CN104267209B - Method and system for expressway video speed measurement based on virtual coils - Google Patents
Method and system for expressway video speed measurement based on virtual coils Download PDFInfo
- Publication number
- CN104267209B CN104267209B CN201410576699.6A CN201410576699A CN104267209B CN 104267209 B CN104267209 B CN 104267209B CN 201410576699 A CN201410576699 A CN 201410576699A CN 104267209 B CN104267209 B CN 104267209B
- Authority
- CN
- China
- Prior art keywords
- edge
- virtual coil
- virtual
- stabilised
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The invention provides a method and a system for expressway video speed measurement based on virtual coils. The method comprises the following steps: setting at least two preset regions in a video picture shot by a speed measurement camera as virtual coils, and detecting whether a picture formed by a specific target vehicle passes the virtual coils or not; measuring the speed of the vehicle according to time difference that the specific target vehicle sequentially triggers the virtual coils. The core of the invention lies in the improvement over a virtual coil triggering detection mechanism. The stable edge features of a plurality of key regions are adaptively recognized in the virtual coil regions, and the triggering of the virtual coils is realized through the detection of the variation, caused by the specific target vehicle, of the inner edge features of the virtual coil regions. As the stable edge features have strong robustness relative to the gradual change and the sudden change of external environment factors such as sun exposure, shadow and vehicle light radiation, error triggering caused by external interference can be effectively avoided. The reliable, accurate and high-real-timeliness virtual coil triggering detection mechanism is built.
Description
Technical field
The present invention relates to highway velocity measuring technique field, public more particularly, to a kind of high speed based on virtual coil
Road video frequency speed-measuring method and system.
Background technology
Traditional high way super speed monitoring uses ground sensing coil speed measuring, laser velocimeter and radar velocity measurement.Three of the above is surveyed
Speed method all has respective weak point.Ground sensing coil speed measuring is to have vehicle for a pair by sensing at road surface pre-plugged
The coil of ability, record object vehicle passes in succession through the time of two coils, and then calculates with coil-span according to time difference
Speed.But the installation of ground induction coil can destroy the road surface of highway, and for the vehicle speed measuring travelled along non-linear traces
Precision is relatively low.Laser velocimeter utilizes laser pulse repeatedly to find range to sailing vehicle, and then calculates speed.But laser velocimeter is wanted
Asking instrument just to sending a car, not can exceed that 10 degree with the misalignment angle of vehicle heading, installation site condition is harsher, and
The problem that measurement error is big is equally existed for travelling the vehicle of trail change in speed trial ground.Radar velocity measurement is to utilize Doppler
Principle tests the speed, but similar with laser velocimeter, equally exists the misalignment angle with vehicle heading and to be maintained at 10 degree
Within problem, bring significant limitation to its scope of application.
Along with high-speed camera and the progress of image procossing identification technology, can each from the shooting of freeway surveillance and control video camera
Frame video pictures positions target vehicle, and calculates speed according to the shooting time of its driving trace and each frame video pictures.
Highway velocity measuring technique based on this principle is referred to as video frequency speed-measuring.Compare traditional velocity measuring technique presented hereinbefore, depending on
Frequently the following aspects that tests the speed can show obvious advantage.First, it is only necessary to there is the high-speed camera of high-resolution
I.e. can realize highway and the shooting of associated vehicle picture, signal acquisition process and equipment are all simplified, to the most public
The road surface on road and other facility not adversely affect, it is easy to rebuilding construction；Secondly, compared to radar and laser velocimeter for deviation
For the sensitivity of angle, video frequency speed-measuring is the loosest for the requirement of shooting angle；3rd, by calculating video frequency speed-measuring is relevant
The improvement of method, the adaptability for vehicle driving trace is greatly improved, and also can during for vehicle lane change even turning driving
Enough realizations are accurately tested the speed；4th, whether ground induction coil, laser or radar velocity measurement, it is necessary to additional shooting is to identify car
Board and evidence obtaining, and video frequency speed-measuring can by testing the speed, Car license recognition, the step such as record evidence obtaining concentrated, system integrated and
Response speed has obtained the biggest lifting.
The core of video frequency speed-measuring technology is the process to video pictures, identification and analytical calculation, and therefore its algorithm determines
The accuracy of whole system and real-time.The basis of video frequency speed-measuring algorithm is that in video pictures, the identification of specific objective vehicle is taken out
Taking, target is the part identification representing specific objective vehicle in video pictures to be extracted and be different from its in video pictures
Remaining part is divided.Specific objective vehicle identification abstracting method includes that optical flow method, frame differential method, background subtraction, vehicle lamp area position
Method, License Plate method etc..For the part representing specific objective vehicle extracted, it is also performed to the process of necessity and removes the moon
The impurity such as shadow.
On this basis, the algorithm performing to test the speed to specific objective vehicle can be divided into location tracking method and virtual coil method two
Type.
Location tracking method is by location described specific objective vehicle location in continuous some frame video pictures, meter
Calculate this specific objective vehicle movement velocity in video pictures, and then the actual speed of conversion vehicle.Location tracking method tests the speed
Degree of accuracy is higher, and for the better adaptability of complicated vehicle driving trace, but algorithm is extremely complex, real-time operation Performance comparision
Difference, can run into bottleneck during the many vehicles real time speed measuring problem under big traffic conditions in actual applications.
Virtual coil method is video pictures presumptive area to be set as virtual coil, as the picture area pair of virtual coil
Answer a certain position in the middle of highway, when vehicle is by the described position of highway, can cause in video pictures virtual
Picture signal change at coil, this is referred to as triggering virtual coil.True between the highway location that two virtual coils are corresponding
Real spacing L can be obtained in advance by field survey or calculating.Thus, trigger two void by described specific objective vehicle
Intend the time difference △ t of coil, the speed v=L/ △ t of this specific objective vehicle can be calculated；Wherein, △ t is generally with determining this spy
The frame-to-frame differences of former and later two video pictures that the vehicle that sets the goal triggers each virtual coil represents.The algorithm of virtual coil method is more
For simple, the speed of service faster, effective monitoring region more than location tracking method, thus the most universal.
