CN109398533A - A kind of mobile platform and the method for mobile platform tracking for a long time - Google Patents
A kind of mobile platform and the method for mobile platform tracking for a long time Download PDFInfo
- Publication number
- CN109398533A CN109398533A CN201811395714.1A CN201811395714A CN109398533A CN 109398533 A CN109398533 A CN 109398533A CN 201811395714 A CN201811395714 A CN 201811395714A CN 109398533 A CN109398533 A CN 109398533A
- Authority
- CN
- China
- Prior art keywords
- target
- tracking
- mobile platform
- preset condition
- frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000001514 detection method Methods 0.000 claims abstract description 40
- 230000000007 visual effect Effects 0.000 claims abstract description 22
- 230000008859 change Effects 0.000 claims description 19
- 239000013598 vector Substances 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 4
- 230000009471 action Effects 0.000 claims description 3
- 230000000903 blocking effect Effects 0.000 description 5
- 230000007547 defect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 241000270295 Serpentes Species 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000009194 climbing Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B62—LAND VEHICLES FOR TRAVELLING OTHERWISE THAN ON RAILS
- B62D—MOTOR VEHICLES; TRAILERS
- B62D63/00—Motor vehicles or trailers not otherwise provided for
- B62D63/02—Motor vehicles
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R11/00—Arrangements for holding or mounting articles, not otherwise provided for
- B60R11/04—Mounting of cameras operative during drive; Arrangement of controls thereof relative to the vehicle
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B62—LAND VEHICLES FOR TRAVELLING OTHERWISE THAN ON RAILS
- B62D—MOTOR VEHICLES; TRAILERS
- B62D63/00—Motor vehicles or trailers not otherwise provided for
- B62D63/02—Motor vehicles
- B62D63/04—Component parts or accessories
-
- 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
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R11/00—Arrangements for holding or mounting articles, not otherwise provided for
- B60R2011/0001—Arrangements for holding or mounting articles, not otherwise provided for characterised by position
- B60R2011/004—Arrangements for holding or mounting articles, not otherwise provided for characterised by position outside the vehicle
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R11/00—Arrangements for holding or mounting articles, not otherwise provided for
- B60R2011/0042—Arrangements for holding or mounting articles, not otherwise provided for characterised by mounting means
- B60R2011/008—Adjustable or movable supports
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Transportation (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of methods of mobile platform and mobile platform tracking for a long time, wherein, a kind of mobile platform with long-time following function, it include: visual identifying system and platform control system, visual identifying system includes panorama reflecting mirror and the image collecting device for acquiring the image information on panorama reflecting mirror；Platform control system includes: detection module, tracking module and prediction module, and detection module detects target according to the image information that visual identifying system receives；Tracking module tracking module tracks target according to the image information that visual identifying system receives；Target motion track when prediction module is blocked for target movement pattern before being blocked when target is blocked according to target.The present invention so that practical moving-vision platform is tracked for a long time and be unlikely to lose target, solve tracking during easily occur block, target kinematic nonlinearities, noise non-gaussian the problems such as.
Description
Technical field
The present invention relates to field of intelligent control, in particular to the side of a kind of mobile platform and mobile platform tracking for a long time
Method.
Background technique
Mobile mechanism's technical solution of existing mobile robot has crawler type, leg formula, snake formula, great-jump-forward, wheeled etc. several
Kind.Wherein crawler type ground adhesion property and passage capacity are good, but speed compared with slow, power consumption is larger, turn to when to ground failure
Degree is big.Though legged mobile robot can satisfy certain special performance requirements, adapt to complicated landform, certainly due to its structure
Too many, mechanism is complicated by spending, and causes it to be difficult to control, movement speed is slow, power consumption is big.Although snake formula and great-jump-forward are in certain sides
Face, such as complex environment, particular surroundings, mobility have its unique superiority, but there is also some fatal defects, such as hold
Loading capability, robust motion etc..In contrast, wheeled mobile robot have from heavy and light, carrying is big, mechanism is simple, driving and
Control relatively convenient, the speed of travel fast, maneuverability, the advantages that work efficiency is high, thus be widely used in industry, agricultural,
The fields such as anti-terrorism is explosion-proof, family, space exploration.
