CN106778503B - A kind of method and system of detection and tracking based on circulation frame buffer zone - Google Patents
A kind of method and system of detection and tracking based on circulation frame buffer zone Download PDFInfo
- Publication number
- CN106778503B CN106778503B CN201611042801.XA CN201611042801A CN106778503B CN 106778503 B CN106778503 B CN 106778503B CN 201611042801 A CN201611042801 A CN 201611042801A CN 106778503 B CN106778503 B CN 106778503B
- Authority
- CN
- China
- Prior art keywords
- frame
- target
- detection
- buffering area
- tracking
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
Abstract
The present invention relates to a kind of method of detection and tracking based on buffering, including several detectors, comprise the following steps：A, cyclic buffer is established on a timeline；B, extract a frame of video by strategy from the continuous frame sequence inputted in real time and record frame number；C, the frame chosen is sent in several object detectors and carries out parallel target detection；D, buffering area is addressed to read according to the source frame sequence number of detection and frame corresponding to by its source frame number is arranged to start frame, if start frame is StartFrame (a), StartFrame (b), StartFrame (c), StartFrame (d) ... .etc；E, the target and corresponding start frame obtained in buffering area to each detection carries out initialization process and processing is tracked to target.The time accounting that tracking process can be used for actually calculating is brought up to, from the performance for improving tracking, supports bigger tracking object set.
Description
Technical field
The invention belongs to the detecting and tracking technical field of frame data, more particularly to a kind of detection based on circulation frame buffer zone
With the method and system of tracking.
Background technology
Video intelligent analysis and processing is currently based on more and more widely to be applied among every field.Should at these
It is in, the problem of an emphasis：For the video and/or image of continuous multiple frames, how therefrom efficiently to detect interested
Target, and the target that these are detected constantly is tracked；The detect and track problem of target i.e. in video flowing.
Under normal circumstances, the framework of the universal solution of such issues that is as shown in Figure 1.
Detection more than is handled in framework with middle tracking, and the algorithm that the detection and tracking of usual target are used all disappears very much
Computing resource is consumed, the real-time processing of data can not be generally realized in the single thread of single processor.It is especially complex at some
Algorithm in（Such as the detection of human face target）, due to calculating consumption, the input generally also to successive frame carries out taking out frame inspection
The strategy of survey, it is subsequent calculating is tracked to target frame by frame during, generally all targets that target tightening are entered simultaneously
Row processing.Above is carry out parallel processing solution when being commonly used in target detection tracking.
But above-mentioned conventional or existing solution is without solving a crucial bottleneck problem, so as to cause
Degree of parallelism is greatly limited, and reason is as follows：
Although although 1, tracking calculating frame by frame is the progress that can compare high concurrent, understood in narration described above,
During target due to tracking calculating frame by frame comes from target detection, when in detection frame, tracking, which will calculate, to block, until
After target detection is partially completed detection and has safeguarded tracking object set, it can just continue.2nd, because target detection operation is consumption
When, the obstruction of long period is tracked into operation.3rd, simultaneously because detection operation and the maintenance of tracking object set are serial, mesh
Mark detection operation originally also can not be parallel.Whole processing framework can not be run parallel at full speed, and limited speed is in the serial of target detection
Process.4th, simultaneously because target detection serial operation could meet real-time processing requirement, it is necessary to introduce and take out frame, this considerably increases
The probability of target missing inspection.
Based on above mentioned problem, if it is considered that the completion status of target detection, continues being tracked for non-obstruction.Then when a certain
After the completion of the target detection of frame, due to detecting time delay, detection frame is ahead of by the frame time axle of target following, made
The targeted failure obtained into detection, can not be to this detection knot target line setting and follow-up tracking.As shown in Figure 2.
As shown in Fig. 2 when n-th frame, target detection process detects to the target in frame of video, but due to inspection
The reason for surveying calculation delay, finish until m frames just detect.From the figure, it can be seen that the target in the upper left corner of n-th frame is through excessive
After the motion of frame, along certain line-of-road movement to the upper right corner of picture frame.So for tracking process, detect
Target failed at the very start.
