CN109523571A - A kind of the motion profile optimization method and system of non-characteristic matching - Google Patents
A kind of the motion profile optimization method and system of non-characteristic matching Download PDFInfo
- Publication number
- CN109523571A CN109523571A CN201811250104.2A CN201811250104A CN109523571A CN 109523571 A CN109523571 A CN 109523571A CN 201811250104 A CN201811250104 A CN 201811250104A CN 109523571 A CN109523571 A CN 109523571A
- Authority
- CN
- China
- Prior art keywords
- motion profile
- track
- point
- sequence
- moving direction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of motion profile optimization method of non-characteristic matching and systems, described method includes following steps: step S1, according to viewing field of camera angle and operating distance, calculate transverse field width, the matching frame per second of screen ratio, mankind's classics walking speed and motion track is accounted for according to mankind's moving distance, calculate track window N, to motion profile arrangement set, with N frame be calculate width calculate each point of motion profile sequence movement speed and moving direction, and further the sequence of calculation moving direction change rate;Step S2 determines the abnormal inflection point in track according to the movement speed of motion profile sequence and moving direction change rate, and motion profile sequence is cut to multistage motion profile sequence according to abnormal inflection point, and from the motion profile arrangement set of Combination nova Cheng Xin;Step S3 carries out disconnection judgement to motion profile according to new motion profile arrangement set, is overlapped judgement and separation judgement, the motion profile arrangement set after being repaired.
Description
Technical field
The present invention relates to stream of people's statistical technique fields in video image processing, count more particularly to one kind for the stream of people
Non- characteristic matching motion profile optimization method and system.
Background technique
At present in video image processing, method for tracking target mainly has: matching process based on target quiescent feature and
Matching process based on target range and geometrical characteristic.Since the method based on target quiescent characteristic matching needs to calculate all mesh
The characteristic information being marked on inside entire video sequence, and needs and all characteristics carry out circulation matching, so having matching
Property it is strong, not because local time because target lose due to cause tracking fail, can effectively be distinguished in the case where multiple target tracking each
The characteristics of target, meanwhile, strong matched method global in this way also results in the huge problem of calculation amount;And it is based on distance and geometry
The matching process of feature has matching speed fast, but in the case where multiple target tracking, due to lacking the feature letter of different target
Breath, leads to the problem that matching accuracy rate is low.
Field is counted in the stream of people at present, computing terminal prepositionization is generally used, causes the computing capability of computing terminal weaker,
So can only be tracked and be counted using the matching process of distance and geometrical characteristic, all accuracys rate are not high.
Summary of the invention
In order to overcome the deficiencies of the above existing technologies, purpose of the present invention is to provide a kind of movements of non-characteristic matching
Track optimizing method and system, to introduce the behavior of the mankind after obtaining using distance and the matched motion profile of geometrical characteristic
Matching Model optimizes motion profile, improves the accuracy rate of tracing path.
In view of the above and other objects, the present invention proposes a kind of motion profile optimization method of non-characteristic matching, including such as
Lower step:
Step S1 calculates transverse field width FW according to viewing field of camera angle FOV and operating distance WD, mobile according to the mankind
Distance accounts for the matching frame per second FPS shielded than SL, mankind classics walking speed VR and motion track, calculates track window N, to movement rail
Mark arrangement set TR, it is the movement speed and moving direction for calculating width and calculating each point of motion profile sequence with N frame, goes forward side by side one
Walk the moving direction change rate Δ D of the sequence of calculationp;
Step S2 determines that the exception in track turns according to the movement speed of motion profile sequence and moving direction change rate
Motion profile sequence is cut to multistage motion profile sequence according to abnormal inflection point by point, and from the motion profile of Combination nova Cheng Xin
Arrangement set TB;
Step S3, according to new motion profile arrangement set TBTo motion profile carry out disconnection judgement, be overlapped judgement and
Separation judgement, the motion profile arrangement set after being repaired.
Preferably, in step S1, the transverse field width FW and track window N calculation method are as follows:
FW=2 × WD × tan (FOV)
Wherein FOV is viewing field of camera angle, and WD is operating distance, and SL is that window accounts for screen ratio, and VR is mankind's classics walking speed,
FPS is the matching frame per second of motion track.
Preferably, in step S1, to Mr. Yu motion profile sequence Tr∈TR, Tr={ { xi,yi,ti, ∈ (0~M) },
Wherein xiAnd yiFor the coordinate of motion profile point, tiFor the time of the motion profile point, M is sequence length, movement speed viAnd shifting
Dynamic direction diWith moving direction change rate Δ diCalculation method it is as follows:
Δdi=| da-db|
Wherein subscript a and b are respectively as follows:
Preferably, in step S2, the Rule of judgment of the exception inflection point are as follows:
Δdi>DL
vi>VL
Such as meet conditions above, then by motion profile sequence using inflection point as separation, single-row respectively is new motion profile
Sequence, all new motion profile combined sequences are at a new motion profile arrangement set TB,
Above-mentioned DLChange angle, V for maximumLFor nonstatic speed.
Preferably, the method for the track disconnection judgement is as follows:
Compare track Tb1End point Pb1With track Tb2Starting point Pb2If time tb2Greater than tb1And there is tb2And tb1's
Difference is less than tLThen think the continuity in two sections of track having times, wherein tLThe maximum tolerance time lost for target;
If moving direction db1And db2Difference be less than DL, it is believed that two sections of tracks have orientation consistency, wherein DLFor target
Maximum tolerance turn round limitation, i.e., maximum variation angle, for can setting parameter, usually
Calculate track Tb1End point Pb1With track Tb2Starting point Pb2Distance, if Pb1And Pb2Distance be less than Pb1
According to vb2In tLThe distance of time movement, then it is assumed that two sections of curves have the continuity in space,
If meeting three above condition simultaneously, then it is assumed that two tracks are identical strip path curve.
Preferably, the method for being overlapped judgement is as follows:
The first step, to Mr. Yu's section track Tb1, according to its end point Pb1Mobile trend, i.e. vb1And db1, it is calculated in tLAfterwards
Position, i.e. Pb′1, tLThe maximum tolerance time lost for target;
Second step compares other one section of track Tb2In all the points, if track Tb1Prolongation and Tb2Intersection, then recognize
For Tb1And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that Tb1With
Tb2With temporal continuity;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2Have
Consistency on direction;
Third step, if meeting all conditions in second step, T is thought in surveyb2Part after middle intersection point also belongs to Tb1, after merging
Obtain new Tb′1。
Preferably, the method for the separation judgement are as follows:
First, to Mr. Yu's section track Tb1, according to its starting point Pb1Mobile trend, i.e. vb1And db1, retrospectively calculate its in tL
Preceding position, i.e. Pb′1, tLThe maximum tolerance time lost for target;
Second step compares other one section of track Tb2In all the points, if track Tb1Reverse extending part and Tb2Intersection,
Then think Tb1And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that
Tb1And Tb2With temporal continuity;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2
With the consistency on direction;
Third step, if meeting all conditions in second step, then it is assumed that Tb2Part before middle intersection point also belongs to Tb1, after merging
Obtain new T 'b1。
In order to achieve the above objectives, the present invention also provides a kind of motion profile optimization systems of non-characteristic matching, comprising:
Window calculation unit, for calculating transverse field width FW, root according to viewing field of camera angle FOV and operating distance WD
The matching frame per second FPS shielded than SL, mankind classics walking speed VR and motion track is accounted for according to mankind's moving distance, calculates track window
N, to motion profile arrangement set TR, it is the movement speed and movement for calculating width and calculating each point of motion profile sequence with N frame
Direction, and the further moving direction change rate Δ D of the sequence of calculationp;
Inflection point identifying processing unit is determined for the movement speed and moving direction change rate according to motion profile sequence
Motion profile sequence is cut to multistage motion profile sequence according to abnormal inflection point by the abnormal inflection point in track, and from Combination nova
The motion profile arrangement set T of Cheng XinB;
Unit is repaired in track, for according to new motion profile arrangement set TBTo motion profile progress disconnection judgement, again
Close judgement and separation judgement, the motion profile arrangement set after being repaired.
Preferably, the transverse field width FW and track window N calculation method are as follows:
FW=2 × WD × tan (FOV)
Wherein FOV is viewing field of camera angle, and WD is operating distance, and SL is that window accounts for screen ratio, and VR is mankind's classics walking speed,
FPS is the matching frame per second of motion track
Preferably for certain motion profile sequence Tr∈TR, Tr={ { xi,yi,ti, i ∈ (0~M) }, wherein xiAnd yi
For the coordinate of motion profile point, tiFor the time of the motion profile point, M is sequence length, movement speed viWith moving direction diWith
Moving direction change rate Δ diCalculation method it is as follows:
Δdi=| da-db|
Wherein subscript a and b are respectively as follows:
Compared with prior art, a kind of motion profile optimization method of non-characteristic matching of the present invention and system are regarded according to camera
Rink corner and operating distance determine visual field width, and obtain trajectory calculation window width by stream of people's conventional speeds and matching frame per second,
And then then speed, direction and the direction change rate for calculating track are passed through with judging whether track has unreasonable inflection point
Broken string judgement, is overlapped judgement and separation judges that three modules repair track set, obtains meeting stream of people's movement law
Track set, the present invention repair after the result of non-characteristic matching, can repair part symbol without carrying out a large amount of calculate
Close stream of people's movement law as a result, motion profile is optimized.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of the motion profile optimization method of non-characteristic matching of the present invention;
Fig. 2 is a kind of system architecture diagram of the motion profile optimization system of non-characteristic matching of the present invention;
Fig. 3 is the system construction drawing of the motion profile optimization system of the non-characteristic matching of the specific embodiment of the invention;
Fig. 4 is that schematic diagram is repaired in the track disconnection of disconnection judgment module in the specific embodiment of the invention.
Fig. 5 is the coincidence track reparation schematic diagram that judgment module is overlapped in the specific embodiment of the invention.
Fig. 6 is the separated track reparation schematic diagram that judgment module is separated in the specific embodiment of the invention.
Fig. 7 is that alternately schematic diagram is repaired in track in the specific embodiment of the invention.
Specific embodiment
Below by way of specific specific example and embodiments of the present invention are described with reference to the drawings, those skilled in the art can
Understand further advantage and effect of the invention easily by content disclosed in the present specification.The present invention can also pass through other differences
Specific example implemented or applied, details in this specification can also be based on different perspectives and applications, without departing substantially from
Various modifications and change are carried out under spirit of the invention.
Fig. 1 is a kind of step flow chart of the motion profile optimization method of non-characteristic matching of the present invention.As shown in Figure 1, this
The motion profile optimization method for inventing a kind of non-characteristic matching, includes the following steps:
Step S1 calculates transverse field width FW according to viewing field of camera angle FOV and operating distance WD, mobile according to the mankind
Distance accounts for the matching frame per second FPS shielded than SL, mankind classics walking speed VR and motion track, calculates track window N, to movement rail
Mark arrangement set TR, it is the movement speed and moving direction for calculating width and calculating each point of motion profile sequence with N frame, further
The moving direction change rate Δ D of the sequence of calculationp。
Specifically, the calculation method of transverse field width FW and track window N are as follows:
FW=2 × WD × tan (FOV)
Wherein FOV is viewing field of camera angle;WD is operating distance, i.e. physical distance of the camera to central region object;SL is
Window accounts for screen ratio, is normally defined 0.05;VR is mankind's classics walking speed, classical value 4m/s;FPS is of motion track
With frame per second.
To Mr. Yu motion profile sequence Tr∈TRIs defined as:
Tr={ { xi,yi,ti, i ∈ (0~M) }
Wherein xiAnd yiFor the coordinate of motion profile point, tiFor the time of the motion profile point, M is sequence length.
Its movement speed viWith moving direction diWith moving direction change rate Δ diCalculation method are as follows:
Δdi=| da-db|
Wherein subscript a and b are respectively as follows:
According to the above calculated result, then all motion profile sequence T are updatedr∈TR:
Tr={ { xi,yi,ti,vi,di,Δdi, i ∈ (0~M) }
Step S2 determines that the exception in track turns according to the movement speed of motion profile sequence and moving direction change rate
Motion profile sequence is cut to multistage motion profile sequence according to abnormal inflection point by point, and from the motion profile of Combination nova Cheng Xin
Arrangement set TB。
In the specific embodiment of the invention, the Rule of judgment of abnormal inflection point are as follows:
Δdi>DL
vi>VL
Wherein DLChange angle for maximum, usuallyVLFor nonstatic speed, usually 1m/s;
Such as meet conditions above, then by motion profile sequence using inflection point as separation, single-row respectively is new motion profile
Sequence, all new motion profile combined sequences are at a new motion profile arrangement set TB
Step S3, according to new motion profile arrangement set TBDisconnection judgement is carried out to motion profile, be overlapped judgement and is divided
Motion profile arrangement set from judgement, after being repaired.
In the specific embodiment of the invention, in step S3, the judgment method of the track disconnection are as follows:
(1) track T is comparedb1End point Pb1With track Tb2Starting point Pb2If tb2Greater than tb1And there is tb2And tb1Difference
Less than tLThen think the continuity in two sections of track having times, wherein tLIt is that can set ginseng for the maximum tolerance time that target is lost
It counts, usually 2s;
(2) if db1And db2Difference be less than DL, it is believed that two sections of tracks have orientation consistency, wherein DLMost for target
Big tolerance is turned round limitation, i.e., maximum variation angle, for can setting parameter, usually
(3) P is calculatedb1And Pb2Distance, if Pb1And Pb2Distance be less than Pb1According to vb2In tLThe distance of time movement,
Then think that two sections of curves have the continuity in space.
If meeting three above condition simultaneously, then it is assumed that two track (track Tb1With track Tb2) actually same
Track.
Specifically, track disconnection the specific implementation process is as follows:
There is Tb1,Tb2∈TB, have
Pb1={ xb1,yb1,tb1,vb1,db1,Δdb1It is Tb1End point;
Pb2={ xb2,yb2,tb2,vb2,db2,Δdb2It is Tb2Starting point;
If meeting:
Wherein, tLFor connection time difference limit, i.e., target lose the maximum tolerance time, for can setting parameter, be generally set to
2s;
Loop through TBIn all motion profile sequences;
In step S3, the method for being overlapped judgement are as follows:
The first step, to Mr. Yu's section track Tb1, according to its end point Pb1Mobile trend, i.e. vb1And db1, it is calculated in tLAfterwards
Position, i.e. P 'b1;
Second step compares other one section of track Tb2In all the points, if track Tb1Prolongation and Tb2Intersection, then recognize
For Tb1And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that Tb1With
Tb2With temporal continuity;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2Have
Consistency on direction.
Third step, if meeting all conditions in second step, T is thought in surveyb2Part after middle intersection point also belongs to Tb1, after merging
Obtain new T 'b1.
Specifically, it is described be overlapped judgement the specific implementation process is as follows:
There is Tb1∈TB, wherein Pb1={ xb1,yb1,tb1,vb1,db1,Δdb1It is Tb1End point;
According to Pb1Mobile trend, predict tLThe position P of track after timeb′1={ x 'b1,y′b1,t′b1, have:
x′b1=xb1+sin(db1)×vb1×tL
y′b1=yb1+cos(db1)×vb1×tL
t′b1=tb1+tL
There is Tb2∈TB, wherein Pb2i={ xb2j,yb2j,tb2j,vb2j,db2j,Δdb2jIt is Tb1Jth point
If having:
tb1<tb2j<t′b1
And there is line segment (Pb1,P′b1) and line segment (Pb2j,Pb2(j-1)) intersection, then it is assumed that Tb1And Tb2For the motion profile of coincidence
Sequence needs to split intersection,
There is new track T 'b1=Tb1+Tb2(i>j)
In step S3, the method for the separation judgement are as follows:
First, to Mr. Yu's section track Tb1, according to its starting point Pb1Mobile trend, i.e. vb1And db1, retrospectively calculate its in tL
Preceding position, i.e. P 'b1;
Second step compares other one section of track Tb2In all the points, if track Tb1Reverse extending part and Tb2Intersection,
Then think Tb1And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that
Tb1And Tb2With temporal continuity;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2
With the consistency on direction;
Third step, if meeting all conditions in second step, then it is assumed that Tb2Part before middle intersection point also belongs to Tb1, after merging
Obtain new T 'b1。
Specifically, it is described separation judgement the specific implementation process is as follows:
There is Tb1∈TB, wherein Pb1={ xb1,yb1,tb1,vb1,db1,Δdb1It is Tb1Starting point;
According to Pb1Mobile trend, predict tLThe position P ' of time preceding trackb1={ x 'b1,y′b1,t′b1, have:
x′b1=xb1-sin(db1)×vb1×tL
y′b1=yb1-cos(db1)×vb1×tL
t′b1=tb1-tL
There is Tb2∈TB, wherein Pb2i={ xb2j,yb2j,tb2j,vb2j,db2j,Δdb2jIt is Tb1Jth point
If having:
t′b1<tb2j<tb1
And there is line segment (P 'b1, Pb1) and line segment (Pb2j,Pb2(j+1)) intersection, then it is assumed that Tb1And Tb2For the motion profile of coincidence
Sequence needs to split intersection.
There is new track T 'b1=Tb1+Tb2(i≤j)。
By above-mentioned steps, new motion profile arrangement set T ' can be obtainedB。
Fig. 2 is a kind of system architecture diagram of the motion profile optimization system of non-characteristic matching of the present invention.As shown in Fig. 2, this
Invent a kind of motion profile optimization system of non-characteristic matching, comprising:
Window calculation unit 201, for calculating transverse field width FW according to viewing field of camera angle FOV and operating distance WD,
The matching frame per second FPS shielded than SL, mankind classics walking speed VR and motion track is accounted for according to mankind's moving distance, calculates Track View
Mouth N, and to motion profile arrangement set TR, with N frame be calculate width calculate each point of motion profile sequence movement speed and
Moving direction further calculates the moving direction change rate Δ D of sequencep。
Inflection point identifying processing unit 202, for calculating the shifting of the motion profile sequence obtained according to window calculation unit 201
Dynamic speed and moving direction change rate, determine the abnormal inflection point in track, are cut to motion profile sequence according to abnormal inflection point
Multistage motion profile sequence, and from the motion profile arrangement set T of Combination nova Cheng XinB。
Unit 203 is repaired in track, for according to new motion profile arrangement set TBBroken string judgement is carried out to motion profile,
It is overlapped judgement and separation judgement, the motion profile arrangement set after being repaired.
Fig. 3 is the structure chart of the motion profile optimization system of the non-characteristic matching of the specific embodiment of the invention.It includes window
Mouth computing unit (1-window calculator), inflection point identifying processing unit (2-Outlier recognizer) and rail
Mark repairs unit (3-Combiner).
In the specific embodiment of the invention, since there may be transient jitters for motion track, in the speed for calculating motion track
When degree and direction, need to obtain more stable data using certain calculating time window N.
In (1-window calculator), using known viewing field of camera angle FOV and the obtained camera of measurement to regarding
Lateral physical width, i.e. transverse field width FW is calculated in the visual field in the physical distance WD of wild central object.We need people
Work defined parameters window accounts for screen than SL, and usually 0.05, in this way, the physical width of calculation window in turn, the i.e. product of SL and FW.
In addition, it would be desirable to which defined parameters mankind classics walking speed VR is normally defined 4m/s.According to real in above data and video
The frame per second FPS of border tracking, can be calculated time window N.
In (2-Outlier recognizer), T can be judgedrWhether there is inflection point, the judgement of inflection point in each track
Need to calculate the movement speed v of each point in trackiWith the change rate Δ d of moving directioni.Movement speed viCalculating be to pass through
Target point i finds front with window NIt puts and subsequentPoint can calculate the moving distance in window, according to shifting
Movement speed v of the track at i-th point can be obtained in dynamic distance and time window Ni, further according to the coordinate put before and after window, calculate
Obtain moving direction d of the track at i-th pointi, and then calculate the change rate Δ d of moving directioni。
Wherein due to the beginning of track afterBefore point and endIt is unable to satisfy the requirement of window calculation, therefore thinks this portion
The movement speed and moving direction divided are constant.
The judgement of inflection point needs to judge whether the point is nonstatic point, because when the point is rest point, due to number
According to the shake of itself, lead to the change rate Δ d of moving directioniVariation is very big, influences to determine effect.
If there is movement speed viGreater than nonstatic speed VLWhen as nonstatic point, wherein VLFor can setting parameter, usually
1m/s;
Since people's moving direction is continuous, it is impossible to carry out wide-angle in the case where keeping certain movement speed
It turns round, therefore thinks the change rate Δ d of moving directioniGreater than DLNonstatic point be abnormal inflection point, usual inflection point is multiple mobile mesh
Mark is because being overlapped alternating etc. caused by reasons, therefore by track using inflection point as separation, motion track is divided into multistage track and from new
It is combined into a new set TBTo carry out the calculating of next step.
Include 3 repair modules in (3-Combiner), be disconnection judgment module respectively, is overlapped judgment module and separation
Judgment module.
Disconnection judgment module be in order to solve motion profile due to block or the factors such as missing inspection caused by with track fracture simultaneously
It is judged as two tracks.
Specifically, the judgment rule of disconnection judgment module is as follows:
First, compare a track Tb1End point Pb1With an other track Tb2Starting point Pb2If tb2Greater than tb1
And there is tb2And tb1Difference be less than tLThen think the continuity in two sections of track having times, wherein tLThe maximum tolerance lost for target
Time, for can setting parameter, usually 2s.
Second, if db1With | db2Difference be less than DL, it is believed that two sections of tracks have orientation consistency, wherein DLFor target
Maximum tolerance turn round limitation, for can setting parameter, usually
Third calculates Pb1And Pb2Distance, if Pb1And Pb2Distance be less than Pb1According to vb2In tLTime it is mobile away from
From, then it is assumed that two sections of curves have the continuity in space.
As met three above condition simultaneously, then it is assumed that two tracks are actually identical strip path curve.
Specific example such as Fig. 4, two sections of track then combinable reparations for meeting time, direction and Space Consistency simultaneously.
In addition special circumstances such as Fig. 7 causes the mistake of Fig. 7, but pass through since two movements are intersected when may tracking
Inflection point judgement, track is split after disconnection, then can obtain correct two tracks.
Be overlapped judgment module and be to solve motion profile due to starting target and separate, it is parallel later in the case where track it is disconnected
The problem of splitting, specific such as Fig. 5, deterministic process are as follows:
The first step has certain section of track Tb1, according to its end point Pb1Mobile trend, i.e. vb1And db1, it is calculated in tLAfterwards
Position, i.e. P 'b1。
Second step compares other one section of track Tb2In all the points, if track Tb1Prolongation and Tb2Intersection, then recognize
For Tb1And Tb2With continuity spatially, if comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that Tb1With
Tb2With temporal continuity, if comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2Have
Consistency on direction.
Third step, if meeting all conditions in second step, then it is assumed that Tb2Part after middle intersection point also belongs to Tb1, after merging
Obtain new T 'b1。
Separation judgment module is to solve motion profile due to starting target and be overlapped, and individual traces are endless after separating later
Whole problem is specific such as Fig. 6.Its deterministic process is as follows:
The first step has certain section of track Tb1, according to its starting point Pb1Mobile trend, i.e. vb1And db1, retrospectively calculate its in tL
Preceding position, i.e. P 'b1。
Second step compares other one section of track Tb2In all the points, if track Tb1Reverse extending part and Tb2Intersection,
Then think Tb1And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that
Tb1And Tb2With temporal continuity;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2
With the consistency on direction.
Third step, if meeting all conditions in second step, then it is assumed that Tb2Part before middle intersection point also belongs to Tb1, after merging
Obtain new T 'b1.
By three above module, then new motion profile arrangement set T ' can be obtainedB。
In conclusion the motion profile optimization method and system of a kind of non-characteristic matching of the present invention according to viewing field of camera angle and
Operating distance determines visual field width, and obtains trajectory calculation window width, Jin Erji by stream of people's conventional speeds and matching frame per second
Speed, direction and the direction change rate of track are calculated, to judge whether track has unreasonable inflection point, is then sentenced by broken string
It is disconnected, it is overlapped judgement and separation judges that three modules repair track set, obtain the track collection for meeting stream of people's movement law
It closes, the present invention repairs after the result of non-characteristic matching, partially meets the stream of people without carrying out largely calculating to repair
Movement law as a result, motion profile is optimized.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.Any
Without departing from the spirit and scope of the present invention, modifications and changes are made to the above embodiments by field technical staff.Therefore,
The scope of the present invention, should be as listed in the claims.
Claims (10)
1. a kind of motion profile optimization method of non-characteristic matching, includes the following steps:
Step S1 calculates transverse field width FW, according to mankind's moving distance according to viewing field of camera angle FOV and operating distance WD
The matching frame per second FPS of Zhan Ping ratio SL, mankind classics walking speed VR and motion track calculate track window N, to motion profile sequence
Arrange set TR, it is to calculate width to calculate the movement speed and moving direction of each point of motion profile sequence, and further count with N frame
Calculate the moving direction change rate Δ D of sequencep;
Step S2 determines the abnormal inflection point in track, root according to the movement speed of motion profile sequence and moving direction change rate
Motion profile sequence is cut to multistage motion profile sequence according to abnormal inflection point, and from the motion profile sequence sets of Combination nova Cheng Xin
Close TB;
Step S3, according to new motion profile arrangement set TBDisconnection judgement is carried out to motion profile, judgement is overlapped and separates and sentence
It is disconnected, the motion profile arrangement set after being repaired.
2. a kind of motion profile optimization method of non-characteristic matching as described in claim 1, which is characterized in that in step S1
In, the transverse field width FW and track window N calculation method are as follows:
FW=2 × WD × tan (FOV)
Wherein FOV is viewing field of camera angle, and WD is operating distance, and SL is that window accounts for screen ratio, and VR is mankind's classics walking speed, FPS
For the matching frame per second of motion track.
3. a kind of motion profile optimization method of non-characteristic matching as claimed in claim 2, which is characterized in that in step S1
In, to Mr. Yu motion profile sequence Tr∈TR, Tr={ { xi, yi, ti, i ∈ (0~M) }, wherein xiAnd yiFor motion profile point
Coordinate, tiFor the time of the motion profile point, M is sequence length, movement speed viWith moving direction diChange with moving direction
Rate Δ diCalculation method it is as follows:
Δdi=| da-db|
Wherein subscript a and b are respectively as follows:
4. a kind of motion profile optimization method of non-characteristic matching as claimed in claim 3, feature exist, in, in step S2,
The Rule of judgment of the exception inflection point are as follows:
Δdi> DL
vi> VL
Such as meet conditions above, then by motion profile sequence using inflection point as separation, single-row respectively is new motion profile sequence,
All new motion profile combined sequences are at a new motion profile arrangement set TB,
Above-mentioned DLChange angle, V for maximumLFor nonstatic speed.
5. a kind of motion profile optimization method of non-characteristic matching as claimed in claim 4, feature exists, in step S3
In, the method for the track disconnection judgement is as follows:
(1) track T is comparedb1End point Pb1With track Tb2Starting point Pb2If time tb2Greater than tb1And there is tb2And tb1Difference
Less than tLThen think the continuity in two sections of track having times, wherein tLThe maximum tolerance time lost for target;
(2) if moving direction db1And db2Difference be less than DL, it is believed that two sections of tracks have orientation consistency, wherein DLFor target
Maximum tolerance is turned round limitation, i.e., maximum variation angle, for can setting parameter, usually
(3) track T is calculatedb1End point Pb1With track Tb2Starting point Pb2Distance, if Pb1And Pb2Distance be less than Pb1
According to vb2In tLThe distance of time movement, then it is assumed that two sections of curves have the continuity in space,
If meeting three above condition simultaneously, then it is assumed that two tracks are identical strip path curve.
6. a kind of motion profile optimization method of non-characteristic matching as claimed in claim 4, feature exists, in the coincidence
The method of judgement is as follows:
The first step, to Mr. Yu's section track Tb1, according to its end point Pb1Mobile trend, i.e. vb1And db1, it is calculated in tLPosition afterwards
It sets, i.e. P 'b1, tLThe maximum tolerance time lost for target;
Second step compares other one section of track Tb2In all the points, if track Tb1Prolongation and Tb2Intersection, then it is assumed that Tb1
And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that Tb1And Tb2Tool
Continuity in having time;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2With on direction
Consistency;
Third step, if meeting all conditions in second step, T is thought in surveyb2Part after middle intersection point also belongs to Tb1, obtained after merging new
T 'b1。
7. a kind of motion profile optimization method of non-characteristic matching as claimed in claim 4, feature exists, in the separation
The method of judgement are as follows:
The first step, to Mr. Yu's section track Tb1, according to its starting point Pb1Mobile trend, i.e. vb1And db1, retrospectively calculate its in tLBefore
Position, i.e. P 'b1, tLThe maximum tolerance time lost for target;
Second step compares other one section of track Tb2In all the points, if track Tb1Reverse extending part and Tb2Intersection, then recognize
For Tb1And Tb2With continuity spatially;If comparing Tb2Intersection point time tb2And tb1Time difference be less than tL, then it is assumed that Tb1With
Tb2With temporal continuity;If comparing Tb2Intersection point direction db2And db1Differential seat angle be less than DL, then it is assumed that Tb1And Tb2Have
Consistency on direction;
Third step, if meeting all conditions in second step, then it is assumed that Tb2Part before middle intersection point also belongs to Tb1, obtained after merging new
T 'b1。
8. a kind of motion profile optimization system of non-characteristic matching, comprising:
Window calculation unit, for transverse field width FW being calculated, according to people according to viewing field of camera angle FOV and operating distance WD
Class moving distance accounts for the matching frame per second FPS shielded than SL, mankind classics walking speed VR and motion track, calculates track window N, right
Motion profile arrangement set TR, it is the movement speed and moving direction for calculating width and calculating each point of motion profile sequence with N frame,
And the moving direction change rate Δ D of the further sequence of calculationp;
Inflection point identifying processing unit determines track for the movement speed and moving direction change rate according to motion profile sequence
In abnormal inflection point, motion profile sequence is cut to by multistage motion profile sequence according to abnormal inflection point, and from Combination nova Cheng Xin
Motion profile arrangement set TB;
Unit is repaired in track, for according to new motion profile arrangement set TBDisconnection judgement is carried out to motion profile, is overlapped judgement
And separation judgement, the motion profile arrangement set after being repaired.
9. a kind of motion profile optimization system of non-characteristic matching as claimed in claim 8, which is characterized in that the laterally view
Field width degree FW and track window N calculation method are as follows:
FW=2 × WD × tan (FOV)
Wherein FOV is viewing field of camera angle, and WD is operating distance, and SL is that window accounts for screen ratio, and VR is mankind's classics walking speed, FPS
For the matching frame per second of motion track.
10. a kind of motion profile optimization system of non-characteristic matching as claimed in claim 9, which is characterized in that Mr. Yu's item
Motion profile sequence Tr∈TR, Tr={ { xi, yi, ti, i ∈ (0~M) }, wherein xiAnd yiFor the coordinate of motion profile point, tiFor
The time of the motion profile point, M are sequence length, movement speed viWith moving direction diWith moving direction change rate Δ diMeter
Calculation method is as follows:
Δdi=| da-db|
Wherein subscript a and b are respectively as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811250104.2A CN109523571B (en) | 2018-10-25 | 2018-10-25 | Non-feature matching motion trajectory optimization method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811250104.2A CN109523571B (en) | 2018-10-25 | 2018-10-25 | Non-feature matching motion trajectory optimization method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109523571A true CN109523571A (en) | 2019-03-26 |
CN109523571B CN109523571B (en) | 2020-11-17 |
Family
ID=65774093
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811250104.2A Active CN109523571B (en) | 2018-10-25 | 2018-10-25 | Non-feature matching motion trajectory optimization method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109523571B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110660095A (en) * | 2019-09-27 | 2020-01-07 | 中国科学院自动化研究所 | Visual SLAM (simultaneous localization and mapping) initialization method, system and device in dynamic environment |
CN111161305A (en) * | 2019-12-18 | 2020-05-15 | 任子行网络技术股份有限公司 | Intelligent unmanned aerial vehicle identification tracking method and system |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090082955A1 (en) * | 2007-05-15 | 2009-03-26 | Thales | Method of plotting a portion of trajectory of an aircraft comprising a circular arc of constant radius |
CN102663452A (en) * | 2012-04-14 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Suspicious act detecting method based on video analysis |
CN103510855A (en) * | 2013-10-17 | 2014-01-15 | 中国石油集团渤海钻探工程有限公司 | Horizontal well logging guiding track optimization method |
CN103927763A (en) * | 2014-03-24 | 2014-07-16 | 河海大学 | Identification processing method for multi-target tracking tracks of image sequences |
CN106652439A (en) * | 2015-10-28 | 2017-05-10 | 华为数字技术(苏州)有限公司 | System for determining traffic congestion coefficient and server and method thereof |
CN106780538A (en) * | 2016-12-30 | 2017-05-31 | 南京仁光电子科技有限公司 | A kind of method of error hiding during solution image trace |
CN106898034A (en) * | 2017-02-28 | 2017-06-27 | 北京润科通用技术有限公司 | A kind of target trajectory analogy method and device |
-
2018
- 2018-10-25 CN CN201811250104.2A patent/CN109523571B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090082955A1 (en) * | 2007-05-15 | 2009-03-26 | Thales | Method of plotting a portion of trajectory of an aircraft comprising a circular arc of constant radius |
CN102663452A (en) * | 2012-04-14 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Suspicious act detecting method based on video analysis |
CN103510855A (en) * | 2013-10-17 | 2014-01-15 | 中国石油集团渤海钻探工程有限公司 | Horizontal well logging guiding track optimization method |
CN103927763A (en) * | 2014-03-24 | 2014-07-16 | 河海大学 | Identification processing method for multi-target tracking tracks of image sequences |
CN106652439A (en) * | 2015-10-28 | 2017-05-10 | 华为数字技术(苏州)有限公司 | System for determining traffic congestion coefficient and server and method thereof |
CN106780538A (en) * | 2016-12-30 | 2017-05-31 | 南京仁光电子科技有限公司 | A kind of method of error hiding during solution image trace |
CN106898034A (en) * | 2017-02-28 | 2017-06-27 | 北京润科通用技术有限公司 | A kind of target trajectory analogy method and device |
Non-Patent Citations (3)
Title |
---|
晏细兰,石坤泉: "一种用于人流统计的目标跟踪优化系统", 《现代信息科技》 * |
王江峰: "基于轨迹片段关联的目标跟踪与事件检测方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
袁冠: "移动对象轨迹数据挖掘方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110660095A (en) * | 2019-09-27 | 2020-01-07 | 中国科学院自动化研究所 | Visual SLAM (simultaneous localization and mapping) initialization method, system and device in dynamic environment |
CN110660095B (en) * | 2019-09-27 | 2022-03-25 | 中国科学院自动化研究所 | Visual SLAM (simultaneous localization and mapping) initialization method, system and device in dynamic environment |
CN111161305A (en) * | 2019-12-18 | 2020-05-15 | 任子行网络技术股份有限公司 | Intelligent unmanned aerial vehicle identification tracking method and system |
Also Published As
Publication number | Publication date |
---|---|
CN109523571B (en) | 2020-11-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106780557A (en) | A kind of motion target tracking method based on optical flow method and crucial point feature | |
CN105069408A (en) | Video portrait tracking method based on human face identification in complex scenario | |
CN108133172B (en) | Method for classifying moving objects in video and method and device for analyzing traffic flow | |
CN102222346B (en) | Vehicle detecting and tracking method | |
CN113221677B (en) | Track abnormality detection method and device, road side equipment and cloud control platform | |
CN107506702A (en) | Human face recognition model training and test system and method based on multi-angle | |
CN113032502A (en) | Ship anomaly detection method based on improved track segment DBSCAN clustering | |
WO2017035663A1 (en) | System and method for detecting and tracking objects | |
CN104933710A (en) | Intelligent analysis method of store people stream track on the basis of surveillance video | |
CN104504897A (en) | Intersection traffic flow characteristic analysis and vehicle moving prediction method based on trajectory data | |
CN108764167A (en) | A kind of target of space time correlation recognition methods and system again | |
CN103235933A (en) | Vehicle abnormal behavior detection method based on Hidden Markov Model | |
CN103164706A (en) | Object counting method and device based on video signal analysis | |
CN102254428B (en) | Traffic jam detection method based on video processing | |
CN105513349A (en) | Double-perspective learning-based mountainous area highway vehicle event detection method | |
CN103577875A (en) | CAD (computer-aided design) people counting method based on FAST (features from accelerated segment test) | |
CN109523571A (en) | A kind of the motion profile optimization method and system of non-characteristic matching | |
CN113936465B (en) | Traffic event detection method and device | |
CN106682573A (en) | Pedestrian tracking method of single camera | |
CN110046212B (en) | Road condition change information determining method and device, computer equipment and storage medium | |
Meng et al. | Video‐Based Vehicle Counting for Expressway: A Novel Approach Based on Vehicle Detection and Correlation‐Matched Tracking Using Image Data from PTZ Cameras | |
CN109711267A (en) | A kind of pedestrian identifies again, pedestrian movement's orbit generation method and device | |
CN108898612A (en) | Multi-object tracking method based on the enhancing study of multiple agent depth | |
CN108764338A (en) | A kind of pedestrian tracking algorithm applied to video analysis | |
Zhang et al. | Bidirectional multiple object tracking based on trajectory criteria in satellite videos |
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 | ||
EE01 | Entry into force of recordation of patent licensing contract | ||
EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20190326 Assignee: Guangzhou Jingyu Electronic Information Technology Co.,Ltd. Assignor: GUANGZHOU PANYU POLYTECHNIC Contract record no.: X2023980033739 Denomination of invention: A non feature matching method and system for motion trajectory optimization Granted publication date: 20201117 License type: Common License Record date: 20230317 |