Virtual coil method needs the major issue solved to be to set up reliable virtual coil detection trigger in actual applications
Mechanism.The mark that virtual coil triggers be in video pictures virtual coil region the picture signal generation amplitude of pixel threshold value with
On steps change.Typically use the grey scale difference signal of pixel for triggering the picture signal of judgement, computational methods be by
In current video picture, the pixel grey scale signal in virtual coil region is carried out with the grey scale signal in this region in default background frame
Subtraction, and then judge that grey scale difference signal value, whether more than described threshold value, then thinks empty in the case of more than this threshold value
Intend coil to be triggered.
But, the reliability ratio of above-mentioned detection trigger mechanism is relatively low, it is easy to the situation of false triggering occurs.Impact detection
The factor of reliability is in addition to noise signal common in video pictures shooting process, and also include in highway actual environment is various
Factor, such as: sunshine on daytime and shade change can produce interference to detection trigger；Car light light beam road pavement was irradiated and was brought figure night
Image signal gray scale is suddenlyd change, and virtual coil can be caused to be triggered ahead of time before vehicle is actually passed through；Under particular light state, certain
Under-effected to cause triggering to virtual coil area pixel gray scale of the most light yellow or light green color automobile.
Solving the dysgenic Main Means of above-mentioned factor in prior art is that structure Gaussian Background model is various to adapt to
Environmental factors changes.Gaussian Background model is thought, does not has in most of frame background frames of moving target, and pixel grey scale can be obeyed
Gauss distribution N (μ, σ2), μ is pixel grey scale average, and σ2It it is variance.Gaussian Background model is for the some accumulated before
Background frame in each grey scale pixel value carry out adding up thus calculate μ and σ2Value, thus construct background model, and based on reality
Time shooting video pictures constantly μ and σ to background model2Value is updated；By by the pixel value of current video picture with
Background model compares, and extracts the foreground pixel being wherein not belonging to background frame, and then realizes sentencing virtual coil triggering
Disconnected.
Triggering of Gaussian Background model realization is utilized to judge that can actually adapt to background slowly changes the impact brought,
Such as can overcome sunshine on daytime and the interference of shade change, but, car light light beam at night above-mentioned is irradiated or
For sudden changes such as particular color vehicle can not react, its false triggering problem brought still cannot be solved.
Summary of the invention
For drawbacks described above of the prior art, the invention provides a kind of highway video based on virtual coil and survey
Speed method and system.The core of the present invention is to be improved, virtual coil detection trigger mechanism in virtual coil region
Identify the stabilised edge feature that some key areas are had the most adaptively, and then caused by specific objective vehicle by detection
Virtual coil region inner margin changing features realize triggering to virtual coil.
A kind of highway video frequency speed-measuring method based on virtual coil provided by the present invention, including: virtual coil touches
Send out detecting step, at least two presumptive area in the video pictures of the shot by camera that tests the speed is set as virtual coil, and
And whether the picture that detection is formed by specific objective vehicle is through described virtual coil；Test the speed step, according to specific objective vehicle
The time difference triggering each described virtual coil successively carries out vehicle speed measuring；It is characterized in that, described virtual coil detection trigger
Step specifically includes:
Key area stabilised edge characteristic extraction step, for extracting without the virtual coil district of video pictures under vehicle-state
Spatial distribution and time-domain stability qualified stabilised edge information in territory, form the virtual coil comprising described stabilised edge
Region template；
Edge feature change-detection step, for the real-time limit of virtual coil extracted region for captured in real-time video pictures
Edge information, and described stabilised edge information and real-time marginal information are carried out matching operation, it is judged that whether the similarity of the two
Less than minimum similar threshold value；In the case of less than described minimum similar threshold value, determine that virtual coil is triggered.
Preferably, described key area stabilised edge characteristic extraction step includes following sub-step:
Edge extracting step, for performing Gaussian smoothing filter for virtual coil region, then utilizes Canny operator to enter
Row edge detection algorithm, it is thus achieved that the binary image of marginal information represents in the presence of virtual coil region；
Key area identification step, calculate edge in the presence of described virtual coil region image specific axis upwards
Run length, whether more than distance of swimming threshold value, is more than the edge of distance of swimming threshold value as time by run length, relatively described run length
Select edge；And the candidate edge of at least two frame video pictures being spaced specific duration is performed calculus of differences, it is thus achieved that edge difference
Partial image；The nonzero value region in edge difference partial image is got rid of, it is thus achieved that stabilised edge feature in candidate edge；
Edge feature generation step, it is thus achieved that and preserve described stabilised edge feature, thus formed and there is stabilised edge feature
Virtual coil region template.
It may further be preferable that described edge extracting step includes following sub-step: step 1, utilize Canny rim detection
Device carries out edge extracting for the first time to filtering image；Step 2, after filtering image is carried out top cap conversion, uses Canny rim detection
Device carries out second time edge extracting；Step 3, after the image after converting top cap carries out logarithmic transformation, uses Canny edge detector
Carry out third time edge extracting；Step 4, by the result images superposition of three edge extractings；Step 5, enters the image after superposition
Row skeletonizing processes, and obtains marginal information.
It may further be preferable that in described key area identification step, for edge in the presence of virtual coil region, point
Do not calculate the number N of the uninterrupted pixel that each edge is had in the X-axis and Y direction of imagexAnd Ny, and then be NxAnd Ny
Distribute different weights, thus by each edge of weighted calculation at described specific axis run length H=α 1 N upwardsx+α2·
Ny, wherein α 1 and α 2 represents weight coefficient.
It may further be preferable that in described key area identification step, described candidate edge is first carried out Extension algorithm
It is the marginal zone with bigger line thickness by edge expansion, then performs described calculus of differences.
Preferably, in described edge feature change-detection step, described stabilised edge information and real-time edge are believed
Breath performs Hausdorff distance and calculates, and judges that whether Hausdorff distance is less than minimum similar threshold value.
It may further be preferable that described Hausdorff distance calculates and includes: stablizing in virtual coil region template will be constituted
The pixel of marginal information as a point set P1, and using form described real-time marginal information pixel as another point set P2,
Calculate Hausdorff distance H (P1, the P2)=max{h (P1, P2), h (P2, P1) between point set P1 and point set P2 }, wherein h (P1,
P2) maximum to the Euclidean distance of point set P2 of the pixel in point set P1 is represented,Its
Middle a and b is belonging to the pixel of point set P1 and point set P2 respectively, and (a b) represents the Euclidean distance between a and b to D；h(P2,P1)
Represent the maximum to the Euclidean distance of point set P1 of the pixel in point set P2,Wherein a
Be belonging to the pixel of point set P2 and point set P1 respectively with b, (a b) represents the Euclidean distance between a and b to D.
The present invention and then provide a kind of highway video frequency speed-measuring system based on virtual coil, including: virtual coil
Detection trigger module, for being set as dummy line by least two presumptive area in the video pictures of the shot by camera that tests the speed
Circle, and whether detect the picture formed by specific objective vehicle through described virtual coil；Speed measuring module, for according to specific
Target vehicle triggers the time difference of each described virtual coil successively and carries out vehicle speed measuring；It is characterized in that, described virtual coil
Detection trigger module specifically includes:
Key area stabilised edge characteristic extracting module, for extracting without the virtual coil district of video pictures under vehicle-state
Spatial distribution and time-domain stability qualified stabilised edge information in territory, form the virtual coil comprising described stabilised edge
Region template；
Edge feature change detection module, for the real-time limit of virtual coil extracted region for captured in real-time video pictures
Edge information, and described stabilised edge information and real-time marginal information are carried out matching operation, it is judged that whether the similarity of the two
Less than minimum similar threshold value；In the case of less than described minimum similar threshold value, determine that virtual coil is triggered.
Preferably, described key area stabilised edge characteristic extracting module includes:
Edge extracting module, for performing Gaussian smoothing filter for virtual coil region, then utilizes Canny operator to enter
Row edge detection algorithm, it is thus achieved that the binary image of marginal information represents in the presence of virtual coil region；
Key area identification module, calculate edge in the presence of described virtual coil region image specific axis upwards
Run length, whether more than distance of swimming threshold value, is more than the edge of distance of swimming threshold value as time by run length, relatively described run length
Select edge；And the candidate edge of at least two frame video pictures being spaced specific duration is performed calculus of differences, it is thus achieved that edge difference
Partial image；The nonzero value region in edge difference partial image is got rid of, it is thus achieved that stabilised edge feature in candidate edge；
Edge feature generation module, it is thus achieved that and preserve described stabilised edge feature, thus formed and there is stabilised edge feature
Virtual coil region template.
Preferably, described edge feature change detection module includes:
Edge extraction module in real time, for the video pictures for captured in real-time, according to default boundary line coordinate
Segmentation virtual coil region, and then extract the real-time marginal information in the presence of it by Boundary extracting algorithm；
Edge matching module, performs Hausdorff distance to described stabilised edge information and real-time marginal information and calculates, and
And judge that whether Hausdorff distance is less than minimum similar threshold value.
Visible, the present invention is in the middle of highway video frequency speed-measuring method and system, for virtual coil detection trigger mechanism
Improved, by virtual coil without the stabilised edge feature extraction in car status screen out, and will be due to occlusion etc.
The virtual coil region inner margin changing features that reason is caused is as the foundation judging triggering.Owing to stabilised edge feature is relative
Gradual change and sudden change in outside environmental elements such as sunshine, shade, car light irradiations all have the strongest robustness, can be prevented effectively from
The false triggering that these external interference are caused, whether daytime and night can keep reliable detection trigger performance；And
Occlusion is also stable to the change of virtual coil region inner margin feature, compared to gray scale detection, does not haves Yin Te
Fixed illumination condition and vehicle color and the triggering failure that causes.The present invention is above-mentioned reliable, accurate and high by constructing
The virtual coil trigger mechanism of real-time, has fully ensured that the highway video frequency speed-measuring method realized based on virtual coil principle
And the high-quality of system runs.
Accompanying drawing explanation
The present invention is further detailed explanation with detailed description of the invention below in conjunction with the accompanying drawings:
Fig. 1 is without coil region schematic diagram virtual under car state in the embodiment of the present invention；
Fig. 2 is the method flow diagram extracting key area stabilised edge feature in the embodiment of the present invention；
Fig. 3 is the schematic diagram in the embodiment of the present invention to the edge calculations run length in virtual coil region；
Fig. 4 is the schematic diagram of the candidate edge execution Extension algorithm in the embodiment of the present invention to virtual coil region；
Fig. 5 is the virtual coil region template schematic diagram preserving stabilised edge feature in the embodiment of the present invention；
Fig. 6 is vehicle virtual coil region schematic diagram under state in the embodiment of the present invention；
Fig. 7 is the method flow diagram by detecting edge feature change triggers virtual coil.
Detailed description of the invention
In order to make those skilled in the art be more fully understood that technical scheme, and make the above-mentioned mesh of the present invention
, feature and advantage can become apparent from understandable, below in conjunction with embodiment and embodiment accompanying drawing, the present invention is made the most in detail
Explanation.
Highway video frequency speed-measuring method of the present invention and system use Computational Method of Velocity Measurement based on virtual coil.This
Bright presumptive area in the video pictures of the shot by camera that tests the speed being set as virtual coil, general to set at least two virtual
Coil, the time difference triggering each described virtual coil according to specific objective vehicle successively realizes testing the speed calculating.
The core of the present invention is to be improved virtual coil detection trigger mechanism, is no longer rely on detecting current video
Whether picture produces the steps change of pixel grey scale to determine touching of virtual coil relative to background frame or background model
Send out, but in virtual coil region, identify some key areas adaptively, determine that the stabilised edge of these key areas is special
Levy, and then the edge feature change caused by specific objective vehicle by detection realizes the triggering to virtual coil.