In summary, all solutions, parallelization is realized without in the processing framework as shown in Figure 1 of solution at all
The key issue of processing.Even if there is the parallel processing of part, also offset in the serial process of remainder.And simply to detection
Parallelization processing is carried out, then brings the problems such as detecting targeted failure.
In order to solve the above problems, the object detecting and tracking of complete parallel is realized, the present invention is directed to existing general side
The bottleneck of case, redesigned calculating processing scheme, realize complete parallel target detection and full decoupled detection with
The PipeLine processing of tracking.And solves following problem on this basis：
1：Carried out target caused by taking out frame detection as detecting not parallel enough and lost omission problem.
2：The calculating time can be used because the obstruction of detection have compressed tracking, causes speed/scale of the processing of target following
Limited problem.
In the technical scheme such as above said, some crucial bottleneck problems be present, cause the parallel of overall operational and be limited,
And the related problem of system will be triggered such as because degree of parallelism is limited：
1：Simultaneously because target detection serial operation could meet real-time processing requirement, it is necessary to introduce and take out frame, this increases
The probability of target missing inspection is added.
2：The calculating time can be used because the obstruction of detection have compressed tracking, causes speed/scale of the processing of target following
Limited problem.
3：Due to detecting targeted failure problem caused by detection time delay.
In order to solve the above problems, the object detecting and tracking of complete parallel is realized, the present invention is directed to existing general side
The bottleneck of case, redesigned calculating processing scheme, realize complete parallel target detection and full decoupled detection with
The PipeLine processing of tracking.
The content of the invention
It is an object of the invention to provide a kind of method of detection and tracking based on circulation frame buffer zone, it is intended to solution
Problem.
The present invention is achieved in that a kind of method of detection and tracking based on circulation frame buffer zone, is wrapped in this method
Several detectors are included, the described method comprises the following steps：
A, cyclic buffer is established on a timeline；
B, extract a frame of video by strategy from the continuous frame sequence inputted in real time and record frame number；
C, the frame chosen is sent in several object detectors and carries out parallel target detection；
D, buffering area is addressed to read according to the source frame sequence number of detection and frame corresponding to by its source frame number has been arranged to
Beginning frame, if start frame is StartFrame (a), StartFrame (b), StartFrame (c), StartFrame (d)
....etc；
E, the target and corresponding start frame obtained in buffering area to each detection carries out initialization process and to mesh
Mark is tracked processing.
The present invention further technical scheme be：It is further comprising the steps of in the step E：
E1, detect that new object is set to Object (a) and looked in buffering area according to StartFrame parameters in buffering area
Go to start；
E2, tracing positional started according to Object (a) parameter initialization target；
E3, tracking and amendment using position progress of the target following mode frame by frame to target frame by frame；
E4, frame is being carried out continuing processing before the frame of buffering area runs out of to incite somebody to action untill ruing out of the frame in buffering area
It switchs to calculate in real time.
The present invention further technical scheme be：Cyclic buffer in the step A, the over time propulsion of axle new one
Frame will cover a frame data oldest in buffering area；All tracking targets share same buffering area；Buffering area is meeting simultaneously
The demand of data access supports sequential access and random access both of which during row.
The present invention further technical scheme be：Frame in the step B in continuous frame sequence does not weigh in time shaft sequence
It is folded；Different Object, safeguards the time shaft being each independently of the others, and each Object is tracked fortune by respective time shaft
Calculate.
The present invention further technical scheme be：In the step C between some object detectors independently of each other carry out and
Concurrently detect；The time that multiple detectors carry out parallel target detection completion is random on a timeline；Multiple detectors exist
After the completion of parallel detection, the target for detecting to obtain transfers to merger module to synchronize, collect, and the operation such as merger, is tracked
Object gathers.
Another object of the present invention is to provide a kind of system of detection and tracking based on circulation frame buffer zone, the system
Include several detectors, the system includes：
Buffering area establishes module, for establishing cyclic buffer on a timeline；
Frame number setup module, for extracting a frame of video by strategy from the continuous frame sequence inputted in real time and recording frame
Sequence number；
Parallel detection module, parallel target inspection is carried out for the frame chosen to be sent in several object detectors
Survey；
Frame setup module, buffering area is addressed to read and by its source frame correspondence for the source frame sequence number according to detection
Frame be arranged to start frame, if start frame is StartFrame (a), StartFrame (b), StartFrame (c),
StartFrame(d)....etc；
Initialize and tracking module, target and corresponding start frame for being obtained in buffering area to each detection are entered
Row initialization process is simultaneously tracked processing to target.
The present invention further technical scheme be：Simultaneously tracking module also includes for the initialization：
Search unit, existed for detecting that new object is set to Object (a) in buffering area according to StartFrame parameters
Start frame is found in buffering area；
Position tracking unit, start tracing positional for the parameter initialization target according to Object (a)；
Tracking and amending unit, for carrying out tracking frame by frame to the position of target using target following mode frame by frame and repairing
Just；
Converting unit is handled, for continue processing until ruing out of buffering to frame before the frame of buffering area runs out of
Switched to calculate in real time untill frame in area.
The present invention further technical scheme be：The cyclic buffer that the buffering area is established in module, over time axle
The new frame of propulsion will cover a frame data oldest in buffering area；All tracking targets share same buffering area；Buffering
Sequential access and random access both of which are supported in the needs of data access in meeting parallel procedure in area.
The present invention further technical scheme be：Frame in the frame number setup module in continuous frame sequence is in time shaft
Sequence is not overlapping；Different Object, safeguards the time shaft being each independently of the others, and each Object is entered by respective time shaft
Line trace computing.
The present invention further technical scheme be：In the parallel detection module between some object detectors independently of each other
Carry out and concurrently detect；The time that multiple detectors carry out parallel target detection completion is random on a timeline；Multiple inspections
Device is surveyed after the completion of parallel detection, the target for detecting to obtain transfers to merger module to synchronize, collect, and the operation such as merger, obtains
The Object set of tracking.
The beneficial effects of the invention are as follows：Eliminate target detection be connected with object tracking process between obstruction, realize inspection
Survey with tracking complete parallel, eliminate the obstruction between each target detection, realize different frame detection between it is complete simultaneously
OK, on the basis of realizing complete parallel detection and calculating, the performance of detection is improved, reduces the frequency that frame is taken out in detection, so as to
The probability of target missing inspection is reduced, is being eliminated between target detection and target following on the basis of obstruction, bringing up to tracking process can
For the time accounting actually calculated, from the performance for improving tracking, bigger tracking object set is supported.
Brief description of the drawings
Fig. 1 is the general procedure framework of video flowing target detection tracking.
Fig. 2 is that detection time delay causes to detect targeted failure schematic diagram when completing.
Fig. 3 is the detection provided in an embodiment of the present invention based on circulation frame buffer zone high concurrent and tracking schematic diagram.
Fig. 4 is the detection provided in an embodiment of the present invention based on circulation frame buffer zone high concurrent and tracking flow chart.
Fig. 5 is detection and the structural representation of the system of tracking provided in an embodiment of the present invention based on circulation frame buffer zone
Figure.
Embodiment
Detection and the flow chart of the method for tracking provided by the invention based on circulation frame buffer zone, details are as follows for it：
Step S1, establishes cyclic buffer on a timeline；Cyclic buffer, it is assumed that present frame m, the capacity of buffering area
For m-n, then the content of buffering area is video requency frame data of the n-th frame to m frames；The propulsion of axle over time, a new frame will cover
An oldest frame data in buffering area；Buffering area is supported to share in multiple concurrent target detections and target following；Buffering area branch
Random read-write is held, supports to read since any one frame.
Step S2, by the system video frame of policy selection one and frame number is set from the continuous frame sequence inputted in real time；From
A system video frame is selected by tactful X in the continuous frame sequence inputted in real time, wherein tactful X can be formulated flexibly,
Set frame serial number Frame (x), Frame (y), Frame (z) ... .etc. as shown in figure 3, these frames in time shaft sequence
It is not overlapping.
Step S3, the frame chosen is sent in several object detectors and carries out parallel target detection；It will select
The frame come, which is sent in several object detectors, carries out parallel target detection.Assuming that respectively obtain a series of target Object
(a), Object (b), Object (c), Object (d) ... .etc. due to being parallel target detection, complete by these target detections
Time is random in time shaft.
Step S4, is read out according to the source frame sequence number of detection to buffering area and frame corresponding to by its source frame number is start frame
It is set to StartFrame (a), StartFrame (b), StartFrame (c), StartFrame (d) ... .etc；For each inspection
The target that device obtains is surveyed, according to the source frame sequence number recorded（As Object (a) by Frame (x) frames is detected to obtain, then
StartFrame (a)=Frame (x), other the like）, buffering area is read out, finds corresponding source in buffering area
Frame corresponding to frame number is that start frame is set to StartFrame (a), StartFrame (b), StartFrame (c), StartFrame
(d)....etc。
Step S5, the target and corresponding start frame obtained in buffering area to each detection carry out initialization process simultaneously
Processing is tracked to target.The target Objecct and corresponding start frame StarFrame obtained for each detection, just
Beginningization is some to be based on frame buffer zone object tracking process, each tracking process processing target, between each tracking process simultaneously
Row operation.
Method for tracking target therein includes：First, looked in buffering area according to StartFrame parameters in buffering area
Go to start.Secondly, tracing positional is started according to Object parameter initialization target.Again, target following side frame by frame is utilized
Formula carries out tracking and amendment frame by frame to the position of target；Since start frame, be successively read frame data, using target frame by frame with
Track method, can flexibly it select such as based on optical flow approach, the tracking and amendment of position progress frame by frame to target.Finally, buffering
The frame in area carries out continuing to handle being switched to calculate in real time untill ruing out of the frame in buffering area before ruing out of to frame；
Before the frame of buffering area runs out of, the frame of buffering area is carried out continuing processing, untill ruing out of the frame in buffering, switchs to reality
When calculate.
Another object of the present invention is to provide a kind of system of detection and tracking based on circulation frame buffer zone, the system
Include several detectors, the system includes：
Buffering area establishes module, for establishing cyclic buffer on a timeline；
Frame number setup module, for extracting a frame of video by strategy from the continuous frame sequence inputted in real time and recording frame
Sequence number；
Parallel detection module, parallel target inspection is carried out for the frame chosen to be sent in several object detectors
Survey；
Frame setup module, buffering area is addressed to read and by its source frame correspondence for the source frame sequence number according to detection
Frame be arranged to start frame, if start frame is StartFrame (a), StartFrame (b), StartFrame (c),
StartFrame(d)....etc；
Initialize and tracking module, target and corresponding start frame for being obtained in buffering area to each detection are entered
Row initialization process is simultaneously tracked processing to target.
Simultaneously tracking module also includes for the initialization：
Search unit, existed for detecting that new object is set to Object (a) in buffering area according to StartFrame parameters
Start frame is found in buffering area；
Position tracking unit, start tracing positional for the parameter initialization target according to Object (a)；
Tracking and amending unit, for carrying out tracking frame by frame to the position of target using target following mode frame by frame and repairing
Just；
Converting unit is handled, for continue processing until ruing out of buffering to frame before the frame of buffering area runs out of
Switched to calculate in real time untill frame in area.
The cyclic buffer that the buffering area is established in module, the new frame of the propulsion of axle is by covering buffering area over time
An oldest frame data；All tracking targets share same buffering area；Buffering area data access in parallel procedure is met
Demand support sequential access and random access both of which.
Frame in the frame number setup module in continuous frame sequence is not overlapping in time shaft sequence；Different Object,
The time shaft being each independently of the others is safeguarded, each Object is tracked computing by respective time shaft.
Carry out between some object detectors in the parallel detection module and concurrently detect independently of each other；Multiple detections
The time that device carries out parallel target detection completion is random on a timeline；Multiple detectors are after the completion of parallel detection, inspection
The target measured transfers to merger module to synchronize, collect, the operation such as merger, the Object set tracked.
Eliminate target detection be connected with object tracking process between obstruction, realize detection with tracking complete parallel,
The obstruction between each target detection is eliminated, the complete parallel between different frame detection is realized, is realizing complete parallel
Detection is with the basis of calculating, improving the performance of detection, reducing the frequency for detecting and taking out frame, so as to reduce the probability of target missing inspection,
On the basis of blocking between elimination target detection and target following, bring up to tracking process and accounted for available for the time actually calculated
Than from the performance for improving tracking, supporting bigger tracking object set.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement made within refreshing and principle etc., should be included in the scope of the protection.
Claims (10)
1. a kind of method of detection and tracking based on circulation frame buffer zone, this method include several detectors, its feature
It is, the described method comprises the following steps：
A, cyclic buffer is established on a timeline；
B, extract a frame of video by strategy from the continuous frame sequence inputted in real time and record frame number；
C, the frame chosen is sent in several object detectors and carries out parallel target detection；
D, buffering area is addressed to read according to the source frame sequence number of detection and frame corresponding to by its source frame number is arranged to start frame,
If start frame is StartFrame (a), StartFrame (b), StartFrame (c), StartFrame (d) ... .etc.；
E, the target and corresponding start frame obtained in buffering area to each detection carries out initialization process and target is entered
Line trace is handled.
2. according to the method for claim 1, it is characterised in that further comprising the steps of in the step E：
E1, detect that new object is set to Object (R), found in buffering area according to StartFrame parameters in buffering area
Beginning frame；
E2, tracing positional started according to Object (R) parameter initialization target；
E3, tracking and amendment using position progress of the target following mode frame by frame to target frame by frame；
E4, frame is being carried out continuing processing untill ruing out of the frame in buffering area by its turn before the frame of buffering area runs out of
To calculate in real time;
Wherein, R ∈（a,b,c,d….etc.）.
3. according to the method for claim 2, it is characterised in that cyclic buffer in the step A, over time axle
A new frame is promoted to cover a frame data oldest in buffering area；All tracking targets share same buffering area；Buffering area
Sequential access and random access both of which are supported in the needs of data access in meeting parallel procedure.
4. according to the method for claim 3, it is characterised in that the frame in the step B in continuous frame sequence is in time shaft
Sequence is not overlapping；Different destination objects, safeguards the time shaft being each independently of the others, and each destination object presses the respective time
Axle is tracked computing.
5. according to the method for claim 4, it is characterised in that in the step C between some object detectors mutually solely
It is vertical to carry out and concurrently detect；The time that multiple detectors carry out parallel target detection completion is random on a timeline；It is multiple
For detector after the completion of parallel detection, the target for detecting to obtain transfers to merger module to synchronize, collect and the operation of merger, obtains
To the goal set of tracking.
6. a kind of system of detection and tracking based on circulation frame buffer zone, the system include several detectors, its feature
It is, the system includes：
Buffering area establishes module, for establishing cyclic buffer on a timeline；
Frame number setup module, for extracting a frame of video by strategy from the continuous frame sequence inputted in real time and recording frame sequence
Number；
Parallel detection module, parallel target detection is carried out for the frame chosen to be sent in several object detectors；
Frame setup module, reading and frame corresponding to by its source frame number are addressed to buffering area for the source frame sequence number according to detection
Start frame is arranged to, if start frame is StartFrame (a), StartFrame (b), StartFrame (c), StartFrame
(d)....etc.；
Initialize and tracking module, target and corresponding start frame for being obtained in buffering area to each detection are carried out just
Beginningization handles and processing is tracked to target.
7. system according to claim 6, it is characterised in that simultaneously tracking module also includes for the initialization：
Search unit, for detecting that new object is set to Object (R), buffered in buffering area according to StartFrame parameters
Start frame is found in area；
Position tracking unit, for starting tracing positional according to Object (R) parameter initialization target；
Tracking and amending unit, for carrying out tracking and amendment frame by frame to the position of target using target following mode frame by frame；
Converting unit is handled, for continue processing until ruing out of in buffering area to frame before the frame of buffering area runs out of
Frame untill switched to calculate in real time；
Wherein, R ∈（a,b,c,d….etc.）.
8. system according to claim 7, it is characterised in that the cyclic buffer that the buffering area is established in module, with
A frame data oldest in buffering area will be covered by the new frame of propulsion of time shaft；All tracking targets share same buffering
Area；Buffering area supports sequential access and random access both of which in the needs of data access in meeting parallel procedure.
9. system according to claim 8, it is characterised in that the frame in the frame number setup module in continuous frame sequence
It is not overlapping in time shaft sequence；Different destination objects, the time shaft being each independently of the others is safeguarded, each destination object is by each
From time shaft be tracked computing.
10. system according to claim 9, it is characterised in that in the parallel detection module some object detectors it
Between independently of each other carry out and concurrently detect；The time that multiple detectors carry out parallel target detection completion be on a timeline with
Machine；Multiple detectors are after the completion of parallel detection, and the target for detecting to obtain transfers to merger module to synchronize, collect and merger
Operation, the goal set tracked.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611042801.XA CN106778503B (en) | 2016-11-11 | 2016-11-11 | A kind of method and system of detection and tracking based on circulation frame buffer zone |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611042801.XA CN106778503B (en) | 2016-11-11 | 2016-11-11 | A kind of method and system of detection and tracking based on circulation frame buffer zone |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106778503A CN106778503A (en) | 2017-05-31 |
| CN106778503B true CN106778503B (en) | 2018-04-03 |
Family
ID=58973963
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611042801.XA Active CN106778503B (en) | 2016-11-11 | 2016-11-11 | A kind of method and system of detection and tracking based on circulation frame buffer zone |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106778503B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107992366B (en) * | 2017-12-26 | 2021-12-31 | 杭州易现先进科技有限公司 | Method, system and electronic equipment for detecting and tracking multiple target objects |
| CN110503663B (en) * | 2019-07-22 | 2022-10-14 | 电子科技大学 | Random multi-target automatic detection tracking method based on frame extraction detection |
-
2016
- 2016-11-11 CN CN201611042801.XA patent/CN106778503B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106778503A (en) | 2017-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109413392B (en) | Embedded multi-channel video image acquisition and parallel processing system and method | |
| CN106778503B (en) | A kind of method and system of detection and tracking based on circulation frame buffer zone | |
| CN106534692A (en) | Video image stabilization method and device | |
| CN101846511B (en) | Detection system of space debris | |
| CN105308618B (en) | Face recognition by means of parallel detection and tracking and/or grouped feature motion shift tracking | |
| CN101929859A (en) | Image full-frame scanning based space debris detecting method | |
| GB2605727A (en) | Stream object tracking with delayed object detection | |
| US8234454B2 (en) | Method and system of numerical analysis for continuous data | |
| CN102625110A (en) | Caching system and caching method for video data | |
| Wang et al. | A semi-automatic video labeling tool for autonomous driving based on multi-object detector and tracker | |
| Mahmoudi et al. | Multi-CPU/multi-GPU based framework for multimedia processing | |
| CN103873779A (en) | Method for controlling intelligent camera for parking lot | |
| CN101615296B (en) | Tracking point detection apparatus and method, program, and recording medium | |
| JP2010146522A (en) | Face image tracking device, face image tracking method, and program | |
| JP7157330B2 (en) | Monitoring device, monitoring method and monitoring program | |
| CN111078195A (en) | Target capture parallel acceleration method based on OPENCL | |
| US10726284B2 (en) | Device for selecting and describing points of interest in a sequence of images, for example for the pairing of points of interest | |
| CN102494674A (en) | High precision positioning method of dark space debris | |
| JP2022025878A (en) | Object detection device and program | |
| JP7207014B2 (en) | Display control device, display control program and inspection system | |
| Doerr et al. | A methodology for image-based tracking of seismic-induced motions | |
| JPS62262185A (en) | Pipe line control system | |
| Tingting | Temporal Mapping and Temporal Prediction based Ultra-low Delay Object Tracking Computing Architecture for Visual Feedback System | |
| Hu | Temporal Mapping and Temporal Prediction based Ultra-low Delay Object Tracking Computing Architecture for Visual Feedback System | |
| CN101251811B (en) | Method and device for associating threads within non-related processes based on memory paging behaviors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |