CN104809437B - A kind of moving vehicles detection and tracking method based on real-time video - Google Patents
A kind of moving vehicles detection and tracking method based on real-time video Download PDFInfo
- Publication number
- CN104809437B CN104809437B CN201510210294.5A CN201510210294A CN104809437B CN 104809437 B CN104809437 B CN 104809437B CN 201510210294 A CN201510210294 A CN 201510210294A CN 104809437 B CN104809437 B CN 104809437B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- result
- target
- tracking
- real
- 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
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of moving vehicles detection and tracking method based on real-time video, and this method comprises the following steps：S101, obtain original training set；S102, training preliminary classification device；S103, repetitive exercise grader；S104, the generation of vehicle candidate region；S105, vehicle confirm and tracking；S106, wagon flow statistics；S107, algorithm interaction.The detect and track that the present invention can stablize the vehicle occurred in real-time video, and vehicle flowrate is counted, not only robustness is high, it can adapt to noise, illumination and Changes in weather, also there is certain adaptability to blocking, and processing speed is fast, accuracy is high, can meet the operation demand of real-time system.
Description
Technical field
The present invention relates to moving vehicles detection and tracking technical field, more particularly to a kind of vehicle detection based on real-time video with
Tracking.
Background technology
Moving vehicles detection and tracking refers to obtain the information of vehicles occurred in picture automatically from by analyzing sequence of video images,
And tenacious tracking is carried out to same vehicle target in continuous videos.Moving vehicles detection and tracking technology is intelligent transportation system, intelligence
The key technology in the fields such as energy security protection, can be very good the work of auxiliary related personnel and improves work efficiency, therefore as number
The research hotspot of word image processing field.Common moving vehicles detection and tracking has road vehicle detection and tracking, entrance vehicle
Detection and tracking, the moving vehicles detection and tracking of satellite top view etc..
Moving vehicles detection and tracking based on video has higher practical value and is widely used, and many scholars cause in recent years
Power is in research related algorithm.Many vehicle checking methods are emerged so far, the state university's Reynolds branch school of Nevada ,Usa
Doctor Zehang (IEEE member) et al. is doing existing method numerous studies and summary, and in 2006 years《IEEE
TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》On publish thesis " On-Road
Vehicle Detection:A Review ", are generally summarized as generation candidate region by existing method in literary and candidate region are tested
Demonstrate,prove two steps, i.e., in the picture fast positioning be probably vehicle region, then further confirm that whether it is vehicle.Although
Researcher was also constantly improved vehicle detecting algorithm in recent years, but Integral Thought still follows the two steps.Wait
The method of the generation of favored area mainly has three classes, is based on priori, based on stereoscopic vision and based drive method respectively.
Based on the method for priori by analyzing the features such as the symmetry in image, color, shade, angle point, edge, texture, car light
To obtain vehicle candidate region.Method based on stereoscopic vision mainly obtains the disparity map of image, against thoroughly by two cameras
The stereoscopic features of vehicle are analyzed depending on conversion etc..Based drive method mainly by optical flow method, motion vector method, frame difference method and
The methods of background subtraction method, obtains the information of moving object so as to be used as vehicle candidate.The method of candidate region verification is mainly divided
For the two class methods based on template and based on appearance.Method based on template is the mould with obtaining in advance by the image of candidate region
Plate image confirms that this kind of method is had a great influence by template compared to relatively, cannot be by when candidate vehicle and template differ greatly
Correct detection.Method based on appearance is mainly the method study vehicle characteristics using machine learning, and by the spy of candidate region
Levy and relatively confirm compared with vehicle characteristics, this kind of methods and results are more stable, and in most cases performance is good, but detect knot
Fruit is had a great influence by training sample set.The method for tracking target of mainstream has kalman filter method, Meanshift methods at present
With Camshift methods etc., these methods are monotrack method, and the method for multiple target tracking is all based on greatly these sides
The improvement of method.It is worth noting that these trackings need to specify tracking target first, then carry out in the video sequence with
Track, is usually only to carry out vehicle detection at the beginning when vehicle checking method is combined, detects just stopping inspection after target
Survey and start to track the target, the advantage of doing so is that time overhead is few, and error detection can be suppressed.But this mode
Also there is the drawbacks of its is very important, error detection result will be directly passed to tracker, and have no chance to be corrected, if
Testing result is bad, can directly injure the reliability of tracker.
