (static the small-scale obstacle thing, the quick motion barrier across track
Hinder thing) or it is dangerous (moving obstacle that static large obstacle, influence train pass through).
The present invention has following beneficial effects compared with existing system:
Because algorithm is relatively easy, complexity reduction, process time is shortened, can quickly process orbital image;It is based on
Top cap is converted and the rail flanges extraction algorithm of Otsu threshold values more effectively can more accurately extract rail flanges;Using based on form
The mixed method of conversion carries out target detection, effectively improves frame difference method adjoint big because of the limitation chosen in time interval
The shortcoming of amount noise and breakpoint, plays the role of denoising and smoothed profile to frame difference method result, overcomes background subtraction sentencing
Disconnected object by it is quiet to dynamic being moved when, there is " ghost " defect, optimize the effect of moving object detection;So to track
The detection of obstacles degree of accuracy is also higher.
Brief description of the drawings
In order to illustrate more clearly the embodiments of the present invention, below will be to be used needed for embodiment or description of the prior art
Accompanying drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, to ability
For the those of ordinary skill of domain, on the premise of not paying creative work, can also obtain other attached according to these accompanying drawings
Figure.
Fig. 1 is straight way of the invention, class straight way system detectio flow chart;
Fig. 2 is bend system framework flow chart of the invention;
Fig. 3 is bend monitoring model schematic diagram;
Fig. 4 is straight way, class straight way is static, moving obstacle skimulated motion track schematic diagram;
Fig. 5 frame differential methods and mathematical morphology blending algorithm block diagram;
Fig. 6 moving object detection algorithm block diagrams.
Specific embodiment
Referring to Fig. 1-Fig. 6, a kind of track obstacle detecting system based on machine vision, the system is by installed in train
The vehicle-mounted vidicon of head obtains black and white video image, judges that track is class straight way or bend.Bend is using fixed high-definition camera
Wireless image is received with wireless launcher realize manual detection bend situation.The real-time image sequences that class straight way will get enter
Row analysis, using the rail flanges extraction algorithm based on top cap conversion and Otsu threshold values, based on the interframe that mathematical morphology improves
Calculus of finite differences is combined and Corner Feature Matching pursuitalgorithm with background subtraction, and preceding object thing is divided into without danger in real time
(static large obstacle, influence train pass through for (static the small-scale obstacle thing, the quick moving obstacle across track) and danger
Moving obstacle) two classes.The rear result for testing and analyzing identification is notified to driver and conductor, is reached and is prevented effectively from erroneous judgement or traffic thing
Therefore the purpose for occurring.Comprise the following steps that:
1st, realtime graphic is gathered
Using monocular-camera mode, optical anti-vibration, super remote focal length B/W camera, in train traveling process in real time
Collection image sequence.Grand electric H3225A-K-G MDVR car video recorders are used in the present embodiment, with optical anti-vibration, super remote Jiao
Away from B/W camera can apply.The image memory device of collection is vehicle-mounted industrial computer:Beijing China is safe double 19 cun
Panel computer P1903.
2nd, classification of track is judged
When the vehicle-mounted vidicon excess of stroke, it is judged as bend (wide-angle turning track, right angle track);Then using fixed high definition
Video camera and wireless launcher receive wireless image and realize manual detection bend track condition, referring to figs. 2 and 3;This implementation
High-definition camera is fixed in example using bend high-definition probe, brand Ansky model Ask-3510-67, wireless launcher is used
Radio transmission device:Technical grade wireless simulation amount signal transmission DTD110FC.Bend high-definition probe fixed position can be except Fig. 3
Outer other position.
If vehicle-mounted vidicon does not have the excess of stroke, it is judged to class straight way (rectilinear orbit, small radian turning track), continues to adopt
Realtime graphic is obtained with vehicle-mounted vidicon, into next step, using vehicle-mounted industrial computer:The safe double 19 cun of flat board electricity of Beijing China
Brain P1903 processes image.
3rd, image preprocessing
First, enhancing treatment is done to image, top cap computing, image X is done using a diameter of 4 " diamond " structural element
Top cap computing on structural element B is designated as X ° of B, is defined asDual-rail part is highlighted, suppresses background high
Bright part;
Then, Otsu threshold calculations optimal threshold T are used to enhanced image, by image binaryzation, obtains binary map
Picture;Multiple discontinuity zones are included by the image after Threshold segmentation;
Finally, the connectedness of each pixel pixel adjacent thereto in discontinuity zone is checked, mark extracts region, by connection
Field mark retains double track characteristic straight line.
4th, detection window is set up
Double track characteristic straight line is retained according to connected component labeling, using fitting a straight line, is realized to track using least square method
Edge positioning is extracted, by two rail linears, linear equation is obtained;The position of detection window window is determined with the particular location of rail
Put, size and dimension, detection window rail portion is confined in the range of it.
5th, barrier is determined whether
The characteristics of having excluded background based on image in class straight way track rule consistency and detection window, can ignore track
Motion, when there is suspicious barrier intrusion detection window, will certainly cause characteristics of image to change, and this step chooses histogram feature
Threshold value T is defined based on experience value1And T2For differentiating histogram variances variation delta σk 2With histogram burr number
The value changes situation of BurrNum, by two-value monochrome pixels ratio r ati o, Δ σk 2With tri- monitorings of parameter of BurrNum,
Determining whether barrier invasion may.Algorithm is as follows:
Histogram average:
(wherein rkIt is histogram feature data, pkProbability corresponding to it, L is characterized the number of data)
Histogram variances:
Two-value monochrome pixels ratio:
(nbIt is the quantity of black pixel, nkIt is the quantity of white pixel)
From the point of view of significance level and reliability, ratio nb/nkIt is more directly perceived, actual conditions can be more showed, therefore enumerate
Basis for estimation:
(1)ratio<When 4.5, no matter Δ σk 2With the value of BurrNum how, all judge there is invasion barrier;
(2)ratio>When 4.5, then Δ σ is observedk 2How value with BurrNum changes, if both exceed threshold value, judges
There is invasion barrier;
(3)ratio>When 4.5, and Δ σk 2With the value of BurrNum all without departing from threshold value or one of them exceed threshold value, then
Judge no barrier, it may be possible to the influence of the external interference such as illumination condition.
If without barrier, return starts anew;If there is barrier, next step is carried out.
6th, detection of obstacles
Stationary obstruction also has relative motion to train, so it is static going back to be differentiated by movement locus and motion feature
It is moving obstacle.With reference to Fig. 3.
With reference to Fig. 5, using the frame differential method improved by mathematical morphology, difference fortune is carried out to the image in detection window
Calculate, extract segmentation moving target;Frame differential method is exactly in sequence of video images, to extract the adjacent frame of two frame three or multiframe figure
Picture, and to extract two field picture carry out the image operation that pixel is subtracted each other.If the difference of adjacent pixel is less than the threshold value of setting,
Think that the pixel is static background, conversely, then extracting the pixel as moving target, according to this principle, meet all
The pixel of moving target threshold value joins together that the moving target in scene just can be extracted, and removes jamming pattern, is reached with this
To the purpose for extracting segmentation moving target.
Frame differential method principle is as follows:
Dk(x, y)=| Fk(x,y)-Fk-1(x,y)|
Fk(x, y) represents continuous image, D in videok(x, y) represents the differential chart obtained by two continuous frames image subtraction
Picture, then by Dk(x, y) does following treatment:
(P is given threshold, and by experiment, simulation is obtained repeatedly)
Rk(x, y) is to judge the foundation whether target moves:If 1, then target discrimination is motion;Conversely, being static.
