CN103927878B - A kind of automatic shooting device for parking offense and automatically grasp shoot method - Google Patents

A kind of automatic shooting device for parking offense and automatically grasp shoot method Download PDF

Info

Publication number
CN103927878B
CN103927878B CN201410143566.XA CN201410143566A CN103927878B CN 103927878 B CN103927878 B CN 103927878B CN 201410143566 A CN201410143566 A CN 201410143566A CN 103927878 B CN103927878 B CN 103927878B
Authority
CN
China
Prior art keywords
point
prime
angle
image
parking
Prior art date
Application number
CN201410143566.XA
Other languages
Chinese (zh)
Other versions
CN103927878A (en
Inventor
张勇
王韬
王军群
谭中慧
袁彬
Original Assignee
中海网络科技股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中海网络科技股份有限公司 filed Critical 中海网络科技股份有限公司
Priority to CN201410143566.XA priority Critical patent/CN103927878B/en
Publication of CN103927878A publication Critical patent/CN103927878A/en
Application granted granted Critical
Publication of CN103927878B publication Critical patent/CN103927878B/en

Links

Abstract

The invention discloses a kind of automatic shooting device for parking offense and automatic grasp shoot method, comprise front end candid photograph unit, transmission network and management server, front end is captured unit and is comprised electronic zoom camera lens, smart camera and The Cloud Terrace, electronic zoom camera lens is located at the front end of smart camera and is connected with smart camera, smart camera is located on The Cloud Terrace, and smart camera is also connected with management server by transmission network. Front end is captured unit and is adopted wide-angle lens to monitor whole guarded region, the movement locus of each car in tracing and monitoring region, when finding after Parking by the mode of track following, determine center of gravity and the area of stopped vehicle, determine the angle of cloud platform rotation, pitching by geometric operation, controlling The Cloud Terrace makes target be positioned at the center of image, by controlling lens zoom, the size of adjustment aim in image, making it be applicable to doing graphical analysis is license plate number identification, and license plate number has been identified the electronic zoom camera lens of rear smart camera control and returned to wide-angle monitoring pattern.

Description

A kind of automatic shooting device for parking offense and automatically grasp shoot method

Technical field

The present invention relates to a kind of supervising device of monitoring the parking offense in visual field, more particularly,Relate to a kind of automatic shooting device for parking offense and automatic grasp shoot method.

Background technology

Along with social development, follow the continuous rising of vehicle guaranteeding organic quantity, various violating the regulations stoppingThe phenomenon of car is also more and more general, abroad, generally all directly investigates and prosecutes and breaks rules and regulations by traffic-police is on-the-spotParking behavior, also has the aid police such as Countries mandate traffic management assistant, cleaner to investigate and prosecute violating the regulationsStop, seldom adopt the mode of electronic monitoring to investigate and prosecute parking offense behavior. And at home, have partThe mode that city pilot is installed automatic snapshot controller during process in CCTV camera violating the regulations rear end is captured violating the regulationsStop, but spread not yet. Existing tripod head type video traffic event detection device is by settingPresetting bit, pre-sets scene parameter, by the image of focusing video camera collection is analyzedProcess, judged whether parking offense, but there is following defect in current device: be 1) fixedZoom lens can not convergent-divergent, and this will cause the scope of capturing greatly to be dwindled, and has brought also to the identification of taking picturesCertain difficulty. 2) most tripod head type video traffic event detection devices are output as SD signal, aobviousSo this does not meet video monitoring towards high definition, web development direction. 3) video detecting deviceImage processing apparatus is placed on camera rear end by majority, and this pattern has reduced the integrated level of system.

Summary of the invention

Order of the present invention is for the defect existing in prior art, provides one automatically to knowParking offense event in not whole monitoring visual field, identifies the vehicles peccancy number-plate number automatically, and willVehicles peccancy information and evidence violating the regulations upload onto the server and process for traffic control personnel, and whole process is wholeAutomatically complete the automatic shooting device for parking offense and automatic candid photograph side without manual interventionMethod.

For achieving the above object, the inventive system comprises: comprise front end candid photograph unit and pass throughTransmission network captures with front end the management server that unit is connected, and described front end is captured unit and comprised electricityMoving zoom camera lens, smart camera and The Cloud Terrace, described electronic zoom camera lens be located at smart camera beforeEnd is also connected with smart camera, and described smart camera is located on The Cloud Terrace.

Described smart camera comprises control module, and described control module is compiled with video respectively by busDecoder module, Parking analysis module, car plate identification module, image capture module andTransport module is connected; The input of described image capture module is connected with electronic zoom camera lens, described inThe output of image capture module also respectively with coding and decoding video module, Parking video analysis mouldPiece and car plate identification module are connected.

Described transmission network adopts wire transmission network or wireless-transmission network.

Described The Cloud Terrace adopts thread screw type of drive, and angle control precision reaches 0.1 °.

Described electronic zoom camera lens adopts the wide-angle lens of zooming range in 20 times.

Described front end is captured unit and is also comprised the light filling illuminating lamp being connected with smart camera, described benefitIllumination bright lamp is distributed flash lamp.

Automatic grasp shoot method of the present invention is as follows:

First, front end is captured unit and is monitored whole guarded region, each in tracing and monitoring regionThe movement locus of car, when finding, after Parking, to determine stopped vehicle by the mode of track followingCenter of gravity and size, determine the angle of cloud platform rotation, pitching by geometric operation, control cloudPlatform makes target be positioned at the center of image, by controlling lens zoom, and large in image of adjustment aimLittle, making it be applicable to doing graphical analysis is license plate number identification, and it is extensive that license plate number has been identified rear candid photograph cameraContinuing to do Parking to wide-angle monitoring pattern again detects.

The mode concrete steps of described track following are as follows:

1) according to the video image size IMAGE_WIDTH, the IMAGE_HEIGHT that capture unit collection,Every two field picture is divided into piece BLOCK_WIDTH, BLOCK_HEIGHT by required required precision, dividesCutting rear image size is

WID = IMAGE _ WIDTH BLOCK _ WIDTH

HEI = IMAGE _ HEIGHT BLOCK _ HEIGHT

2) select frame period to carry out inter-frame difference by required required precision, and by the threshold value of difference result and settingPhase, TwoVal represents the result after binarization segmentation, and DifGray represents inter-frame difference result, and Th is that binaryzation is dividedCut threshold value, this value is by the study analysis of several scenes, and the empirical value of acquisition, when the poor result of frame is greater than settingAfter threshold value, binarization segmentation result is made as to 255 and is white, otherwise binaryzation result is 0 to be black,And then carry out binarization segmentation

TwoVal = 255 , DifGray > Th 0 , DifGray < Th ,

3) for the image after cutting apart, selected 3 × 3 window size, the sad value of each in calculation window,Determine the most obvious point of gray-value variation in each window with this, as the Corner in this detection range,As the foundation of tracking and matching;

4) in the time series of each frame, find using all direction search method the point that approaches the most with Corner asJoin a little, with the tracking of this realize target, the relevant information of a record point, and then the motion rail of acquisition target vehicleMark;

The described pursuit path that utilizes is found the concrete grammar stopping:

For the moving target in video scene, its movement locus mainly stops or disappears in following situation:

1) moving vehicle exceedes video detection range: this kind of situation is by video detection range is set, when track surpassesWhile crossing the region of setting, think that vehicle exceedes detection range, respective objects is not processed;

2) following the tracks of angle point drops on road surface: when road surface exists the more or shooting of ponding, the strong road surface of illumination shadeWhen the situations such as in machine shake and road track cut-off rule and zebra stripes are more occur, will in object matching processAngle point is chosen on road surface and causes the termination of pursuit path, and when this kind of situation occurs, its track is around without fixingThe segmentation object of size, does not process this type of track;

3) moving target stagnation of movement has Parking to occur: when this kind of situation occurs, in video detection range, nothing is newThe appearance of trail angle point, in vehicle matching process, angle point is longer in certain position time of staying, and exists around trackA certain size Target Segmentation result, when this kind of situation occurs, thinks and has Parking to occur.

Described geometric operation is determined the method for the angle of cloud platform rotation, pitching: mobile camera moving positionParameter, i.e. level angle α, vertically faces upward (bowing) angle θ,

O is video camera position, and O' is that video camera is projected in ground point, and A is the former focus point of video camera,B is parking offense event occurrence positions, and in the time utilizing track following to obtain the generation of parking offense event, B point is satMark is B point center of gravity position, establishes it for B (x, y), and A point coordinates can obtain by camera, and it is former cameraFocus position, establishes it for A (x0,yo);

Horizontally rotate in process at video camera, O'A=O'C, because video camera sets up location aware, and A, B pointLocation aware, therefore calculate the actual distance of O'A and O'B to real coordinate position transforming relationship according to pixel coordinateFrom, in like manner can calculate the actual range that AM and BM are corresponding, therefore in Δ O'MB, can calculate video cameraLevel angleThe angle of depression that in like manner can calculate video camera is &theta; = arctan O &prime; B O O &prime; - arctan O &prime; C O O &prime; = arctan O &prime; B O O &prime; - arctan O &prime; A O O &prime; , Can obtain thus two, video camera and rotate ginsengNumber, can move to assigned address by video camera;

&alpha; = arctan BM O &prime; M &theta; = arctan O &prime; B O O &prime; - arctan O &prime; A O O &prime;

Due in actual scene, the relation of B and A has several as follows,

1) pixel coordinate is to real coordinate position transforming relationship

If in 3D rendering, the coordinate of certain point in actual coordinates is W (X, Y, Z), by the ratio of similar trianglesRelation can obtain its subpoint I (x', y'), wherein

x &prime; = f * X f - Z , y &prime; = f * Y f - Z - - - ( 1.1 )

Owing to having comprised the division arithmetic of variable Z in above formula, therefore it belongs to nonlinear change, adopt homogeneous coordinatesMethod converts thereof into linear matrix and calculates;

The homogeneous coordinates that any point in space (X, Y, Z) is corresponding are defined as (kX, kY, kZ, k), k be transformation of scale because ofSon, can be any non-zero constant, and realizing its space transformational relation by homogeneous coordinates can be expressed as

(kX,kY,kZ,k)T=T*(x,y,z,1)T(1.2)

Wherein transformation matrix T is the square formation of 4 × 4;

Adopt the method for two-dimensional transform to realize the conversion of pixel distance and actual range, two-dimensional transform matrix is one 3 × 3Square formation, the pixel coordinate of establishing certain point is (x, y), the coordinate of corresponding actual range is (X, Y), in homogeneous coordinatesTransformational relation under system can be expressed as:

KX KY K = a 00 a 01 a 02 a 10 a 11 a 12 a 20 a 21 1 x y 1 - - - ( 1.3 )

The expression formula that can obtain (X, Y) by above formula is

X = a 00 x + a 01 y + a 02 a 20 x + a 21 y + 1 Y = a 10 x + a 11 y + a 12 a 20 x + a 21 y + 1 - - - ( 1.4 )

From formula (1.3) homogeneous coordinates equation, in transformation matrix, contain eight unknown numbers, as long as calibrateFour known points, can solve this transformation matrix, in actual scene, due to urban road, highwayTrack cut-off rule length and lane width have clear and definite regulation, and the pixel coordinate of calibration point can be direct by programObtain, can set up thus pixel-distance map relation, hereinafter to be referred as mapping table, for obtaining above-mentioned mapping relations,Set cut-off rule the behavior of place, regulation cut-off rule top and be expert at zero point, the track of known distance in each trackCut-off rule set positions calibration point, brings demarcation each point into formula (1.4) and can obtain pixel-range conversion matrixIn each coefficient, and then obtain the mapping relations of pixel-distance in the horizontal direction and in vertical direction.

Electronic zoom camera lens that described front end is captured unit will collect through image capture moduleImage offers coding and decoding video module, Parking analysis module and car plate identification mouldPiece, Parking analysis module is carried out analyzing and processing by the traffic video image to continuous,Be tested with without Parking and occur;

In the time having detected that Parking occurs, camera focus is arrived suspected locations, the driving of going forward side by side by cloud platform rotationBoard is searched, and in the time that video camera turns to assigned address but do not find car plate, thinks false alarm, not to this thingPart is processed, and video camera turns to former presetting bit;

When video camera finds after car plate, this vehicle is processed, and in database, searched in a few minutes and beNo have this car plate record of breaking rules and regulations, if had, and not duplicate record, video camera moves to presetting bit, if somewhatIn clock, without this vehicle registration, capture 3 minutes, write down this car plate, and take 4 differences according to different focalPicture, records this vehicle peccancy docking process,

Be angle and the lens parameters that control module is adjusted The Cloud Terrace, obtain the spy of stop vehicle violating the regulationsWrite image, identify vehicles peccancy car plate and captured after, by transmission network by wide-angle fieldScape image and close-up image transfer to management server;

Finally, control module control The Cloud Terrace and automatic zoom camera lens parameter return to wide-angle state,Break rules and regulations automatically to capture next time.

Compared with prior art, adopt the present invention to there is following technique effect:

1, large scene parking offense detects

Electronic zoom camera lens, smart camera are selected high-definition camera, ensure car in whole monitoring rangeClear picture, ensures that on distal-most end vehicle image, comprising enough pixels sentences for event detectionDisconnected; On detection algorithm, adopt the mode of vehicle tracking, follow the tracks of the movement locus of each car,Judge vehicle-state according to movement locus, the generation that this mode can effectively reduce erroneous judgement, fail to judge.

2, vehicle is accurately located

To parking offense vehicle punishment according to comprising vehicle distant view picture and license plate number car clearlyClose shot picture, automatically captures front end when parking offense and finds after parking offense, needs to obtain and stopsThe precise position information of vehicle can be controlled The Cloud Terrace upset, lens zoom is taken the near of vehicles peccancyScape picture, if vehicle location inaccuracy just cannot be captured effective vehicle close shot picture. ThisBright automatic shooting device adopts threedimensional model to shine upon whole monitoring visual field, ensures vehicle close shot imageCapture precision.

3, The Cloud Terrace, camera lens are accurately controlled

In 100 meters of monitoring ranges, take the close shot image of vehicle, angle of visual field scope is about 2 °~23 °, zooming range is 20 times of left and right, take be of moderate size, the vehicle close shot of clear picturePicture needs accurately control The Cloud Terrace and camera lens, particularly takes visual field distal view picture, and the whole angle of visual field onlyHave 2 °, cradle head control slightly error will lose objects. Automatic shooting device of the present invention, adoptsWith the high accuracy The Cloud Terrace of thread screw type of drive, angle control precision reaches 0.1 °, and adopts pre-Set camera lens coordinates with The Cloud Terrace, can accurately catch all vehicle targets in visual field.

4, the processing of invalid record

Full-automatic parking offense detection system based on video unavoidably can produce invalid record,As the record of the record of number-plate number None-identified, the incomplete record of car plate and repetition, when thisRecord a bit quantity and reach certain ratio, bring huge work can to recording processing personnel violating the regulationsAmount, whole system will lose its availability. It is single that automatic shooting device of the present invention takes front end to captureUnit and management server dual logging filter type, filtering rule can arrange flexibly, can effectively controlThe quantity of invalid record, ensures that the candid photograph of whole system is efficient.

Brief description of the drawings

Fig. 1 is the principle schematic of a kind of automatic shooting device for parking offense of the present invention;

Fig. 2 track following design sketch;

Fig. 3 geometrical model figure;

Fig. 4 location diagram;

Fig. 5 coordinate system conversion schematic diagram;

Fig. 6 camera calibration schematic diagram.

Detailed description of the invention

Further illustrate technical scheme of the present invention below in conjunction with drawings and Examples.

Referring to Fig. 1, the present invention includes front end and capture unit 11, transmission network 12 and management clothesBusiness device 13, front end capture unit 11 comprise electronic zoom camera lens 111, smart camera 112 andThe Cloud Terrace 113, smart camera 112 comprises control module 1121, control module 1121 is passed through bus1122 respectively with coding and decoding video module 1123, Parking analysis module 1124, car plateIdentification module 1125, image capture module 1126 and transport module 1127 are connected, and control mouldPiece 1121 is connected with electronic zoom camera lens 111 and The Cloud Terrace 113 respectively by RS485 cable,Image capture module 1126 is also connected with electronic zoom camera lens 111, for receiving imaging signal;Image capture module 1126 also respectively with coding and decoding video module 1123, Parking video analysisModule 1124 and car plate identification module 1125 are connected. Electronic zoom camera lens 111 is located at intelligent phaseThe front end of machine 112 is also connected with smart camera, the zooming range of electronic zoom camera lens 20 times withIn, smart camera 112 is located on The Cloud Terrace 113, and smart camera 112 is also by transmission network 12Be connected with management server 13.

Implementation method of the present invention is as follows: first, front end is captured unit 11 and monitored whole monitored spaceTerritory, the movement locus of each car in tracing and monitoring region, when the mode by track following is sent outAfter existing Parking, determine center of gravity and the size of stopped vehicle, determine cloud by geometric operationPlatform 113 rotates, the angle of pitching, controls The Cloud Terrace 113 and makes target be positioned at the center of image, passes throughControl lens zoom, the size of adjustment aim in image, making it be applicable to doing graphical analysis is car plateNumber identification, license plate number has been identified rear candid photograph camera and has been returned to wide-angle monitoring pattern and continue to do parking thingPart detects.

For accurately representing accurately to reflect state of motion of vehicle by vehicle movement track, the present invention adopts based on angle pointMoving body track method, specific implementation process is as follows:

Traditional moving Object Segmentation method have powerful connections calculus of finite differences and frame differential method, background subtraction point-score is to backgroundDependence is large and time complexity is larger, detects and has certain influence, for improving system for the real-time of systemSystem real-time, reduces operand, and the present invention adopts frame differential method to carry out moving Object Segmentation, and to being partitioned intoTarget carry out SAD calculating, in the certain limit of moving target determine its optimum angle point, it is followed the tracks of,And the movement locus of definite moving target, concrete steps are as follows:

1) according to the big or small IMAGE_WIDTH, the IMAGE_HEIGHT that gather video image, by every frameImage is divided into a certain size piece BLOCK_WIDTH, BLOCK_HEIGHT, cuts apart rear image sizeFor

WID = IMAGE _ WIDTH BLOCK _ WIDTH

HEI = IMAGE _ HEIGHT BLOCK _ HEIGHT

2) select certain frame period (as 3 frames, 5 frames) to carry out inter-frame difference, and by difference result and settingThreshold, and then carry out binarization segmentation

TwoVal = 255 , DifGray > Th 0 , DifGray < Th

3) for the image after cutting apart, the window size of selected 3*3, the sad value of each in calculation window,Determine the most obvious point of gray-value variation in each window with this, as the Corner in this detection range,As the Main Basis of tracking and matching;

4) in the time series of each frame, find using all direction search method the point that approaches the most with Corner asJoin a little, with the tracking of this realize target, the relevant information of a record point, and then the motion rail of acquisition target vehicleMark;

5) because the normal transport condition of vehicle on road is straight-line travelling, in tracing process due to feature angle pointDrift, may cause pursuit path can not accurately reflect state of motion of vehicle, thus this type of track should be rejected,To facilitate the judgement of follow-up trajectory analysis.

Use this method follows the tracks of formation pursuit path as shown in Figure 2.

The described pursuit path that utilizes is found the concrete grammar stopping:

For the moving target in video scene, its movement locus mainly stops or disappears in following situation:

1) moving vehicle exceedes video detection range: this kind of situation is by video detection range is set, when track surpassesWhile crossing the region of setting, think that vehicle exceedes detection range, respective objects is not processed;

2) following the tracks of angle point drops on road surface: when road surface exists the more or shooting of ponding, the strong road surface of illumination shadeWhen the situations such as in machine shake and road track cut-off rule and zebra stripes are more occur, will in object matching processAngle point is chosen on road surface and causes the termination of pursuit path, and when this kind of situation occurs, its track is around without fixingThe segmentation object of size, does not process this type of track;

3) moving target stagnation of movement has Parking to occur: when this kind of situation occurs, in video detection range, nothing is newThe appearance of trail angle point, in vehicle matching process, angle point is longer in certain position time of staying, and exists around trackA certain size Target Segmentation result, when this kind of situation occurs, thinks and has Parking to occur.

Described geometric operation determines that The Cloud Terrace 113 rotates, the method for the angle of pitching: mobile camera moving positionThe parameter of putting, i.e. level angle α, vertically faces upward (bowing) angle θ,

Referring to Fig. 3, O is video camera position, and O' is that video camera is projected in ground point, and A is video cameraFormer focus point, B is parking offense event occurrence positions, is utilizing track following to obtain the generation of parking offense eventTime, B point coordinates is B point center of gravity position, establishes it for B (x, y), and A point coordinates can obtain by camera,It is former camera focus position, establishes it for A (x0,yo);

Horizontally rotate in process at video camera, O'A=O'C, because video camera sets up location aware, and A, B pointLocation aware, therefore calculate the actual distance of O'A and O'B to real coordinate position transforming relationship according to pixel coordinateFrom, in like manner can calculate the actual range that AM and BM are corresponding, therefore in Δ O'MB, can calculate video cameraLevel angleThe angle of depression that in like manner can calculate video camera is &theta; = arctan O &prime; B O O &prime; - arctan O &prime; C O O &prime; = arctan O &prime; B O O &prime; - arctan O &prime; A O O &prime; , Can obtain thus two, video camera and rotate ginsengNumber, can move to assigned address by video camera;

&alpha; = arctan BM O &prime; M &theta; = arctan O &prime; B O O &prime; - arctan O &prime; A O O &prime;

Referring to Fig. 4, due in actual scene, the relation of B and A has several as follows, due in actual scene,The relation of B and A has several as follows, only the interior position relationship in region 1 is described to calculating herein, other position relationship computational methodsTo repeat no more.

1) pixel coordinate is to real coordinate position transforming relationship

Referring to Fig. 5, establishing the coordinate of certain point in actual coordinates in 3D rendering is W (X, Y, Z), by similar triangleThe proportionate relationship of shape can obtain its subpoint I (x', y'), wherein

x &prime; = f * X f - Z , y &prime; = f * Y f - Z - - - ( 1.1 )

Owing to having comprised the division arithmetic of variable Z in above formula, therefore it belongs to nonlinear change, adopt homogeneous coordinatesMethod converts thereof into linear matrix and calculates;

The homogeneous coordinates that any point in space (X, Y, Z) is corresponding are defined as (kX, kY, kZ, k), k be transformation of scale because ofSon, can be any non-zero constant, and realizing its space transformational relation by homogeneous coordinates can be expressed as

(kX,kY,kZ,k)T=T*(x,y,z,1)T(1.2)

Wherein transformation matrix T is the square formation of 4 × 4;

Adopt the method for two-dimensional transform to realize the conversion of pixel distance and actual range, two-dimensional transform matrix is one 3 × 3Square formation, the pixel coordinate of establishing certain point is (x, y), the coordinate of corresponding actual range is (X, Y), in homogeneous coordinatesTransformational relation under system can be expressed as:

KX KY K = a 00 a 01 a 02 a 10 a 11 a 12 a 20 a 21 1 x y 1 - - - ( 1.3 )

The expression formula that can obtain (X, Y) by above formula is

X = a 00 x + a 01 y + a 02 a 20 x + a 21 y + 1 Y = a 10 x + a 11 y + a 12 a 20 x + a 21 y + 1 - - - ( 1.4 )

From formula (1.3) homogeneous coordinates equation, in transformation matrix, contain eight unknown numbers, as long as calibrateFour known points, can solve this transformation matrix, in actual scene, due to urban road, highwayTrack cut-off rule length and lane width have clear and definite regulation, and the pixel coordinate of calibration point can be direct by programObtain, can set up thus pixel-distance map relation, hereinafter to be referred as mapping table, in the horizontal direction and Vertical SquareThe demarcation of upwards selecting is counted more, and the mapping table accuracy rate simulating in respective direction is higher, but too muchCalibration point increased the amount of calculation of program, reduced the speed of service of program. Therefore should select the mark of right quantityFixed point is calculated. In Fig. 6 regulation below the behavior of place, white cut-off rule top be expert at zero point and (schemeThe horizontal direction distance of middle AD place straight line corresponding points is zero), the track cut-off rule of known distance in each trackSet positions calibration point.

Bring demarcation each point into formula (1.4) and can obtain each coefficient in pixel-range conversion matrix, and then obtainThe mapping relations of pixel-distance in the horizontal direction and in vertical direction.

Described front end is captured the electronic zoom camera lens 111 of unit 11 through image capture module 1126The image collecting is offered to coding and decoding video module 1123, Parking analysis module1124 and car plate identification module 1125, Parking analysis module 1124 is by continuouslyTraffic video image carry out analyzing and processing, be tested with without Parking and occur;

In the time having detected that Parking occurs, camera focus is arrived suspected locations, the driving of going forward side by side by cloud platform rotationBoard is searched, and in the time that video camera turns to assigned address but do not find car plate, thinks false alarm, not to this thingPart is processed, and video camera turns to former presetting bit;

When video camera finds after car plate, this vehicle is processed, and in database, searched in a few minutes and beNo have this car plate record of breaking rules and regulations, if had, and not duplicate record, video camera moves to presetting bit, if somewhatIn clock, without this vehicle registration, capture 3 minutes, write down this car plate, and take 4 differences according to different focalPicture, records this vehicle peccancy docking process,

Be angle and the lens parameters that control module 1121 is adjusted The Cloud Terrace 113, obtain violating the regulations stoppingThe only close-up image of vehicle, after identifying the car plate of vehicles peccancy and having captured, passes through transmission networkWide-angle scene image and close-up image are transferred to management server 13 by network 12;

Finally, control module 1121 is controlled The Cloud Terrace 113 and automatic zoom camera lens 111 parameters are extensiveArrive again wide-angle state, break rules and regulations automatically to capture next time.

Electronic zoom camera lens 111 is that focal length, focusing, aperture can motorized adjustment camera lenses, and light is logicalCross on the sensor devices that this camera lens is radiated at smart camera, the image capture module of smart camera will be adoptedCollection to image offer coding and decoding video module 1123, Parking analysis module 1124And car plate identification module 1125.

Parking analysis module 1124 is by analyzing continuous traffic video imageProcess, algorithm routine is automatically tested with without Parking and occurs. In the time that event occurs, controlModule 1121 is adjusted The Cloud Terrace angle and lens parameters, obtains the close-up image of stop vehicle violating the regulations,Identify vehicles peccancy car plate and captured after, by transmission network 12 by wide-angle scene graphPicture and close-up image transfer to management server. Finally, control module control The Cloud Terrace and camera lens ginsengNumber returns to wide-angle state, breaks rules and regulations automatically to capture next time.

The Cloud Terrace 113 is fastenings of smart camera 112; Parking analysis module 1124The view data that electronic zoom camera lens 111 is collected is processed, and break in traffic rules and regulations automatically detectedAfter event, event information is passed to control module 1121; Control module 1121 is to The Cloud Terrace 113Angle and smart camera 112 parameters are adjusted and are implemented automatically to capture, video recording. Front end is captured unit11 realize parking offense event detection, car plate identification violating the regulations, the functions such as evidence obtains of breaking rules and regulations, pipeReason server 13 realize record storage violating the regulations, record management violating the regulations, event interlock violating the regulations process,Capture the functions such as front-end control management.

Front end is captured between unit 11 and management server 13 and is connected by transmission network 12, canTo adopt cable network to connect as Ethernet, also can adopt wireless network to connect as GPRS, 3GDeng, can also be that wire transmission network and wireless-transmission network use simultaneously, by two netsThe mutually redundant use of network, guarantees to capture unit at the situation lower front end of a network congestion or suspension11 can also realize by another network and management server 13 transmission of signal.

Automatic shooting device of the present invention also comprises light filling illuminating lamp (not shown), and light filling shinesBright lamp is connected with smart camera. For example automatic shooting device of the present invention can utilize street lighting to realizeParking offense detected and captured night, and for without illumination or the too dim section of light, can adoptDistributed flash lamp carries out light filling.

The key technical indexes is as follows:

A) front end is captured unit

1, detection range: 100m*8m;

2, Parking Detection accuracy: >=85%;

3, car plate accuracy of identification: >=90%;

4, communication interface: 10M/100M self adaptation Ethernet interface, or optional 3G communication;

5, power supply input: 24VAC, 50~60Hz;

6, power consumption :≤120W(comprises The Cloud Terrace, heating, smart camera and light filling illuminating lamp);

7, working environment: temperature-30 DEG C~70 DEG C, humidity 20%~90%;

8, degree of protection: IP65;

B) coding and decoding video module:

1, video compression standard: H.264;

2, transmission of video agreement: RTP, UDP;

3, video image resolution ratio: D1,720P, 1080i, 1080p;

4, video frequency coding rate: 128Kbps~12Mbps;

6, communication interface: 10M/100M self adaptation Ethernet interface;

7, power consumption :≤5W/ road video;

8, working environment: temperature-30 DEG C~70 DEG C, humidity 20%~90%;

Those of ordinary skill in the art will be appreciated that, above embodiment is only used forObject of the present invention is described, and not as limitation of the invention, as long as in essence of the present inventionIn scope, variation to the above embodiment, modification all will drop on claim of the present inventionIn scope.

Claims (3)

1. an automatic grasp shoot method for parking offense, is characterized in that:
First, front end is captured unit (11) and is monitored whole guarded region, in tracing and monitoring regionThe movement locus of each car, when finding after Parking by the mode of track following, determines and stopsSail center of gravity and the size of vehicle, determine that by geometric operation The Cloud Terrace (113) rotates, pitchingAngle, control The Cloud Terrace (113) and make target be positioned at the center of image, by controlling lens zoom,The size of adjustment aim in image, making it be applicable to doing graphical analysis is license plate number identification, license plate numberIdentified rear candid photograph camera return to wide-angle monitoring pattern continue to do Parking detect;
The mode concrete steps of described track following are as follows:
1) according to capture video image size IMAGE_WIDTH that unit (11) gathers,IMAGE_HEIGHT, by every two field picture by required required precision be divided into piece BLOCK_WIDTH,BLOCK_HEIGHT, cuts apart rear image size and is
W I D = I M A G E _ W I D T H B L O C K _ W I D T H
H E I = I M A G E _ H E I G H T B L O C K _ H E I G H T
2) select frame period to carry out inter-frame difference by required required precision, and by the threshold value of difference result and settingCompare, TwoVal represents the result after binarization segmentation, and DifGray represents inter-frame difference result, and Th is two-valueChange segmentation threshold, this value is by the study analysis of several scenes, and the empirical value of acquisition, establishes when the poor result of frame is greater thanAfter fixed threshold value, binarization segmentation result is made as to 255 and is white, otherwise binaryzation result is 0 blackLook, and then carry out binarization segmentation
T w o V a l = 255 , D i f G r a y > T h 0 , D i f G r a y < T h ,
3) for the image after cutting apart, selected 3 × 3 window size, the sad value of each in calculation window,Determine the most obvious point of gray-value variation in each window with this, as the Corner in detection range, doFor the foundation of tracking and matching;
4) in the time series of each frame, find using all direction search method the point that approaches the most with Corner asJoin a little, with the tracking of this realize target, the relevant information of a record point, and then the motion rail of acquisition target vehicleMark;
The described mode of passing through track following is found the concrete grammar stopping:
For the moving target in video scene, its movement locus mainly stops or disappears in following situation:
1) moving vehicle exceedes video detection range: this kind of situation is by video detection range is set, when track surpassesWhile crossing the region of setting, think that vehicle exceedes detection range, respective objects is not processed;
2) following the tracks of angle point drops on road surface: when road surface exists the more or shooting of ponding, the strong road surface of illumination shadeWhen in machine shake and road, the more situation of track cut-off rule and zebra stripes occurs, in object matching process by anglePoint selection is on road surface and cause the termination of pursuit path, and when this kind of situation occurs, its track is around without fixing largeLittle segmentation object, does not process this type of track;
3) moving target stagnation of movement has Parking to occur: when this kind of situation occurs, in video detection range, nothing is newThe appearance of trail angle point, in vehicle matching process, angle point is longer in certain position time of staying, and exists around trackA certain size Target Segmentation result, when this kind of situation occurs, thinks and has Parking to occur.
2. the automatic grasp shoot method for parking offense according to claim 1, its featureBe: described geometric operation determines that The Cloud Terrace (113) rotates, the method for the angle of pitching: cameraThe parameter of shift position, i.e. level angle α, vertically face upward/angle of depression θ,
O is video camera position, and O' is that video camera is projected in ground point, and A is the former focus point of video camera,B is parking offense event occurrence positions, and in the time utilizing track following to obtain the generation of parking offense event, B point is satMark is B point center of gravity position, establishes it for B (x, y), and A point coordinates can obtain by camera, and it is former cameraFocus position, establishes it for A (x0,yo);
Horizontally rotate in process at video camera, O'A=O'C, because video camera sets up location aware, and A, B pointLocation aware, therefore calculate the actual distance of O'A and O'B to real coordinate position transforming relationship according to pixel coordinateFrom, in like manner can calculate the actual range that AM and BM are corresponding, therefore in Δ O'MB, can calculate video cameraLevel angleThe angle of depression that in like manner can calculate video camera is &theta; = arctan O &prime; B OO &prime; - arctan O &prime; C OO &prime; = arctan O &prime; B OO &prime; - arctan O &prime; A OO &prime; , Can obtain thus two, video camera and rotate ginsengNumber, can move to assigned address by video camera;
&alpha; = a r c t a n B M O &prime; M &theta; = a r c t a n O &prime; B OO &prime; - a r c t a n O &prime; A OO &prime;
Due in actual scene, the relation of B and A has several as follows,
1) pixel coordinate is to real coordinate position transforming relationship
If in 3D rendering, the coordinate of certain point in actual coordinates is W (X, Y, Z), by the ratio of similar trianglesRelation can obtain its subpoint I (x', y'), wherein
x &prime; = f * X f - Z , y &prime; = f * Y f - Z - - - ( 1.1 )
Owing to having comprised the division arithmetic of variable Z in above formula, therefore it belongs to nonlinear change, adopt homogeneous coordinatesMethod converts thereof into linear matrix and calculates;
The homogeneous coordinates that any point in space (X, Y, Z) is corresponding are defined as (kX, kY, kZ, k), k be transformation of scale because ofSon, can be any non-zero constant, and realizing its space transformational relation by homogeneous coordinates can be expressed as
(kX,kY,kZ,k)T=T*(x,y,z,1)T(1.2)
Wherein transformation matrix T is the square formation of 4 × 4;
Adopt the method for two-dimensional transform to realize the conversion of pixel distance and actual range, two-dimensional transform matrix is one 3 × 3Square formation, the pixel coordinate of establishing certain point is (x, y), the coordinate of corresponding actual range is (X, Y), in homogeneous coordinatesTransformational relation under system can be expressed as:
K X K Y K = a 00 a 01 a 02 a 10 a 11 a 12 a 20 a 21 1 x y 1 - - - ( 1.3 )
The expression formula that can obtain (X, Y) by above formula is
X = a 00 x + a 01 y + a 02 a 20 x + a 21 y + 1 Y = a 10 x + a 11 y + a 12 a 20 x + a 21 y + 1 - - - ( 1.4 )
From formula (1.3) homogeneous coordinates equation, in transformation matrix, contain eight unknown numbers, as long as calibrateFour known points, can solve this transformation matrix, in actual scene, due to urban road, highwayTrack cut-off rule length and lane width have clear and definite regulation, and the pixel coordinate of calibration point can be direct by programObtain, can set up thus pixel-distance map relation, hereinafter to be referred as mapping table, for obtaining above-mentioned mapping relations,Set cut-off rule the behavior of place, regulation cut-off rule top and be expert at zero point, the track of known distance in each trackCut-off rule set positions calibration point, brings demarcation each point into formula (1.4) and can obtain pixel-range conversion matrixIn each coefficient, and then obtain the mapping relations of pixel-distance in the horizontal direction and in vertical direction.
3. the automatic grasp shoot method for parking offense according to claim 1, its featureBe:
Described front end is captured the electronic zoom camera lens (111) of unit (11) through IMAQ mouldThe image collecting is offered coding and decoding video module (1123), Parking by piece (1126)Analysis module (1124) and car plate identification module (1125), Parking video analysisModule (1124) is carried out analyzing and processing by the traffic video image to continuous, is tested with without stoppingCar event occurs;
In the time having detected that Parking occurs, camera focus is arrived suspected locations, the driving of going forward side by side by cloud platform rotationBoard is searched, and in the time that video camera turns to assigned address but do not find car plate, thinks false alarm, not to this thingPart is processed, and video camera turns to former presetting bit;
When video camera finds after car plate, this vehicle is processed, and in database, searched in a few minutes and beNo have this car plate record of breaking rules and regulations, if had, and not duplicate record, video camera moves to presetting bit, if somewhatIn clock, without this vehicle registration, capture 3 minutes, write down this car plate, and take 4 differences according to different focalPicture, records this vehicle peccancy docking process,
Be angle and the lens parameters that control module (1121) is adjusted The Cloud Terrace (113), obtainThe close-up image of stop vehicle violating the regulations, after identifying the car plate of vehicles peccancy and having captured, passes throughWide-angle scene image and close-up image are transferred to management server (13) by transmission network (12);
Finally, control module (1121) is controlled The Cloud Terrace (113) and automatic zoom camera lens (111)Parameter returns to wide-angle state, breaks rules and regulations automatically to capture next time.
CN201410143566.XA 2014-04-10 2014-04-10 A kind of automatic shooting device for parking offense and automatically grasp shoot method CN103927878B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410143566.XA CN103927878B (en) 2014-04-10 2014-04-10 A kind of automatic shooting device for parking offense and automatically grasp shoot method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410143566.XA CN103927878B (en) 2014-04-10 2014-04-10 A kind of automatic shooting device for parking offense and automatically grasp shoot method

Publications (2)

Publication Number Publication Date
CN103927878A CN103927878A (en) 2014-07-16
CN103927878B true CN103927878B (en) 2016-05-25

Family

ID=51146087

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410143566.XA CN103927878B (en) 2014-04-10 2014-04-10 A kind of automatic shooting device for parking offense and automatically grasp shoot method

Country Status (1)

Country Link
CN (1) CN103927878B (en)

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376554B (en) * 2014-10-16 2017-07-18 中海网络科技股份有限公司 A kind of parking offense detection method based on image texture
CN104464290B (en) * 2014-12-08 2018-12-21 无锡加视诚智能科技有限公司 A kind of road traffic parameter acquisition based on embedded dual core chip and capturing system violating the regulations
CN104637051A (en) * 2014-12-31 2015-05-20 广东电网有限责任公司中山供电局 High-variable precision photographing method based on mobile carrier
CN104504908A (en) * 2015-01-14 2015-04-08 苏州科达科技股份有限公司 Monitoring method and system for illegal parking
CN105894817B (en) * 2015-01-26 2019-04-30 杭州海康威视数字技术股份有限公司 The evidence collecting method and its device of vehicle violation parking
CN104809876A (en) * 2015-05-07 2015-07-29 吴柯维 Illegal vehicle parking detection linkage capture system and integrated camera
CN105023439A (en) * 2015-07-02 2015-11-04 湖北贝力思智能科技有限公司 Intelligent dynamic license plate recognition alarm system
CN105096596A (en) * 2015-07-03 2015-11-25 北京润光泰力科技发展有限公司 Traffic violation detecting method and system
CN105185109A (en) * 2015-08-06 2015-12-23 苏州市世跃智能科技有限公司 Double-dome-camera-based automatic detection and snapshot apparatus for illegal parking
CN107305627B (en) * 2016-04-22 2020-03-27 杭州海康威视数字技术股份有限公司 Vehicle video monitoring method, server and system
CN107945519B (en) * 2017-09-18 2019-12-27 孙健鹏 Method and device for realizing traffic information processing
CN107566727A (en) * 2017-09-19 2018-01-09 深圳云停智能交通技术研究院有限公司 A kind of video camera Automatic adjustment method
CN107705574A (en) * 2017-10-09 2018-02-16 荆门程远电子科技有限公司 A kind of precisely full-automatic capturing system of quick road violation parking
CN110415526A (en) * 2019-07-29 2019-11-05 北京精英路通科技有限公司 A kind of parking monitoring device

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN202771619U (en) * 2012-07-18 2013-03-06 中海网络科技股份有限公司 Automatic snapshot device for illegal parking

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20130097868A (en) * 2012-02-27 2013-09-04 주식회사 레이스전자 Intelligent parking management method and system based on camera
KR20130100615A (en) * 2012-03-02 2013-09-11 경기대학교 산학협력단 Method for warning illegal parking

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN202771619U (en) * 2012-07-18 2013-03-06 中海网络科技股份有限公司 Automatic snapshot device for illegal parking

Also Published As

Publication number Publication date
CN103927878A (en) 2014-07-16

Similar Documents

Publication Publication Date Title
Bojarski et al. End to end learning for self-driving cars
CN101799987B (en) Self-adaptive intelligent traffic light and control method thereof
CN103714538B (en) road edge detection method, device and vehicle
CN102036054B (en) Intelligent video monitoring system based on three-dimensional virtual scene
CN102521983B (en) Vehicle violation detection system based on high definition video technology and method thereof
CN103176185B (en) Method and system for detecting road barrier
ES2352300T3 (en) System to detect vehicles.
CN103413313B (en) The binocular vision navigation system of electrically-based robot and method
CN102693632B (en) Monitoring device and monitoring method based on visual image
Jog et al. Pothole properties measurement through visual 2D recognition and 3D reconstruction
CN100459704C (en) Intelligent tunnel safety monitoring apparatus based on omnibearing computer vision
CN106845547B (en) A kind of intelligent automobile positioning and road markings identifying system and method based on camera
CN105015419B (en) Automatic parking method with building figure is positioned based on stereoscopic vision
CN104508727B (en) Three-dimensional body detection device and three-dimensional body detection method
CN100419813C (en) Omnibearing visual sensor based road monitoring apparatus
US8284996B2 (en) Multiple object speed tracking system
CN201213038Y (en) Traffic information collecting system based on linear array CCD video camera
CN102768803B (en) Vehicle intelligent monitoring and recording system and method based on radar and video detection
CN103676829B (en) Intelligent city based on video total management system and method thereof
CN100507970C (en) Red light overriding detection system and method based on digital video camera
Zhang et al. Video-based vehicle detection and classification system for real-time traffic data collection using uncalibrated video cameras
CN105744232B (en) A kind of method of the transmission line of electricity video external force damage prevention of Behavior-based control analytical technology
CN101593422B (en) Traffic violation detection method based on video tracking and pattern recognition
CN102592450B (en) System and method for snapshotting behavior of motor vehicles not giving precedence to pedestrians on zebra stripes
US20200041284A1 (en) Map road marking and road quality collecting apparatus and method based on adas system

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CP03 Change of name, title or address
CP03 Change of name, title or address

Address after: 200135 No. 600 Minsheng Road, Shanghai, China (Shanghai) free trade zone

Patentee after: COSCO marine Polytron Technologies Inc

Address before: 200135 building 600, 21 Minsheng Road, Shanghai, Pudong New Area

Patentee before: CHINA SHIPPING NETWORK TECHNOLOGY CO., LTD.