That weighs vehicle detecting algorithm performance refers mainly to indicate robustness, accuracy and processing speed.But existing side
Method is difficult to take into account this three indexs at the same time, although such as frame difference method and its robustness of edge detection method it is higher, processing speed also compared with
It hurry up, but the accuracy of both approaches is relatively low；And although optical flow method and motion vector method accuracy are fine, robustness compared with
Poor and processing speed is difficult that can not meet the application demand of real-time system；Background subtraction method is because have higher accuracy
With faster processing speed, application is relatively broad, but due to sensitive to noise, illumination and weather, its robustness is also one
The problem of can not be ignored.
The content of the invention
It is an object of the invention to by a kind of moving vehicles detection and tracking method based on real-time video, be carried on the back to solve the above
The problem of scape technology segment is mentioned.
For this purpose, the present invention uses following technical scheme：
A kind of moving vehicles detection and tracking method based on real-time video, it includes the following steps：
S101, obtain original training set：
Classify to the image results in file system, be respectively placed in different files, obtain initial sample
Pictures；The absolute path list of samples pictures in each file is respectively obtained using autoexec, and to each file
File in folder assigns different marks, and finally the sample absolute path list of tape label is incorporated into a text；
S102, training preliminary classification device：
The original training set obtained using step S101 trains grader, reads in each sample and its mark successively, analyzes
Its feature obtains the feature vector of each sample, carries out Supervised classification to feature vector, obtains preliminary classification device；
S103, repetitive exercise grader：
Classified using the step S102 preliminary classification devices obtained to the moving object in video sequence, and by moving object
Local area image where body adds timestamp name storage into file system by classification results mark, if user is to classification
As a result satisfaction need not then be iterated training, and unsatisfied misclassification is found out as a result, and adding it to just if dissatisfied
In the sample set file of true class, new training sample set text is obtained using step S101 the methods, is then utilized
Step S102 the methods are iterated training, so repeatedly until user is satisfied with classification results the iteration for then completing grader
Training；Grader write-in file is saved in file system；
S104, the generation of vehicle candidate region：
Grader file is read in, the moving region image occurred in video sequence is detected successively, analyzes moving region image
Feature obtain its feature vector, feature vector feeding grader is classified and obtains classification results, if the result is that car
Then think that the moving region image includes vehicle, the generation of vehicle candidate region is completed with this；
S105, vehicle confirm and tracking：
Vehicle detection in having been obtained in step S104 per two field picture is as a result, first using present frame as base in real-time video
Standard, to all vehicle detections in each vehicle detection result traversal previous frame of present frame as a result, will meet similarity indices
An and matching result for being considered as current goal of similarity measurement result minimum；Likewise, on the basis of previous frame, to upper
One frame each testing result traversal present frame as a result, will meet similarity indices and similarity measurement result it is minimum one
A matching result for being considered as previous frame target；Finally, if previous frame exists, present frame lost target is considered as target disappearance,
Be considered as fresh target if target that present frame occurs if previous frame does not occur and distribute new mark for it, if present frame and
Two targets of previous frame match each other, and are considered as same target and with same tag into line trace.
Especially, the moving vehicles detection and tracking method based on real-time video further includes：
S106, wagon flow statistics：
Measurement type is divided into three kinds：The vehicle fleet occurred in picture, drives into vehicle fleet, outgoing vehicles sum；Vehicle
The statistics of sum is carried out according to the result of vehicle tracking in step S105, is only just counted when fresh target occurs；Drive into
Counted with outgoing vehicles according to the mode of dummy line, when original state is that vehicle's current condition outside line is then to think it in line
To drive into vehicle, counted and changed its original state and be set in line, wherein, the statistical method of outgoing vehicles and drive into vehicle
Statistical method it is identical.