Fig. 1 is the virtual coil area schematic of the embodiment of the present invention.Fig. 1 shows that video camera shoots regarding of region of testing the speed
Frequently picture, the video pictures shown in Fig. 1 represents that this shooting tests the speed region without vehicle-state.Video frequency speed-measuring according to the present invention is calculated
Video pictures region within method, dotted line frame A, B and described dotted line frame is redefined for virtual coil A and B.Lead to when there being vehicle
When crossing the relevant position of highway, picture signal change in dotted line frame A, B region of video pictures can be caused, thus trigger void
Intend loop A and B.In the virtual coil detection trigger mechanism of the present invention, the change of described picture signal will appear as dotted line frame A, B
Within the change of edge feature.
As it is shown in figure 1, in the middle of video pictures region corresponding for virtual coil A, exist by partial pixel formed some
Key area, in these key areas the gray scale of pixel and/or chrominance information and pixel grey scale in video pictures background area and/
Or chrominance information compares and there is obvious difference, so that the intersection between key area with background area has stronger
Strong edge feature.Corresponding to the pavement structure of highway, described background area mainly reflects the general road of highway
Top layer, face, and the traffic marking C on highway pavement shown in FIG, different the demarcation line D of road crust, pavement facilities
Some irregular projections on (such as vertical shaft well lid etc.) E even road surface or damaged F all can be formed in the middle of video pictures as described in pass
Key range, thus show described edge feature at the intersection of these key areas Yu general road top layer.
The edge feature that key area is formed is easy to be detected by image analysis algorithm, and relative to day
According to, the gradual change of the outside environmental elements such as shade, car light irradiation and sudden change, all there is the strongest robustness.And when vehicle crosses,
Due to blocking and the introducing of vehicle own edges of vehicle road pavement so that the edge feature in virtual coil region can present bright
Aobvious change.Accordingly, with respect to prior art testing mechanism based on grey scale difference, detection based on edge feature can fully carry
The reliability that high virtual coil triggers, effectively prevents false triggering.
Fig. 2 shows the method flow diagram extracting key area stabilised edge feature of the embodiment of the present invention.For Fig. 1 institute
Show without coil region picture virtual under vehicle-state, the present invention performs the edge extracting step shown in Fig. 2, key area successively
Identification step and edge feature generation step, thus define the stabilised edge characteristic parameter without vehicle-state virtual coil.First
First it should be explicitly made clear at this point, what edge extracting step, key area identification step and edge feature generation step were targeted is all
Part within dotted line frame A, B in the middle of video pictures shown in Fig. 1, i.e. virtual coil region, according to default expression dotted line frame
The coordinate of boundary line, is partitioned into above-mentioned zone in the middle of video pictures.
In edge extracting step, Signal Pretreatment is first carried out, the effect of Signal Pretreatment be eliminate white noise and
Strengthen picture contrast.The mode of Signal Pretreatment is carried out Gaussian smoothing filter, selects mould according to the shape of Gaussian function
Plate is filtered, and template formula isBy this formula selected pixels point (i,
J) around the adjacent pixel regions of (2K+1) × (2L+1) size is weighted averagely, and (m n) represents in this adjacent pixel regions W
Point (m, n) weight coefficient at place.The noise of normal distribution can be eliminated by Signal Pretreatment, and edge therein is had ratio
Preferably protective effect, it is fuzzy that minimizing image produces because filtering high fdrequency component.For the image after Signal Pretreatment, hold
Row edge detection algorithm.Edge is that in the middle of image, pixel grey scale exists obvious discontinuous region, it is possible to use ask single order and
The method of second dervative detects edge, and the process differentiated can be completed by convolution by spatial domain differential operator.Commonly use
Spatial domain differential operator includes Roberts operator, Prewitt operator, Sobel operator, Laplace operator.Wherein, Roberts calculates
The edge definition of son location is high, but does not has noise removal function；Image can be smoothed by Sobel and Prewitt operator,
But easily manufacture false edge；Laplace operator is the most sensitive to noise, but anti-noise ability is more weak, easily causes edge and does not connects
Pass through.Therefore, the present invention have employed Canny operator in edge detection process.Canny operator finds the local maxima of image gradient
Value, the influence of noise being subject to due to different images is different, and Canny operator follows optimal edge detection, is a kind of anti-noise and location
Accurate compromise selection.Canny operator uses the finite difference of first-order partial derivative to calculate amplitude and the direction of gradient, and right
Gradient magnitude carries out non-maxima suppression, by the detection of dual threshold algorithm and adjoining edge.But Canny operator is to some gray scale differences very
Little weak rim detection still has some limitations, and is easily lost Small object details while suppression noise.The present invention
Described in edge detection algorithm, based on Canny operator, by the image border after original image, top cap conversion and logarithmic transformation
Testing result superposition also carries out skeletonizing process, it is achieved that the extraction at edge the most weak to image border.Based on Canny operator
Rim detection specifically include: step 1, utilize Canny edge detector that filtering image carries out for the first time edge extracting.Step
2, after filtering image is carried out top cap conversion, carry out second time edge extracting with Canny edge detector.Top cap conversion be based on
A kind of image procossing mode of mathematical morphology, it is the image after deducting opening operation from artwork, and wherein, opening operation can be used
In compensating uneven background luminance.Step 3, after the image after converting top cap carries out logarithmic transformation, uses Canny rim detection
Device carries out third time edge extracting.Step 4, by the result images superposition of three edge extractings.Step 5, to the image after superposition
Carry out skeletonizing process, obtain edge image.Skeletonizing is to be one group of thin skeleton by the object reduction in bianry image, and these are thin
Skeleton still retains the important information of primary object shape.Skeletonizing can extract the characteristic information of pattern from image, disappears in a large number
Except redundant data.Pass through edge extracting, it is thus achieved that the binary image of marginal information represents in the presence of virtual coil region.
For the marginal information obtained by edge extracting step, key area identification step is believed from whole described edges
Breath is selected the marginal information that the key area with virtual coil is associated, thus realizes passing through dummy line under state to without vehicle
The form of collar region, texture have the identification of the key area of significant effect.Video pictures at display express highway pavement
In, the edge extracted by edge extracting step represents the boundary between non-background area and the background area on road surface.But,
Described non-background area may not all can be as above-mentioned for indicating without vehicle by the form of virtual coil region, stricture of vagina under state
The key area of reason characteristic.As key area, first its edge formed should be provided with foot under various environmental conditions
Enough stability, under optimum state only when there is vehicle by virtual coil corresponding road surface region, the edge of key area
Significant change just can occur, and its edge keeps constant under other environmental change；Secondly, key area and the limit of formation thereof
Edge should occupy sufficiently large space, thus can be because sheltering from key area when vehicle is by virtual coil corresponding road surface region
Territory all or part of and cause the abundant change at edge；On the contrary, if the area of space that edge is distributed is too small, then having can
Cannot be able to be affected when vehicle passes through.Therefore, the marginal information obtained for edge extracting step, described key area
Identification step needs to be selected it based on above both sides factor.
In key area identification step, first marginal information is carried out spatial shape analysis, therefrom extract and be distributed in
The sufficiently large edge within spatial dimension.The spatial shape analysis of marginal information include to edge image specific axis upwards
Run length calculate, and judge whether this specific axis run length upwards is all higher than predetermined distance of swimming threshold value,
I.e. choose described edge as candidate edge in the case of being more than.L is from dummy line by edge extracting step as shown in Figure 3
The closed edge extracted in collar region, then, in described key area identification step, calculate edge L the most respectively at figure
The number N of the uninterrupted pixel being had in the X-axis of picture and Y directionxAnd Ny, and then according to special orientation axes O relative to X-axis and Y
The angle of axle, for NxAnd NyDistribute different weights, thus the distance of swimming being projected on specific axial O by weighted calculation edge L is long
Degree H=α 1 Nx+α2·Ny, wherein α 1 and α 2 represents weight coefficient.Generally, specific axial O should be set as and highway
The approximately perpendicular direction of direction of traffic, and according to the angle initialization relative to X-axis and the Y-axis above-mentioned weight system of specific axial O
Number.After obtaining edge L run length on specific axial O, whether compare run length H more than distance of swimming threshold value
Hthreshold, using edge L as candidate edge in the case of more than this distance of swimming threshold value.
For candidate edge, key area identification step and then candidate edge change in time domain is detected, sentence
Whether disconnected its possesses enough stability.By the detection of stability, be temporarily stored in short-and-medium in virtual coil region can be got rid of
Astable edge (edge of junk is lost on such as road surface).First, the video being spaced specific duration for extraction from video is drawn
The candidate edge obtained according to approach presented above in the middle of face, obtains edge difference partial image by calculus of differences.Owing to depositing
In the error that correction error and edge extracting calculate, it is possible to cause two frame video pictures all keep constant target to be given birth to
Candidate edge is become to equally exist certain difference.Therefore, first to the time in two frame video pictures in key area identification step
Selecting edge to carry out Extension algorithm, as shown in Figure 4, edge L through Extension algorithm, the edge that line thickness is W expansion is originally
Width is the marginal zone of W ', and then performs the calculus of differences between the candidate edge of two frame video pictures again.Can be for difference
The video pictures of decimation in time performs the most above-mentioned calculus of differences repeatedly, and determines each edge that these calculus of differencess are obtained
Nonzero value region common in the middle of difference image.Nonzero value region in the edge difference partial image obtained reflects virtual coil
The edge changed in time in the middle of region.Therefore, by getting rid of the non-zero in edge difference partial image in candidate edge
Value region, obtained be video pictures virtual coil region when present in stabilised edge feature, i.e. constitute stabilised edge
Pixel point set.Stabilised edge feature all has relative to gradual change and the sudden change of the outside environmental elements such as sunshine, shade, car light irradiation
There is the strongest robustness.Therefore these stabilised edge features obtained based on above-mentioned calculating may be used for triggering virtual region
Detection.Edge feature generation step obtains and preserves described stabilised edge feature, thus is formed and there is stabilised edge feature
Virtual coil region template, as follow-up, virtual region is triggered the foundation that judges.Fig. 5 shows through above-mentioned calculating side
After method processes, described virtual coil A, B are formed as virtual coil region template, the binary picture of the most described pixel point set
As representing, wherein comprise the stabilised edge feature that key area is had.From fig. 5, it is seen that key area C, D, E institute shape
The edge become is extracted and is stored in virtual coil region template as described stabilised edge feature, and the region G in Fig. 1
The edge formed requires or does not have stability to be filtered out owing to it does not meets spatial distribution.
On this basis, the present invention so that by detection by specific objective vehicle in the virtual coil region of video pictures
The edge feature change caused, it is achieved the triggering to virtual coil.When the picture of vehicle enters within virtual coil region,
Will necessarily cause the marginal information within virtual coil that significantly change occurs.Vehicle includes through the impact brought: vehicle
Self and car light light beam thereof, shade can make virtual coil region produce the marginal portion originally not having, and vehicle through out-of-date entirely
Key area described in portion or partial occlusion and stabilize it marginal information and change, such as in Fig. 6 automobile M picture pass through
Virtual coil A causes key area E all to be blocked and key area C is at least partially obscured.To vehicle self and car light thereof
Bundle, extraction and the detection at the produced edge of shade also are able to the effect reaching to trigger virtual coil to a certain extent, but examine
The reliability surveyed is strong, such as at automobile M itself and when being introduced into road surface corresponding for virtual coil A, and the irradiation of its Herba Plantaginis light beam
Likely have resulted in the marginal information in virtual coil A and produce change.But, in this case, owing to not hidden
Gear, in virtual coil A, the stabilised edge information of key area will not occur substantial variations, is only actually passed through dummy line at vehicle
The substantial variation of stabilised edge information just can be caused when enclosing region thus block key area, thus can be by by real
Time video pictures and described virtual coil region template match computing and the basis for estimation that triggers as virtual coil.
Fig. 7 shows the method flow diagram by detecting edge feature change triggers virtual coil.Detection edge feature becomes
Change to judge that the method for virtual coil comprises the following steps: step 1, for the video pictures of captured in real-time, according to default limit
Coordinate segmentation virtual coil region, boundary line.And then, in step 2, for the virtual coil region of real-time video picture, pass through limit
Edge extraction algorithm extracts the real-time marginal information in the presence of it, it is thus achieved that the binary image of marginal information represents in real time, its tool
Body algorithm includes Signal Pretreatment and the edge extracting utilizing Canny operator to realize.Step 3, according to virtual coil region mould
Stabilised edge information in plate, and the real-time edge letter in the virtual coil region of the real-time video picture obtained in step 2
Breath, carries out matching operation to the two, it is judged that whether the similarity of the two is less than minimum similar threshold value, less than this minimum similar threshold
Then thinking in the case of value that the edge that virtual coil region is comprised there occurs to significantly change, this virtual coil is triggered.In order to
Improve the degree of accuracy of matching operation in step 3, perform to calculate Hausdorff to described stabilised edge information and real-time marginal information
The matching process of distance.
Hausdorff distance is for measuring the matching degree of two point sets.Virtual coil region mould will be constituted in the present invention
The pixel of the stabilised edge in plate as a point set P1, and using form described real-time edge pixel as another point set
P2, thus Hausdorff distance H (P1, the P2)=max{h (P1, P2) between point set P1 and point set P2 can be calculated, h (P2, P1) }, h
(P1, P2) represents the maximum to the Euclidean distance of point set P2 of the pixel in point set P1,
Wherein a and b is belonging to the pixel of point set P1 and point set P2 respectively, and (a b) represents the Euclidean distance between a and b to D.Similar
, h (P2, P1) represents the maximum to the Euclidean distance of point set P1 of the pixel in point set P2.Hausdorff distance H (P1,
P2) represent the maximum among h (P1, P2) and h (P2, P1), composition virtual coil can be reflected by this Hausdorff distance
Matching degree between pixel point set and the pixel point set constituting described real-time edge of the stabilised edge in region template.When
Hausdorff distance H (P1, P2) is less than minimum similar threshold value Hthreshold, then it is assumed that the edge that virtual coil region is comprised is sent out
Having given birth to and significantly changed, this virtual coil is triggered.
On the basis of judging that virtual coil is triggered, highway video frequency speed-measuring method provided by the present invention and system
The time difference being triggered successively according to two virtual coils, carries out the calculating of testing the speed for specific objective vehicle, and it is concrete
Speed measuring method, because of essentially identical with prior art, does not repeats them here.
The present invention and then provide a kind of highway video frequency speed-measuring system based on virtual coil, including: virtual coil
Detection trigger module, for being set as dummy line by least two presumptive area in the video pictures of the shot by camera that tests the speed
Circle, and whether detect the picture formed by specific objective vehicle through described virtual coil；Speed measuring module, for according to specific
Target vehicle triggers the time difference of each described virtual coil successively and carries out vehicle speed measuring.Wherein, described virtual coil triggers inspection
Survey module specifically includes: key area stabilised edge characteristic extracting module, for extracting without the void of video pictures under vehicle-state
Intend spatial distribution and time-domain stability qualified stabilised edge information in coil region, formed and comprise described stabilised edge
Virtual coil region template；Edge feature change detection module, for the virtual coil region for captured in real-time video pictures
Extract real-time marginal information, and described stabilised edge information and real-time marginal information are carried out matching operation, it is judged that the two
Whether similarity is less than minimum similar threshold value；In the case of less than described minimum similar threshold value, determine that virtual coil is triggered.
Described key area stabilised edge characteristic extracting module includes: edge extracting module, for performing height for virtual coil region
This smothing filtering, then utilizes Canny operator to carry out edge detection algorithm, it is thus achieved that edge letter in the presence of virtual coil region
The binary image of breath represents；Key area identification module, in the presence of calculating described virtual coil region, edge is at image
Whether specific axis run length upwards, relatively described run length be more than distance of swimming threshold value, by run length more than distance of swimming threshold value
Edge as candidate edge；And the candidate edge of at least two frame video pictures being spaced specific duration is performed difference fortune
Calculate, it is thus achieved that edge difference partial image；The nonzero value region in edge difference partial image is got rid of, it is thus achieved that stabilised edge is special in candidate edge
Levy；Edge feature generation module, it is thus achieved that and preserve described stabilised edge feature, thus formed and there is the virtual of stabilised edge feature
Coil region template.Described edge feature change detection module includes: edge extraction module in real time, for for clapping in real time
The video pictures taken the photograph, according to default coordinate segmentation virtual coil region, boundary line, and then extracts it by Boundary extracting algorithm
In the presence of real-time marginal information；Edge matching module, performs described stabilised edge information and real-time marginal information
Hausdorff distance calculates, and judges that whether Hausdorff distance is less than minimum similar threshold value.
Visible, the present invention is in the middle of highway video frequency speed-measuring method and system, for virtual coil detection trigger mechanism
Improved, by virtual coil without the stabilised edge feature extraction in car status screen out, and will be due to occlusion etc.
The virtual coil region inner margin changing features that reason is caused is as the foundation judging triggering.Owing to stabilised edge feature is relative
Gradual change and sudden change in outside environmental elements such as sunshine, shade, car light irradiations all have the strongest robustness, can be prevented effectively from
The false triggering that these external interference are caused, whether daytime and night can keep reliable detection trigger performance；And
Occlusion is also stable to the change of virtual coil region inner margin feature, compared to gray scale detection, does not haves Yin Te
Fixed illumination condition and vehicle color and the triggering failure that causes.The present invention is above-mentioned reliable, accurate and high by constructing
The virtual coil trigger mechanism of real-time, has fully ensured that the highway video frequency speed-measuring method realized based on virtual coil principle
And the high-quality of system runs.
The above, the only detailed description of the invention of the present invention, the present invention can be applied in miscellaneous equipment；More than retouch
Size and quantity in stating are the most informative, and those skilled in the art can select suitable application chi according to actual needs
Very little, without deviating from the scope of the present invention.Protection scope of the present invention is not limited thereto, any technology being familiar with the art
Personnel in the technical scope that the invention discloses, the change that can readily occur in or replacement, all should contain the protection model in the present invention
Within enclosing.Therefore, protection scope of the present invention should be as the criterion with the protection domain that claim is defined.
Claims (7)
1. a highway video frequency speed-measuring method based on virtual coil, including: virtual coil detection trigger step, will test the speed
At least two presumptive area in the video pictures of shot by camera is set as virtual coil, and detects by specific objective car
Formed picture whether through described virtual coil；Test the speed step, triggers each described void successively according to specific objective vehicle
The time difference intending coil carries out vehicle speed measuring；It is characterized in that, described virtual coil detection trigger step specifically includes:
Key area stabilised edge characteristic extraction step, for extracting without in the virtual coil region of video pictures under vehicle-state
Spatial distribution and time-domain stability qualified stabilised edge information, form the virtual coil region comprising described stabilised edge
Template；
Edge feature change-detection step, believes for the real-time edge of virtual coil extracted region for captured in real-time video pictures
Breath, performs Hausdorff distance and calculates described stabilised edge information and real-time marginal information, and judge Hausdorff away from
From whether less than minimum similar threshold value；In the case of less than described minimum similar threshold value, determine that virtual coil is triggered；
Wherein, described key area stabilised edge characteristic extraction step includes following sub-step:
Edge extracting step, for performing Gaussian smoothing filter for virtual coil region, then utilizes Canny operator to carry out limit
Edge detection algorithm, it is thus achieved that the binary image of marginal information represents in the presence of virtual coil region；
Key area identification step, calculates the specific axis distance of swimming upwards at image of edge in the presence of described virtual coil region
Run length, whether more than distance of swimming threshold value, is more than the edge of distance of swimming threshold value as candidate limit by length, relatively described run length
Edge；And the candidate edge of at least two frame video pictures being spaced specific duration is performed calculus of differences, it is thus achieved that edge difference component
Picture；The nonzero value region in edge difference partial image is got rid of, it is thus achieved that stabilised edge feature in candidate edge；
Edge feature generation step, it is thus achieved that and preserve described stabilised edge feature, thus form the void with stabilised edge feature
Intend coil region template.
Highway video frequency speed-measuring method based on virtual coil the most according to claim 1, it is characterised in that described limit
Edge extraction step includes following sub-step: step 1, utilizes Canny edge detector that filtering image carries out edge for the first time and carries
Take；Step 2, after filtering image is carried out top cap conversion, carries out second time edge extracting with Canny edge detector；Step 3,
After image after converting top cap carries out logarithmic transformation, carry out third time edge extracting with Canny edge detector；Step 4, will
The result images superposition of three edge extractings；Step 5, carries out skeletonizing process to the image after superposition, obtains marginal information.
Highway video frequency speed-measuring method based on virtual coil the most according to claim 1, it is characterised in that described pass
In the identification step of key range, for edge in the presence of virtual coil region, calculate each edge respectively in the X-axis of image and Y-axis
The number N of the uninterrupted pixel being had on directionxAnd Ny, and then be NxAnd NyDistribute different weights, thus counted by weighting
Calculate each edge at described specific axis run length H=α 1 N upwardsx+α2·Ny, wherein α 1 and α 2 represents weight coefficient.
Highway video frequency speed-measuring method based on virtual coil the most according to claim 1, it is characterised in that described pass
In the identification step of key range, described candidate edge is first carried out Extension algorithm edge is expanded for have bigger line thickness
Marginal zone, then perform described calculus of differences.
Highway video frequency speed-measuring method based on virtual coil the most according to claim 1, it is characterised in that described
Hausdorff distance calculates and includes: using the pixel of the stabilised edge information in composition virtual coil region template as one
Point set P1, and using form described real-time marginal information pixel as another point set P2, calculate point set P1 and point set P2 it
Between Hausdorff distance H (P1, P2)=max{h (P1, P2), h (P2, P1), wherein h (P1, P2) represents in point set P1
Pixel to the maximum of the Euclidean distance of point set P2,Wherein a and b is belonging to respectively
The pixel of point set P1 and point set P2, (a b) represents the Euclidean distance between a and b to D；H (P2, P1) represents the picture in point set P2
Vegetarian refreshments to the maximum of the Euclidean distance of point set P1,Wherein a and b is belonging to a little respectively
The pixel of collection P2 and point set P1, (a b) represents the Euclidean distance between a and b to D.
6. a highway video frequency speed-measuring system based on virtual coil, including: virtual coil detection trigger module, being used for will
At least two presumptive area in the video pictures of the shot by camera that tests the speed is set as virtual coil, and detects by specific mesh
Whether the picture that mark vehicle is formed is through described virtual coil；Speed measuring module, for triggering successively respectively according to specific objective vehicle
The time difference of individual described virtual coil carries out vehicle speed measuring；It is characterized in that, described virtual coil detection trigger module is specifically wrapped
Include:
Key area stabilised edge characteristic extracting module, for extracting without in the virtual coil region of video pictures under vehicle-state
Spatial distribution and time-domain stability qualified stabilised edge information, form the virtual coil region comprising described stabilised edge
Template；
Edge feature change detection module, believes for the real-time edge of virtual coil extracted region for captured in real-time video pictures
Breath, and described stabilised edge information and real-time marginal information are performed Hausdorff distance calculating, and judge
Whether Hausdorff distance is less than minimum similar threshold value；Virtual coil is determined in the case of less than described minimum similar threshold value
It is triggered；
Wherein, described key area stabilised edge characteristic extracting module includes following submodule:
Edge extracting module, for performing Gaussian smoothing filter for virtual coil region, then utilizes Canny operator to carry out limit
Edge detection algorithm, it is thus achieved that the binary image of marginal information represents in the presence of virtual coil region；
Key area identification module, calculates the specific axis distance of swimming upwards at image of edge in the presence of described virtual coil region
Run length, whether more than distance of swimming threshold value, is more than the edge of distance of swimming threshold value as candidate limit by length, relatively described run length
Edge；And the candidate edge of at least two frame video pictures being spaced specific duration is performed calculus of differences, it is thus achieved that edge difference component
Picture；The nonzero value region in edge difference partial image is got rid of, it is thus achieved that stabilised edge feature in candidate edge；
Edge feature generation module, it is thus achieved that and preserve described stabilised edge feature, thus form the void with stabilised edge feature
Intend coil region template.
Highway video frequency speed-measuring system based on virtual coil the most according to claim 6, it is characterised in that described limit
Edge changing features detection module includes:
Edge extraction module in real time, for the video pictures for captured in real-time, according to default boundary line coordinate segmentation
Virtual coil region, and then extract the real-time marginal information in the presence of it by Boundary extracting algorithm；
Edge matching module, performs Hausdorff distance to described stabilised edge information and real-time marginal information and calculates, and sentence
Whether disconnected Hausdorff distance is less than minimum similar threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410576699.6A CN104267209B (en) | 2014-10-24 | 2014-10-24 | Method and system for expressway video speed measurement based on virtual coils |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410576699.6A CN104267209B (en) | 2014-10-24 | 2014-10-24 | Method and system for expressway video speed measurement based on virtual coils |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104267209A CN104267209A (en) | 2015-01-07 |
| CN104267209B true CN104267209B (en) | 2017-01-11 |
Family
ID=52158750
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410576699.6A Active CN104267209B (en) | 2014-10-24 | 2014-10-24 | Method and system for expressway video speed measurement based on virtual coils |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104267209B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105551265B (en) * | 2015-02-09 | 2017-10-20 | 南京蓝泰交通设施有限责任公司 | A kind of magnitude of traffic flow detection method based on virtual detection band |
| CN105448086A (en) * | 2015-07-22 | 2016-03-30 | 南通大学 | Traffic flow detection method based on virtual detection bands |
| CN105354573B (en) * | 2015-12-15 | 2019-03-22 | 重庆凯泽科技股份有限公司 | A kind of container licence plate recognition method and system |
| CN111650392A (en) * | 2020-07-03 | 2020-09-11 | 东北大学 | Metal sheet movement speed detection method based on linear array camera stereoscopic vision |
| CN112557812B (en) * | 2020-11-24 | 2022-06-03 | 山东理工大学 | Small current ground fault positioning method and system based on Hausdorff distance |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1984236A (en) * | 2005-12-14 | 2007-06-20 | 浙江工业大学 | Method for collecting characteristics in telecommunication flow information video detection |
| CN102136196A (en) * | 2011-03-10 | 2011-07-27 | 北京大学深圳研究生院 | Vehicle velocity measurement method based on image characteristics |
| CN102324183A (en) * | 2011-09-19 | 2012-01-18 | 华中科技大学 | Vehicle detection and grasp shoot method based on compound virtual coil |
-
2014
- 2014-10-24 CN CN201410576699.6A patent/CN104267209B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1984236A (en) * | 2005-12-14 | 2007-06-20 | 浙江工业大学 | Method for collecting characteristics in telecommunication flow information video detection |
| CN102136196A (en) * | 2011-03-10 | 2011-07-27 | 北京大学深圳研究生院 | Vehicle velocity measurement method based on image characteristics |
| CN102324183A (en) * | 2011-09-19 | 2012-01-18 | 华中科技大学 | Vehicle detection and grasp shoot method based on compound virtual coil |
Non-Patent Citations (1)
| Title |
|---|
| 基于视频虚拟线圈的交通流参数检测;尹朝征;《中国优秀博硕士学位论文全文数据库 (硕士) 信息科技辑》;20040615(第02期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104267209A (en) | 2015-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110178167B (en) | Intersection violation video identification method based on cooperative relay of cameras | |
| CN110285793A (en) | A kind of Vehicular intelligent survey track approach based on Binocular Stereo Vision System | |
| CN106373394B (en) | Vehicle detection method and system based on video and radar | |
| CN103778786B (en) | A kind of break in traffic rules and regulations detection method based on remarkable vehicle part model | |
| CN104267209B (en) | Method and system for expressway video speed measurement based on virtual coils | |
| CN110322702A (en) | A kind of Vehicular intelligent speed-measuring method based on Binocular Stereo Vision System | |
| CN107463890B (en) | A kind of Foregut fermenters and tracking based on monocular forward sight camera | |
| CN102867416B (en) | Vehicle part feature-based vehicle detection and tracking method | |
| CN110175576A (en) | A kind of driving vehicle visible detection method of combination laser point cloud data | |
| CN108615358A (en) | A kind of congestion in road detection method and device | |
| CN106845364B (en) | Rapid automatic target detection method | |
| CN109064495A (en) | A kind of bridge floor vehicle space time information acquisition methods based on Faster R-CNN and video technique | |
| CN106096525A (en) | A kind of compound lane recognition system and method | |
| CN105844959A (en) | Method for determining entering of vehicles to parking spaces, device, method for determining exiting of vehicles from parking spaces, and device | |
| CN111563469A (en) | Method and device for identifying irregular parking behaviors | |
| CN103686083B (en) | Real-time speed measurement method based on vehicle-mounted sensor video streaming matching | |
| CN107066968A (en) | The vehicle-mounted pedestrian detection method of convergence strategy based on target recognition and tracking | |
| CN107315095A (en) | Many vehicle automatic speed-measuring methods with illumination adaptability based on Video processing | |
| CN107031661A (en) | A kind of lane change method for early warning and system based on blind area camera input | |
| CN106446807A (en) | Well lid theft detection method | |
| CN104063882A (en) | Vehicle video speed measuring method based on binocular camera | |
| CN106778540A (en) | Parking detection is accurately based on the parking event detecting method of background double layer | |
| Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
| CN110718061A (en) | Traffic intersection vehicle flow statistical method and device, storage medium and electronic equipment | |
| CN107808524A (en) | A kind of intersection vehicle checking method based on unmanned plane |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Method and system for expressway video speed measurement based on virtual coils Effective date of registration: 20171219 Granted publication date: 20170111 Pledgee: Hangzhou hi tech Company limited by guarantee Pledgor: Zhejiang Li Shi Science and Technology Co., Ltd. Registration number: 2017330000310 |