Meanwhile, with reference to Fig. 6, using background subtraction, will detect that the detection image in window carries out difference with background image, so
Afterwards moving region is detected with threshold value.If b (x, y) is background image, image sequence f (x, y, i) is defined, wherein (x, y) is figure
Image position coordinate, i is number of image frames.The gray scale that the gray value of each two field picture subtracts background is worth to error image:
Id (x, y, i)=f (x, y, i)-b (x, y)
A binaryzation difference image is can obtain by setting threshold value T:
(threshold value T is obtained by experimental simulation)
Finally, pixel is carried out to the image obtained by two kinds of algorithms and takes common factor, more accurately target is extracted, obtained
The size of target, the shape facility of target.
7th, the identification and classification of barrier
Barrier size, speed and direction discernment are carried out using Corner Feature Matching pursuitalgorithm, feature extraction is adopted
Harris Corner Detection Algorithms are used, is that a kind of point feature based on signal extracts operator, in this method, will be with adjoint point brightness contrast
(intensity contrast of pixel neighborhoods point) sufficiently large point is defined as angle point, and pixel correlation function used is as follows:
Wherein, w is the smoothing windows for carrying out noise reduction process, and (u, v) is offset coordinates, and I is image pixel matrix, I (x, y) value
The pixel value at image midpoint (x, y).IxAnd IyRepresentative image pixel single order partial differential in the horizontal direction, in vertical direction respectively,
Ix 2And Iy 2Then it is respectively the second order Grad in both direction.Then figure can just be detected by calculating angle point receptance function
Angle point as in:
R=det (M)-k*tr2(M)
Wherein, tr (M) and det (M) represent the mark and determinant of matrix M respectively, and Harris Corner Detection Algorithms recommend k
Take 0.04.
The characteristic matching Feature Points Matching algorithm related using gray scale is based on, by taking the extraction of three frames as an example:
Firstly, for each characteristic point pi∈E1, pj∈E3, respectively centered on characteristic point, construct N × N size
Window, W is designated as respectivelyi,Wj, the correlation function of each window is then calculated respectively
Can be according to CcorThe value of (i, j) is the correlation that can determine that individual features point, and the value is bigger, and corresponding characteristic point is adjacent
Domain gray scale closer to.During match point is looked for, matched by the way of mutually matching, i.e., as the characteristic point pi in E1
Optimal match point be the p in E3i, while pjOptimal match point also be piWhen, it is considered as (pipj) it is a pair of best match spy
Levy a little.Final corners Matching result is obtained accordingly.
The type of preceding object thing is judged with this, is without dangerous (static the small-scale obstacle thing, the quick motion barrier across track
Hinder thing) or it is dangerous (moving obstacle that static large obstacle, influence train pass through).According to detection recognition result adjustment train
Operation.
Claims (5)
1. a kind of track obstacle detecting system based on machine vision, it is characterised in that:Comprise the following steps that:
Step 1:Collection realtime graphic:Black and white video image is obtained by the vehicle-mounted vidicon installed in train head, using monocular
Video camera mode, optical anti-vibration, the B/W camera of super remote focal length;
Step 2:Judge classification of track:According to vehicle-mounted vidicon, whether the excess of stroke judges that track is class straight way or bend;Bend track
Wireless image is received using fixed high-definition camera and wireless launcher and realizes manual detection bend situation;Class straight way track enters
Enter step 3;
Step 3:Image preprocessing:The real-time image sequences that class straight way will get are analyzed, using based on top cap conversion and
The rail flanges extraction algorithm of Otsu threshold values, obtains track double track characteristic straight line;
Step 4:Set up detection window:Using fitting a straight line, edge is extracted to track using least square method and is positioned, by rail linear
Change, obtain linear equation;Detection window is determined with the particular location of rail, detection window confines in the range of it rail;
Step 5:Determine whether barrier:Based on detection window in histogram feature change, based on experience value define threshold value T1 and
T2 is used to differentiate histogram variances variation delta σ k2With the value changes situation of histogram burr number BurrNum, by two-value
Monochrome pixels ratio r atio, Δ σ k2With tri- monitorings of parameter of BurrNum, determining whether barrier invasion may;
Step 6:Detection of obstacles:The frame differential method improved based on mathematical morphology is combined with background subtraction, to obstacle
Thing feature is extracted, and obtains size, the shape facility of barrier;
Step 7:The identification and classification of barrier:Barrier speed is carried out using Corner Feature Matching pursuitalgorithm and direction is known
Not, preceding object thing is divided into without dangerous and dangerous two class in real time;
The detailed process of step 7 is as follows:
Barrier size, speed and direction discernment are carried out using Corner Feature Matching pursuitalgorithm, feature extraction is used
Harris Corner Detection Algorithms, are that a kind of point feature based on signal extracts operator, in the track obstacle based on machine vision
By the intensity contrast of pixel neighborhoods point in detecting system, sufficiently large point is defined as angle point, and pixel correlation function used is as follows
It is shown:
Wherein, w is the smoothing windows for carrying out noise reduction process, and (u, v) is offset coordinates, and I is image pixel matrix, I (x, y) value image
The pixel value at midpoint (x, y);Ix and Iy difference representative image pixel single order partial differential in the horizontal direction, in vertical direction,
WithThen it is respectively the second order Grad in both direction;Then in can just detecting image by calculating angle point receptance function
Angle point;
R=det (M)-k*tr2(M)
Wherein, tr (M) and det (M) represent the mark and determinant of matrix M respectively, and Harris Corner Detection Algorithms recommend k to take
0.04;
The characteristic matching Feature Points Matching algorithm related using gray scale is based on, by taking the extraction of three frames as an example:
Firstly, for each characteristic point pi∈E1, pj∈E3, respectively centered on characteristic point, construct a window for N × N sizes
Mouthful, Wi, Wj are designated as respectively, the correlation function of each window is then calculated respectively
Can be according to CcorThe value of (i, j) is the correlation that can determine that individual features point, and the value is bigger, corresponding feature vertex neighborhood ash
Degree closer to;During match point is looked for, matched by the way of mutually matching, that is, worked as E1In characteristic point piMost
Good match point is E3In pi, while pjOptimal match point also be piWhen, it is considered as (pi, pj) it is a pair of best match features
Point;Final corners Matching result is obtained accordingly;
The type of preceding object thing is judged with this, is without dangerous or dangerous.
2. a kind of track obstacle detecting system based on machine vision according to claim 1, it is characterised in that:It is described
The detailed process of step 3 is as follows:
First, enhancing treatment is done to image, top cap computing is done using a diameter of 4 " diamond " structural element, image X on
The top cap computing of structural element B is designated as X ° of B, is defined asDual-rail part is highlighted, suppresses background highlights high
Point;
Then, Otsu threshold calculations optimal threshold T are used to enhanced image, by image binaryzation, obtains bianry image;Through
The image crossed after Threshold segmentation includes multiple discontinuity zones;
Finally, the connectedness of each pixel pixel adjacent thereto in discontinuity zone is checked, mark extracts region, by connected domain mark
Note retains double track characteristic straight line.
3. a kind of track obstacle detecting system based on machine vision according to claim 1, it is characterised in that:It is described
The detailed process of step 4 is:Double track characteristic straight line is retained according to connected component labeling, using fitting a straight line, using least square method
Realize extracting track edge positioning, by two rail linears, obtain linear equation;Detection window is determined with the particular location of rail
The position of window, size and dimension, detection window confine in the range of it rail portion.
4. a kind of track obstacle detecting system based on machine vision according to claim 1, it is characterised in that:It is described
The detailed process of step 5 is as follows:
The characteristics of having excluded background based on image in class straight way track rule consistency and detection window, can ignore track fortune
It is dynamic, when there is suspicious barrier intrusion detection window, characteristics of image will certainly be caused to change, in the track obstacle based on machine vision
Histogram feature is chosen in quality testing examining system;
Defining threshold value T1 and T2 based on experience value is used to differentiate histogram variances variation delta σ k2With histogram burr number
The value changes situation of BurrNum, by two-value monochrome pixels ratio r atio, Δ σ k2With tri- monitorings of parameter of BurrNum,
Determining whether barrier invasion may;Algorithm is as follows:
Histogram average:
Wherein rkIt is histogram feature data, pkProbability corresponding to it, L is characterized the number of data;
Histogram variances:
Two-value monochrome pixels ratio:
Wherein, nbIt is the quantity of black pixel, nkIt is the quantity of white pixel;
From the point of view of significance level and reliability, ratio nb/nk is more directly perceived, can more show actual conditions, therefore enumerates and sentence
Disconnected foundation:
(1) during ratio < 4.5, no matter Δ σ k2With the value of BurrNum how, all judge there is invasion barrier;
(2) during ratio > 4.5, then Δ σ k are observed2How value with BurrNum changes, if both exceed threshold value, judgement have into
Invade barrier;
(3) during ratio > 4.5, and Δ σ k2With the value of BurrNum all without departing from threshold value or one of them exceed threshold value, then sentence
Break no barrier, is the influence of the external interferences such as illumination condition;
If without barrier, return starts anew;If there is barrier, next step is carried out.
5. a kind of track obstacle detecting system based on machine vision according to claim 1, it is characterised in that:It is described
The detailed process of step 6 is as follows:
Stationary obstruction also has relative motion to train, so differentiating static or fortune by movement locus and motion feature
Dynamic barrier;
Using the frame differential method improved by mathematical morphology, calculus of differences is carried out to the image in detection window, extract segmentation
Moving target;Frame differential method is exactly in sequence of video images, to extract the adjacent frame of two frame three or multiple image, and to extracting
Two field picture carry out the image operation that pixel is subtracted each other;If the difference of adjacent pixel is less than the threshold value of setting, then it is assumed that the pixel
It is static background, conversely, then extracting the pixel as moving target, according to this principle, moving target threshold is met by all
The pixel of value joins together that the moving target in scene just can be extracted, and removes jamming pattern, and extraction segmentation is reached with this
The purpose of moving target;
Frame differential method principle is as follows:
DK(x, y)=| Fk(x, y)-Fk-1(x, y) |
Fk(x, y) represents continuous image, D in videoK(x, y) represents the error image obtained by two continuous frames image subtraction, so
Afterwards by DK(x, y) does following treatment:
Wherein, P is given threshold, and by experiment, simulation is obtained repeatedly;
Rk(x, y) is to judge the foundation whether target moves:If 1, then target discrimination is motion;Conversely, being static;
Meanwhile, using background subtraction, will detect that the detection image in window carries out difference with background image, then examined with threshold value
Survey moving region;If b (x, y) is background image, image sequence f (x, y, i) is defined, wherein (x, y) is image position coordinates, i
It is number of image frames;The gray scale that the gray value of each two field picture subtracts background is worth to error image:
Id (x, y, i)=f (x, y, i)-b (x, y)
A binaryzation difference image is can obtain by setting threshold value T:
Wherein, threshold value T is obtained by experimental simulation;
Finally, pixel is carried out to the image obtained by two kinds of algorithms and takes common factor, more accurately target is extracted, obtain target
Size, the shape of target, differentiation be it is static or motion.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410681599.XA CN104331910B (en) | 2014-11-24 | 2014-11-24 | A kind of track obstacle detecting system based on machine vision |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410681599.XA CN104331910B (en) | 2014-11-24 | 2014-11-24 | A kind of track obstacle detecting system based on machine vision |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104331910A CN104331910A (en) | 2015-02-04 |
CN104331910B true CN104331910B (en) | 2017-06-16 |
Family
ID=52406630
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410681599.XA Expired - Fee Related CN104331910B (en) | 2014-11-24 | 2014-11-24 | A kind of track obstacle detecting system based on machine vision |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104331910B (en) |
Families Citing this family (58)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103957387B (en) * | 2014-05-09 | 2019-11-22 | 江苏万友消防安全远程监控系统有限公司 | A kind of security protection channel sundries detection method and detection system |
EP3091370B1 (en) * | 2015-05-05 | 2021-01-06 | Volvo Car Corporation | Method and arrangement for determining safe vehicle trajectories |
CN106808482B (en) * | 2015-12-02 | 2019-07-19 | 中国科学院沈阳自动化研究所 | A kind of crusing robot multisensor syste and method for inspecting |
CN105809679B (en) * | 2016-03-04 | 2019-06-18 | 李云栋 | Mountain railway side slope rockfall detection method based on visual analysis |
DE102016205330A1 (en) | 2016-03-31 | 2017-10-05 | Siemens Aktiengesellschaft | Method and system for detecting obstacles in a danger area in front of a rail vehicle |
DE102016205339A1 (en) * | 2016-03-31 | 2017-10-05 | Siemens Aktiengesellschaft | Method and system for detecting obstacles in a danger area in front of a rail vehicle |
CN105882683B (en) * | 2016-04-13 | 2018-04-17 | 北京康拓红外技术股份有限公司 | Railroad train check-up detecting system and method based on machine vision |
JP6633458B2 (en) * | 2016-06-02 | 2020-01-22 | 株式会社日立製作所 | Vehicle control system |
CN106156742A (en) * | 2016-07-06 | 2016-11-23 | 尚艳燕 | A kind of balance car barrier bypassing method and device |
CN107622666B (en) * | 2016-07-13 | 2020-11-24 | 尚艳燕 | Balance car cruise law enforcement method and device |
CN106254823A (en) * | 2016-07-28 | 2016-12-21 | 广州紫川电子科技有限公司 | Object based on thermal infrared imager swarms into method for detecting, Apparatus and system |
CN107780324A (en) * | 2016-08-28 | 2018-03-09 | 上海华测导航技术股份有限公司 | A kind of airfield pavement method for inspecting and system |
CN107784606A (en) * | 2016-08-30 | 2018-03-09 | 源渠(上海)信息技术有限公司 | A kind of training organization's commending system and method |
US11021177B2 (en) | 2016-10-20 | 2021-06-01 | Rail Vision Ltd | System and method for object and obstacle detection and classification in collision avoidance of railway applications |
CN107985335A (en) * | 2016-10-26 | 2018-05-04 | 深圳金澜汉源科技有限公司 | track traffic safety detection method |
CN106778569A (en) * | 2016-12-05 | 2017-05-31 | 河海大学 | Train preceding object object detecting method based on video image |
CN106709928B (en) * | 2016-12-22 | 2019-12-10 | 湖北工业大学 | fast two-dimensional maximum inter-class variance threshold method for noisy images |
CN107636680B (en) * | 2016-12-30 | 2021-07-27 | 达闼机器人有限公司 | Obstacle detection method and device |
CN107169401B (en) * | 2017-02-10 | 2020-05-05 | 北京交通大学 | Rail invader detection method based on rail visual feature spectrum |
JP6743750B2 (en) * | 2017-04-14 | 2020-08-19 | 株式会社ダイフク | Goods transport facility |
CN108803588A (en) * | 2017-04-28 | 2018-11-13 | 深圳乐动机器人有限公司 | The control system of robot |
CN109521757B (en) * | 2017-09-18 | 2021-12-31 | 阿波罗智能技术(北京)有限公司 | Static obstacle identification method and device |
CN107831760A (en) * | 2017-09-27 | 2018-03-23 | 安徽硕威智能科技有限公司 | Robot barrier thing processing system and method |
CN108491758B (en) * | 2018-02-08 | 2020-11-20 | 深圳市睿灵创新科技开发有限公司 | Track detection method and robot |
CN108536141A (en) * | 2018-03-08 | 2018-09-14 | 杭州晶智能科技有限公司 | A kind of stochastic path planning method of automatic dust absorption machine people |
CN108427931B (en) * | 2018-03-21 | 2019-09-10 | 合肥工业大学 | The detection method of barrier before a kind of mine locomotive based on machine vision |
CN110443819B (en) * | 2018-05-03 | 2022-04-15 | 比亚迪股份有限公司 | Method and device for detecting track of monorail train |
CN108830257A (en) * | 2018-06-29 | 2018-11-16 | 电子科技大学 | A kind of potential obstacle detection method based on monocular light stream |
CN109389617A (en) * | 2018-08-27 | 2019-02-26 | 深圳大学 | A kind of motion estimate based on piece heterogeneous system and method for tracing and system |
CN109165600B (en) * | 2018-08-27 | 2021-11-26 | 浙江大丰实业股份有限公司 | Intelligent search platform for stage performance personnel |
CN109215136B (en) | 2018-09-07 | 2020-03-20 | 百度在线网络技术(北京)有限公司 | Real data enhancement method and device and terminal |
CN109143242B (en) | 2018-09-07 | 2020-04-14 | 百度在线网络技术(北京)有限公司 | Obstacle absolute velocity estimation method, system, computer device, and storage medium |
CN109255181B (en) | 2018-09-07 | 2019-12-24 | 百度在线网络技术(北京)有限公司 | Obstacle distribution simulation method and device based on multiple models and terminal |
CN110375659B (en) | 2018-09-11 | 2021-07-27 | 百度在线网络技术(北京)有限公司 | Method, device, equipment and storage medium for detecting height of obstacle |
CN109165629B (en) | 2018-09-13 | 2019-08-23 | 百度在线网络技术(北京)有限公司 | It is multifocal away from visual barrier cognitive method, device, equipment and storage medium |
CN109188460A (en) * | 2018-09-25 | 2019-01-11 | 北京华开领航科技有限责任公司 | Unmanned foreign matter detection system and method |
CN109522847A (en) * | 2018-11-20 | 2019-03-26 | 中车株洲电力机车有限公司 | A kind of track and road barricade object detecting method based on depth map |
CN109709843B (en) * | 2018-12-13 | 2021-01-05 | 西安电子科技大学 | Train water filling port detection and positioning method |
CN109591850A (en) * | 2018-12-24 | 2019-04-09 | 郑州畅想高科股份有限公司 | A kind of track foreign matter detecting method and device |
CN109766839A (en) * | 2019-01-11 | 2019-05-17 | 上海泽高电子工程技术有限公司 | A kind of track foreign matter detecting method based on image |
CN109703568B (en) | 2019-02-19 | 2020-08-18 | 百度在线网络技术(北京)有限公司 | Method, device and server for learning driving strategy of automatic driving vehicle in real time |
CN109857112A (en) * | 2019-02-21 | 2019-06-07 | 广东智吉科技有限公司 | Obstacle Avoidance and device |
CN109712421B (en) | 2019-02-22 | 2021-06-04 | 百度在线网络技术(北京)有限公司 | Method, apparatus and storage medium for speed planning of autonomous vehicles |
CN110287897B (en) * | 2019-06-27 | 2021-05-14 | 清华大学 | Rail train visual positioning system based on deep learning |
CN110435689B (en) * | 2019-09-09 | 2020-12-08 | 中铁电气化局集团西安电气化工程有限公司 | Detection method of intelligent obstacle detection vehicle based on modular control |
CN110717477A (en) * | 2019-10-22 | 2020-01-21 | 深圳市微度数字技术有限公司 | Intelligent early warning and analyzing device for track |
CN111127534A (en) * | 2019-11-05 | 2020-05-08 | 深圳市三宝创新智能有限公司 | Obstacle detection method |
CN110716212B (en) * | 2019-11-14 | 2021-07-27 | 吉林大学 | Method and system for detecting road surface obstacle |
CN110974088B (en) * | 2019-11-29 | 2021-09-24 | 深圳市杉川机器人有限公司 | Sweeping robot control method, sweeping robot and storage medium |
CN112989883B (en) * | 2019-12-16 | 2024-02-02 | 中国科学院沈阳计算技术研究所有限公司 | Method for identifying obstacle in front of train |
CN112183337A (en) * | 2020-09-28 | 2021-01-05 | 华北电力大学(保定) | Detection method and device for preventing mechanical invasion of power transmission line |
CN113465621B (en) * | 2021-06-22 | 2022-09-20 | 同济大学 | Dijkstra path planning method and device considering conflict probability and storage medium |
CN113642453A (en) * | 2021-08-11 | 2021-11-12 | 北京京东乾石科技有限公司 | Obstacle detection method, device and system |
CN115100633B (en) * | 2022-08-24 | 2022-12-13 | 广东中科凯泽信息科技有限公司 | Obstacle identification method based on machine learning |
CN116101275A (en) * | 2023-04-12 | 2023-05-12 | 禾多科技(北京)有限公司 | Obstacle avoidance method and system based on automatic driving |
CN116612119B (en) * | 2023-07-20 | 2023-09-19 | 山东行创科技有限公司 | Machine vision-based method for detecting working state image of drill bit for machine tool |
CN116778532B (en) * | 2023-08-24 | 2023-11-07 | 汶上义桥煤矿有限责任公司 | Underground coal mine personnel target tracking method |
CN116862910B (en) * | 2023-09-04 | 2023-11-21 | 山东经典印务有限责任公司 | Visual detection method based on automatic cutting production |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE10113413A1 (en) * | 2001-03-20 | 2002-09-26 | Alfred Spitzley | Method for opto-electronic monitoring of spatial regions, especially for use with driverless rail and agricultural vehicles using an opto-electronic obstruction detection and collision avoidance system |
JP2011076214A (en) * | 2009-09-29 | 2011-04-14 | Alps Electric Co Ltd | Obstacle detection device |
CN103116743B (en) * | 2013-02-01 | 2016-08-10 | 浙江捷尚视觉科技股份有限公司 | A kind of railway obstacle detection method based on on-line study |
-
2014
- 2014-11-24 CN CN201410681599.XA patent/CN104331910B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN104331910A (en) | 2015-02-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104331910B (en) | A kind of track obstacle detecting system based on machine vision | |
CN110261436B (en) | Rail fault detection method and system based on infrared thermal imaging and computer vision | |
CN107738612B (en) | Automatic parking space detection and identification system based on panoramic vision auxiliary system | |
CN107341454B (en) | Method and device for detecting obstacles in scene and electronic equipment | |
CN101030256B (en) | Method and apparatus for cutting vehicle image | |
CN102765365B (en) | Pedestrian detection method based on machine vision and pedestrian anti-collision warning system based on machine vision | |
US7046822B1 (en) | Method of detecting objects within a wide range of a road vehicle | |
CN101739829B (en) | Video-based vehicle overspeed monitoring method and system | |
CN103021177B (en) | Method and system for processing traffic monitoring video image in foggy day | |
RU2698610C2 (en) | Method and processing unit for control over traffic monitoring system | |
CN110210363A (en) | A kind of target vehicle crimping detection method based on vehicle-mounted image | |
CN111198371A (en) | Forward-looking obstacle detection system | |
CN103984961A (en) | Image detection method for detecting foreign matter at bottom of vehicle | |
CN105488454A (en) | Monocular vision based front vehicle detection and ranging method | |
CN104237881A (en) | FMCW anti-collision radar multi-target detecting and tracking system and method | |
CN104778444A (en) | Method for analyzing apparent characteristic of vehicle image in road scene | |
WO2016129301A1 (en) | Entry possibility determining device for vehicle | |
CN103065307B (en) | The area merges detection method of SAR/SPOT image under non-precision registration | |
CN107886034B (en) | Driving reminding method and device and vehicle | |
CN107274678B (en) | A kind of night vehicle flowrate and model recognizing method based on Kinect | |
CN110298300A (en) | A method of detection vehicle violation crimping | |
JP6139088B2 (en) | Vehicle detection device | |
CN109522847A (en) | A kind of track and road barricade object detecting method based on depth map | |
KR101406316B1 (en) | Apparatus and method for detecting lane | |
CN106778569A (en) | Train preceding object object detecting method based on video image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170616 Termination date: 20171124 |
|
CF01 | Termination of patent right due to non-payment of annual fee |