Especially, the moving vehicles detection and tracking method based on real-time video further includes：
S107, algorithm interaction：
Since the local area image of the vehicle target in each two field picture can be all stored into file system, while should
Information write service device data of the testing result including camera number, record time, vehicle target image results path
Storehouse；Every time when reaching vehicle flowrate time interval, by present period data write into Databasce and by counter O reset；Interaction journey
Sequence can carry out rendering display result when there is data insertion with the state in real-time monitoring data storehouse according to insertion data；Whenever up to
During to running state of programs daily record interval, program current state is write in journal file, to break down when has good grounds,
On the other hand, program provides configuration file and changes parameter for user, and user can be changed including vehicle flowrate time interval, program
Supplemental characteristic including operating status daily record interval, frame picture processing duty cycle；
Especially, the step S105 is further included：Improve vehicle detection effect using vehicle tracking result, tenacious tracking is pre-
If the target of number just confirms as vehicle, in this, as vehicle detection as a result, preserving the picture of regional area where vehicle at the same time
As a result file system is arrived and by testing result insertion of data into data storehouse.
Especially, the step S105 further comprises：For tenacious tracking as a result, giving if suddenly disappearing pre-
Survey the default frame number of compensation, if within default frame number target occur again if continue to track, otherwise it is assumed that target is lost.
Moving vehicles detection and tracking method proposed by the present invention based on real-time video has the following advantages that：First, robustness is high：
The amendment carried out using tracking result to testing result effectively overcomes the influence of noise, and the repetitive exercise of grader effectively overcomes
The influence of the change such as illumination and weather, and at any time more preferable testing result can be obtained with exptended sample collection.2nd, accuracy
It is high：Such as passing through 4 repetitive exercises, when sample set quantity is 4146, the size of grader is 89.1M, continuous testing results
There was only 3415 mistakes point in 62448 obtained classification results as a result, accuracy has reached 94.53%, and continuation iteration can
To reach the accuracy of higher.3rd, processing speed is fast, and program scale is controllable：Such as program differentiates all video inputs
Rate is adjusted to 640*360 and is handled again, therefore picture quality change has no effect on processing speed, average to be per frame processing speed
20ms.It is 4M in video code flow bandwidth, resolution ratio 1280*720, frame per second 25, when frame processing duty cycle is 5ms, CPU is used
Rate is 45% or so, and memory usage is 142M or so.And when frame processing duty cycle be changed into 200ms other specifications it is constant when, CPU
Utilization rate is reduced to 24% or so, and memory usage is 137M or so.When parameter is arranged to bandwidth 512K, resolution ratio 640*360,
Frame per second 15, during duty cycle 200ms, CPU usage is 17% or so, and memory usage is 115M or so, can be matched somebody with somebody completely low
Put smooth operation in computer.4th, there is good interactivity：Program performs different function moulds according to different input instructions
Block, operating parameter can be changed by user, and variation output result facilitates Front End to render display effect, stores to database
Mode the program is completely independent with Front End, front end can even use different programming languages.
Brief description of the drawings
Fig. 1 is the moving vehicles detection and tracking method flow diagram provided in an embodiment of the present invention based on real-time video；
Fig. 2 is vehicle candidate region generation method flow chart provided in an embodiment of the present invention；
Fig. 3 confirms and tracking flow chart for vehicle provided in an embodiment of the present invention；
Fig. 4 is vehicle tracking process schematic provided in an embodiment of the present invention.
Embodiment
The invention will be further described with reference to the accompanying drawings and examples.It is understood that tool described herein
Body embodiment is used only for explaining the present invention, rather than limitation of the invention.It also should be noted that for the ease of retouching
State, part related to the present invention rather than full content are illustrate only in attached drawing, it is unless otherwise defined, used herein all
Technical and scientific term is identical with the normally understood implication of those skilled in the art for belonging to the present invention.It is used herein
Term be intended merely to description specific embodiment, it is not intended that in limitation the present invention.
It refer to shown in Fig. 1, Fig. 1 is the moving vehicles detection and tracking method provided in an embodiment of the present invention based on real-time video
Flow chart.
The moving vehicles detection and tracking method based on real-time video specifically comprises the following steps in the present embodiment：
S101, obtain original training set：
All moving objects occurred in capture video sequence, and the image storage of regional area where moving object is arrived
In file system, manual sort is carried out to the image results in file system by user, is respectively placed in different files,
Obtain initial samples pictures collection.Such as it is classified as follows：0 non-vehicle class, the positive and negative noodles of 1 vehicle, 2 vehicle at night classes, 3 vehicular sideviews
There are shielding automobile class in class, 4 local vehicle classes, 5.The absolute of samples pictures in each file is respectively obtained using autoexec
Path list, and corresponding numeral mark is assigned to the file in each file, it is finally that the sample of these tape labels is absolute
Path list is incorporated into same text.
S102, training preliminary classification device：
Training preliminary classification device.The original training set obtained using step S101 trains grader, reads in each sample successively
Originally and its mark, such as first can uniformly adjust image to the size of 64*64, then analyze its histograms of oriented gradients and obtain
To 1764 dimensional feature vectors of each sample, Supervised classification is finally carried out to these feature vectors using support vector machines, is obtained
To preliminary classification device.
S103, repetitive exercise grader：
The preliminary classification device obtained using step S102 classifies the moving object in video sequence, and by moving object
Local area image where body adds timestamp name storage into file system by classification results mark, if user is to classification
As a result satisfaction need not then be iterated training, need to find out unsatisfied misclassification as a result, and being added if dissatisfied
Into the sample set file of correct class, new training sample set text is obtained using step S101 the methods, then
Training is iterated using step S102 the methods, so repeatedly until user, which is satisfied with classification results, then completes grader
Repetitive exercise.Grader write-in file is saved in file system.
S104, the generation of vehicle candidate region：
Grader file is read in first, is detected the moving region image occurred in video sequence successively, is analyzed moving region
The feature of image obtains its feature vector, and feature vector feeding grader is classified and obtains classification results, if result
It is that vehicle then thinks that the moving region image includes vehicle, the generation of vehicle candidate region is completed with this.It is specific as shown in Figure 2：
Initiation parameter；It is loaded into grader；Obtain next two field picture；Judge whether image is empty, if it is empty, then terminates, otherwise ought
Preceding result is copied to previous frame；Detect moving region；Moving region key words sorting is obtained using grader；Determine whether car
, if it is not, next two field picture is then obtained, if so, then preserving the information of candidate region；Vehicle confirms and tracking；Judge whether to move back
Go out, if so, then terminating, if it is not, then obtaining next two field picture.
S105, vehicle confirm and tracking：
As shown in Figure 3 and Figure 4, it is specific as follows：Obtained in step S104 per the vehicle candidate region in two field picture as a result,
In real-time video first on the basis of present frame, to all in each vehicle detection result Ci traversal previous frames of present frame
A Pj for meeting similarity indices and similarity measurement result minimum as a result, is considered as the matching of current goal by vehicle detection
As a result it is Ci->Pj.It is similar, on the basis of previous frame, to each testing result Pi traversal present frames of previous frame as a result,
A Cj for meeting similarity indices and similarity measurement result minimum is considered as to the matching result i.e. Pi- of previous frame target>
Cj.If last previous frame exists and present frame lost target is considered as target disappearance, if previous frame does not occur and present frame
The target of appearance is then considered as fresh target and new mark is distributed for it, if mutual of two targets of present frame and previous frame
Match somebody with somebody, that is, meet Ci->Pj and Pj->Ci, then Ci and Pj be considered as same target, and to Ci with the mark identical with Pj carry out with
Track.C1, C2, C3 of present frame are respectively the tracking result of P1, P2, P3 of previous frame in Fig. 4, and target P4 disappears, and C4 is new
Target.Improving vehicle detection effect using vehicle tracking result at the same time, the target that tenacious tracking is more than m times just confirms as vehicle,
In this, as vehicle detection as a result, preserving the image results of regional area where vehicle at the same time to file system and by testing result
Insertion of data into data storehouse, effectively eliminates partial noise interference；And give for the result of tenacious tracking if suddenly disappearing
Predictive compensation n frames, continue to track if target within n frames occurs again, otherwise it is assumed that target is lost, efficiently solve detection
The problem of unstable interruption.
S106, wagon flow statistics：
Measurement type is divided into three kinds, is the vehicle fleet occurred in timing statistics in picture respectively, drive into vehicle fleet with
And outgoing vehicles sum.The statistics of vehicle fleet is carried out according to the result of vehicle tracking in step S105, only when new mesh
Mark and just counted now, the count flag position of the target is changed after counting, the target that each frame occurs afterwards is all inherited
The flag bit prevents repeat count.Entering and exiting vehicle is counted according to the mode of dummy line, and vehicle is examined in order to prevent
The fluctuation of result coordinate is surveyed, dummy line sets certain width.According to its relative position with dummy line when target first appears
Its original state is set, when the online outer vehicle's current condition of original state then to think that it drives into line, is counted and incited somebody to action
Its original state, which changes, to be set in line, and outgoing vehicles are similarly.
S107, algorithm interaction：
As described above, the local area image of the vehicle target in each two field picture can be all stored into file system, together
When by the camera number of the testing result, record time, the information write service device data such as vehicle target image results path
Storehouse.Every time when reaching the time interval of vehicle flowrate, by present period data write into Databasce and by counter O reset.Interaction
Program can carry out rendering display result when there is data insertion with the state in real-time monitoring data storehouse according to insertion data.Whenever
When reaching running state of programs daily record interval, program current state is write in journal file, is had when ensuring to break down with this
According to can look into.On the other hand, program provide configuration file for user change parameter, user can change vehicle flowrate time interval,
Running state of programs daily record interval, frame picture processing duty cycle, video camera ID, video camera and bayonet relative position, dummy line are sat
The parameter such as mark, image results store path, Database Properties, video address.Modification frame picture processing duty cycle can control journey
Sequence scale, effectively controls the CPU usage and memory usage of the program, such as frame handles duty cycle under default parameters state
For 5ms when, CPU usage is 45% or so, and memory usage is 142M or so.And when frame processing duty cycle be changed into 200ms its
During his parameter constant, CPU usage is reduced to 24% or so, and memory usage is 137M or so, is ensured with this in different configurations
Computer in can reach optimal operational condition.
Technical scheme advantage is as follows：First, robustness is high：The amendment carried out using tracking result to testing result
Effectively overcoming the influence of noise, the repetitive exercise of grader effectively overcomes the influence of the changes such as illumination and weather, and with
When more preferable testing result can be obtained with exptended sample collection.2nd, accuracy is high：Such as passing through 4 repetitive exercises, sample set number
Measure for 4146 when, the size of grader is 89.1M, only 3415 in 62448 classification results that continuous testing results obtain
Mistake point is as a result, accuracy has reached 94.53%, and continuation iteration can reach the accuracy of higher.3rd, processing speed is fast,
And program scale is controllable：Such as program is handled all video input resolution adjustments for 640*360 again, therefore image
Quality, which changes, has no effect on processing speed, and average every frame processing speed is 20ms.It is 4M in video code flow bandwidth, resolution ratio is
1280*720, frame per second 25, when frame processing duty cycle is 5ms, CPU usage is 45% or so, and memory usage is left for 142M
It is right.And when frame processing duty cycle be changed into 200ms other specifications it is constant when, CPU usage is reduced to 24% or so, memory usage
For 137M or so.When parameter is arranged to bandwidth 512K, resolution ratio 640*360, frame per second 15, during duty cycle 200ms, CPU usage
For 17% or so, memory usage is 115M or so, completely can the smooth operation in low configuration computer.4th, have good
Interactivity：Program performs different function modules, such as collecting sample picture module, training classification according to different input instructions
Device module, on-line operation module etc..Operating parameter can be changed by user, and it is aobvious that variation output result facilitates Front End to render
Show effect, store to the mode of database and the program is completely independent with Front End, front end can even use not
Same programming language.
Note that it above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that
The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious changes,
Readjust and substitute without departing from protection scope of the present invention.Therefore, although being carried out by above example to the present invention
It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also
It can include other more equivalent embodiments, and the scope of the present invention is determined by scope of the appended claims.
Claims (5)
- A kind of 1. moving vehicles detection and tracking method based on real-time video, it is characterised in that include the following steps：S101, obtain original training set：Classify to the image results in file system, be respectively placed in different files, obtain initial samples pictures Collection；The absolute path list of samples pictures in each file is respectively obtained using autoexec, and in each file File assign different mark, finally the sample absolute path list of tape label is incorporated into a text；S102, training preliminary classification device：The original training set obtained using step S101 trains grader, reads in each sample and its mark successively, and it is special to analyze it The feature vector of each sample is obtained, Supervised classification is carried out to feature vector, obtains preliminary classification device；S103, repetitive exercise grader：Classified using the step S102 preliminary classification devices obtained to the moving object in video sequence, and by moving object institute Local area image by classification results mark add timestamp name storage into file system, if user is to classification results It is satisfied then training need not be iterated, if dissatisfied unsatisfied misclassification is found out as a result, and adding it to correct class Sample set file in, obtain new training sample set text using step S101 the methods, then utilize step S102 the methods are iterated training, so repeatedly until user is satisfied with classification results the iteration instruction for then completing grader Practice；Grader write-in file is saved in file system；S104, the generation of vehicle candidate region：Grader file is read in, detects the moving region image occurred in video sequence successively, analyzes the spy of moving region image Its feature vector is obtained, feature vector feeding grader is classified and obtains classification results, if the result is that if vehicle Think that the moving region image includes vehicle, the generation of vehicle candidate region is completed with this；S105, vehicle confirm and tracking：Obtained in step S104 per the vehicle detection in two field picture as a result, in real-time video first on the basis of present frame, To all vehicle detections in each vehicle detection result of present frame traversal previous frame as a result, will meet similarity indices and One matching result for being considered as current goal of similarity measurement result minimum；Likewise, on the basis of previous frame, to previous frame It is each testing result traversal present frame as a result, will meet that minimum one of similarity indices and similarity measurement result regards For the matching result of previous frame target；Finally, if previous frame exists, present frame lost target is considered as target disappearance, if Previous frame does not occur and the target of present frame appearance is then considered as fresh target and new mark is distributed for it, if present frame and upper one Two targets of frame match each other, and are considered as same target and with same tag into line trace.
- 2. the moving vehicles detection and tracking method according to claim 1 based on real-time video, it is characterised in that further include：S106, wagon flow statistics：Measurement type is divided into three kinds：The vehicle fleet occurred in picture, drives into vehicle fleet, outgoing vehicles sum；Vehicle fleet Statistics according in step S105 vehicle tracking result carry out, only just counted when fresh target occurs；Drive into and sail Go out vehicle according to the mode of dummy line to count, when original state is that vehicle's current condition outside line is that it is then thought in line to sail Enter vehicle, counted and change its original state and be set in line, wherein, the statistical method of outgoing vehicles and the system for driving into vehicle Meter method is identical.
- 3. the moving vehicles detection and tracking method according to claim 2 based on real-time video, it is characterised in that further include：S107, algorithm interaction：Since the local area image of the vehicle target in each two field picture can be all stored into file system, while this is detected As a result the information write service device database including camera number, record time, vehicle target image results path；Often It is secondary when reaching vehicle flowrate time interval, by present period data write into Databasce and by counter O reset；Interactive program can With the state in real-time monitoring data storehouse, carry out rendering display result according to insertion data when there is data insertion；Whenever reaching journey During sort run status log interval, program current state is write in journal file, to break down when has good grounds, another Aspect, program provide configuration file and change parameter for user, and user can be changed including vehicle flowrate time interval, program operation Supplemental characteristic including status log interval, frame picture processing duty cycle.
- 4. the moving vehicles detection and tracking method based on real-time video according to one of claims 1 to 3, it is characterised in that The step S105 is further included：Improve vehicle detection effect using vehicle tracking result, the target of tenacious tracking preset times is Vehicle is confirmed as, in this, as vehicle detection as a result, preserving the image results of regional area where vehicle at the same time to file system And by testing result insertion of data into data storehouse.
- 5. the moving vehicles detection and tracking method according to claim 4 based on real-time video, it is characterised in that the step S105 further comprises：For tenacious tracking as a result, giving predictive compensation if suddenly disappearing presets frame number, if pre- If target occurs again within frame number, continue to track, otherwise it is assumed that target is lost.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510210294.5A CN104809437B (en) | 2015-04-28 | 2015-04-28 | A kind of moving vehicles detection and tracking method based on real-time video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510210294.5A CN104809437B (en) | 2015-04-28 | 2015-04-28 | A kind of moving vehicles detection and tracking method based on real-time video |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104809437A CN104809437A (en) | 2015-07-29 |
| CN104809437B true CN104809437B (en) | 2018-04-13 |
Family
ID=53694248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510210294.5A Active CN104809437B (en) | 2015-04-28 | 2015-04-28 | A kind of moving vehicles detection and tracking method based on real-time video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104809437B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105844222B (en) * | 2016-03-18 | 2019-07-30 | 上海欧菲智能车联科技有限公司 | The front vehicles collision warning systems and method of view-based access control model |
| CN106462762B (en) * | 2016-09-16 | 2019-07-19 | 香港应用科技研究院有限公司 | Based on vehicle detection, tracking and the positioning for enhancing anti-perspective transform |
| CN107038423B (en) * | 2017-04-20 | 2020-03-27 | 常州智行科技有限公司 | Vehicle real-time detection and tracking method |
| CN107330365A (en) * | 2017-05-27 | 2017-11-07 | 深圳市美好幸福生活安全系统有限公司 | Traffic sign recognition method based on maximum stable extremal region and SVM |
| CN107506753B (en) * | 2017-09-19 | 2021-01-05 | 北京工业大学 | Multi-vehicle tracking method for dynamic video monitoring |
| CN107909012B (en) * | 2017-10-30 | 2022-03-18 | 北京中科慧眼科技有限公司 | Real-time vehicle tracking detection method and device based on disparity map |
| CN107992900B (en) * | 2017-12-18 | 2020-09-11 | 深圳市盛波光电科技有限公司 | Sample acquisition method, training method, device, medium and equipment for defect detection |
| CN108830332A (en) * | 2018-06-22 | 2018-11-16 | 安徽江淮汽车集团股份有限公司 | A kind of vision vehicle checking method and system |
| CN108985233B (en) * | 2018-07-19 | 2021-08-03 | 常州智行科技有限公司 | High-precision vehicle tracking method based on digital image correlation |
| CN109191839A (en) * | 2018-09-13 | 2019-01-11 | 江南大学 | Vehicle classification and method of counting based on deep learning |
| CN109726670B (en) * | 2018-12-26 | 2021-03-26 | 浙江捷尚视觉科技股份有限公司 | Method for extracting target detection sample set from video |
| CN110765940B (en) * | 2019-10-22 | 2022-12-30 | 杭州姿感科技有限公司 | Target object statistical method and device |
| CN111046820A (en) * | 2019-12-17 | 2020-04-21 | 上海船舶研究设计院（中国船舶工业集团公司第六0四研究院） | Statistical method and device for vehicles in automobile roll-on-roll-off ship and intelligent terminal |
| CN112130792B (en) * | 2020-09-11 | 2023-12-15 | 安徽中科新辰技术有限公司 | Method for realizing multi-terminal synchronization of visual view |
| CN112950961B (en) * | 2021-01-27 | 2022-07-08 | 苏州智芯控联信息科技有限公司 | Traffic flow statistical method, device, equipment and storage medium |
| CN112967276B (en) * | 2021-03-31 | 2023-09-05 | 杭州海康威视数字技术股份有限公司 | Object detection method, object detection device, endoscope system, electronic device, and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102142090A (en) * | 2011-03-15 | 2011-08-03 | 中国科学技术大学 | Vehicle detection method and system |
| CN103020590A (en) * | 2012-11-20 | 2013-04-03 | 北京航空航天大学深圳研究院 | Vehicle identification system and method based on three-dimensional model and image matching |
| CN103246896A (en) * | 2013-05-24 | 2013-08-14 | 成都方米科技有限公司 | Robust real-time vehicle detection and tracking method |
| CN103258427A (en) * | 2013-04-24 | 2013-08-21 | 北京工业大学 | Urban expressway traffic real-time monitoring system and method based on information physical network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0218982D0 (en) * | 2002-08-15 | 2002-09-25 | Roke Manor Research | Video motion anomaly detector |
-
2015
- 2015-04-28 CN CN201510210294.5A patent/CN104809437B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102142090A (en) * | 2011-03-15 | 2011-08-03 | 中国科学技术大学 | Vehicle detection method and system |
| CN103020590A (en) * | 2012-11-20 | 2013-04-03 | 北京航空航天大学深圳研究院 | Vehicle identification system and method based on three-dimensional model and image matching |
| CN103258427A (en) * | 2013-04-24 | 2013-08-21 | 北京工业大学 | Urban expressway traffic real-time monitoring system and method based on information physical network |
| CN103246896A (en) * | 2013-05-24 | 2013-08-14 | 成都方米科技有限公司 | Robust real-time vehicle detection and tracking method |
Non-Patent Citations (1)
| Title |
|---|
| 车辆检测跟踪系统研究;吕万修;《中国优秀硕士学位论文全文数据库 信息科技辑》;20140115;第2014年卷(第01期);第I138-1742页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104809437A (en) | 2015-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104809437B (en) | A kind of moving vehicles detection and tracking method based on real-time video | |
| CN109711320B (en) | Method and system for detecting violation behaviors of staff on duty | |
| CN103324937B (en) | The method and apparatus of label target | |
| González et al. | Automatic traffic signs and panels inspection system using computer vision | |
| US8509526B2 (en) | Detection of objects in digital images | |
| Mallikarjuna et al. | Traffic data collection under mixed traffic conditions using video image processing | |
| CN108830332A (en) | A kind of vision vehicle checking method and system | |
| Fachrie | A simple vehicle counting system using deep learning with YOLOv3 model | |
| CN104615986A (en) | Method for utilizing multiple detectors to conduct pedestrian detection on video images of scene change | |
| CN102254428A (en) | Traffic jam detection method based on video processing | |
| CN103632247A (en) | Meter information identification system and method based on intelligent storage system | |
| CN105184229A (en) | Online learning based real-time pedestrian detection method in dynamic scene | |
| CN104123714B (en) | A kind of generation method of optimal objective detection yardstick in people flow rate statistical | |
| Lisanti et al. | A multi-camera image processing and visualization system for train safety assessment | |
| CN109848070B (en) | Product sorting system and method | |
| CN112102443A (en) | Marking system and marking method suitable for substation equipment inspection image | |
| CN108288020A (en) | Video shelter detecting system based on contextual information and method | |
| CN111951328A (en) | Object position detection method, device, equipment and storage medium | |
| Zhao et al. | A real time intelligent detection and counting method of cells based on YOLOv5 | |
| CN115359306A (en) | Intelligent identification method and system for high-definition images of railway freight inspection | |
| CN111626992B (en) | Intelligent crossing box inspection method | |
| Jensen et al. | A framework for automated traffic safety analysis from video using modern computer vision | |
| CN111401104B (en) | Classification model training method, classification method, device, equipment and storage medium | |
| JP2023529644A (en) | Method and system for training neural networks to classify objects in bulk flow | |
| Pu et al. | A robust and real-time approach for license plate detection |
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 |