CN109754409A - A kind of monitor video pedestrian target matched jamming System and method for - Google Patents
A kind of monitor video pedestrian target matched jamming System and method for Download PDFInfo
- Publication number
- CN109754409A CN109754409A CN201711076378.XA CN201711076378A CN109754409A CN 109754409 A CN109754409 A CN 109754409A CN 201711076378 A CN201711076378 A CN 201711076378A CN 109754409 A CN109754409 A CN 109754409A
- Authority
- CN
- China
- Prior art keywords
- tracking
- target
- pedestrian
- monitor video
- matching
- 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.)
- Pending
Links
Landscapes
- Closed-Circuit Television Systems (AREA)
- Image Analysis (AREA)
Abstract
A kind of monitor video pedestrian target matched jamming system, including monitor video pedestrian detection and tracking system, detecting and tracking frame feature judge system, pedestrian tracking matching treatment system；The monitor video pedestrian detection and tracking system realize the pedestrian received in monitor video code stream and detecting and tracking video, the detecting and tracking frame feature judges that system judges that the pedestrian tracking matching treatment system determines whether fresh target according to the matching confidence level result of each detecting and tracking frame by spatial informations such as geometric positions to detection block each in present frame and existing tracking box.
Description
Technical field
The present invention relates to digital image processing techniques fields, more particularly to multiple pedestrian targets under a kind of monitor video
Matched jamming is realized.
Background technique
Pedestrian detection tracks blending image processing, pattern-recognition, Theory of Automatic Control, in the monitoring of suspected crime molecule, walks
Losing object searching, staff attendance of showing up notice etc. has important application prospect.Existing pedestrian's matching tracking method is to pass through
Euclidean distance calculates current detection target and has the matching confidence level of tracking target, is assigned by Hungary Algorithm.This
Kind processing method is also easy to produce the appointment combination of mistake when tracking target is blocked and fresh target occurs, differs greatly in calculating
The matching confidence level of a certain tracking target and another detection target also increases processor workload.
Summary of the invention
Against the above deficiency, the present invention proposes a kind of monitor video pedestrian target matched jamming system and method, to improve
For existing pedestrian's matching tracking method during monitor video pedestrian's matched jamming, it is inefficient that matching process calculates complicated and processing
Problem.
For achieving the above object, the present invention provides a kind of monitor video pedestrian target matched jamming system, feature
It is: judges system, pedestrian tracking matching treatment system including monitor video pedestrian detection and tracking system, detecting and tracking frame feature
System；The monitor video pedestrian detection and tracking system realize the pedestrian received in monitor video code stream and detecting and tracking video,
The detecting and tracking frame feature judges that system is empty by geometric position to detection block each in present frame and existing tracking box etc.
Between information judged that the pedestrian tracking matching treatment system is according to the matching confidence level result determination of each detecting and tracking frame
It is no fresh target occur.
In above-mentioned technical proposal, the monitor video pedestrian detection and tracking system include that monitor video code stream is received, gone
People's detection, pedestrian tracking；The monitor video code stream includes the offline video code of live camera real-time video code stream and storage
Stream；The pedestrian detection is realized to the global detection of current video frame, the bounding box coordinate of pedestrian target occurs；
The pedestrian tracking realizes existing tracking box in the position prediction of present frame.
In above-mentioned technical proposal, the detecting and tracking frame feature judges system according to the feature of pedestrian target in monitor video
The multitarget relation distributed with error-prone, using the lesser matching of the width of detection block and tracking box destination probability more same than exclusion
Combination；The positional relationship that the two is determined using detection block and tracking box overlapping region size, further excludes same destination probability
Lesser matching combination.
In above-mentioned technical proposal, the pedestrian tracking matching treatment system includes matching confidence calculations, successful match pair
As modules such as processing, fresh target processing.Matching confidence calculations module determines optimal by comparing various combined Euclidean distances
The combination of matched form, successful match will be updated to corresponding tracker and subsequent identifying processing, the pedestrian of non-successful match
Target will open tracker as fresh target and track.It is described matching confidence level calculation method be calculate current detection frame and
The Euclidean distance of tracking box centre coordinate and width, or use mahalanobis distance, Pasteur's distance calculating method.
Purpose to realize the present invention, the present invention also provides a kind of monitor video pedestrian target matching tracking method, features
Be the following steps are included:
(1) the monitor video code stream that system docking receives is analyzed and processed, and obtains present frame by pedestrian detection algorithm
In target；At this time to these target start pedestrian trackings if not tracked target；As have tracking target if need by
The target of detection and tracked target carry out matching primitives；
(2) it treats matched detection block and tracking carries out geometrical characteristic and positional relationship compares, it is less for overlapping region
The target in even non-overlapping region can determine whether that, for different target, the combination for meeting centre distance and overlapping area condition is also answered
The case where differing greatly in geometric size is avoided, the matching confidence level of various combination then can be calculated；
(3) if target assigns current existing tracking box by matching primitives, the tracking box state is updated；If mesh
Mark is judged as fresh target by matching process, then opens a new pedestrian tracking device to it, be achieved in more under monitor video
The matched jamming of a pedestrian target.
Further, in the above method, the specific implementation of step (2) is as follows:
(21) matrix is set, line number is the target numbers that present frame detects, columns is currently to have tracking target
Number；
(22) normal distribution principle should be met according to the same target width variation of consecutive frame, calculates each detection block phase first
Compared with the width ratio of tracking box, the combination that reduced value differs greatly is set as to match, and terminates the subsequent calculating stream of the combination
Journey；
(23) greater overlap region principle should be had according to the same target of consecutive frame, then calculate each detection block and each with
The intersection ratio of track frame, the lesser combination of reduced value are set as to match, and terminate the subsequent calculation process of the combination；
(24) combination to (22) (23) Rule of judgment is met simultaneously, calculates the Euclidean distance of centre coordinate and width, most
It is selected afterwards by matching algorithm apart from optimal combining form, and executes tracker and update and fresh target tracking unlatching.
Present invention improves over execution processes and processing speed that pedestrian's object matching under monitor video tracks, there is preferable reality
With property and stability.The present invention uses under monitor video pedestrian's motion feature and feature calculation amount as multilayer judgment mechanism, and
Efficient object matching is realized in conjunction with matching confidence level, can effectively improve the tracking accuracy of pedestrian target and place under monitor video
Speed is managed, practical execution performance is stablized, attains the results expected.
Detailed description of the invention
Fig. 1 is realization principle figure of the invention.
Specific embodiment
As shown in Figure 1, the present invention includes monitor video pedestrian detection and tracking system, the judgement of detecting and tracking frame feature are
System, pedestrian tracking matching treatment system；The monitor video pedestrian detection and tracking system, which are realized, receives monitor video code stream simultaneously
Pedestrian in detecting and tracking video, the detecting and tracking frame feature judge that system passes through to detection block each in present frame and existing
The spatial informations such as the geometric position of tracking box are judged that the pedestrian tracking matching treatment system is according to each detecting and tracking frame
Matching confidence level result determines whether fresh target.
The monitor video pedestrian detection and tracking system include the reception of monitor video code stream, pedestrian detection, pedestrian tracking；
The monitor video code stream includes the offline video code stream of live camera real-time video code stream and storage；The pedestrian detection is real
Now to the global detection of current video frame, there is the bounding box coordinate of pedestrian target；The pedestrian tracking is realized
Position prediction of the existing tracking box in present frame.
The detecting and tracking frame feature judges what system was distributed according to the feature and error-prone of pedestrian target in monitor video
Multitarget relation is combined with the width of tracking box than excluding lesser match of same destination probability using detection block；Using detection
Frame and tracking box overlapping region size determine the positional relationship of the two, further exclude the lesser match group of same destination probability
It closes.
The pedestrian tracking matching treatment system includes matching confidence calculations, successful match object handles, at fresh target
The modules such as reason.Matching confidence calculations module determines Optimum Matching form by comparing various combined Euclidean distances, matching at
The combination of function will be updated to corresponding tracker and subsequent identifying processing, the pedestrian target of non-successful match will be used as fresh target
Tracker is opened to be tracked.The calculation method of the matching confidence level be calculate current detection frame and tracking box centre coordinate and
The Euclidean distance of width, can also be using calculation methods such as mahalanobis distance, Pasteur's distances.
The groundwork process of above system is as follows:
(1) the monitor video code stream that system docking receives is analyzed and processed, and obtains present frame by pedestrian detection algorithm
In target.At this time to these target start pedestrian trackings if not tracked target；As have tracking target if need by
The target of detection and tracked target carry out matching primitives；
(2) it treats matched detection block and tracking carries out geometrical characteristic and positional relationship compares, it is less for overlapping region
The target in even non-overlapping region can determine whether that, for different target, the combination for meeting centre distance and overlapping area condition is also answered
The case where differing greatly in geometric size is avoided, the matching confidence level of various combination then can be calculated.Specific steps are as follows:
(21) matrix is set, line number is the target numbers that present frame detects, columns is currently to have tracking target
Number；
(22) normal distribution principle should be met according to the same target width variation of consecutive frame, calculates each detection block phase first
Compared with the width ratio of tracking box, the combination that reduced value differs greatly is set as to match, and terminates the subsequent calculating stream of the combination
Journey；
(23) greater overlap region principle should be had according to the same target of consecutive frame, then calculate each detection block and each with
The intersection ratio of track frame, the lesser combination of reduced value are set as to match, and terminate the subsequent calculation process of the combination；
(24) Euclidean distance of centre coordinate and width is calculated the combination for meeting (22) (23) Rule of judgment simultaneously, finally
It is selected by matching algorithm apart from optimal combining form, and executes tracker and update and fresh target tracking unlatching；
(3) if target assigns current existing tracking box by matching primitives, the tracking box state is updated；If mesh
Mark is judged as fresh target by matching process, then opens a new pedestrian tracking device to it, be achieved in more under monitor video
The matched jamming of a pedestrian target.
Claims (7)
1. a kind of monitor video pedestrian target matched jamming system, it is characterised in that: including monitor video pedestrian detection and tracking
System, detecting and tracking frame feature judge system, pedestrian tracking matching treatment system；The monitor video pedestrian detection and tracking are
System realizes the pedestrian received in monitor video code stream and detecting and tracking video, and the detecting and tracking frame feature judge that system passes through pair
The spatial informations such as geometric position of each detection block and existing tracking box are judged in present frame, at the pedestrian tracking matching
Reason system determines whether fresh target according to the matching confidence level result of each detecting and tracking frame.
2. monitor video pedestrian target matched jamming system according to claim 1, it is characterised in that: the monitor video
Pedestrian detection and tracking system include the reception of monitor video code stream, pedestrian detection, pedestrian tracking；The monitor video code stream includes
The offline video code stream of live camera real-time video code stream and storage；The pedestrian detection realizes the overall situation to current video frame
, there is the bounding box coordinate of pedestrian target in detection；The pedestrian tracking realizes existing tracking box in present frame
Position prediction.
3. monitor video pedestrian target matched jamming system according to claim 1, it is characterised in that: the detecting and tracking
Frame feature judges the multitarget relation that system is distributed according to the feature and error-prone of pedestrian target in monitor video, using detection block
It is combined with the width of tracking box than excluding lesser match of same destination probability；Using detection block and tracking box overlapping region size
The positional relationship for determining the two further excludes the lesser matching combination of same destination probability.
4. monitor video pedestrian target matched jamming system according to claim 1, it is characterised in that: the pedestrian tracking
Matching treatment system includes the modules such as matching confidence calculations, successful match object handles, fresh target processing.Match confidence level meter
It calculates module and determines Optimum Matching form by comparing various combined Euclidean distances, the combination of successful match will be to corresponding tracker
It is updated and subsequent identifying processing, the pedestrian target of non-successful match will open tracker as fresh target and track.
5. monitor video pedestrian target matched jamming system according to claim 4, it is characterised in that: the matching confidence
The calculation method of degree be calculate current detection frame and tracking box centre coordinate and width Euclidean distance, or using mahalanobis distance,
Pasteur's distance calculating method.
6. a kind of monitor video pedestrian target matching tracking method, it is characterised in that the following steps are included:
(1) the monitor video code stream that system docking receives is analyzed and processed, and is obtained in present frame by pedestrian detection algorithm
Target；At this time to these target start pedestrian trackings if not tracked target；It needs will test if having tracking target
Target and tracked target carry out matching primitives；
(2) treat matched detection block and tracking carry out geometrical characteristic and positional relationship and compares, it is less for overlapping region even
The target in non-overlapping region can determine whether that, for different target, the combination for meeting centre distance and overlapping area condition should also be avoided
Then the case where differing greatly in geometric size can calculate the matching confidence level of various combination；
(3) if target assigns current existing tracking box by matching primitives, the tracking box state is updated；If target is logical
Overmatching method is judged as fresh target, then opens a new pedestrian tracking device to it, be achieved in multiple rows under monitor video
The matched jamming of people's target.
7. monitor video pedestrian target matching tracking method according to claim 6, which is characterized in that the tool of step (2)
Body implementation is as follows:
(21) matrix is set, line number is the target numbers that present frame detects, columns is currently to have tracking target numbers；
(22) normal distribution principle should be met according to consecutive frame same target width variation, calculate first each detection block compared to
The width ratio of tracking box, the combination that reduced value differs greatly are set as to match, and terminate the subsequent calculation process of the combination；
(23) greater overlap region principle should be had according to the same target of consecutive frame, then calculates each detection block and each tracking box
Intersection ratio, reduced value it is lesser combination be set as to match, terminate the subsequent calculation process of the combination；
(24) combination to (22) (23) Rule of judgment is met simultaneously, calculates the Euclidean distance of centre coordinate and width, finally leads to
Overmatching algorithms selection executes tracker and updates and fresh target tracking unlatching apart from optimal combining form.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711076378.XA CN109754409A (en) | 2017-11-06 | 2017-11-06 | A kind of monitor video pedestrian target matched jamming System and method for |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711076378.XA CN109754409A (en) | 2017-11-06 | 2017-11-06 | A kind of monitor video pedestrian target matched jamming System and method for |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109754409A true CN109754409A (en) | 2019-05-14 |
Family
ID=66399716
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711076378.XA Pending CN109754409A (en) | 2017-11-06 | 2017-11-06 | A kind of monitor video pedestrian target matched jamming System and method for |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109754409A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517293A (en) * | 2019-08-29 | 2019-11-29 | 京东方科技集团股份有限公司 | Method for tracking target, device, system and computer readable storage medium |
| CN110853076A (en) * | 2019-11-08 | 2020-02-28 | 重庆市亿飞智联科技有限公司 | Target tracking method, device, equipment and storage medium |
| WO2021147826A1 (en) * | 2020-01-21 | 2021-07-29 | 北京工业大学 | Edge end-oriented pedestrian tracking and accurate people counting method |
| WO2021189825A1 (en) * | 2020-03-25 | 2021-09-30 | 苏州科达科技股份有限公司 | Multi-target tracking method and apparatus, and storage medium |
| WO2021223367A1 (en) * | 2020-05-06 | 2021-11-11 | 佳都新太科技股份有限公司 | Single lens-based multi-pedestrian online tracking method and apparatus, device, and storage medium |
-
2017
- 2017-11-06 CN CN201711076378.XA patent/CN109754409A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517293A (en) * | 2019-08-29 | 2019-11-29 | 京东方科技集团股份有限公司 | Method for tracking target, device, system and computer readable storage medium |
| US11393103B2 (en) | 2019-08-29 | 2022-07-19 | Boe Technology Group Co., Ltd. | Target tracking method, device, system and non-transitory computer readable medium |
| CN110853076A (en) * | 2019-11-08 | 2020-02-28 | 重庆市亿飞智联科技有限公司 | Target tracking method, device, equipment and storage medium |
| WO2021147826A1 (en) * | 2020-01-21 | 2021-07-29 | 北京工业大学 | Edge end-oriented pedestrian tracking and accurate people counting method |
| WO2021189825A1 (en) * | 2020-03-25 | 2021-09-30 | 苏州科达科技股份有限公司 | Multi-target tracking method and apparatus, and storage medium |
| WO2021223367A1 (en) * | 2020-05-06 | 2021-11-11 | 佳都新太科技股份有限公司 | Single lens-based multi-pedestrian online tracking method and apparatus, device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109754409A (en) | A kind of monitor video pedestrian target matched jamming System and method for | |
| CN105405154B (en) | Target object tracking based on color-structure feature | |
| Xing et al. | Multi-object tracking through occlusions by local tracklets filtering and global tracklets association with detection responses | |
| US7706571B2 (en) | Flexible layer tracking with weak online appearance model | |
| Walia et al. | Intelligent video target tracking using an evolutionary particle filter based upon improved cuckoo search | |
| Aeschliman et al. | A probabilistic framework for joint segmentation and tracking | |
| Chou et al. | Multi-robot cooperation based human tracking system using laser range finder | |
| CN103400120B (en) | Video analysis-based bank self-service area push behavior detection method | |
| CN105354540A (en) | Video analysis based method for implementing person fall-down behavior detection | |
| CN113192105A (en) | Method and device for tracking multiple persons and estimating postures indoors | |
| CN111986237A (en) | Real-time multi-target tracking algorithm irrelevant to number of people | |
| Chia et al. | Kernel-based object tracking via particle filter and mean shift algorithm | |
| CN114926859A (en) | Pedestrian multi-target tracking method in dense scene combined with head tracking | |
| CN109977796A (en) | Trail current detection method and device | |
| Mittal et al. | Pedestrian detection and tracking using deformable part models and Kalman filtering | |
| Jiang et al. | Tracking pedestrians using smoothed colour histograms in an interacting multiple model framework | |
| Fan et al. | Pedestrian detection and tracking using deformable part models and kalman filtering | |
| Wang | Particle filters for visual tracking | |
| CN116342645A (en) | Multi-target tracking method for natatorium scene | |
| Lim et al. | Non-overlapping distributed tracking system utilizing particle filter | |
| Cheng et al. | An end-to-end framework of road user detection, tracking, and prediction from monocular images | |
| Manzoor et al. | Qualitative Analysis of Single Object and Multi Object Tracking Models | |
| Zhang et al. | Robust pedestrian tracking in crowd scenarios using an adaptive GMM-based framework | |
| Wang et al. | Multiple feature fusion for tracking of moving objects in video surveillance | |
| Konstantinou | Vision-based construction worker task productivity monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190514 |