It is scouted in the mobile platform that trolley and unmanned plane etc. use visual sensor unmanned, to the intelligence of specific objective
Tracking is also always the key points and difficulties of a research.It is existing major part method be consider monitoring scene under video target with
The characteristics of track, the scene, is that camera is fixed and only needs to track target in short-term.And it is unmanned scout trolley and nobody
In the vision system application scenarios of machine, the tracking of specific objective is not only belonged to track for a long time, and its imaging sensor
It needs when target will be disengaged from the visual field, it is mobile to continue to track target, such imaging sensor figure collected by holder
As being also easier to by noise pollution.Occlusion issue is also often considered in the conventional method, but is to be directed to block in short-term mostly,
Existing method is suitable only for being detached from situation of the target blocked still in range of video, and unsuitable sheet blocks so that target
It is detached from the situation of field range.Existing method also is only applicable to target and moves the feelings that linear and noise meets Gaussian Profile
Shape, this situation are not present in actual scene.
Summary of the invention
For the technical problems in the prior art, the object of the present invention is to provide a kind of mobile platform and movement are flat
The method of director's time tracking.
The purpose of the invention is achieved by the following technical solution: a kind of mobile platform with long-time following function, packet
Include: visual identifying system and platform control system, visual identifying system include panorama reflecting mirror and anti-for acquiring panorama
Penetrate the image collecting device of the image information on mirror；
Platform control system includes: detection module, the image information pair that detection module is received according to visual identifying system
Target is detected；
Tracking module, tracking module track target according to the image information that visual identifying system receives；And
Prediction module, for mesh when target movement pattern is blocked before being blocked when target is blocked according to target
Mark motion track.
Preferably, mobile platform has car body, wheel, connector and steering engine, and connector one end is articulated on car body, vehicle
Wheel is set to the connector other end, and the steering engine is connect with connector, for controlling the wobbling action of connector to adjust vehicle
The chassis of body between ground at a distance from；
It further include lifting device, the panorama reflecting mirror is set on lifting device；
Visual identifying system further includes four cameras, and four cameras are separately positioned on the front, rear portion, left part of car body
And right part, the front of car body, the atlas information at rear, left and right are acquired respectively；
It further include the second steering engine, the camera positioned at the front of car body is set on the second steering engine, and the second steering engine is for adjusting
Section is in the pitch angle of the camera of Vehicular body front.
A method of long-time tracking is carried out using the above-mentioned mobile platform with long-time following function, comprising:
Video first frame image is read in, and target following frame is selected and to tracking module, detection module, prediction module and detection
Label is initialized；
After determining target following frame, next frame image is read, tracks target using tracking module；
Judge whether tracking succeeds, if tracking failure, re-detection is carried out to target using detection module, if detecting mesh
Mark, then done the calculating of target's center position and frame contour with testing result, otherwise carry out predicting tracing；
The calculating of target's center position and frame contour size is carried out using tracking module if tracking successfully；
After obtaining frame contour, judge whether the area change factor for tracking frame contour meets the first preset condition, if not
Meet the first preset condition, then adjusts mobile platform state until meeting the first preset condition；If meeting the first preset condition,
After obtaining target's center position, judge whether target's center position meets the second preset condition, if not meeting the second default item
Part then adjusts mobile platform state until meeting the second preset condition, if meeting the second preset condition, then judges whether to receive
To the order for terminating tracking, terminate to track if receiving, otherwise, is back to and reads next frame image.
Preferably, exist during tracking and be judged as tracking failure if blocking.
Preferably, judge tracking whether successful condition are as follows: according to tracking module calculate image block conservative similarity, if
Then it is to track successfully more than or equal to preset conservative similarity threshold, is tracking failure if being less than conservative similarity threshold.
Preferably, first preset condition are as follows: track area of the area change factor in tracking frame contour of frame contour
In change rate setting range；
Target following frame is selected and records the area α of initial tracking box1, calculate the target following frame of each frame of input
Area be α2, pass through one reference area change rate C of formula:
Formula one:
Set upper bound threshold value C1, lower bound threshold value C2If area change rate C is in C1To C2When in range, target's center position is judged
It sets and whether meets the second preset condition.
Preferably, the second preset condition are as follows: target's center position is without departing from offset threshold range；
Target's center position is without departing from offset threshold range specifically: setting left margin threshold value L1, right margin threshold value R1, on
Boundary threshold U1, lower boundary threshold value D1If target's center position (x, y) exceeds offset threshold range, that is, meet: x < L1Or x > R1
Or y < D1Or y > U1, then adjust mobile platform state until meet the first preset condition, when target's center position (x, y) without departing from
Offset threshold range meets: L1<x<R1And D1<y<U1, then judge whether to receive the order for terminating tracking, if receiving
Terminate tracking, is back to if not and reads next frame image.
Preferably, the predicting tracing is to carry out predicting tracing using support vector regression.
Preferably, when determination will carry out predicting tracing, according to the mesh collected before predicting tracing in setting number of frames
Mark movement pattern next frame target shift position.
Preferably, after target following frame is selected, setting detection label initial value is 0, then, judge detection mark whether as
1, if it is not, tracking target using tracking module；If judgement detection be labeled as 1, re-detection target, if detecting target,
The calculating of target's center position and frame contour is done with testing result, and changes detection labeled as 0, otherwise, carries out predicting tracing simultaneously
It keeps detection to be labeled as 1 and carries out the calculating of target's center position and frame contour size；
Then judge whether that receive the order for terminating tracking terminates to track if receiving meeting the second preset condition, it is no
It is then back to and reads next frame image.
The present invention has the following advantages and effects with respect to the prior art:
1, the present invention enables practical moving-vision platform to track for a long time and is unlikely to lose target, solves during tracking
Easily occur block, target kinematic nonlinearities, noise non-gaussian the problems such as.
2, compared with the prior art, long time-tracking both may be implemented in the present invention, can also overcome to block etc. to a certain extent and ask
Topic is realized using the support vector regression device for being adapted to real world system noise non-gaussian, the nonlinear feature of motion profile
The trajectory predictions of target point after blocking completely, to solve to occur to block loss tracking target problem entirely during tracking.
3, compared to when large area occur and blocking, target is detached from the visual field by blocking shielding and no longer returns the former visual field, existing
The case where having technology detector that can not detect target again, failing so as to cause tracking, the present invention are utilizing support vector regression
On the basis of device carries out trajectory predictions, the corresponding Motion of platform is formulated, is lost to reduce and be tracked when above situation occurs
The probability lost.
4, present invention is primarily based on the improvement that existing wheeled technical solution has carried out chassis and image passback, in chassis side
Existing technical solution is compared in face, and for the present invention using the structure and control method on variable chassis, mainly overcoming former technology can not
Take into account the defect of obstacle climbing ability and concealment.In terms of image passback, the existing one direction low-angle for scouting trolley is compared
Image Intelligence acquisition and passback, the present invention realize 360 ° of panoramic picture information collections using lifting device and panorama reflecting mirror etc.
And passback, it overcomes original technology scheme and collects the unilateral defect of information.
Detailed description of the invention
Fig. 1 is the structural schematic diagram for the mobile platform that the present invention has long-time following function；
Fig. 2 is the front view for the mobile platform that the present invention has long-time following function；
Fig. 3 is the flow diagram for the method that mobile platform of the present invention carries out long-time tracking.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
A kind of mobile platform with long-time following function, comprising: visual identifying system and platform control system, depending on
Feel that identifying system includes panorama reflecting mirror 1 and the image collecting device 2 for acquiring the image information on panorama reflecting mirror 1；
Platform control system includes: detection module, the image information pair that detection module is received according to visual identifying system
Target is detected；
Tracking module, tracking module track target according to the image information that visual identifying system receives；And
Prediction module, for mesh when target movement pattern is blocked before being blocked when target is blocked according to target
Mark motion track.
Preferably, mobile platform has car body 3, wheel 4, connector 5 and steering engine 6, and 5 one end of connector is articulated with car body
On 3, wheel 4 is set to 5 other end of connector, and the steering engine 6 is connect with connector 5, for controlling the wobbling action of connector 5
To adjust the chassis of car body 3 between ground at a distance from；
It further include lifting device 7, the panorama reflecting mirror 1 is set on lifting device 7；
Visual identifying system further includes four cameras 8, four cameras 8 be separately positioned on the front of car body 3, rear portion,
Left part and right part acquire the front of car body 3, the image information at rear, left and right respectively, further include the second steering engine,
Camera positioned at the front of car body is set on the second steering engine, and the second steering engine is used to adjust the camera positioned at Vehicular body front
Pitch angle.
Preferably, further include electron speed regulator for controlling mobile platform movement speed.
Preferably, the mobile platform is remotely pilotless trolley.
Preferably, steering engine 6 is fixedly installed on connector 5, and the shaft of steering engine 6 is connect with car body 3.
The mirror surface of panorama reflecting mirror 1 is the hyperboloidal mirror face protruded to image collecting device 2.
It preferably, further include driving motor and shaft coupling, the driving motor is installed on connector, and wheel passes through connection
Axis device is installed on driving motor.
A method of long-time tracking is carried out using the above-mentioned mobile platform with long-time following function, comprising:
Step 1, target following frame is selected: the image arrived according to the cameras capture for scouting trolley, using existing target
Detection module is for example: YOLOv3 algorithm, carries out real-time target detection to image, and select all people in interactive interface highlight box
Class target, user can choose interested target frame by interactive interface, and computer receives the location information of target frame, simultaneously
Record the area a of initial tracking box1, tracking operation can be started；
Step 2, after determining target following frame, target is tracked using tracking module: for example: using dimension self-adaption KCF
Algorithm carries out the tracking of dimension self-adaption to next frame image.
Step 3, judge whether tracking succeeds, if tracking failure, re-detection target, if detecting target, with detection
As a result it does target's center position and frame contour calculates, otherwise carry out predicting tracing；
The calculating of target's center position and frame contour size is carried out using tracking module if tracking successfully；
Step 4, after obtaining frame contour, judge whether the area change factor for tracking frame contour meets the first default item
Part adjusts mobile platform state until meeting the first preset condition, then executes step 5 if not meeting the first preset condition；
Step 5, if meeting the first preset condition, after obtaining target's center position, whether judge target's center position
Meet the second preset condition, if not meeting the second preset condition, adjust mobile platform state until meet the second preset condition,
If meeting the second preset condition, then judges whether to receive the order for terminating tracking, terminate to track if receiving, otherwise, return
Step 2 is back to handle next frame.
Preferably, exist during tracking and be judged as tracking failure if blocking.
Preferably, judge tracking whether successful condition are as follows: the conservative similarity of image block is determined according to tracking module, if
Then it is to track successfully more than or equal to preset conservative similarity threshold, is tracking failure if being less than conservative similarity threshold.
Preferably, first preset condition are as follows: track area of the area change factor in tracking frame contour of frame contour
In change rate setting range.
Preferably, the second preset condition are as follows: target's center position is without departing from preset offset threshold range.
Preferably, the predicting tracing is to carry out predicting tracing using support vector regression.
Preferably, when determination will carry out predicting tracing, according to the mesh collected before predicting tracing in setting number of frames
Mark movement pattern next frame target shift position.
The present invention proposes to carry out predicting tracing to target with support vector regression, in this example, track it is unobstructed and
It scouts small car state and does not change the stage, save the two-dimensional coordinate of target's center's point of newest 10 frame, to be predicted in determination
When tracking, it is utilized respectively the horizontal axis coordinate and ordinate of orthogonal axes of 10 coordinates based on time series saved, training is supported
Vector regression.The support vector regression that the present invention uses is trained using the data of single-input single-output, when input is
Between, output is coordinate value, and kernel function uses gaussian kernel function.That completes two groups of data can obtain two supporting vectors after training
Regression machine brings the newest time into, can obtain the abscissa value and ordinate value of the target predicted position of a new frame respectively, in advance
During surveying tracking, the size of target frame is remained unchanged, and is only predicted target's center's point position.
Preferably, after target following frame is selected, initial frame is read in, detection module detects target, initialization tracking module, pre-
Module and re-detection module are surveyed, setting detection label initial value is 0, then, reads a new frame image, judges that detection label is
No is 1, if it is not, target then is tracked using tracking module, if 1, then re-detection target；If detecting target, with detection
As a result the calculating of target's center position and frame contour is done, change detection is labeled as 0, otherwise, is predicted using support vector regression
Target trajectory (predicting tracing) simultaneously keeps detection labeled as 1, then, carries out the calculating of target's center position and frame contour size；
When meeting the second preset condition and not receiving the order for terminating tracking, it is back to and reads at next frame image,
Circulation carries out above-mentioned steps.
Preferably, target following frame is selected and records the area α of initial tracking box1, calculate the mesh of each frame of input
The area for marking tracking box is α2, pass through one reference area change rate C of formula:
Formula one:
Set upper bound threshold value C1, lower bound threshold value C2If area change rate C is in C1To C2When in range, it may be assumed that i.e. C1<C<C2, sentence
Whether disconnected target's center position meets the second preset condition.
Mobile platform state is adjusted until meeting the first preset condition specifically: work as C > C2When, needed for mobile platform retreats
Distance makes area change rate C meet C1<C<C2, as C < C1When, mobile platform advance required distance meets area change rate C
C1<C<C2。
Preferably, target's center position is without departing from offset threshold range, specifically: setting left margin threshold value L1, right margin
Threshold value R1, coboundary threshold value U1, lower boundary threshold value D1If target's center position (x, y) exceeds offset threshold range, that is, meet: x <
L1Or x > R1Or y < D1Or y > U1, then mobile platform state is adjusted until meeting the first preset condition.When target's center position (x,
Y) without departing from offset threshold range, that is, meet: L1<x<R1And D1<y<U1, then then judge whether to receive the life for terminating tracking
It enables, receives, terminate algorithm, be back to if not and read next frame image, circulation carries out above sequence of steps.
Mobile platform state is adjusted until meeting the second preset condition, specially as x < L1, then needed for mobile platform turns left
Angle makes x meet L1<x<R1；Work as x > R1, then angle needed for mobile platform right-hand rotation, makes x meet L1<x<R1；Work as y > U1, then move
Platform will control the second steering engine and the camera 8 positioned at Vehicular body front is faced upward required angle, and y is made to meet D1<y<U1；As y < D1,
Then mobile platform makes the angle needed for 8 nutation of camera of Vehicular body front for the second steering engine is controlled, and y is made to meet D1<y<U1。
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (10)
1. a kind of mobile platform with long-time following function, characterized by comprising: visual identifying system and platform control
System processed, visual identifying system include panorama reflecting mirror and the Image Acquisition for acquiring the image information on panorama reflecting mirror
Device；
Platform control system includes: detection module, and the image information that detection module is received according to visual identifying system is to target
It is detected；
Tracking module, tracking module track target according to the image information that visual identifying system receives；And
Prediction module is moved for target when target movement pattern is blocked before being blocked when target is blocked according to target
Dynamic rail mark.
2. the mobile platform according to claim 1 with long-time following function, which is characterized in that mobile platform has
Car body, wheel, connector and steering engine, connector one end are articulated on car body, and wheel is set to the connector other end, the rudder
Machine is connect with connector, for controlling the wobbling action of connector to adjust the chassis of car body between ground at a distance from；
It further include lifting device, the panorama reflecting mirror is set on lifting device；
Visual identifying system further includes four cameras, four cameras be separately positioned on the front of car body, rear portion, left part and
Right part acquires the front of car body, the atlas information at rear, left and right respectively；
It further include the second steering engine, the camera positioned at the front of car body is set on the second steering engine, and the second steering engine is for adjusting position
In the pitch angle of the camera of Vehicular body front.
3. a kind of mobile platform using described in as claimed in claim 1 or 22 with long-time following function carries out long-time tracking
Method, characterized by comprising:
Video first frame image is read in, and target following frame is selected and marks to tracking module, detection module, prediction module and detection
It is initialized；
After determining target following frame, next frame image is read, tracks target using tracking module；
Judge whether tracking succeeds, if tracking failure, re-detection is carried out to target using detection module, if detecting target,
The calculating that target's center position and frame contour are then done with testing result, otherwise carries out predicting tracing；
The calculating of target's center position and frame contour size is carried out using tracking module if tracking successfully；
After obtaining frame contour, judge whether the area change factor for tracking frame contour meets the first preset condition, if not meeting
First preset condition then adjusts mobile platform state until meeting the first preset condition；If meeting the first preset condition, obtaining
After taking target's center position, judge whether target's center position meets the second preset condition, if not meeting the second preset condition,
Mobile platform state is adjusted until meeting the second preset condition, if meeting the second preset condition, then judges whether to receive knot
The order of beam tracking, terminates to track if receiving, and otherwise, is back to and reads next frame image.
4. the method for mobile platform according to claim 1 tracking for a long time, which is characterized in that if existing during tracking
It blocks, is judged as tracking failure.
5. the method for mobile platform according to claim 1 tracking for a long time, which is characterized in that judge whether tracking succeeds
Condition are as follows: according to tracking module calculate image block conservative similarity, if more than or be equal to preset conservative similarity threshold
It is tracking failure if being less than conservative similarity threshold then to track successfully.
6. the method for mobile platform according to claim 1 tracking for a long time, which is characterized in that first preset condition
Are as follows: the area change factor of frame contour is tracked in the area change rate setting range of tracking frame contour；
Target following frame is selected and records the area α of initial tracking box1, calculate the face of the target following frame of each frame of input
Product is α2, pass through one reference area change rate C of formula:
Formula one:
Set upper bound threshold value C1, lower bound threshold value C2If area change rate C is in C1To C2When in range, judge that target's center position is
It is no to meet the second preset condition.
7. the method for mobile platform according to claim 1 tracking for a long time, which is characterized in that the second preset condition are as follows:
Target's center position is without departing from offset threshold range；
Target's center position is without departing from offset threshold range specifically: setting left margin threshold value L1, right margin threshold value R1, coboundary
Threshold value U1, lower boundary threshold value D1If target's center position (x, y) exceeds offset threshold range, that is, meet: x < L1Or x > R1Or y < D1
Or y > U1, then mobile platform state is adjusted until meeting the first preset condition, when target's center position (x, y) is without departing from offset threshold
It is worth range, that is, meets: L1<x<R1And D1<y<U1, then judge whether to receive the order for terminating tracking, terminate if receiving with
Track is back to if not and reads next frame image.
8. the method for mobile platform according to claim 1 tracking for a long time, which is characterized in that the predicting tracing is to make
Predicting tracing is carried out with support vector regression.
9. the method for the tracking for a long time of mobile platform according to claim 1 or 8, which is characterized in that be carried out in determination
It is mobile according to the target movement pattern next frame target collected before predicting tracing in setting number of frames when predicting tracing
Position.
10. the method for mobile platform according to claim 1 tracking for a long time, which is characterized in that target following frame is selected
Afterwards, setting detection label initial value is 0, then, judges that detection marks whether as 1, if it is not, tracking module is used to track mesh
Mark；If judgement detection is labeled as 1, re-detection target does target's center position and wheel if detecting target with testing result
The calculating of wide frame, and detection is changed labeled as 0, otherwise, carries out predicting tracing and keep detection labeled as 1 and carry out target's center
The calculating of position and frame contour size；
Then judge whether that receive the order for terminating tracking terminates to track, otherwise return if receiving meeting the second preset condition
It is back to and reads next frame image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811395714.1A CN109398533B (en) | 2018-11-22 | 2018-11-22 | Mobile platform and long-time tracking method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811395714.1A CN109398533B (en) | 2018-11-22 | 2018-11-22 | Mobile platform and long-time tracking method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109398533A true CN109398533A (en) | 2019-03-01 |
| CN109398533B CN109398533B (en) | 2024-04-09 |
Family
ID=65474530
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811395714.1A Active CN109398533B (en) | 2018-11-22 | 2018-11-22 | Mobile platform and long-time tracking method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109398533B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112598703A (en) * | 2020-12-14 | 2021-04-02 | 北京爱笔科技有限公司 | Article tracking method and device |
| WO2023072290A1 (en) * | 2021-10-29 | 2023-05-04 | 中移(成都)信息通信科技有限公司 | Tracking target occlusion determination method and apparatus, device and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU97329U1 (en) * | 2010-04-20 | 2010-09-10 | Государственное образовательное учреждение высшего профессионального образования Курский государственный университет | THREE-WHEEL MOBILE ROBOT WITH A CAMERA |
| CN102881022A (en) * | 2012-07-20 | 2013-01-16 | 西安电子科技大学 | Concealed-target tracking method based on on-line learning |
| CN103325126A (en) * | 2013-07-09 | 2013-09-25 | 中国石油大学(华东) | Video target tracking method under circumstance of scale change and shielding |
| CN104732187A (en) * | 2013-12-18 | 2015-06-24 | 杭州华为企业通信技术有限公司 | Method and equipment for image tracking processing |
| CN204525436U (en) * | 2015-01-30 | 2015-08-05 | 王成财 | A kind of four-wheel drive transfer robot |
| CN105865269A (en) * | 2016-04-19 | 2016-08-17 | 滨州学院 | Invasive four-legged wheel type robot |
| CN106114474A (en) * | 2016-07-30 | 2016-11-16 | 张学衡 | A kind of anti-automobile immersion robot |
| CN106874854A (en) * | 2017-01-19 | 2017-06-20 | 西安电子科技大学 | Unmanned plane wireless vehicle tracking based on embedded platform |
| CN108154524A (en) * | 2018-01-16 | 2018-06-12 | 中国人民解放军陆军装甲兵学院 | Target predicting and tracking method |
| CN209274766U (en) * | 2018-11-22 | 2019-08-20 | 华南理工大学 | A kind of mobile platform |
-
2018
- 2018-11-22 CN CN201811395714.1A patent/CN109398533B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU97329U1 (en) * | 2010-04-20 | 2010-09-10 | Государственное образовательное учреждение высшего профессионального образования Курский государственный университет | THREE-WHEEL MOBILE ROBOT WITH A CAMERA |
| CN102881022A (en) * | 2012-07-20 | 2013-01-16 | 西安电子科技大学 | Concealed-target tracking method based on on-line learning |
| CN103325126A (en) * | 2013-07-09 | 2013-09-25 | 中国石油大学(华东) | Video target tracking method under circumstance of scale change and shielding |
| CN104732187A (en) * | 2013-12-18 | 2015-06-24 | 杭州华为企业通信技术有限公司 | Method and equipment for image tracking processing |
| CN204525436U (en) * | 2015-01-30 | 2015-08-05 | 王成财 | A kind of four-wheel drive transfer robot |
| CN105865269A (en) * | 2016-04-19 | 2016-08-17 | 滨州学院 | Invasive four-legged wheel type robot |
| CN106114474A (en) * | 2016-07-30 | 2016-11-16 | 张学衡 | A kind of anti-automobile immersion robot |
| CN106874854A (en) * | 2017-01-19 | 2017-06-20 | 西安电子科技大学 | Unmanned plane wireless vehicle tracking based on embedded platform |
| CN108154524A (en) * | 2018-01-16 | 2018-06-12 | 中国人民解放军陆军装甲兵学院 | Target predicting and tracking method |
| CN209274766U (en) * | 2018-11-22 | 2019-08-20 | 华南理工大学 | A kind of mobile platform |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112598703A (en) * | 2020-12-14 | 2021-04-02 | 北京爱笔科技有限公司 | Article tracking method and device |
| WO2023072290A1 (en) * | 2021-10-29 | 2023-05-04 | 中移(成都)信息通信科技有限公司 | Tracking target occlusion determination method and apparatus, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109398533B (en) | 2024-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11288818B2 (en) | Methods, systems, and computer readable media for estimation of optical flow, depth, and egomotion using neural network trained using event-based learning | |
| Campbell et al. | Techniques for evaluating optical flow for visual odometry in extreme terrain | |
| WO2019210142A1 (en) | Device and method for determining a center of a trailer tow coupler | |
| CN109583347A (en) | A method of it is tracked for a long time for mobile platform | |
| CN110163963B (en) | Mapping device and mapping method based on SLAM | |
| CN112069969A (en) | Method and system for tracking highway monitoring video mirror-crossing vehicle | |
| CN113111715B (en) | Unmanned aerial vehicle target tracking and information acquisition system and method | |
| Du et al. | Real-time vehicle following through a novel symmetry-based approach | |
| US20050226464A1 (en) | Camera control system to follow moving objects | |
| CN109398533A (en) | A kind of mobile platform and the method for mobile platform tracking for a long time | |
| CN103581536A (en) | Image capturing system and method | |
| Roberts et al. | Memory-based learning for visual odometry | |
| Maanpää et al. | Multimodal end-to-end learning for autonomous steering in adverse road and weather conditions | |
| Dunlay et al. | Obstacle avoidance on roadways using range data | |
| CN209274766U (en) | A kind of mobile platform | |
| Fernandez et al. | Visual odometry for an outdoor mobile robot | |
| Liu | A lidar-inertial-visual slam system with loop detection | |
| CN114429435A (en) | Wide-field-of-view range target searching device, system and method in degraded visual environment | |
| Zhu et al. | PairCon-SLAM: Distributed, online, and real-time RGBD-SLAM in large scenarios | |
| Zhao et al. | An ISVD and SFFSD-based vehicle ego-positioning method and its application on indoor parking guidance | |
| Contreras et al. | A Stereo Visual Odometry Framework with Augmented Perception for Dynamic Urban Environments | |
| Giachetti et al. | The recovery of optical flow for intelligent cruise control | |
| Chang et al. | Omni-directional visual servoing for human-robot interaction | |
| JP7064948B2 (en) | Autonomous mobile devices and autonomous mobile systems | |
| Denman et al. | Multi-view intelligent vehicle surveillance system |
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 |