CN106373430A - Intersection pass early warning method based on computer vision - Google Patents
Intersection pass early warning method based on computer vision Download PDFInfo
- Publication number
- CN106373430A CN106373430A CN201610735587.XA CN201610735587A CN106373430A CN 106373430 A CN106373430 A CN 106373430A CN 201610735587 A CN201610735587 A CN 201610735587A CN 106373430 A CN106373430 A CN 106373430A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- moving target
- early warning
- image
- pedestrian
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/16—Anti-collision systems
- G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses an intersection pass early warning method based on computer vision. The method comprises the following steps: S1) collecting video images of an intersection in real time and capturing close-up images of vehicles; S2) extracting moving objects according to the video images, generating moving object driving information and extracting complete vehicle license plate information according to the close-up images; S3) carrying out classification on the moving objects to obtain a classification result; S4) calculating driving speed of each vehicle according to the classification result and the driving information; S5) storing information of the vehicles passing through the intersection, and predicating the center-of-mass coordinate of each moving object at the next-moment position according to the driving speed and driving information of the vehicle; and S6) according to the driving information, the driving speed and the center-of-mass coordinate of the moving object at the next-moment position, generating an early warning signal and displaying the early warning information for the situations meeting abnormal conditions. The method can carry out analysis on construction intersection pass abnormity conditions through the video image analysis technology and output the early warning signal, and is high in stability and accuracy.
Description
Technical field
The present invention relates to a kind of traffic safety early warning field is and in particular to a kind of intersection based on computer vision leads to
Row method for early warning.
Background technology
In construction site road, especially intersection, is the most dangerous region in road network, because construction site is
Open up, signal lighties and road instruction line are rare, and pedestrian does not concentrate, Construction traffic great majority are oversize vehicle, along with outer temporarily
Carry out sailing into of vehicle, become the Multiple trauma of vehicle accident.Pedestrian, vehicle pass through intersection when, often due to construction building
Thing leads to driver to produce vision dead zone with pedestrian, and external vehicle travels violates construction management regulation to construction specific road section, with
And car speed exceedes construction section safety value, these factors are likely to cause vehicle accident, in construction section traffic signal
Lamp is rare, it is especially desirable to note the generation of traffic safety hidden danger in the case of road is irregular.Traditional traffic prewarning system is simultaneously
It is not properly suited for the such geographical conditions in job site, and early warning range is not wide, technological means are single, and we carry for this
Go out a kind of construction site intersection passing early warning system based on computer vision and method.
Content of the invention
In order to overcome the shortcoming of prior art presence and a kind of not enough, crossing based on computer vision of present invention offer
Current method for early warning, be particularly applicable in the vehicle of construction site intersection passing and pedestrian occur carrying out during abnormal conditions pre-
Alert, the generation trying to forestall traffic accidents.
The present invention adopts the following technical scheme that
A kind of intersection passing method for early warning based on computer vision, comprises the steps:
The video image of s1 Real-time Collection intersection, and capture the close-up image of vehicle；
S2 extracts moving target according to video image, generates moving target travel information and has been extracted according to close-up image
Whole vehicle license information；
S3 classifies to moving target, draws classification results；
S4, according to classification results and travel information, calculates the gait of march of vehicle；
S5 stores information of vehicles, the gait of march according to vehicle and the travel information by intersection, predicts described fortune
The center-of-mass coordinate of moving-target subsequent time position；
S6 according to the center-of-mass coordinate of travel information, gait of march and this moving target subsequent time position, to meeting exception
The situation of condition, generates early warning signal and shows early warning information.
Described moving target includes vehicle and pedestrian, described travel information include vehicles or pedestrians by intersection when
Between section, the center-of-mass coordinate of vehicles or pedestrians and direct of travel.
Described moving target is extracted according to video image, specifically include following steps:
S2.1 sets up mixed Gauss model to the video image of intersection scene, extracts foreground target, generates two-value
The moving target foreground picture changed；
S2.2 counts each moving target pixel number and the image coordinate of pixel, calculates the figure of moving target barycenter
As coordinate
Wherein m, n represent the Breadth Maximum of certain moving target and height in image respectively, and i represents pixel in image
Abscissa, j represents the vertical coordinate of pixel in image, xijRepresent the abscissa value of pixel in moving target, yijRepresent motion
The ordinate value of target pixel points；
S2.3 calculates the motion vector of moving target according to the image coordinate of the barycenter of moving target in adjacent 20 two field pictures,
Determine the direct of travel of moving target by motion vector, direct of travel can carry out universal formulation according to the difference of intersection.
S3 classifies to moving target, draws classification results, and described classification results include engineering truck, non-engineering truck and row
People, concrete steps include:
S3.1 extracts the profile of moving target, and draws, according to profile, the boundary rectangle surrounding profile, according to moving target
The ratio of width to height of boundary rectangle tentatively distinguishes vehicle target and pedestrian target；
If the ratio of width to height of moving target boundary rectangle be less than the empirical value that sets then it is assumed that moving target now as
Pedestrian；
If the ratio of width to height of moving target boundary rectangle be more than the empirical value that sets then it is assumed that moving target now as
The adhesion pedestrian target of engineering truck, non-engineering truck or multiple pedestrian's Adhesion formation；
S3.2, according to the histograms of oriented gradients feature of moving target, sets up grader mould using the algorithm of support vector machine
Type is distinguishing engineering truck, non-engineering truck and adhesion pedestrian target.
Described s3.2 particularly as follows:
S3.2.1 collects the image of engineering truck, non-engineering truck and adhesion pedestrian, the size of unified image, sets up respectively
Sample set, if engineering truck sample set be set a, non-engineering truck sample set be set b, adhesion pedestrian sample collection be set c；
S3.2.2 using the set a in s3.2.1 and set b as just collecting, set c as negative collection, extract respectively and just collecting and bearing
The histograms of oriented gradients feature of collection, as the input of algorithm of support vector machine, generates sorter model one, described grader mould
Type one is used for carrying out two classification to vehicle with pedestrian；
S3.2.3 using the set a in s3.2.1 as just collecting, set b as negative collection, extract the side just collecting with negative collection respectively
To histogram of gradients feature as the input of algorithm of support vector machine, generate sorter model two, described sorter model dual-purpose
In two classification are carried out to engineering truck and non-engineering truck；
S3.2.4 extracts the correspondence original rgb figure of each moving target according to the boundary rectangle of each moving target extracting
Picture, extracts the histograms of oriented gradients feature of each moving target original rgb image；
The histograms of oriented gradients feature of each moving target original rgb image that s3.2.5 extracts is successively as grader
The input of model one, if output result is just then it is assumed that moving target now is vehicle, if output result is negative,
Think that moving target now is adhesion pedestrian；
The histograms of oriented gradients feature of the moving target that classification results in s3.2.5 are vehicle by s3.2.6 is as classification
The input of device model two, if output result is just then it is assumed that moving target now is engineering truck, if output result is
Bear then it is assumed that moving target now is non-engineering truck.
According to classification results and travel information in described s4, calculate the gait of march of vehicle, particularly as follows:
S4.1, according to classification results, chooses vehicle target as the object of calculating speed, does not calculate the gait of march of pedestrian；
S4.2: in video image, two virtual detection lines are set by perpendicular to vehicle target direct of travel；Measure two again
On the corresponding real road of bar virtual detection line apart from δ d；Calculate vehicle in real time video image and successively reach two virtual inspections
The frame number f of survey line；
S4.3: according to sample frequency f of video image, on the corresponding real roads of two virtual detection lines apart from δ d, car
Successively reach the frame number f of two virtual detection lines, calculate gait of march v of vehicle:
Described s5 specifically adopts the center-of-mass coordinate of Kalman filtering algorithm predicted motion target subsequent time position.
Described abnormal condition includes following situation:
The situation of the velocity amplitude that vehicle gait of march specifies beyond job site generates early warning signal
Whether position coordinateses according to vehicle and the center-of-mass coordinate of the vehicle subsequent time position predicted, judge vehicle
There is the special area being located at job site or the situation having sailed this special area into, thus generating early warning signal, described
Special area refers to the non-engineering truck vehicle section that No entry of construction site management regulation；
According to the travel information of vehicle and pedestrian, judge that vehicle and current whether there is of pedestrian, vehicle and vehicle are constituted
The situation of vision dead zone, thus generate early warning signal.
Described early warning information include showing the monitoring image of over-speed vehicles and send early warning voice, to sail job site into special
The monitoring image of non-engineering truck in region simultaneously sends early warning voice and shows there is generation between vehicle and pedestrian, vehicle and vehicle
Collide scene image during probability and send early warning voice display reminding poster.
Panoramic camera gathers the video image of intersection, the close-up image of feature video capture vehicle.
Beneficial effects of the present invention:
(1) effective district can dividing engineering truck and non-engineering truck, thus carrying out automatical and efficient supervision, being easy to difference
Vehicle take different measures of control, for job site safety management provide convenient；
(2) position of moving target subsequent time can be estimated, thus to swarming into job site prohibited area
The vehicle of probability sends alarm, thus avoiding the generation of dangerous situation；
(3) it is capable of detecting when vehicle and the pedestrian of current direction composition vision dead zone, and shows safety instruction information, allow and drive
The person of sailing has grace time to take deceleration or parking measure；
(4) vehicle being passed through in job site, no matter being Construction traffic or external vehicle, carrying out information system management, building
Vertical data base, stores its species by the moment of job site, vehicle, the information such as licence plate.
(5) it is used for detecting that the high-definition camera of construction site scene is erected at the overhead on road surface it is not necessary to imbed equipment
Underground, installs and maintenance cost is relatively low.
Brief description
Fig. 1 is the structural representation of the present invention；
Fig. 2 is the structure in-site installation figure of the embodiment of the present invention；
Fig. 3 is embodiment of the present invention led display screen display content schematic diagram；
Fig. 4 is the workflow diagram of the present invention.
Specific embodiment
With reference to embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not
It is limited to this.
Embodiment
As shown in Figure 1, Figure 2 and Figure 3, a kind of intersection passing early warning system based on computer vision implementing the present invention
System, including front end acquisition module, video information process module, network transmission module, monitoring module and warning module；
Described front end acquisition module includes panoramic camera 2, feature video camera 1 and video encoder, and described panorama is taken the photograph
Camera, feature video camera are connected with video encoder respectively, arrange virtual coil in intersection specific region, and feature images
The visual field align virtual coil 7 of machine, the visual field of panoramic camera includes intersection all directions visual angle；
Described video information process module, specially industrial computer 3, described industrial computer pass through network transmission module respectively with prison
Control module and warning module connect, and described video encoder is connected with industrial computer.
Described monitoring module is made up of information terminal display device 5 and database server 6；
Described warning module includes led display screen 8.
Model ipc610 of described industrial computer.
Model csd-p6-smd3535 of described led display screen, double copies power line, resolution is in more than 720p.
Described network transmission module is wireless network transmissions equipment 4.
Described panoramic camera and feature video camera should be erected at aerial on cross or T-shaped road junction corner road surface.
In the present embodiment, virtual coil is rectangle.
The present embodiment includes two feature video cameras and a panoramic camera, and led has two pieces of display screens, display letter
Breath can facilitate each road vehicle to observe with pedestrian, as shown in Figure 3.
The panoramic camera collection large-scale scene information in intersection simultaneously exports to industrial computer the place carrying out video image
Reason；Virtual coil is set in crossing intersection part specific region, the visual field be aligned crossing intersection part specific region of feature video camera sets
Put virtual coil, capture the high-definition image of vehicle and export and carry out image procossing to industrial computer；The software profit installed on industrial computer
Complete vehicle, the detection of pedestrian, tracking, classification, position anticipation, identification vehicle license with Video processing and mode identification technology
Number, storage information of vehicles, output early warning signal etc.；The outfan of signal has led display screen, database server, information terminal
Display device.Led display screen is used for showing early warning information, and database server is used for storage by construction site intersection
Information of vehicles, information terminal display device is used for the monitored picture of real-time display intersection and the reality synchronous with monitored picture
The design sketch of the functions such as the detection of existing vehicle and pedestrian, tracking, classification, position anticipation, early warning.
As shown in figure 4, a kind of intersection passing method for early warning based on computer vision, comprise the steps:
The vehicle of s1 panoramic camera Real-time Collection intersection and the video image of walk, feature video capture
The close-up image of each vehicle in virtual coil；
S2 extracts moving target according to video image, generates moving target travel information and has been extracted according to close-up image
Whole vehicle license information, described moving target includes vehicle and pedestrian, and described travel information includes vehicles or pedestrians to be passed through to hand over
The time period of cross road mouth, the center-of-mass coordinate of vehicles or pedestrians and direct of travel.
Described vehicle license information is by carrying out License Plate, Character segmentation, character recognition and color to close-up image
Obtained from identification.
S2.1 sets up mixed Gauss model to the video image of intersection scene, extracts foreground target, generates two-value
The moving target foreground picture changed；
S2.2 counts each moving target pixel number and the image coordinate of pixel, calculates the figure of moving target barycenter
As coordinate
Wherein m, n represent the Breadth Maximum of certain moving target and height in image respectively, and i represents pixel in image
Abscissa, j represents the vertical coordinate of pixel in image, xijRepresent the abscissa value of pixel in moving target, yijRepresent motion
The ordinate value of target pixel points；
S2.3 calculates the motion vector of moving target according to the image coordinate of the barycenter of moving target in adjacent 20 two field pictures,
Determine the direct of travel of moving target by motion vector, direct of travel can carry out universal formulation according to the difference of intersection.
S3 classifies to moving target, draws classification results；
Described classification results include engineering truck, non-engineering truck and pedestrian, and concrete steps include:
S3.1 extracts the profile of moving target, and draws, according to profile, the boundary rectangle surrounding profile, according to moving target
The ratio of width to height of boundary rectangle tentatively distinguishes vehicle target and pedestrian target；
If the ratio of width to height of moving target boundary rectangle be less than the empirical value that sets then it is assumed that moving target now as
Pedestrian；
If the ratio of width to height of moving target boundary rectangle be more than the empirical value that sets then it is assumed that moving target now as
The adhesion pedestrian target of engineering truck, non-engineering truck or multiple pedestrian's Adhesion formation；
S3.2, according to the histograms of oriented gradients feature of moving target, sets up grader mould using the algorithm of support vector machine
Type is distinguishing engineering truck, non-engineering truck and adhesion pedestrian target.
S3.2.1 collects the image of engineering truck, non-engineering truck and adhesion pedestrian, the size of unified image, sets up respectively
Sample set, if engineering truck sample set be set a, non-engineering truck sample set be set b, adhesion pedestrian sample collection be set c；
S3.2.2 using the set a in s3.2.1 and set b as just collecting, set c as negative collection, extract respectively and just collecting and bearing
The histograms of oriented gradients feature of collection, as the input of algorithm of support vector machine, generates sorter model one, described grader mould
Type one is used for carrying out two classification to vehicle with pedestrian；
S3.2.3 using the set a in s3.2.1 as just collecting, set b as negative collection, extract the side just collecting with negative collection respectively
To histogram of gradients feature as the input of algorithm of support vector machine, generate sorter model two, described sorter model dual-purpose
In two classification are carried out to engineering truck and non-engineering truck；
S3.2.4 extracts the correspondence original rgb figure of each moving target according to the boundary rectangle of each moving target extracting
Picture, extracts the histograms of oriented gradients feature of each moving target original rgb image；
The histograms of oriented gradients feature of each moving target original rgb image that s3.2.5 extracts is successively as grader
The input of model one, if output result is just then it is assumed that moving target now is vehicle, if output result is negative,
Think that moving target now is adhesion pedestrian；
The histograms of oriented gradients feature of the moving target that classification results in s3.2.5 are vehicle by s3.2.6 is as classification
The input of device model two, if output result is just then it is assumed that moving target now is engineering truck, if output result is
Bear then it is assumed that moving target now is non-engineering truck.
S4, according to classification results and travel information, calculates the gait of march of vehicle；
S4.1, according to classification results, chooses vehicle target as the object of calculating speed, does not calculate the gait of march of pedestrian；
S4.2: in video image, two virtual detection lines are set by perpendicular to vehicle target direct of travel；Measure two again
On the corresponding real road of bar virtual detection line apart from δ d；Calculate vehicle in real time video image and successively reach two virtual inspections
The frame number f of survey line；
S4.3: according to sample frequency f of video image, on the corresponding real roads of two virtual detection lines apart from δ d, car
Successively reach the frame number f of two virtual detection lines, calculate gait of march v of vehicle:
S5 stores information of vehicles, the gait of march according to vehicle and the travel information by intersection, predicts described fortune
The center-of-mass coordinate of moving-target subsequent time position；
Described s5 specifically adopts the center-of-mass coordinate of the position of Kalman filtering algorithm predicted motion target subsequent time.Card
Kalman Filtering is a kind of estimation of recurrence, is divided into forecast period and more new stage, and in forecast period, Kalman filtering algorithm uses
The estimation of current time state, makes the estimation to subsequent time state；In the more new stage, Kalman filtering algorithm using under
The predictive value that the observation optimization of one moment state obtains in forecast period, to obtain a more accurate new estimation value.
Particularly as follows:
Shifting on image for the image coordinate of barycenter and barycenter of a upper moment moving target in s5.1 acquisition video image
Dynamic speed, set up the predictive equation of moving target position it may be assumed that
X (t+1 | t)=ax (t | t)+w (t+1)
In formula: x (t+1 | t) is the state vector of the subsequent time moving target being predicted using current time；x(t|t)
For current time optimal State Estimation vector；A is state-transistion matrix；W (t+1) be process noise it is assumed that be desired for zero white
Noise, its covariance matrix is q (t+1)；
Translational speed v specific formula for calculation on image for the described moving target barycenter is as follows:
Wherein t is the time interval in two moment, and δ d is the distance of t time moving target barycenter movement.
The covariance matrix of s5.2: more new state x (t+1 | t):
P (t+1 | t)=ap (t | t) at+q(t+1)
Wherein: p (t+1 | t) expression x (t+1 | t) corresponding covariance, p (t | t) expression x (t | t) corresponding covariance, at
Represent the transposed matrix of a,
S5.3: the measured value of the moving target state according to subsequent time, in conjunction with the subsequent time moving target predicting
State vector, calculate the optimization estimated value x (t+1 | t+1) of subsequent time moving target state:
X (t+1 | t+1)=x (t+1 | t)+kg (t+1) (z (t+1)-hx (t+1 | t))
Wherein z (t+1) is the measured value of moving target subsequent time state, and h is calculation matrix, and wherein kg (t+1) is card
Germania gain:
Kg (t+1)=p (t+1 | t) ht/(hp(t+1|t)ht+r(t+1))
Wherein r (t+1) is measurement noise covariance matrix, htTransposed matrix for h；
S5.4: update the covariance matrix p (t+1 | t+1) of subsequent time moving target state x (t+1 | t+1):
P (t+1 | t+1)=(i-kg (t+1) h) p (t+1 | t)
Wherein i is unit matrix.
Described abnormal conditions mainly include the speed that the speed of vehicle specifies beyond job site, and non-engineering truck sails construction into
The special area at scene, the current direction between the vehicle of intersection passing and vehicle, between vehicle and pedestrian constitutes vision
Blind area, vehicle and vehicle, vehicle and pedestrian have the probability colliding；
The special area of described job site refers to the non-engineering truck section that No entry of construction site management regulation；
Warning module generates corresponding warning information according to the early warning signal that industrial computer exports, including following several situations:
(1) monitoring image of over-speed vehicles is shown on information terminal display device and sends early warning voice；(2) show in information terminal
On equipment, display is sailed the monitoring image of non-engineering truck of job site special area into and is sent early warning voice；(3) in information
Show on terminal unit and have scene image when colliding probability between vehicle and pedestrian, vehicle and vehicle and send early warning
Voice and on led screen display reminding poster.
Fig. 2 is the led information screen mounting structure schematic diagram of warning module in the embodiment of the present invention, such as the road scene of Fig. 1
Shown in schematic diagram, the current direction of vehicle and pedestrian leads to the appearance of vision dead zone due to the construction area of construction section, now
Warning module output early warning information, to led display screen, reminds vehicle and pedestrian to take care.
Above-described embodiment is the present invention preferably embodiment, but embodiments of the present invention are not subject to described embodiment
Limit, other any spirit without departing from the present invention and the change made under principle, modification, replacement, combine, simplify,
All should be equivalent substitute mode, be included within protection scope of the present invention.
Claims (10)
1. a kind of intersection passing method for early warning based on computer vision is it is characterised in that comprise the steps:
The video image of s1 Real-time Collection intersection, and capture the close-up image of vehicle；
S2 extracts moving target according to video image, generates moving target travel information and is extracted according to close-up image complete
Vehicle license information；
S3 classifies to moving target, draws classification results；
S4, according to classification results and travel information, calculates the gait of march of vehicle；
S5 stores information of vehicles, the gait of march according to vehicle and the travel information by intersection, predicts described motion mesh
The center-of-mass coordinate of mark subsequent time position；
S6 according to the center-of-mass coordinate of travel information, gait of march and this moving target subsequent time position, to meeting abnormal condition
Situation, generate early warning signal and simultaneously show early warning information.
2. intersection passing method for early warning according to claim 1 is it is characterised in that described moving target includes vehicle
And pedestrian, described travel information include vehicles or pedestrians by the time period of intersection, the center-of-mass coordinate of vehicles or pedestrians and
Direct of travel.
3. intersection passing method for early warning according to claim 1 it is characterised in that described according to video image extraction
Moving target, specifically includes following steps:
S2.1 sets up mixed Gauss model to the video image of intersection scene, extracts foreground target, generates binaryzation
Moving target foreground picture；
S2.2 counts each moving target pixel number and the image coordinate of pixel, and the image calculating moving target barycenter is sat
Mark
Wherein m, n represent the Breadth Maximum of certain moving target and height in image respectively, and i represents the horizontal seat of pixel in image
Mark, j represents the vertical coordinate of pixel in image, xijRepresent the abscissa value of pixel in moving target, yijRepresent moving target
The ordinate value of pixel；
S2.3 calculates the motion vector of moving target according to the image coordinate of the barycenter of moving target in adjacent 20 two field pictures, by transporting
Dynamic vector determines the direct of travel of moving target, and direct of travel can carry out universal formulation according to the difference of intersection.
4. intersection passing method for early warning according to claim 1 is it is characterised in that s3 is carried out to moving target point
Class, draws classification results, and described classification results include engineering truck, non-engineering truck and pedestrian, and concrete steps include:
S3.1 extracts the profile of moving target, and draws the boundary rectangle surrounding profile according to profile, external according to moving target
The ratio of width to height of rectangle tentatively distinguishes vehicle target and pedestrian target；
If the ratio of width to height of moving target boundary rectangle is less than the empirical value setting then it is assumed that moving target now is as row
People；
If the ratio of width to height of moving target boundary rectangle is more than the empirical value setting then it is assumed that moving target now is as engineering
The adhesion pedestrian target of car, non-engineering truck or multiple pedestrian's Adhesion formation；
S3.2 according to the histograms of oriented gradients feature of moving target, using the algorithm of support vector machine set up sorter model
Distinguish engineering truck, non-engineering truck and adhesion pedestrian target.
5. intersection passing method for early warning according to claim 4 it is characterised in that described s3.2 particularly as follows:
S3.2.1 collects the image of engineering truck, non-engineering truck and adhesion pedestrian, the size of unified image, sets up sample respectively
Collection, if engineering truck sample set be set a, non-engineering truck sample set be set b, adhesion pedestrian sample collection be set c；
S3.2.2 using the set a in s3.2.1 and set b as just collecting, set c as negative collection, extract respectively just collecting and collect with bearing
Histograms of oriented gradients feature, as the input of algorithm of support vector machine, generates sorter model one, described sorter model one
For two classification are carried out with pedestrian to vehicle；
S3.2.3 using the set a in s3.2.1 as just collecting, set b as negative collection, extract the direction ladder just collecting with negative collection respectively
Degree histogram feature, as the input of algorithm of support vector machine, generates sorter model two, and it is right that described sorter model two is used for
Engineering truck and non-engineering truck carry out two classification；
S3.2.4 extracts the correspondence original rgb image of each moving target according to the boundary rectangle of each moving target extracting,
Extract the histograms of oriented gradients feature of each moving target original rgb image；
The histograms of oriented gradients feature of each moving target original rgb image that s3.2.5 extracts is successively as sorter model
One input, if output result is just then it is assumed that moving target now is vehicle, if output result be negative then it is assumed that
Moving target now is adhesion pedestrian；
The histograms of oriented gradients feature of the moving target that classification results in s3.2.5 are vehicle by s3.2.6 is as grader mould
The input of type two, if output result is just then it is assumed that moving target now is engineering truck, if output result is negative,
Think that moving target now is non-engineering truck.
6. intersection method for early warning according to claim 1 it is characterised in that in described s4 according to classification results and
Travel information, calculates the gait of march of vehicle, particularly as follows:
S4.1, according to classification results, chooses vehicle target as the object of calculating speed, does not calculate the gait of march of pedestrian；
S4.2: in video image, two virtual detection lines are set by perpendicular to vehicle target direct of travel；Measure two void again
Intend on the corresponding real road of detection line apart from △ d；Calculate vehicle in real time video image and successively reach two virtual detection lines
Frame number f；
S4.3: according to sample frequency f of video image, on the corresponding real roads of two virtual detection lines apart from △ d, vehicle is first
Reach the frame number f of two virtual detection lines afterwards, calculate gait of march v of vehicle:
7. intersection passing method for early warning according to claim 1 is it is characterised in that described s5 specifically adopts karr
The center-of-mass coordinate of graceful filtering algorithm predicted motion target subsequent time position.
8. intersection passing method for early warning according to claim 1 it is characterised in that described abnormal condition include as follows
Situation:
The situation of the velocity amplitude that vehicle gait of march specifies beyond job site generates early warning signal
Position coordinateses according to vehicle and prediction vehicle subsequent time position coordinateses, judge whether vehicle is located at job site
Special area or the situation having sailed this special area into, thus generating early warning signal, described special area refers to that construction is existing
The non-engineering truck vehicle section that No entry of field administrative provisions；
According to the travel information of vehicle and pedestrian, judge the feelings of the current whether vision dead zone of vehicle and pedestrian, vehicle and vehicle
Condition, thus generate early warning signal.
9. intersection passing method for early warning according to claim 1 is it is characterised in that described early warning information includes showing
The monitoring image of over-speed vehicles and send early warning voice, sail into job site special area non-engineering truck monitoring image simultaneously
Send and have scene image when colliding probability between early warning voice and display vehicle and pedestrian, vehicle and vehicle and send
Early warning voice display reminding poster.
10. intersection passing method for early warning according to claim 1 is it is characterised in that panoramic camera collection intersects
The video image at crossing, the close-up image of feature video capture vehicle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610735587.XA CN106373430B (en) | 2016-08-26 | 2016-08-26 | Intersection traffic early warning method based on computer vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610735587.XA CN106373430B (en) | 2016-08-26 | 2016-08-26 | Intersection traffic early warning method based on computer vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106373430A true CN106373430A (en) | 2017-02-01 |
| CN106373430B CN106373430B (en) | 2023-03-31 |
Family
ID=57904183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610735587.XA Active CN106373430B (en) | 2016-08-26 | 2016-08-26 | Intersection traffic early warning method based on computer vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106373430B (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108200552A (en) * | 2017-12-14 | 2018-06-22 | 华为技术有限公司 | A kind of V2X communication means and device |
| CN108492567A (en) * | 2018-04-24 | 2018-09-04 | 汪宇明 | Monitor terminal, road traffic method for early warning and system |
| CN108877269A (en) * | 2018-08-20 | 2018-11-23 | 清华大学 | A kind of detection of intersection vehicle-state and V2X broadcasting method |
| CN109147326A (en) * | 2018-09-06 | 2019-01-04 | 北京理工大学 | A kind of Campus transport safety warning system |
| CN109191852A (en) * | 2018-10-25 | 2019-01-11 | 西北工业大学 | Che-road-cloud collaboration traffic flow Tendency Prediction method |
| CN109830123A (en) * | 2019-03-22 | 2019-05-31 | 大陆投资(中国)有限公司 | Crossing anti-collision warning method and system |
| CN110443161A (en) * | 2019-07-19 | 2019-11-12 | 宁波工程学院 | Monitoring method based on artificial intelligence under a kind of scene towards bank |
| CN110610118A (en) * | 2018-06-15 | 2019-12-24 | 杭州海康威视数字技术股份有限公司 | Traffic parameter acquisition method and device |
| CN111260928A (en) * | 2018-11-30 | 2020-06-09 | 浙江宇视科技有限公司 | Method and device for detecting pedestrian without giving way to vehicle |
| CN111462501A (en) * | 2020-05-21 | 2020-07-28 | 山东师范大学 | Super-view area passing system based on 5G network and implementation method thereof |
| CN112154492A (en) * | 2018-03-19 | 2020-12-29 | 德尔克股份有限公司 | Early warning and collision avoidance |
| WO2021004077A1 (en) * | 2019-07-09 | 2021-01-14 | 华为技术有限公司 | Method and apparatus for detecting blind areas of vehicle |
| CN113793514A (en) * | 2021-08-30 | 2021-12-14 | 中冶南方城市建设工程技术有限公司 | Traffic safety warning system for entrances and exits of surrounding plots of construction roads |
| US11443631B2 (en) | 2019-08-29 | 2022-09-13 | Derq Inc. | Enhanced onboard equipment |
| CN115240471A (en) * | 2022-08-09 | 2022-10-25 | 东揽(南京)智能科技有限公司 | Intelligent factory collision avoidance early warning method and system based on image acquisition |
| CN115273479A (en) * | 2022-09-19 | 2022-11-01 | 深圳市博科思智能股份有限公司 | Operation and maintenance management method, device and equipment based on image processing and storage medium |
| CN117789486A (en) * | 2024-02-28 | 2024-03-29 | 南京莱斯信息技术股份有限公司 | Monitoring system and method for right turn stop of intersection of large-sized vehicle |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5590217A (en) * | 1991-04-08 | 1996-12-31 | Matsushita Electric Industrial Co., Ltd. | Vehicle activity measuring apparatus |
| CN101179710A (en) * | 2007-11-30 | 2008-05-14 | 浙江工业大学 | Intelligent video monitoring apparatus of railway crossing |
| CN101388145A (en) * | 2008-11-06 | 2009-03-18 | 北京汇大通业科技有限公司 | Auto alarming method and device for traffic safety |
| CN103345840A (en) * | 2013-05-28 | 2013-10-09 | 南京正保通信网络技术有限公司 | Video detection method of road crossing event at cross road |
| CN103971521A (en) * | 2014-05-19 | 2014-08-06 | 清华大学 | Method and device for detecting road traffic abnormal events in real time |
| CN104050684A (en) * | 2014-05-27 | 2014-09-17 | 华中科技大学 | Video moving object classification method and system based on on-line training |
-
2016
- 2016-08-26 CN CN201610735587.XA patent/CN106373430B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5590217A (en) * | 1991-04-08 | 1996-12-31 | Matsushita Electric Industrial Co., Ltd. | Vehicle activity measuring apparatus |
| CN101179710A (en) * | 2007-11-30 | 2008-05-14 | 浙江工业大学 | Intelligent video monitoring apparatus of railway crossing |
| CN101388145A (en) * | 2008-11-06 | 2009-03-18 | 北京汇大通业科技有限公司 | Auto alarming method and device for traffic safety |
| CN103345840A (en) * | 2013-05-28 | 2013-10-09 | 南京正保通信网络技术有限公司 | Video detection method of road crossing event at cross road |
| CN103971521A (en) * | 2014-05-19 | 2014-08-06 | 清华大学 | Method and device for detecting road traffic abnormal events in real time |
| CN104050684A (en) * | 2014-05-27 | 2014-09-17 | 华中科技大学 | Video moving object classification method and system based on on-line training |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108200552A (en) * | 2017-12-14 | 2018-06-22 | 华为技术有限公司 | A kind of V2X communication means and device |
| CN108200552B (en) * | 2017-12-14 | 2020-08-25 | 华为技术有限公司 | V2X communication method and device |
| CN112154492A (en) * | 2018-03-19 | 2020-12-29 | 德尔克股份有限公司 | Early warning and collision avoidance |
| US11749111B2 (en) | 2018-03-19 | 2023-09-05 | Derq Inc. | Early warning and collision avoidance |
| US11763678B2 (en) | 2018-03-19 | 2023-09-19 | Derq Inc. | Early warning and collision avoidance |
| CN108492567A (en) * | 2018-04-24 | 2018-09-04 | 汪宇明 | Monitor terminal, road traffic method for early warning and system |
| CN110610118A (en) * | 2018-06-15 | 2019-12-24 | 杭州海康威视数字技术股份有限公司 | Traffic parameter acquisition method and device |
| CN108877269A (en) * | 2018-08-20 | 2018-11-23 | 清华大学 | A kind of detection of intersection vehicle-state and V2X broadcasting method |
| CN109147326A (en) * | 2018-09-06 | 2019-01-04 | 北京理工大学 | A kind of Campus transport safety warning system |
| CN109191852A (en) * | 2018-10-25 | 2019-01-11 | 西北工业大学 | Che-road-cloud collaboration traffic flow Tendency Prediction method |
| CN109191852B (en) * | 2018-10-25 | 2021-07-06 | 西北工业大学 | Vehicle-road-cloud cooperative traffic flow situation prediction method |
| CN111260928A (en) * | 2018-11-30 | 2020-06-09 | 浙江宇视科技有限公司 | Method and device for detecting pedestrian without giving way to vehicle |
| CN111260928B (en) * | 2018-11-30 | 2021-07-20 | 浙江宇视科技有限公司 | Method and device for detecting pedestrian without giving way to vehicle |
| CN109830123B (en) * | 2019-03-22 | 2022-01-14 | 大陆投资(中国)有限公司 | Crossing collision early warning method and system |
| CN109830123A (en) * | 2019-03-22 | 2019-05-31 | 大陆投资(中国)有限公司 | Crossing anti-collision warning method and system |
| WO2021004077A1 (en) * | 2019-07-09 | 2021-01-14 | 华为技术有限公司 | Method and apparatus for detecting blind areas of vehicle |
| CN110443161A (en) * | 2019-07-19 | 2019-11-12 | 宁波工程学院 | Monitoring method based on artificial intelligence under a kind of scene towards bank |
| CN110443161B (en) * | 2019-07-19 | 2023-08-29 | 宁波工程学院 | Monitoring method based on artificial intelligence in banking scene |
| US11443631B2 (en) | 2019-08-29 | 2022-09-13 | Derq Inc. | Enhanced onboard equipment |
| US11688282B2 (en) | 2019-08-29 | 2023-06-27 | Derq Inc. | Enhanced onboard equipment |
| CN111462501A (en) * | 2020-05-21 | 2020-07-28 | 山东师范大学 | Super-view area passing system based on 5G network and implementation method thereof |
| CN113793514A (en) * | 2021-08-30 | 2021-12-14 | 中冶南方城市建设工程技术有限公司 | Traffic safety warning system for entrances and exits of surrounding plots of construction roads |
| CN115240471A (en) * | 2022-08-09 | 2022-10-25 | 东揽(南京)智能科技有限公司 | Intelligent factory collision avoidance early warning method and system based on image acquisition |
| CN115240471B (en) * | 2022-08-09 | 2024-03-01 | 东揽(南京)智能科技有限公司 | Intelligent factory collision avoidance early warning method and system based on image acquisition |
| CN115273479A (en) * | 2022-09-19 | 2022-11-01 | 深圳市博科思智能股份有限公司 | Operation and maintenance management method, device and equipment based on image processing and storage medium |
| CN117789486A (en) * | 2024-02-28 | 2024-03-29 | 南京莱斯信息技术股份有限公司 | Monitoring system and method for right turn stop of intersection of large-sized vehicle |
| CN117789486B (en) * | 2024-02-28 | 2024-05-10 | 南京莱斯信息技术股份有限公司 | Monitoring system and method for right turn stop of intersection of large-sized vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106373430B (en) | 2023-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106373430A (en) | Intersection pass early warning method based on computer vision | |
| CN103593678B (en) | A kind of long-span bridge vehicle dynamic load distribution detection method | |
| CN102867417B (en) | Taxi anti-forgery system and taxi anti-forgery method | |
| Yu et al. | Traffic light detection during day and night conditions by a camera | |
| CN103093249B (en) | A kind of taxi identification method based on HD video and system | |
| US20020054210A1 (en) | Method and apparatus for traffic light violation prediction and control | |
| Fernández-Caballero et al. | Road-traffic monitoring by knowledge-driven static and dynamic image analysis | |
| CN107031623A (en) | A kind of road method for early warning based on vehicle-mounted blind area camera | |
| CN103914688A (en) | Urban road obstacle recognition system | |
| CN111800507A (en) | Traffic monitoring method and traffic monitoring system | |
| DE112013001858T5 (en) | Multiple-hint object recognition and analysis | |
| CN111833598B (en) | Automatic traffic incident monitoring method and system for unmanned aerial vehicle on highway | |
| JP4993644B2 (en) | Train operation support device | |
| CN114898296A (en) | Bus lane occupation detection method based on millimeter wave radar and vision fusion | |
| CN113505638B (en) | Method and device for monitoring traffic flow and computer readable storage medium | |
| CN112308032B (en) | Intelligent cooperative linkage control method for underground air door through video detection | |
| CN102163278A (en) | Illegal vehicle intruding detection method for bus lane | |
| Pornpanomchai et al. | Object detection and counting system | |
| KR100532058B1 (en) | Traffic information acquisition method and apparatus using camera calibration | |
| CN103942541A (en) | Electric vehicle automatic detection method based on vehicle-mounted vision within blind zone | |
| CN205983875U (en) | Early warning system passes in intersection based on computer vision | |
| CN202887450U (en) | Taxi anti-fake system | |
| CN109686104A (en) | A kind of vehicle super-high detection device | |
| Yu et al. | A traffic light detection method | |
| CN112686956A (en) | Method for detecting tilt fault of urban road signal lamp post |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |