CN106023650A - Traffic intersection video and computer parallel processing-based real-time pedestrian early-warning method - Google Patents
Traffic intersection video and computer parallel processing-based real-time pedestrian early-warning method Download PDFInfo
- Publication number
- CN106023650A CN106023650A CN201610515762.4A CN201610515762A CN106023650A CN 106023650 A CN106023650 A CN 106023650A CN 201610515762 A CN201610515762 A CN 201610515762A CN 106023650 A CN106023650 A CN 106023650A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- target
- video
- thread
- foreground
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a traffic intersection video and computer parallel processing-based real-time pedestrian early-warning method. The method comprises the steps of extracting the moving foreground of an intersection, extracting pedestrian targets in the classified manner, tracking pedestrian targets in real time and alarming the condition of a pedestrian entering the intersection. During the early-warning process, moving targets in the foreground of a monitoring video is extracted based on the Vibe algorithm. After that, pedestrian targets and non-pedestrian targets, out of all moving targets in the foreground, are classified by using an off-line trained pedestrian linear SVM classification model. Finally, pedestrian targets are tracked based on the joint probabilistic data association (JPDA) tracking algorithm, so that the moving data of pedestrians are acquired. Therefore, the early-warning is conducted. According to the technical scheme of the invention, a traffic intersection video is detected, and a pedestrian target in the traffic intersection video is tracked and processed in real time. In this way, an improved method is provided. The multi-thread parallel processing is conducted and the shared data of queues are buffered dynamically. As a result, on the premise that the calculated amount of the foreground detecting and tracking algorithm is relatively large, the monitoring video can be maximally ensured to be processed in real time.
Description
Technical field
The invention belongs to the video image processing technology utilization in the traffic control system of road vehicle, be for traffic road
The solution of mouth video pedestrian's object real-time tracking early warning.It is mainly used in solving foreground extraction algorithm, track algorithm meter
The problem being difficult to real-time tracking early warning in the case of calculation amount is big.
Background technology
Along with the development of economic society, popularizing of private car, the vehicle that road travels gets more and more, to high-risk section,
The early warning of crossing traffic accident the most increasingly receives publicity.In traffic intersection accident, vehicle and pedestrian, cyclist with
It is easiest to the accident that collides between pedestrian, protects the important topic that the life security of pedestrian is always in traffic safety,
So providing the real-time early warning for crossing pedestrian to be to have certain practical value on the vehicle commuting crossing.Real
Time ground carry out accurate early warning, be this type of scene application important prerequisite.
Number of patent application is CN201110205121.6, entitled " a kind of assistant system of pedestrian safety of intersection "
Patent discloses a kind of method of pedestrian's early warning, uses Single-chip Controlling, laser fence, trips out according to swashing at when red
Whether light barrier light path is interrupted, and detects pedestrian, and vehicular traffic is carried out early warning.The method is very simple, but has
Following shortcoming, needs to add extra device at crossing, interrupts situation for light path and is only judged as pedestrian, to pedestrian's
The direction of motion fails to make a decision.Number of patent application is CN200810175852.9, and entitled " traffic safety is automatic
Alarming method for power and device " patent use scheme be in traffic intersection, the camera video data collected to be carried out
Background modeling, then distinguishes pedestrian and vehicle according to the position of foreground target, size, movement velocity.Method compares
Simply, amount of calculation is little, it is simple to calculate in real time, but accuracy can be affected.And distinguish pedestrian and vehicle time
Wait, owing to being affected by target sizes, need substantial amounts of empirical data, at different crossings, certainly will need a large amount of
Debugging.
For the extraction of foreground target, domestic and international research worker proposes a lot of foreground detection method based on video.?
In numerous foreground detection algorithms, vibe foreground detection is one of conventional method, and vibe algorithm is for the change of illumination
The effects such as the shake with camera are the most sufficiently stable, are relatively suitable for the outdoor such scene of traffic intersection.
For target following, at paper " Tracking-Learning-Detection ", IEEE Pattern
Analysis and Machine Intelligence, 2011, Z.Kalal et al. disclosed targets follow the tracks of calculation for a long time
Method TLD, this algorithm differs from traditional track algorithm and traditional detection algorithm phase with tradition track algorithm
The problems such as combination solves to follow the tracks of the deformation that occurs during tracked of target, partial occlusion are a kind of real-times with
The reasonable method of effective balance.But in the disclosed methods, it is only applicable to the tracking of single goal, it is impossible to adapt to many mesh
Mark follows the tracks of scene.JPDA (Joint Probabilistic Data Association, JPDA)
Being a kind of data association algorithm being applicable to multiple target situation of professor's Bar-Shalom proposition, his advantage is many mesh
Mark is followed the tracks of, and amount of calculation is less, solves measurement and track matching problem that target is intersected, and suitable application area is wide,
In the tracking of Video processing equally applicable.
Summary of the invention
The present invention is used for solving conventional foreground extraction algorithm, track algorithm and is difficult in the case of computationally intensive in real time
The problem of tracking and early warning.
To this end, the technical scheme that the present invention proposes is based on traffic intersection video and the real-time early warning of computer parallel computation
Method, comprises the steps of
A1, pretreatment
The traffic intersection video of real scene is delimited effective monitoring region, delimits the direction entering crossing according to the direction at crossing,
And keeping records；
A2, the extraction of sport foreground
Start one and process thread, use video foreground detection algorithm to extract all of moving target in video, and protected
Deposit to dynamic queue；
A3, the classification of pedestrian target
Start a classification thread, use SVM pedestrian's disaggregated model of off-line training, take out successively from above-mentioned dynamic queue
Moving target filters out pedestrian target, preserves to dynamic queue；
A4, the tracking of pedestrian target
Start a track thread, from dynamic queue, take out pedestrian target successively, use JPDA track algorithm that target is carried out
Follow the tracks of, and analyze the exercise data of each destination path；
A5, exercise data to each above-mentioned destination path calculate each target direction of motion in effective monitoring region,
The direction being compared with the direction entering crossing of setting in pretreatment, having shown that pedestrian enters crossing if analyzing,
Then carry out early warning.
Further, video foreground detection algorithm described in step A2 is Vibe.
Further, all of moving target using video foreground detection algorithm to extract in video in step A2 specifically wraps
Include:
Step 3-1: background model initializing, utilizes single frame video sequence initialization background model, for a pixel,
Have the spatial characteristics of close pixel value in conjunction with neighbor pixel, the random field point pixel value selecting it is made
Model sample value for him；
Step 3-2: to next frame video, calculates the distance of each sample value in new pixel and sample set, if apart from little
In threshold value, then approximation sample point number increases, if approximation sample point number is more than threshold value, then it is assumed that new pixel
Point is background；
Step 3-3: background model is updated, when determining the background model needing to update pixel, with new picture every time
The random sample value replacing this pixel sample set of element value；
Step 3-4: the foreground target obtained is preserved to a dynamic queue by each frame, waits pedestrian target classification line
The use of journey, when this queue is empty, is notified that subsequent thread is hung up, and waiting list capacity reaches some.
Further, in step A3, the classification of pedestrian target includes:
Step 4-1: start a thread, foreground target data in dynamic queue are done two classification of pedestrian；
Step 4-2: foreground target picture is obtained hog feature, judges according to the value of the linear function of Linear SVM
Classification is pedestrian or non-pedestrian；
Step 4-3: pedestrian target SVM classifier obtained preserves to a dynamic queue, waits follow-up tracking
The use of thread, when receiving foreground extraction thread activation notice, then thread activation, when depositing the dynamic of pedestrian target
When state queue is empty, being notified that subsequent thread is hung up, waiting list capacity activates follow-up again after reaching some
Thread.
Further, the tracking of step A4 pedestrian target includes:
Step 5-1: start a thread, when there being pedestrian target, initializes JPDA track algorithm；
Step 5-2:JPDA carries out Model Condition filtering, draws the routing information of each target；
Step 5-3: according to the routing information of target, according to the speed v of its horizontal directionx, vertical direction vy, computer
The turnover direction, crossing delimited in velocity attitude, with pretreatment compares and show whether target enters crossing, if
Enter, then carry out early warning car and notice that there is pedestrian at crossing；
Step 5-4: when all not having pedestrian target to occur in continuous 5 frame data, then follow the tracks of JPDA and reset, flight path
Remove, reinitialize after obtaining new pedestrian target；
Step 5-5: when obtaining pedestrian target classification thread suspension notice, then by thread suspension, wait and regain activation
Signal.
Beneficial effect: present invention is mainly used for solving foreground extraction algorithm, be difficult to reality in the case of track algorithm is computationally intensive
The solution of tracking and early warning problem during reality.Use the multiple computationally intensive algorithm of multi-threading parallel process, utilize dynamic
State queue technology carries out cross-thread Techno-sharing, enables crossing pedestrian's warning algorithm real time execution of complexity.At Duo
I5-4590 processor, the computer of 8g internal memory runs, is provided without algorithm process per second 13 frame of parallel processing, adopts
After parallel processing, algorithm process per second 25 frame, reach requirement of real time.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of the present invention.
Fig. 2 is the multi-thread concurrent work feedback schematic diagram of the present invention.
Detailed description of the invention
In conjunction with accompanying drawing, specific embodiments of the present invention are further described in detail.The present invention uses multithreading also
Row processes multiple computationally intensive algorithms, utilizes dynamic queue's technology to carry out cross-thread Techno-sharing, makes the crossing of complexity
Pedestrian's warning algorithm can real time execution.Mainly comprise the steps of
A1, pretreatment:
The traffic intersection video of real scene is delimited effective monitoring region, delimits the direction entering crossing according to the direction at crossing,
And keeping records
A2, the extraction of sport foreground:
Start one and process thread, use Vibe algorithm to extract all of moving target in video, such as the left side square frame of Fig. 1
Shown in part, and it is saved in dynamic queue, as shown in the left-half of Fig. 2.
A3, the detection of pedestrian target:
Start a classification thread, use SVM pedestrian's disaggregated model of off-line training, from dynamic queue, take out motion successively
Object filtering goes out pedestrian target, as shown in the middle Blocked portion of Fig. 1, preserves to dynamic queue, such as the right side of Fig. 2
Shown in half part.
A4, the tracking of pedestrian target:
Start a track thread, from dynamic queue, take out pedestrian target successively, use JPDA track algorithm that target is carried out
Follow the tracks of, and analyze the exercise data of each destination path, as shown in the right Blocked portion of Fig. 1.
A5, exercise data to each target calculate the direction of motion in effective video region of each target, by the party
Compare to the crossing approach axis of setting in pretreatment, shown that pedestrian enters crossing if analyzing, then carried out pre-
Alert.
For further technical scheme being illustrated, now make an explanation from the following aspect.
One, architecture
Fig. 1 gives based on traffic intersection video, pedestrian's real-time early warning of computer parallel processing system
Flow chart.Here having three modular concurrent to run, by foreground extraction in algorithm, pedestrian target is classified, pedestrian target with
Track separately, if synchronous operation, uses single-threaded working method, and amount of calculation causes the most greatly can not analysis and early warning in real time.Make
By the mode of concurrent operation, effectively utilize the multinuclear process performance of computer, make entirety meet real-time requirement.
Visual Background extractor (Vibe) foreground detection algorithm: compared to the pre-police of other traffic intersections
Method, for background modeling, foreground extraction, uses Vibe detection algorithm.The main thought of this algorithm is concrete thought
Be exactly to store a sample set for each pixel, in sample set sampled value be exactly this pixel past pixel value and
The pixel value of its neighbours point, then is compared to judge whether to belong to background by each new pixel value and sample set
Point.This model mainly includes three aspects: the operation principle of model；The initial method of model；The renewal plan of model
Slightly.
Support vector machines sorting algorithm: being a kind of two classification model, its basic model is defined as on feature space
The linear classifier that interval is maximum, i.e. the learning strategy of support vector machine is margin maximization, finally can be converted into one
Solving of individual convex quadratic programming problem.SVM algorithm, in the case of small sample, has preferable robustness.
JPDA multiple target tracking algorithm: JPDA algorithm, on the basis of PDA algorithm and nearest neighbor algorithm, not only utilizes
Update desired value, but utilize residual error update probability and carry out Federated filter.Realize multiple target tracking, to mobility relatively
Strong target also can preferably be tracked.
Two, method flow
1, pretreatment stage
For specific traffic intersection video, choose two pixel (X in porch, crossing and exit1,Y1),
(X2,Y2), according to slope computing formulaObtain the slope in turnover direction, crossing.And utilize off-line to obtain
Crossing pedestrian's sectional drawing and non-crossing pedestrian's sectional drawing, use Linear SVM two to classify grader, the hog choosing picture is special
Levying, training obtains SVM classifier model.
2, foreground extraction process
Start a thread, first the video obtained is decoded, obtain sequence of frames of video, and carry out foreground extraction.
Step 2-1: background model initializing, utilizes single frame video sequence initialization background model, for a pixel, knot
Closing neighbor pixel and have the spatial characteristics of close pixel value, random selects its field point pixel value as him
Model sample value.
Step 2-2: to next frame video, calculates the distance of each sample value in new pixel and sample set, if distance is less than threshold
Value, then approximation sample point number increases.If approximation sample point number is more than threshold value, then it is assumed that new pixel is the back of the body
Scape.Mainly determined by three parameters during detection: sample set number N, threshold value min and the threshold of closely located judgement
Value R, herein, parameter is set to N=20, min=2, R=20.
Pixel value at X point: V (X)
Background sample set at X, sample set size is N:M (x)={ v1,v2,…,vN}
The region as radius of the R centered by X: SR(v(x))
Judge whether new pixel is background:
{SR(v(x))∩{v1,v2,…,vN}}≥min
Step 2-3: background model is updated, when determining the background model needing to update pixel, with new pixel value every time
The random sample value replacing this pixel sample set.
Step 2-4: the foreground target obtained is preserved to a dynamic queue by each frame, waits pedestrian target classification thread
Use.When this queue is empty, being notified that subsequent thread is hung up, waiting list capacity reaches somes M1, herein
M1=20, again activate subsequent thread.
3, the classification of pedestrian target
Start a thread, foreground target data in dynamic queue are done two classification of pedestrian.
Step 3-1: foreground target picture is obtained hog feature, according to the linear function of Linear SVM: g (x)=wx+b, its
Middle w, b are the parameter of SVM classifier model, and x is the hog characteristic of foreground picture.G (x) ＞ 0, it is judged that
It is pedestrian for classification, otherwise is non-pedestrian.
Step 3-2: pedestrian target SVM classifier obtained preserves to a dynamic queue, waits follow-up track thread
Use.When receiving foreground extraction thread activation notice, then thread activation.When the dynamic queue depositing pedestrian target
During for sky, being notified that subsequent thread is hung up, waiting list capacity reaches somes M2, M herein2=30, again swash
Subsequent thread alive.
4, JPDA pedestrian target is followed the tracks of
Start a thread, pedestrian target in dynamic queue is tracked.
Step 5-1: when there being pedestrian target, initializes JPDA track algorithm.
Step 5-2:JPDA carries out Model Condition filtering, draws the routing information of each target.
Step 5-3: according to the routing information of target, according to the speed v of its horizontal directionx, vertical direction vy, computer speed
The turnover direction, crossing delimited in direction, with pretreatment compares and show whether target enters crossing, if entering,
Then carry out early warning car and notice that there is pedestrian at crossing.
Step 5-4: when all not having pedestrian target to occur in continuous 5 frame data, then follow the tracks of JPDA and reset, and flight path is removed,
Reinitialize after obtaining new pedestrian target.
Step 5-5: when obtaining pedestrian target classification thread suspension notice, then by thread suspension, wait and regain activation signal.
Claims (5)
1. based on traffic intersection video and the real time early warning method of computer parallel computation, it is characterised in that comprise the steps of
A1, pretreatment；
The traffic intersection video of real scene is delimited effective monitoring region, delimits the direction entering crossing according to the direction at crossing, and
And keeping records；
A2, the extraction of sport foreground；
Start one and process thread, use video foreground detection algorithm to extract all of moving target in video, and be saved to
In dynamic queue；
A3, the classification of pedestrian target；
Start a classification thread, use SVM pedestrian's disaggregated model of off-line training, from above-mentioned dynamic queue, take out motion successively
Object filtering goes out pedestrian target, preserves to dynamic queue；
A4, the tracking of pedestrian target；
Start a track thread, from dynamic queue, take out pedestrian target successively, use JPDA track algorithm that target is tracked,
And analyze the exercise data of each destination path；
A5, exercise data to each above-mentioned destination path calculate each target direction of motion in effective monitoring region, should
Direction compares with the direction entering crossing of setting in pretreatment, has shown that pedestrian enters crossing if analyzing, has then carried out pre-
Alert.
The most according to claim 1 based on traffic intersection video and the real time early warning method of computer parallel computation, its feature exists
It is Vibe in video foreground detection algorithm described in step A2.
The most according to claim 1 based on traffic intersection video and the real time early warning method of computer parallel computation, its feature exists
The all of moving target using video foreground detection algorithm to extract in video in step A2 specifically includes:
Step 3-1: background model initializing, utilizes single frame video sequence initialization background model, for a pixel, in conjunction with phase
Adjacent pixel has the spatial characteristics of close pixel value, the random field point pixel value model sample as him selecting it
This value；
Step 3-2: to next frame video, calculates the distance of each sample value in new pixel and sample set, if distance is less than threshold value,
Then approximation sample point number increases, if approximation sample point number is more than threshold value, then it is assumed that new pixel is background；
Step 3-3: background model is updated, when determining the background model needing to update pixel, random with new pixel value every time
Replace a sample value of this pixel sample set；
Step 3-4: the foreground target obtained is preserved to a dynamic queue by each frame, waits the use of pedestrian target classification thread,
When this queue is empty, being notified that subsequent thread is hung up, waiting list capacity reaches some.
The most according to claim 1 based on traffic intersection video and the real time early warning method of computer parallel computation, its feature exists
In step A3, the classification of pedestrian target includes:
Step 4-1: start a thread, foreground target data in dynamic queue are done two classification of pedestrian；
According to the value of the linear function of Linear SVM, step 4-2: foreground target picture is obtained hog feature, judges that classification is as row
People or non-pedestrian；
Step 4-3: pedestrian target SVM classifier obtained preserves to a dynamic queue, waits making of follow-up track thread
With, when receiving foreground extraction thread activation notice, then thread activation, when the dynamic queue depositing pedestrian target is empty,
Being notified that subsequent thread is hung up, waiting list capacity activates subsequent thread after reaching some again.
The most according to claim 1 based on traffic intersection video and the real time early warning method of computer parallel computation, its feature exists
Tracking in step A4 pedestrian target includes:
Step 5-1: start a thread, when there being pedestrian target, initializes JPDA track algorithm；
Step 5-2:JPDA carries out Model Condition filtering, draws the routing information of each target；
Step 5-3: according to the routing information of target, according to the speed v of its horizontal directionx, vertical direction vy, computer velocity attitude,
Comparing with the turnover direction, crossing delimited in pretreatment and show whether target enters crossing, if entering, then carrying out early warning
Car notices that there is pedestrian at crossing；
Step 5-4: when all not having pedestrian target to occur in continuous 5 frame data, then follow the tracks of JPDA and reset, and flight path is removed, when
Reinitialize after obtaining new pedestrian target；
Step 5-5: when obtaining pedestrian target classification thread suspension notice, then by thread suspension, wait and regain activation signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610515762.4A CN106023650B (en) | 2016-07-01 | 2016-07-01 | Real-time pedestrian's method for early warning based on traffic intersection video and computer parallel processing system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610515762.4A CN106023650B (en) | 2016-07-01 | 2016-07-01 | Real-time pedestrian's method for early warning based on traffic intersection video and computer parallel processing system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106023650A true CN106023650A (en) | 2016-10-12 |
| CN106023650B CN106023650B (en) | 2018-11-30 |
Family
ID=57106260
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610515762.4A Active CN106023650B (en) | 2016-07-01 | 2016-07-01 | Real-time pedestrian's method for early warning based on traffic intersection video and computer parallel processing system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106023650B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530818A (en) * | 2016-12-30 | 2017-03-22 | 北京航空航天大学 | Intelligent parking lot management system based on video processing technology |
| CN106874864A (en) * | 2017-02-09 | 2017-06-20 | 广州中国科学院软件应用技术研究所 | A kind of outdoor pedestrian's real-time detection method |
| CN109544986A (en) * | 2017-09-21 | 2019-03-29 | 帕斯网络有限公司 | Utilize the pedestrian protection system of beacon signal |
| CN110188607A (en) * | 2019-04-23 | 2019-08-30 | 深圳大学 | A kind of the traffic video object detection method and device of multithreads computing |
| CN113223276A (en) * | 2021-03-25 | 2021-08-06 | 桂林电子科技大学 | Pedestrian hurdling behavior alarm method and device based on video identification |
| WO2021184621A1 (en) * | 2020-03-19 | 2021-09-23 | 南京因果人工智能研究院有限公司 | Multi-object vehicle tracking method based on mdp |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509101B (en) * | 2011-11-30 | 2013-06-26 | 昆山市工业技术研究院有限责任公司 | Background updating method and vehicle target extracting method in traffic video monitoring |
| CN103164711B (en) * | 2013-02-25 | 2016-08-03 | 昆山南邮智能科技有限公司 | The method of region based on pixel and support vector machine artificial abortion's density Estimation |
| CN103902976B (en) * | 2014-03-31 | 2017-12-29 | 浙江大学 | A kind of pedestrian detection method based on infrared image |
| CN104008371B (en) * | 2014-05-22 | 2017-02-15 | 南京邮电大学 | Regional suspicious target tracking and recognizing method based on multiple cameras |
| CN104134360A (en) * | 2014-08-14 | 2014-11-05 | 奇瑞汽车股份有限公司 | Intersection pedestrian recognition safety control system and method based on short-range communication |
| CN104217428B (en) * | 2014-08-22 | 2017-07-07 | 南京邮电大学 | A kind of fusion feature matching and the video monitoring multi-object tracking method of data correlation |
| KR101663574B1 (en) * | 2014-12-01 | 2016-10-07 | 계명대학교 산학협력단 | Method and system for detection of sudden pedestrian crossing for safe driving during night time |
-
2016
- 2016-07-01 CN CN201610515762.4A patent/CN106023650B/en active Active
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530818A (en) * | 2016-12-30 | 2017-03-22 | 北京航空航天大学 | Intelligent parking lot management system based on video processing technology |
| CN106874864A (en) * | 2017-02-09 | 2017-06-20 | 广州中国科学院软件应用技术研究所 | A kind of outdoor pedestrian's real-time detection method |
| CN109544986A (en) * | 2017-09-21 | 2019-03-29 | 帕斯网络有限公司 | Utilize the pedestrian protection system of beacon signal |
| CN109544986B (en) * | 2017-09-21 | 2021-06-15 | 帕斯网络有限公司 | Pedestrian protection system using beacon signals |
| CN110188607A (en) * | 2019-04-23 | 2019-08-30 | 深圳大学 | A kind of the traffic video object detection method and device of multithreads computing |
| CN110188607B (en) * | 2019-04-23 | 2022-10-21 | 深圳大学 | Traffic video target detection method and device based on multi-thread parallel computing |
| WO2021184621A1 (en) * | 2020-03-19 | 2021-09-23 | 南京因果人工智能研究院有限公司 | Multi-object vehicle tracking method based on mdp |
| CN113223276A (en) * | 2021-03-25 | 2021-08-06 | 桂林电子科技大学 | Pedestrian hurdling behavior alarm method and device based on video identification |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106023650B (en) | 2018-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106023650A (en) | Traffic intersection video and computer parallel processing-based real-time pedestrian early-warning method | |
| US9805474B1 (en) | Pedestrian tracking at a traffic intersection to identify vulnerable roadway users for traffic signal timing, pedestrian safety, and traffic intersection control | |
| US9460613B1 (en) | Pedestrian counting and detection at a traffic intersection based on object movement within a field of view | |
| CN102819764B (en) | Method for counting pedestrian flow from multiple views under complex scene of traffic junction | |
| CN108230254B (en) | Automatic detection method for high-speed traffic full lane line capable of self-adapting scene switching | |
| US9449506B1 (en) | Pedestrian counting and detection at a traffic intersection based on location of vehicle zones | |
| Rahman et al. | A real-time wrong-way vehicle detection based on YOLO and centroid tracking | |
| CN111932583A (en) | Space-time information integrated intelligent tracking method based on complex background | |
| CN107423679A (en) | A kind of pedestrian is intended to detection method and system | |
| CN102663452A (en) | Suspicious act detecting method based on video analysis | |
| CN104134078B (en) | Automatic selection method for classifiers in people flow counting system | |
| Cui et al. | Abnormal event detection in traffic video surveillance based on local features | |
| CN102768726A (en) | Pedestrian detection method for preventing pedestrian collision | |
| Subaweh et al. | Implementation of pixel based adaptive segmenter method for tracking and counting vehicles in visual surveillance | |
| Malhi et al. | Vision based intelligent traffic management system | |
| CN116153086B (en) | Multi-path traffic accident and congestion detection method and system based on deep learning | |
| Su et al. | A new local-main-gradient-orientation HOG and contour differences based algorithm for object classification | |
| Swathy et al. | Survey on vehicle detection and tracking techniques in video surveillance | |
| WO2017196515A1 (en) | Pedestrian counting and detection at a traffic intersection based on location of vehicle zones | |
| Wang et al. | Vision-based highway traffic accident detection | |
| Deng et al. | Skeleton model based behavior recognition for pedestrians and cyclists from vehicle sce ne camera | |
| CN113537170A (en) | Intelligent traffic road condition monitoring method and computer readable storage medium | |
| BOURJA et al. | Real time vehicle detection, tracking, and inter-vehicle distance estimation based on stereovision and deep learning using YOLOv3 | |
| Al Jarouf et al. | A hybrid method to detect and verify vehicle crash with haar-like features and svm over the web | |
| Yang et al. | A robust vehicle queuing and dissipation detection method based on two cameras |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |