CN103794056B - Based on the vehicle precise classification system and method for real-time two-way video stream - Google Patents
Based on the vehicle precise classification system and method for real-time two-way video stream Download PDFInfo
- Publication number
- CN103794056B CN103794056B CN201410080066.6A CN201410080066A CN103794056B CN 103794056 B CN103794056 B CN 103794056B CN 201410080066 A CN201410080066 A CN 201410080066A CN 103794056 B CN103794056 B CN 103794056B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- video
- type
- camera
- classification
- 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
Landscapes
- Traffic Control Systems (AREA)
- Devices For Checking Fares Or Tickets At Control Points (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of vehicle precise classification system and method based on real-time two-way video stream, system comprises: vehicle body smear camera, high definition snapshot video camera and video vehicle type detector; Method comprises: vehicle body smear camera and high definition snapshot video camera carry out real-time two-way video collection to vehicle; Logical relation is one to one set up after demarcating in virtual pixel space and real physical space by the ken; Vehicle in the ken is carried out target separation; Provide vehicle physical data accurately, carry out Model Reconstruction and three-dimensional measurement, type of vehicle is finally judged.Usefulness of the present invention is: adopt vehicle body smear camera and high definition snapshot video camera two-way high definition video steaming, Embedded Double PASS VIDEO vehicle type detector, use the reconstruction of 3D vision Car body model, fit approach, vehicle many kinds of parameters can be provided, by vehicle accurately typing simultaneously.
Description
Technical field
The present invention relates to a kind of Vehicle Classification System and method, be specifically related to a kind of vehicle precise classification system and method based on real-time two-way video stream, belong to road traffic technical field.
Background technology
Along with the quickening of social economy's develop rapidly and urbanization process, China's road traffic scale is sharply expanded, and a series of traffic administration problems produced therefrom are badly in need of being resolved.Based on the vehicle precise classification management that different traffic administration demands is sought unity of talking to vehicle, it is an effective technological means.
Table 1 " JT-T489-2003 turn pike toll vehicle classification standard "
Vehicle precise classification adopts various dimensions sensor technology, based on traffic industry standard specification " JT-T489-2003 turn pike toll vehicle classification standard " (hereinafter referred to as " standard ", content as shown in table 1), behavior is divided to the kind that the difference of the outward appearance of vehicle, purposes, load-bearing capacity is carried out, vehicle precise classification, as modern intelligent traffic administration system important component part, has great social effect and economic worth.This technology is widely used in multiple field such as blocked road networking charging system, parking lot fee collection management system, emphasis tunnel and bridge tariff management system, vehicle expenses of taxation collection management system, to fair vehicle taxtation burden, scientific and reasonable planning road space has important meaning.
At present, two kinds of methods are mainly adopted to vehicle precise classification.
First method is: adopt a forward acquisition camera, and realize vehicle detection classification based on video detecting method, Vehicle length can be divided into A, B, C, D, E, F six class by system.
This kind of method is single video stream vehicle classification (also known as visible light video detection method), it is on the basis demarcated the ken, orthographic view based on vehicle carries out plane geometry measurement and obtains Vehicle length, and is variety classes according to Vehicle length different demarcation.The method classification foundation simple (only with reference to vehicle commander), realization are easily, but can not meet in " rower " according to index requests such as headstock height, axletree number, wheelbase, wheel numbers, therefore, science and the scientific research fields such as traffic parameter statistics, traffic data mining analysis is only applicable to.
In addition, because the plane geometric figure of the method based on vehicle orthographic view is measured, therefore, very strict to video camera antenna height, angle, in real world applications, difficulty of construction is larger.
Second method is: will feel coil checker circlewise (mainly provides vehicle to exist and trigger, vehicle commander judges), echelette detecting device (mainly provides vehicle ' s contour to scan, Vehicles separation, headstock height), axletree detecting device (mainly provides the vehicle number of axle, wheelbase, wheel number detects), License Plate Identification device (is mainly used in identifying vehicle license numbering, car plate background color) and Data Control main frame (each detecting device real time data is gathered, and based on the time, space synchronously processes, final according to headstock height, axletree number, wheelbase, wheel number, vehicle commander, vehicle license plate characteristic carries out precise classification to vehicle) Integrated predict model (see Fig. 1), to headstock height in foundation " standard ", axletree number, wheelbase, wheel number, carrying or cargo capacity etc. require to be divided into 1-5 type car.In addition, feeling coil checker, echelette detecting device etc. circlewise also can provide separately simple vehicle type classification function, but can not apply as precise classification equipment separately.
The essence of the method is a kind of multi-detector data integration application technology, although vehicle precise classification parameter required in " standard " can be provided comprehensively, but, redundancy too complicated as a kind of its front-end architecture of outdoor utility system, therefore, system Construction is with high costs, environmental adaptation is poor, resistance to overturning is low, and serviceable life is short, later maintenance difficulty is large.Meanwhile, because multi-detector Integrated predict model can not overcome the intrinsic technological deficiency of each Function detection itself, wherein a certain detecting device will cause whole system loss of function.
Summary of the invention
First object of the present invention be to provide a kind of based on track entrance high definition vehicle body smear camera and track just before or the vehicle precise classification system of 45° angle high definition snapshot video camera two-way high definition video steaming before side.
Second object of the present invention is to provide the method for vehicle being carried out to precise classification based on this vehicle precise classification system.
In order to realize first aim, the present invention adopts following technical scheme:
Based on a vehicle precise classification system for real-time two-way video stream, its characteristic is, comprising:
Vehicle body smear camera: be erected at track entrance, antenna height is the side of 1-1.5 rice, vertical scanning vehicle body;
High definition snapshot video camera: be erected at track just before or side before 45° angle direction, antenna height is 1-1.5 rice, realizes vehicle frontal video capture;
Video vehicle type detector: the multiple image transmitted by vehicle body smear camera carries out splicing, reducing vehicle body; According to the image information that high definition snapshot video camera transmits, accurate positioned vehicle position, first carries out Preliminary division to target vehicle simultaneously and accurately divides.
The aforesaid vehicle precise classification system based on real-time two-way video stream, its characteristic is, aforementioned video vehicle type detector comprises with lower module:
Video real-time learning module: be automatically separated the image static background information, the moving target information that obtain needed for systems axiol-ogy, carries out online self study in real time to live video stream simultaneously;
Vehicle detection module: based on two-way video image calibration data, sets up the phasor coordinate space mapped one by one with car, road physical space in systems in which, and the mode of the tracking of employing target trajectory, feature detection, cluster extracts target vehicle information;
Vehicle snapshot identification module: vehicle headstock number plate track and localization is carried out to the target meeting vehicle characteristics, and automatically identify license plate structure, license plate number, car plate background color, can vehicle be provided to there is trigger pip according to the position of car plate in track simultaneously;
Data simultaneous module: the data of vehicle detection module and car plate being captured to identification module gather, and carry out data syn-chronization based on the relation of Time and place, movement velocity, judge whether these group data stem from same vehicle;
Model Reconstruction measurement module: the reconstruction carrying out auto model according to the vehicle data through synchronous process, and carrying out real-time comparison, measurement with history auto model storehouse, the result data of measurement comprises needed for classification: vehicle frontal feature, license plate number, headstock height, the number of axle, wheelbase, vehicle wheel number, vehicle commander;
Vehicle determination module: carry out vehicle classification judgement according to the vehicle technology parameter that Model Reconstruction measurement module exports;
Neural Network Self-learning module: automatically carry out feature extraction and classification to new Shanxi vehicle, records artificial revised categorical data automatically, possesses the automatic detection classification capacity to new model.
The aforesaid vehicle precise classification system based on real-time two-way video stream, its characteristic is, aforementioned video vehicle type detector also comprises:
Video pre-filtering module: carry out pre-service by frame per second sampling analysis to the video flowing not meeting testing requirement, preprocessing process comprises: image noise reduction process, acutance, aberration and saturation degree process, key frame compensate.
In order to realize second target, the present invention adopts following technical scheme:
Based on above-mentioned vehicle precise classification system, vehicle is carried out to a method for precise classification, it is characterized in that, comprise the following steps:
(1) vehicle body smear camera and high definition snapshot video camera carry out real-time two-way video collection to vehicle;
(2) ken demarcation is carried out to the video flowing of vehicle body smear camera and high definition snapshot video camera, logical relation is one to one set up in the virtual pixel space in the video ken and real physical space;
(3) on the basis that the ken is demarcated, based on machine vision algorithm, the vehicle in the ken is carried out target separation;
(4) when target passes through the ken of vehicle body smear camera, record object rises, stop feature and rise, stop the time, calculate target speed, direction of motion, scanning record object side profile, and according to wheel Model Matching locating wheel fetal position and axletree, restore principle by vehicle splicing in multiple image by image overlap, reduction vehicle body, provides vehicle physical data accurately;
(5) when target appears at the ken of high definition snapshot video camera, forward detection, candid photograph are carried out to target vehicle, accurate positioned vehicle position, obtain the number-plate number, background color, licence plate structure, and vehicle width, headstock height;
(6) based on the various dimensions information of vehicles that vehicle body smear camera and high definition snapshot video camera obtain, Model Reconstruction and three-dimensional measurement is carried out;
(7) in conjunction with the result of licence plate type, simple classification definition and three-dimensional measurement, type of vehicle is finally judged, and generate unique result of determination and export to the vehicle classification application system be associated.
Aforesaid method, is characterized in that, in step (7) a metallic, the method that type of vehicle is judged as:
According to vehicle license type, simple classification definition, Preliminary division is carried out to vehicle, hang blue board or simple classification be defined as minibus, minibus vehicle as the first set, hang yellow card or simple classification be defined as passenger vehicle, lorry type as the second set;
In the first aggregate, 1 type car and 2 type cars are accurately marked off according to the difference of headstock height and wheelbase;
In the second set, the difference according to the number of axle, wheel number, wheelbase and vehicle commander accurately marks off 3 type cars, 4 type cars and 5 type cars.
Aforesaid method, is characterized in that, in the first aggregate, headstock is less than 1.3 meters and the vehicle that wheelbase is less than 3 meters is 1 type car, headstock height >=1.3 meter and the vehicle of wheelbase >3 rice is 2 type cars; In the second set, the number of axle is 2 axles, wheel number be 6 to take turns, the vehicle of wheelbase >5 rice is 3 type cars, the number of axle is 3 axles, wheel number be greater than 6 take turns and be less than 10 to take turns, vehicle commander be the vehicle of 15.1 meters-20 meters is 4 type cars, the number of axle is more than 3 axles, wheel number is that 10 to take turns vehicle that is above, vehicle commander more than 20.1 meters be 5 type cars.
Usefulness of the present invention is: 45° angle high definition snapshot video camera two-way high definition video steaming before adopting track entrance high definition vehicle body smear camera and track just or before side, Embedded Double PASS VIDEO vehicle type detector, 3D vision Car body model is used to rebuild, fit approach, headstock height can be provided simultaneously, axletree number, wheelbase, wheel number, vehicle commander, the number-plate number, number plate type (blue board, yellow card), type of vehicle (minibus, minibus, passenger vehicle, lorry), the vehicle parameters such as vehicle brand, vehicle is divided into: 1-5 type car, fully can meet the charge of road closed formula, parking fee collective system, the technical need that country's vehicle expenses of taxation toll collection management etc. detects automatic vehicle classification.
Accompanying drawing explanation
Fig. 1 is the Organization Chart of existing vehicle precise classification system;
Fig. 2 is the Organization Chart of vehicle precise classification system of the present invention;
Fig. 3 is the wiring schematic diagram of the vehicle precise classification system in Fig. 2;
Fig. 4 is the composition schematic diagram of the video vehicle type detector in Fig. 2;
Fig. 5 is the main flow figure obtaining vehicle body characteristic;
Fig. 6 is main flow figure vehicle being carried out to accurately judgement.
Embodiment
Below in conjunction with the drawings and specific embodiments, concrete introduction is done to the present invention.
With reference to Fig. 2 and Fig. 3, vehicle precise classification system based on real-time two-way video stream of the present invention, comprise: vehicle body smear camera, high definition snapshot video camera and video vehicle type detector, digital video is inputed to video vehicle type detector respectively by Ethernet by vehicle body smear camera, high definition snapshot video camera, wherein
1, vehicle body smear camera is erected at track entrance, the side of vertical scanning vehicle body, and adopt the high definition of 2,000,000 pixels, high sensitivity web camera and below 6MM high-definition fixed-focus camera lens, frame per second is up to 25FPS, and antenna height is 1-1.5 rice.The vehicle body scanning camera time shutter is adjustable, and each frame high clear video image all can clear explanation by the automobile body feature of the ken.Consider nighttime imaging effect, and the closely vertical light filling of the visible ray reflective interference that may cause at night, also can adopt the video camera possessing infrared imagery technique and infrared light compensating lamp mode.Vehicle body smear camera is mainly used in catching and locates vehicle body unique point, positioned vehicle tire, axle location and the number of axle, wheelbase.
Consider toll island narrow space problem, therefore, vehicle body scanning camera shooting angle and the ken are selected, should with in image can the most cart of clear, complete display by time its tyre contour outline be most basic demand.Meanwhile, also must meet two frame vehicle body image mosaic in core algorithm, the technical requirement of the picture registration content of more than 10% should be possessed.
2, high definition snapshot camera pedestal be located at track just before or side before 45° angle direction, realize vehicle frontal video capture, adopt the high definition of 2,000,000 pixels, high sensitivity web camera and 12MM high-definition fixed-focus camera lens, frame per second is up to 25FPS, and antenna height is 1-1.5 rice.High definition snapshot video camera adopts captures some left front or right front miter angle tilt, being as the criterion by the common medium and small motor vehicle of complete display in the ken.Night carries out light filling by adopting the synchronous light compensating lamp of LED of specialty to vehicle license, light compensating lamp also adopts miter angle oblique illumination, car plate will be caused to cross quick-fried phenomenon because light filling distance is near, driver's sight line by track can not be impacted simultaneously, high definition snapshot video camera is mainly used in providing candid photograph and the identification of car plate type and car plate numbering, and headstock height, body width are measured.
3, video vehicle type detector adopts embedded high-performance industrial computer, is equipped with Low Power High Performance processor, effectively can processes two-way high-definition network camera video flowing.
The multiple image that vehicle body smear camera transmits splices by video vehicle type detector, reduction vehicle body, provide vehicle physical data accurately, the image information simultaneously transmitted according to high definition snapshot video camera, accurate positioned vehicle position, identify vehicle license information and according to vehicle license information Preliminary division vehicle fundamental type or by high definition snapshot video capture to target vehicle positive feature Preliminary division is carried out to target vehicle, and according to the number of axle on the basis of Preliminary division, wheelbase, wheel number, headstock height, vehicle commander accurately divides target vehicle.
Video vehicle type detector by Ethernet or serial communication mode by vehicle classification real-time data transmission to vehicle classification application systems such as money machine, automatic card dispenser, ETC no-stop charging systems, finally to realize according to the charging respectively of different type of vehicle.Vehicle exists signal can export the automatic gate in track to simultaneously, for controlling the open and close of automatic railing in real time.
As the preferred scheme of one, with reference to Fig. 4, video vehicle type detector comprises with lower module:
Video real-time learning module: be automatically separated the image static background information, the moving target information that obtain needed for systems axiol-ogy, carries out online self study in real time to live video stream simultaneously.
Vehicle detection module: one of nucleus module, based on two-way video image calibration data, sets up the phasor coordinate space mapped one by one with car, road physical space in systems in which, adopts the modes such as target trajectory tracking, feature detection, cluster to extract target vehicle information.
Vehicle snapshot identification module: vehicle headstock number plate track and localization is carried out to the target meeting vehicle characteristics, and automatically identify license plate structure, license plate number, car plate background color, can vehicle be provided to there is trigger pip according to the position of car plate in track simultaneously, and by this signal real-time Transmission to track banister control system, for controlling the open and close of entrance lane railing.
Data simultaneous module: the data of vehicle detection module and car plate being captured to identification module gather, and carry out data syn-chronization based on the relation of Time and place, movement velocity, judge whether these group data stem from same vehicle.
Model Reconstruction measurement module: the vehicle data through synchronous process is committed to this module, the reconstruction of auto model is carried out according to the vehicle data through synchronous process, and carry out real-time comparison, measurement with history auto model storehouse, the result data measured contains the technology essential factor needed for " JT-T489-2003 turn pike toll vehicle classification standard " classification, and this technology essential factor comprises: vehicle frontal feature, license plate number, headstock height, the number of axle, wheelbase, vehicle wheel number, vehicle commander etc.
Vehicle determination module: the vehicle technology parameter exported according to Model Reconstruction measurement module, carries out vehicle classification judgement according to " JT-T489-2003 turn pike toll vehicle classification standard ".
Neural Network Self-learning module: this module mainly provides system self study data accumulation and human assistance vehicle storehouse extended function.Automatically can also carry out feature extraction and classification to new Shanxi vehicle, automatically record artificial revised categorical data, possess the automatic detection classification capacity to new model.
As the preferred scheme of one, video vehicle type detector also comprises: video pre-filtering module, this module carries out pre-service by frame per second sampling analysis to the video flowing not meeting testing requirement, preprocessing process comprises: image noise reduction process, acutance, aberration and saturation degree process, key frame compensation etc., this module is the basic module of video vehicle type detector, mainly guarantee that the high-definition digital video current mass of input system can meet to detect and analyze requirement, thus guarantee the final detection perform stability of system, reliability, universality.
The method of precise classification is carried out in following introduction to vehicle based on above-mentioned vehicle precise classification system.
The first step: vehicle body smear camera and high definition snapshot video camera carry out real-time two-way video collection to vehicle.
Second step: ken demarcation is carried out to the video flowing of vehicle body smear camera and high definition snapshot video camera, logical relation is one to one set up in virtual pixel space in the video ken and real physical space so that can measure correlation model data for different angles, comparison.Namely the ken enters real-time online learning process after having demarcated.
3rd step: on the basis that the ken is demarcated, based on a series of machine vision algorithm, the vehicle in the ken is carried out target separation.
4th step: with reference to Fig. 5, when target passes through the ken of vehicle body smear camera, record object rises, stop feature and rise, stop the time, calculate target speed, direction of motion, scanning record object side profile, and according to wheel Model Matching locating wheel fetal position and axletree, restore principle by vehicle splicing in multiple image by image overlap, reduction vehicle body, provides vehicle physical data accurately.
Video vehicle type detector will carry out synchronously vehicle body characteristic, correct according to time, spatial correlation, and manual synchronizing warehouse-in is carried out to abnormal data, judge to optimize basic data as following vehicle, thus make video vehicle type detector possess the self-learning capability of optic nerve principle.
5th step: with reference to Fig. 5, when target appears at the ken of high definition snapshot video camera, carry out forward detection, candid photograph to target vehicle, synchronous averaging car plate detecting pattern and car test recognition mode double-mode is concrete:
The accurate positioned vehicle position of car plate detecting pattern, obtain the number-plate number, background color and licence plate structure, by license number and background color Preliminary division vehicle fundamental type (blue board: compact car, yellow card: large car), and vehicle is accurately provided to there is trigger pip for driveway controller.
When the stained vehicle of unlicensed or car plate being detected, cannot accurate License Plate be carried out, namely automatically proceed to car test recognition mode.
Car test recognition mode, by being mapped by the coordinate parameters of vehicle orthographic view and high definition snapshot camera calibration, finally obtains the index such as vehicle width, headstock height.
Video vehicle type detector will carry out synchronously headstock characteristic, correct according to time, spatial correlation, and manual synchronizing warehouse-in is carried out to abnormal data, judge to optimize basic data as following vehicle, thus make video vehicle type detector possess the self-learning capability of optic nerve principle.
By the study of a large amount of vehicle sample, the preliminary identification that image carries out vehicle class, brand can be captured based on vehicle forward, and artificial progressively intervene under set up property data base (comprising: car plate numbering, type of vehicle definition, brand or series, the vehicle number of axle, vehicle wheel number etc.) and complete online real-time learning accumulation, lay the first stone for next step system accurately identifies automatically.This identifying information automatically can also be merged with other identifying informations, strengthen the confidence level of vehicle cab recognition, finally realize the detection of accurate vehicle classification.
6th step: the various dimensions information of vehicles obtained based on vehicle body smear camera and high definition snapshot video camera, carries out Model Reconstruction and three-dimensional measurement.
First headstock picture and vehicle body scanned picture are carried out vehicle splicing, reduction according to key elements such as capture time, unique point, spatial relationships, and compare with the video camera physics information of demarcation, measure, the classification support sizes certificates such as final acquisition vehicle headstock height, tire location, the number of axle, wheelbase, Vehicle length.
7th step: in conjunction with the result of licence plate type, simple classification definition and three-dimensional measurement, type of vehicle is finally judged, and generate unique result of determination and export to the vehicle classification application system be associated.
Type of vehicle is judged, is divided into Preliminary division and accurately divides two steps, see Fig. 6.
Preliminary division: using hang blue board or simple classification be defined as minibus, minibus vehicle as the first set, using hang yellow card or simple classification be defined as passenger vehicle, lorry type as the second set.
Accurate division: in the first aggregate, difference according to headstock height and wheelbase accurately marks off 1 type car and 2 type cars, concrete, headstock is less than 1.3 meters and the vehicle that wheelbase is less than 3 meters is 1 type car, headstock height >=1.3 meter and the vehicle of wheelbase >3 rice is 2 type cars; In the second set, difference according to the number of axle, wheel number, wheelbase and vehicle commander accurately marks off 3 type cars, 4 type cars and 5 type cars, concrete, the number of axle is 2 axles, wheel number be 6 to take turns, the vehicle of wheelbase >5 rice is 3 type cars, the number of axle is 3 axles, wheel number be greater than 6 take turns and be less than 10 to take turns, vehicle commander be the vehicle of 15.1 meters-20 meters is 4 type cars, the number of axle is more than 3 axles, wheel number is that 10 to take turns vehicle that is above, vehicle commander more than 20.1 meters be 5 type cars.
As can be seen here, there is following advantage in the present invention:
1, meet " standard " requirement comprehensively, application layer vehicle precise classification data are provided
The pure dimensional visual measurement technology detected based on headstock, vehicle body provides: vehicle is divided into by headstock height, the number of axle, wheelbase, wheel number, vehicle commander, car plate type, type of vehicle data etc.: 1-5 type car, fully can meet the technical need that the charge of road closed formula, parking fee collective system, national vehicle expenses of taxation toll collection management etc. detect automatic vehicle classification.
Compare traditional single camera video detection, Data mining, echelette detect, the vehicle parameter not only provided enriches, and stability and practicality all must promote.Compared with integrated with the data application layer of multi-detector vehicle precise classification scheme, native system is then fully optimized from the core technology demand of core aspect for vehicle classification, makes system architecture simpler, practical, convenient large-scale application deployment.
2, video vehicle type detector structure is simple, integrated level is high, operational efficiency is high, construction cost is low
Video vehicle type detector, based on vehicle precise classification core data demand, fully to be optimized from video vehicle detection algorithm aspect, cutting.Compared with common many equipment application layer data integrated architecture, its front-end architecture is more simple, and integrated level is higher, and operational efficiency is higher.Native system front end only needs 2 high-definition network cameras and 1 embedded video vehicle type detector just can completion system task, and therefore, cost is far below existing many integration of equipments technical scheme.
3, adopt non-contact detection, resistance to overturning is good, and later maintenance is simple
The video triggering mode adopted belongs to non-contact detection, therefore, without the need to burying underground on road surface, disposing anything.And ground induction coil detecting device, wheel detector all belong to contact triggering in many integration of equipments scheme, need brokenly road to construct when equipment is installed, directly will destroy pavement structure, thus cause road to reduce actual life.And the said equipment critical piece is subject to, and expand with heat and contract with cold in road surface, vehicle rolls and as easy as rolling off a log damage for a long time, originally designed life is that the equipment reality of 5 years only just occurs that large area is damaged for about 3 years, and equipment also becomes very complicated once damage its repair.
The high-definition camera adopted, video vehicle type detector all adopt technical grade, low power dissipation design, and its stabilization of equipment performance is higher, environmental suitability is stronger.High-definition camera is as the direct application apparatus in the outdoor that native system is unique, and its daily servicing is only dedusting to camera protective cover, camera lens and wiping, and in native system, video camera antenna height is only 1.5 meters, and therefore, later maintenance is simple, low cost.
It should be noted that, above-described embodiment does not limit the present invention in any form, the technical scheme that the mode that all employings are equal to replacement or equivalent transformation obtains, and all drops in protection scope of the present invention.
Claims (5)
1., based on the vehicle precise classification system of real-time two-way video stream, its characteristic is, comprising:
Vehicle body smear camera: be erected at track entrance, antenna height is the side of 1-1.5 rice, vertical scanning vehicle body;
High definition snapshot video camera: be erected at track just before or side before 45° angle direction, antenna height is 1-1.5 rice, realizes vehicle frontal video capture;
Video vehicle type detector: the multiple image transmitted by vehicle body smear camera carries out splicing, reducing vehicle body; According to the image information that high definition snapshot video camera transmits, accurate positioned vehicle position, first carries out Preliminary division to target vehicle simultaneously and accurately divides;
Described video vehicle type detector comprises with lower module:
Video real-time learning module: be automatically separated the image static background information, the moving target information that obtain needed for systems axiol-ogy, carries out online self study in real time to live video stream simultaneously;
Vehicle detection module: based on two-way video image calibration data, sets up the phasor coordinate space mapped one by one with car, road physical space in systems in which, and the mode of the tracking of employing target trajectory, feature detection, cluster extracts target vehicle information;
Vehicle snapshot identification module: vehicle headstock number plate track and localization is carried out to the target meeting vehicle characteristics, and automatically identify license plate structure, license plate number, car plate background color, can vehicle be provided to there is trigger pip according to the position of car plate in track simultaneously;
Data simultaneous module: the data of vehicle detection module and car plate being captured to identification module gather, and carry out data syn-chronization based on the relation of Time and place, movement velocity, judge whether these group data stem from same vehicle;
Model Reconstruction measurement module: the reconstruction carrying out auto model according to the vehicle data through synchronous process, and carrying out real-time comparison, measurement with history auto model storehouse, the result data of measurement comprises needed for classification: vehicle frontal feature, license plate number, headstock height, the number of axle, wheelbase, vehicle wheel number, vehicle commander;
Vehicle determination module: carry out vehicle classification judgement according to the vehicle technology parameter that Model Reconstruction measurement module exports;
Neural Network Self-learning module: automatically carry out feature extraction and classification to new Shanxi vehicle, records artificial revised categorical data automatically, possesses the automatic detection classification capacity to new model.
2. the vehicle precise classification system based on real-time two-way video stream according to claim 1, its characteristic is, described video vehicle type detector also comprises:
Video pre-filtering module: carry out pre-service by frame per second sampling analysis to the video flowing not meeting testing requirement, preprocessing process comprises: image noise reduction process, acutance, aberration and saturation degree process, key frame compensate.
3. based on the vehicle precise classification system described in claim 1 or 2, vehicle is carried out to the method for precise classification, it is characterized in that, comprise the following steps:
(1) vehicle body smear camera and high definition snapshot video camera carry out real-time two-way video collection to vehicle;
(2) ken demarcation is carried out to the video flowing of vehicle body smear camera and high definition snapshot video camera, logical relation is one to one set up in the virtual pixel space in the video ken and real physical space;
(3) on the basis that the ken is demarcated, based on machine vision algorithm, the vehicle in the ken is carried out target separation;
(4) when target passes through the ken of vehicle body smear camera, record object rises, stop feature and rise, stop the time, calculate target speed, direction of motion, scanning record object side profile, and according to wheel Model Matching locating wheel fetal position and axletree, restore principle by vehicle splicing in multiple image by image overlap, reduction vehicle body, provides vehicle physical data accurately;
(5) when target appears at the ken of high definition snapshot video camera, forward detection, candid photograph are carried out to target vehicle, accurate positioned vehicle position, obtain the number-plate number, background color, licence plate structure, and vehicle width, headstock height;
(6) based on the various dimensions information of vehicles that vehicle body smear camera and high definition snapshot video camera obtain, Model Reconstruction and three-dimensional measurement is carried out;
(7) in conjunction with the result of licence plate type, simple classification definition and three-dimensional measurement, type of vehicle is finally judged, and generate unique result of determination and export to the vehicle classification application system be associated.
4. method according to claim 3, is characterized in that, in step (7), the method that type of vehicle is judged as:
According to vehicle license type, simple classification definition, Preliminary division is carried out to vehicle, hang blue board or simple classification be defined as minibus, minibus vehicle as the first set, hang yellow card or simple classification be defined as passenger vehicle, lorry type as the second set;
In the first aggregate, 1 type car and 2 type cars are accurately marked off according to the difference of headstock height and wheelbase;
In the second set, the difference according to the number of axle, wheel number, wheelbase and vehicle commander accurately marks off 3 type cars, 4 type cars and 5 type cars.
5. method according to claim 4, is characterized in that, in the first aggregate, headstock is less than 1.3 meters and the vehicle that wheelbase is less than 3 meters is 1 type car, headstock height >=1.3 meter and the vehicle of wheelbase >3 rice is 2 type cars; In the second set, the number of axle is 2 axles, wheel number be 6 to take turns, the vehicle of wheelbase >5 rice is 3 type cars, the number of axle is 3 axles, wheel number be greater than 6 take turns and be less than 10 to take turns, vehicle commander be the vehicle of 15.1 meters-20 meters is 4 type cars, the number of axle is more than 3 axles, wheel number is that 10 to take turns vehicle that is above, vehicle commander more than 20.1 meters be 5 type cars.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410080066.6A CN103794056B (en) | 2014-03-06 | 2014-03-06 | Based on the vehicle precise classification system and method for real-time two-way video stream |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410080066.6A CN103794056B (en) | 2014-03-06 | 2014-03-06 | Based on the vehicle precise classification system and method for real-time two-way video stream |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103794056A CN103794056A (en) | 2014-05-14 |
| CN103794056B true CN103794056B (en) | 2015-09-30 |
Family
ID=50669667
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410080066.6A Active CN103794056B (en) | 2014-03-06 | 2014-03-06 | Based on the vehicle precise classification system and method for real-time two-way video stream |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103794056B (en) |
Families Citing this family (45)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021375B (en) * | 2014-05-29 | 2017-11-07 | 银江股份有限公司 | A kind of model recognizing method based on machine learning |
| DE102014012285A1 (en) * | 2014-08-22 | 2016-02-25 | Jenoptik Robot Gmbh | Method and axle counting device for non-contact axle counting of a vehicle and axle counting system for road traffic |
| CN104866858A (en) * | 2015-05-27 | 2015-08-26 | 北京信路威科技股份有限公司 | Vehicle model characteristic analysis method |
| CN105427614A (en) * | 2015-08-28 | 2016-03-23 | 北京动视元科技有限公司 | Model classification system and method |
| CN105224951B (en) * | 2015-09-30 | 2019-05-03 | 深圳市华尊科技股份有限公司 | A kind of vehicle type classification method and sorter |
| CN105426922B (en) | 2015-12-08 | 2019-02-12 | 同方威视技术股份有限公司 | Rain model recognition methods and system and safety detection method and system |
| CN105719482B (en) * | 2016-04-20 | 2018-07-10 | 太原市高远时代科技有限公司 | Dynamic vehicle information collecting device |
| CN106127107A (en) * | 2016-06-14 | 2016-11-16 | 宁波熵联信息技术有限公司 | The model recognizing method that multi-channel video information based on license board information and vehicle's contour merges |
| CN106441529A (en) * | 2016-08-30 | 2017-02-22 | 山西万立科技有限公司 | Vehicle type recognition and cheating diagnosis system based on video image technology |
| CN106446949B (en) * | 2016-09-26 | 2019-05-21 | 成都通甲优博科技有限责任公司 | A kind of vehicle model recognition methods and device |
| CN106600722A (en) * | 2016-11-14 | 2017-04-26 | 南京积图网络科技有限公司 | Toll service device, method and system |
| CN107730617A (en) * | 2017-02-04 | 2018-02-23 | 西安艾润物联网技术服务有限责任公司 | Unlicensed vehicle toll method and apparatus |
| US10671873B2 (en) | 2017-03-10 | 2020-06-02 | Tusimple, Inc. | System and method for vehicle wheel detection |
| CN106980837A (en) * | 2017-03-29 | 2017-07-25 | 广西信路威科技发展有限公司 | A kind of model recognizing method analyzed based on vehicle characteristic information and system |
| CN107122747B (en) * | 2017-04-28 | 2020-06-09 | 北京理工大学 | Non-contact detection device and method for railway vehicle carriage state |
| CN107886728B (en) * | 2017-11-27 | 2024-05-10 | 山西国强高科股份有限公司 | Video recognition device for truck model |
| CN107895156A (en) * | 2017-11-29 | 2018-04-10 | 合肥寰景信息技术有限公司 | For the extraction of multiple target vehicle characteristics and analysis system |
| CN108320517A (en) * | 2017-12-28 | 2018-07-24 | 浙江中新长清信息科技有限公司 | Car plate and vehicle identification system and monitoring server |
| AU2019241892B2 (en) * | 2018-03-09 | 2024-03-28 | Tusimple, Inc. | System and method for vehicle wheel detection |
| CN110874932B (en) * | 2018-08-29 | 2021-06-15 | 北京万集科技股份有限公司 | Method, device and system for identifying number of axles |
| CN111161542B (en) * | 2018-11-08 | 2021-09-28 | 杭州海康威视数字技术股份有限公司 | Vehicle identification method and device |
| CN109637153A (en) * | 2019-01-25 | 2019-04-16 | 合肥市智信汽车科技有限公司 | A kind of vehicle-mounted mobile violation snap-shooting system based on machine vision |
| CN109815933A (en) * | 2019-02-11 | 2019-05-28 | 山东省公安厅高速公路交通警察总队 | A kind of multiaxle trucks automatic identification apparatus for obtaining evidence, system and method |
| CN109767507A (en) * | 2019-03-01 | 2019-05-17 | 河北交通职业技术学院 | A kind of highway non-parking charge payment system based on fence |
| CN110047276B (en) * | 2019-03-11 | 2020-11-27 | 广州文远知行科技有限公司 | Method and device for determining congestion state of obstacle vehicle and related product |
| CN111833469B (en) * | 2019-04-18 | 2022-06-28 | 杭州海康威视数字技术股份有限公司 | Vehicle charging method and system applied to charging station |
| CN110232324A (en) * | 2019-05-16 | 2019-09-13 | 苏州思卡信息***有限公司 | A kind of automatic recognition system and method for expressway tol lcollection vehicle |
| CN110550038B (en) * | 2019-09-30 | 2021-02-09 | 重庆元韩汽车技术设计研究院有限公司 | Intelligent driving decision information acquisition system and method |
| CN113129598B (en) * | 2019-12-31 | 2022-09-20 | 武汉万集信息技术有限公司 | Free flow vehicle type identification method, system and equipment based on infrared camera |
| CN111292432A (en) * | 2020-01-14 | 2020-06-16 | 北京巨视科技有限公司 | Vehicle charging type distinguishing method and device based on vehicle type recognition and wheel axle detection |
| CN111667688B (en) * | 2020-05-06 | 2021-07-09 | 北京科技大学 | Comprehensive traffic information analysis method and system based on pavement vibration signals |
| CN111507427A (en) * | 2020-05-08 | 2020-08-07 | 湖南车路协同智能科技有限公司 | Vehicle classification method |
| CN111860201B (en) * | 2020-06-28 | 2023-07-25 | 中铁大桥科学研究院有限公司 | Ramp heavy vehicle identification method and system combining image identification and bridge monitoring |
| CN111860384B (en) * | 2020-07-27 | 2021-12-14 | 上海福赛特智能科技有限公司 | Vehicle type recognition method |
| CN111860381A (en) * | 2020-07-27 | 2020-10-30 | 上海福赛特智能科技有限公司 | Axle counting device and method for lane openings |
| CN111796472A (en) * | 2020-07-31 | 2020-10-20 | 广东优尼克智能照明科技有限公司 | ETC buckle snapshot lamp and assembling and using method |
| CN112002027A (en) * | 2020-09-04 | 2020-11-27 | 侯晓峰 | Highway charging method, device, terminal and storage medium |
| CN112329747B (en) * | 2021-01-04 | 2021-04-23 | 湖南大学 | Vehicle parameter detection method based on video identification and deep learning and related device |
| CN113111884B (en) * | 2021-03-26 | 2024-05-24 | 沈阳天眼智云智能技术研究院有限公司 | Video detection method of special dangerous chemical transportation vehicle |
| CN113537167B (en) * | 2021-09-15 | 2021-12-03 | 成都数联云算科技有限公司 | Vehicle appearance recognition method, system, device and medium |
| CN114120661B (en) * | 2021-11-17 | 2023-07-25 | 常州大学 | Vehicle detection system and method based on target detection |
| CN114199132B (en) * | 2021-12-16 | 2024-02-27 | 上海应用技术大学 | Laser three-dimensional scanner and scanning method based on machine vision |
| CN114624726B (en) * | 2022-03-17 | 2023-05-16 | 南通探维光电科技有限公司 | Axle identification system and axle identification method |
| CN115083170B (en) * | 2022-07-06 | 2024-06-07 | 深圳市创安视迅技术有限公司 | Image capturing method and device of camera, electronic equipment and storage medium |
| CN117877265A (en) * | 2024-01-24 | 2024-04-12 | 特微乐行(广州)技术有限公司 | Expressway monitoring early warning and disposal management and control system based on video AI analysis |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1052204A (en) * | 1989-11-25 | 1991-06-12 | 西安公路研究所 | Automotive type statistical method of automatic recognition and device |
| MXPA04000393A (en) * | 2004-01-14 | 2005-07-15 | Americas Resources S C | System and process for identifying and classifying vehicles. |
| CN101783076A (en) * | 2010-02-04 | 2010-07-21 | 西安理工大学 | Method for quick vehicle type recognition under video monitoring mode |
| CN102708385A (en) * | 2012-04-06 | 2012-10-03 | 张丛喆 | Method and system for comparison and recognition of three-dimensional vehicle types in video monitoring scenes |
| CN103048697A (en) * | 2012-12-17 | 2013-04-17 | 华南理工大学 | Optical curtain device capable of measuring advancing direction, axial number and two-dimensional shape of vehicle |
| CN202995057U (en) * | 2012-12-17 | 2013-06-12 | 华南理工大学 | Light curtain device for vehicle detection |
| CN103324920A (en) * | 2013-06-27 | 2013-09-25 | 华南理工大学 | Method for automatically identifying vehicle type based on vehicle frontal image and template matching |
| CN203225009U (en) * | 2012-12-12 | 2013-10-02 | 北京万集科技股份有限公司 | Laser type traffic condition investigation system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| ES2384750T3 (en) * | 2009-10-01 | 2012-07-11 | Kapsch Trafficcom Ag | Devices and procedures for vehicle classification |
-
2014
- 2014-03-06 CN CN201410080066.6A patent/CN103794056B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1052204A (en) * | 1989-11-25 | 1991-06-12 | 西安公路研究所 | Automotive type statistical method of automatic recognition and device |
| MXPA04000393A (en) * | 2004-01-14 | 2005-07-15 | Americas Resources S C | System and process for identifying and classifying vehicles. |
| CN101783076A (en) * | 2010-02-04 | 2010-07-21 | 西安理工大学 | Method for quick vehicle type recognition under video monitoring mode |
| CN102708385A (en) * | 2012-04-06 | 2012-10-03 | 张丛喆 | Method and system for comparison and recognition of three-dimensional vehicle types in video monitoring scenes |
| CN203225009U (en) * | 2012-12-12 | 2013-10-02 | 北京万集科技股份有限公司 | Laser type traffic condition investigation system |
| CN103048697A (en) * | 2012-12-17 | 2013-04-17 | 华南理工大学 | Optical curtain device capable of measuring advancing direction, axial number and two-dimensional shape of vehicle |
| CN202995057U (en) * | 2012-12-17 | 2013-06-12 | 华南理工大学 | Light curtain device for vehicle detection |
| CN103324920A (en) * | 2013-06-27 | 2013-09-25 | 华南理工大学 | Method for automatically identifying vehicle type based on vehicle frontal image and template matching |
Non-Patent Citations (4)
| Title |
|---|
| 基于计算机视觉的汽车整车尺寸测量***的研究;翟乃斌;《中国博士学位论文全文数据库》;20070415(第4期);参见论文第24-27页第2.3节、第52-57页3.4节、第65-74页、第77-78页、第85-87页、第113-114页6.3节、图2-5、图2-6、图3-12—图3-17 * |
| 新型高速公路入口自动发卡***的研发与应用;蒋新花;《交通工程》;20131031(第10期);参见论文第72-75页 * |
| 李晋惠,等.一种图像处理的汽车类型识别算法.《西安工业大学学报》.2009,第29卷(第3期), * |
| 翟乃斌,等.基于计算机视觉的汽车整车尺寸测量***.《交通与计算机》.2006,第24卷(第3期), * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103794056A (en) | 2014-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103794056B (en) | Based on the vehicle precise classification system and method for real-time two-way video stream | |
| CN105427614A (en) | Model classification system and method | |
| CN109870456B (en) | Rapid detection system and method for road surface health condition | |
| CN105512720B (en) | A kind of public transit vehicle passenger flow statistics method and system | |
| CN103854320B (en) | Vehicle automatic identifying method based on laser radar | |
| CN103279996B (en) | Information of vehicles in a kind of multilane situation detects and recognition system | |
| CN102867417B (en) | Taxi anti-forgery system and taxi anti-forgery method | |
| CN103123722B (en) | Road object detection method and system | |
| CN100437660C (en) | Device for monitoring vehicle breaking regulation based on all-position visual sensor | |
| CN102867414B (en) | Vehicle queue length measurement method based on PTZ (Pan/Tilt/Zoom) camera fast calibration | |
| CN103714363A (en) | Motor vehicle exhaust smoke video identification system | |
| CN107256636A (en) | A kind of traffic flow acquisition methods for merging laser scanning and video technique | |
| CN106127107A (en) | The model recognizing method that multi-channel video information based on license board information and vehicle's contour merges | |
| CN103021059A (en) | Video-monitoring-based public transport passenger flow counting method | |
| CN203260072U (en) | Vehicle information detection and identification system under multiple-lane condition | |
| CN107274678B (en) | A kind of night vehicle flowrate and model recognizing method based on Kinect | |
| CN106767854A (en) | mobile device, garage map forming method and system | |
| CN103308056A (en) | Road marking detection method | |
| CN103810854B (en) | A kind of based on the artificial intelligent transportation parameter detection method demarcated | |
| CN107447699A (en) | A kind of high-adaptability cycle path intelligent gate system | |
| CN207993028U (en) | A kind of passenger flow device based on infrared binocular camera ranging | |
| CN101226637B (en) | Method for detecting automatically contact point of vehicle wheel and ground | |
| CN115273028B (en) | Intelligent parking lot semantic map construction method and system based on global perception | |
| CN102496281A (en) | Vehicle red-light violation detection method based on combination of tracking and virtual loop | |
| CN103164958A (en) | Method and system for vehicle monitoring |
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 |