CN102253391B - Multi-laser-radar-based pedestrian target tracking method - Google Patents
Multi-laser-radar-based pedestrian target tracking method Download PDFInfo
- Publication number
- CN102253391B CN102253391B CN2011101071373A CN201110107137A CN102253391B CN 102253391 B CN102253391 B CN 102253391B CN 2011101071373 A CN2011101071373 A CN 2011101071373A CN 201110107137 A CN201110107137 A CN 201110107137A CN 102253391 B CN102253391 B CN 102253391B
- Authority
- CN
- China
- Prior art keywords
- radar
- pedestrian
- point
- representes
- feet
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000033001 locomotion Effects 0.000 claims abstract description 34
- 238000001914 filtration Methods 0.000 claims abstract description 24
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 230000001133 acceleration Effects 0.000 claims description 36
- 230000009471 action Effects 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 15
- 230000004927 fusion Effects 0.000 claims description 13
- 230000007704 transition Effects 0.000 claims description 12
- 210000001699 lower leg Anatomy 0.000 claims description 9
- 230000008878 coupling Effects 0.000 claims description 8
- 238000010168 coupling process Methods 0.000 claims description 8
- 238000005859 coupling reaction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 5
- 230000003287 optical effect Effects 0.000 claims description 5
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000011160 research Methods 0.000 claims description 3
- 230000003068 static effect Effects 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 230000009897 systematic effect Effects 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000005286 illumination Methods 0.000 abstract description 3
- 230000007547 defect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Landscapes
- Radar Systems Or Details Thereof (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a multi-laser-radar-based pedestrian target tracking method. The multi-laser-radar-based pedestrian target tracking method comprises the following steps that: a sensor network consisting of a plurality of single-line laser radars monitors a whole scene; at every sampling moment, the radars output footstep point distance data of pedestrians, and the data among different radars are unified at a coordinate system through system calibration; a footstep movement model of the pedestrians is built and pedestrian movement states are tracked through Kalman filtering so as to finally acquire a stable target trace and relevant statistical information. In the method, the laser radars are used as sensors, so that the defects of influence of illumination variance on an image sensor in the tracking method can be overcome and reliability is enhanced greatly; furthermore, the tracking range is wide, the algorithm speed is high, the number of tracked pedestrians is large, all movement modes of the pedestrians are compatible, such as standing, walking and jogging, and positions of the pedestrians can be detected effectively and movement traces of the pedestrians can be tracked in real time. The multi-laser-radar-based pedestrian target tracking method is applicable to occasions in which main movement targets are pedestrians.
Description
Technical field
The present invention relates to a kind of pedestrian's method for tracking target, especially relate to a kind of pedestrian's method for tracking target based on multilasered optical radar.
The present invention relates to a kind of pedestrian's method for tracking target, be applicable to and in various public arenas, realize a plurality of pedestrians are being stood, walking, the tracking under the motor pattern such as jog based on multilasered optical radar.
Background technology
At subway, the station, in the crowded public arena such as department store, statistical informations such as pedestrian's flow, direction of motion can help engineering design side to carry out more reasonably gateway layout, and safe emergency prediction schemes reasonable in design etc. have huge commercial value.Want to obtain automatically these real-time statistics information, just need follow the tracks of accurately all pedestrian's targets in the scene.Present target following technology is followed the tracks of the moving target in the image through certain image processing algorithm mainly based on vision sensors such as video cameras.Yet the tracking technique based on image processing algorithm is all relatively responsive to the variation of ambient lighting, can't adapt to the requirement of round-the-clock tracking.The visual field of a video camera covering is limited simultaneously, though can lean on a plurality of video cameras to realize covering on a large scale, the target following of striding camera realizes difficulty, and is still at the experimental stage.By contrast, laser radar can directly obtain the range information of target, and does not receive the influence of illumination variation.Extensive gradually along with what use, price also progressively reduces.
Summary of the invention
The object of the present invention is to provide a kind of pedestrian's method for tracking target based on multilasered optical radar; Not only following range is big, and accuracy is high, and algorithm speed is fast; And the compatible pedestrian's of ability various motor patterns; Comprise and standing that walking and jog etc. can very detect pedestrian's position and real-time follow-up pedestrian's movement locus effectively.
The technical scheme that the present invention adopts is may further comprise the steps:
(1) setting of radar site and system calibrating;
(2) data acquisition and fusion;
(3) pairing of the cluster in motor point and step point;
(4) based on the target following of manikin and Kalman filtering.
The setting and the system calibrating of said step (1) radar site: be about to the diverse location that a plurality of radars are arranged on scene, the position choice principle is set is: (a) make any position of target in scene all be not easy to be blocked; (b) evenly arrange laser radar at the scene edge, with the possibility that the raising target is detected, radar scanning plan range floor level 23~25cm, this guarantees that highly radar in most of the cases can both detect and obtain the range data of two pin of pedestrian; Demarcation to a plurality of radars realizes through following method: some bar-shaped objects are set in scene, and the coordinate of establishing i the bar-shaped object that from N radar, obtains is:
And suppose that the coordinate with the 1st radar is a reference coordinates, on two dimensional surface, establish the transition matrix R of 1 radar fix of a n radar fix to the
nWith translational movement T
nFor:
Then obtain i o'clock transfer equation from 1 radar fix of a n radar fix to the:
Transition matrix and translational movement are unique when i>3 confirms, through demarcating data-switching that different radar fixs system is obtained down under same radar fix, in the formula: x representes abscissa value a little, and y representes ordinate value a little.
(2) data acquisition of said step and fusion: promptly through on data transmission to a main frame of hub with a plurality of radars collections, the time of while record acquisition; Data acquisition between a plurality of radars is by unified outer synchronous triggering; Extract earlier the background data of each radar automatically through the data of gathering the above frame number of 50 frames, the method for distilling of background data has adopted histogram method, promptly adds up in institute's acquisition frame the range information of data point on each direction; And do statistics with histogram; The frequency of occurrences is the highest is the background dot data on this direction, preserves this background data, and range data on all directions that then the same time collected and background data are relatively; If then think background dot less than the threshold quantity that sets and remove apart from difference, the foreground point that remaining promptly is in the scene is the motor point; Coordinate conversion is carried out through calibrating parameters in the foreground point that each radar of synchronization obtains, promptly obtain the exercise data point after this fusion that collects constantly under the unified global coordinate system.
The cluster in said step (3) motor point and the pairing of step point: the exercise data point that promptly fusion is obtained; The phase mutual edge distance is carried out cluster less than the point of first distance threshold; And with the position of type center of interior point as candidate's step; Then the step candidate point is carried out the coupling on time domain and the spatial domain, to confirm whether certain two candidate's step point belongs to same pedestrian, and two kinds of situation are arranged: near (a) search matched two step point prediction positions of the existing target that Kalman filtering obtains; If two candidate's step points in this frame respectively with two predicted positions less than the second distance threshold value, judge that then two candidate's step points are the observation station that has target; (b) if current candidate's step point fails in existing target, to find coupling, then belong to fresh target or noise point; The step candidate point that in present frame, does not mate with the search of the 3rd distance threshold then is if find then with match point associating becoming together two the step points that belong to same fresh target that find.
Said step (4) is based on the target following of modelling of human body motion and Kalman filtering: promptly the pedestrian has been set up motion model, and adopted Kalman filtering to carry out target following according to this model:
1) foundation of pedestrian dummy
Research shows that pedestrian's state is divided into: stand, walking, jog, the pedestrian always has a feet in the walking and the motion process of jogging; A swing pin; And both periodically exchange, and feet was always not liftoff when different was walking, and remained static; And feet can be liftoff and move distance can be arranged in the cycle when jogging, based on above mention stand, the modeling of walking, the three kinds of pedestrian's states of jogging respectively as follows:
The pattern of standing: do not have the driving of physiological action power, the speed of left and right sides pin is zero, and acceleration also is zero; Force=0, v
L, k=0, v
R, k=0, a
L, k=0, a
R, k=0, (1)
Walking and jogging pattern: feet has an invariable speed; The swing pin has an invariable physiological action power; Produce the constant physiological action power of constant magnitude of a forward and negative sense in preceding semiperiod and later half cycle respectively, thereby the motion pin is formed acceleration;
With T is the time interval between two frames, and right crus of diaphragm is in the preceding semiperiod of feet:
force=F,v
L,k=v
L,k-1+A*t,v
R,k=V,a
L,k=A,a
R,k=0,(2)
Right crus of diaphragm is in the later half cycle of feet:
force=-F,v
L,k=v
L,k-1+A*t,v
R,k=V,a
L,k=-A,a
R,k=0,(3)
Left foot is in the preceding semiperiod of feet:
force=F,v
L,k=V,v
R,k=v
R,k-1+A*t?a
L,k=0,a
R,k=A,(4)
Left foot is in the later half cycle of feet:
force=-F,v
L,k=V,v
R,k=v
R,k-1+A*t?a
L,k=0,a
R,k=-A,(5)
Force representes physiological action power in the formula, and F representes the physiological action power of the actual generation of pedestrian, and v representes movement velocity; V representes the speed of generation under the pedestrian F acting force, and a representes acceleration, and A representes the acceleration that the pedestrian produces under the F acting force; L representes left foot, and R representes right crus of diaphragm, and k representes k constantly;
When V=0, the pedestrian promptly is under the walking state, and when V>0, the pedestrian promptly is under the state of jogging;
2) Kalman filtering is followed the tracks of
Be used as state variable to pedestrian's position P and speed V, acceleration A is used as the input quantity of system, and the state equation that then obtains system is:
Wherein S representes state variable, and i representes pedestrian's mark, and Ω representes the state model error;
Represent state-transition matrix with Φ, Ψ representes the transition matrix of acceleration to speed and displacement conversion, and when the radar scanning frequency was 37.5Hz, T was 0.026 second; Then have:
The systematic observation equation is following:
Wherein M is an observation vector, and H is the transition matrix of S to M, and ζ is an observational error;
Equation (10) and equation (11) are the state equation and the observation equation of Kalman filtering; Employing standard Kalman Filter Technology comes state vector is carried out the continuous real-time optimal estimation, obtains the estimated position and the variance of each pedestrian's target under each sampling instant;
3) renewal of speed and acceleration
To convert feet into be one-period to definition swing pin from beginning to swing to; In one-period; The speed of swing pin and the speed of feet are upgraded by Kalman filtering automatically; The then every one-period of crossing of the acceleration of swing pin upgrades once; Upgrade with the average acceleration value in the last cycle, as the acceleration input quantity in next cycle, direction is all the time by swinging toe to feet; The distance of swing pin motion is s in the assumption period, and the time is t, and then more new formula is following for acceleration:
The beneficial effect that has of the present invention is:
The present invention is a sensor with the laser radar, has overcome the shortcoming that is subject to the illumination variation influence that exists in the tracking of imageing sensor, and reliability is greatly improved; Not only following range is big, and algorithm speed is fast, and it is many to follow the tracks of quantity, and various motor patterns that can compatible pedestrian, comprises standing, walking, jogging, and can very detect pedestrian's position and real-time follow-up pedestrian's movement locus effectively.The present invention is applicable to that the main moving target in the scene is pedestrian's a occasion.
Description of drawings
Fig. 1 is that radar system is arranged synoptic diagram.
Fig. 2 is pedestrian's motion model synoptic diagram.
Fig. 3 is pedestrian's target tracking algorism process flow diagram.
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is further described.
Embodiment of the present invention mainly may further comprise the steps:
(1) setting of radar site and system calibrating;
(2) data acquisition and fusion;
(3) pairing of the cluster in motor point and step point;
(4) based on the target following of manikin and Kalman filtering.
The setting and the system calibrating of said step (1) radar site: be about to the diverse location that a plurality of radars are arranged on scene, as shown in Figure 1, through hub with the data transmission of gathering on the different radars to same computing machine, carry out data processing.The position choice principle is set is: (a) make any position of target in scene all be not easy to be blocked; (b) evenly arrange laser radar at the scene edge; To improve the possibility that target is detected; Radar scanning plan range floor level 23~25cm; This guarantees that highly radar in most of the cases can both detect and obtain the range data of two pin of pedestrian, and this excessive height then laser is blocked by the entrained object of pedestrian easily, crosses the low step range data that obtains under the running state of then can't stablizing.Demarcation to a plurality of radars realizes through following method: some bar-shaped objects are set in scene, and the coordinate of establishing i the bar-shaped object that from N radar, obtains is:
And suppose that the coordinate with the 1st radar is a reference coordinates, on two dimensional surface, establish the transition matrix R of 1 radar fix of a n radar fix to the
nWith translational movement T
nFor:
Then obtain i o'clock transfer equation from 1 radar fix of a n radar fix to the:
Transition matrix and translational movement are unique when i>3 confirms, through demarcating data-switching that different radar fixs system is obtained down under same radar fix, in the formula: x representes abscissa value a little, and y representes ordinate value a little.
(2) data acquisition of said step and fusion: promptly through on data transmission to a main frame of hub with a plurality of radars collections, the time of record acquisition is as shown in Figure 1 simultaneously; Data acquisition between a plurality of radars is by unified outer synchronous triggering.During the experiment beginning, through the first background data that extracts each radar automatically of the data of gathering the above frame number of 50 frames, the method for distilling of background data has adopted histogram method; Promptly add up in institute's acquisition frame the range information of data point on each direction; And do statistics with histogram, the frequency of occurrences is the highest is the background dot data on this direction, preserves this background data; Continue experiment then; Range data on all directions that the same time is collected and background data relatively, if then think background dot less than the threshold quantity that sets and remove apart from difference, the foreground point that remaining promptly is in the scene is the motor point; Coordinate conversion is carried out through calibrating parameters R and T in the foreground point that each radar of synchronization obtains, promptly obtain the exercise data point after this fusion that collects constantly under the unified global coordinate system.
The cluster in said step (3) motor point and the pairing of step point: the exercise data point that promptly fusion is obtained; The phase mutual edge distance is carried out cluster less than the point of first distance threshold; And with the position of type center of interior point as candidate's step; As shown in Figure 1, radar scans often more than one of the point that obtains on a pin, and these methods through cluster are found out the step that central point is scanned with representative.Then the step candidate point is carried out the coupling on time domain and the spatial domain, the time domain coupling is meant that with the candidate point in the present frame be the candidate point in the preceding frame of search center search; Promptly search can the matched candidate point in present frame for spatial domain coupling; To confirm whether certain two candidate's step point belongs to same pedestrian; Two kinds of situation are arranged: near (a) search matched two step point prediction positions of the existing target that Kalman filtering obtains; If two candidate's step points in this frame respectively with two predicted positions distances less than the second distance threshold value, judge that then two candidate's step points are the observation station that has target; (b) if current candidate's step point fails in existing target, to find coupling, then belong to fresh target or noise point; The step candidate point that in present frame, does not mate with the search of the 3rd distance threshold then is if find then with match point associating becoming together two the step points that belong to same fresh target that find.
Said step (4) is based on the target following of modelling of human body motion and Kalman filtering: promptly the pedestrian is set up motion model, and adopt Kalman filtering to carry out target following according to this model:
1) foundation of pedestrian dummy
Research shows that pedestrian's state is divided into: stand, walking, jog, hurry up, jogging is meant velocity less than the 3m/ running state of second, because situation is very complicated when hurrying up (velocity is greater than the 3m/ running of second), more difficult modeling puts aside here; The pedestrian always has a feet in the walking and the motion process of jogging, a swing pin, and both periodically exchange, and feet was always not liftoff when different was walking, and remained static; And feet can be liftoff and move distance can be arranged in the cycle when jogging, based on above mention stand, the modeling of walking, the three kinds of pedestrian's states of jogging respectively as follows:
The pattern of standing: do not have the driving of physiological action power, the speed of left and right sides pin is zero, and acceleration also is zero; Force=0, v
L, k=0, v
R, k=0, a
L, k=0, a
R, k=0, (1)
Walking and jogging pattern: feet has an invariable speed; The swing pin has an invariable physiological action power; Produce the constant physiological action power of constant magnitude of a forward and negative sense in preceding semiperiod (the swing pin is near the process of feet) and later half cycle (the swing pin is away from the process of feet) respectively, thereby the motion pin is formed acceleration;
With T is the time interval between two frames, and right crus of diaphragm is in the preceding semiperiod of feet:
force=F,v
L,k=v
L,k-1+A*t,v
R,k=V,a
L,k=A,a
R,k=0,(2)
Right crus of diaphragm is in the later half cycle of feet:
force=-F,v
L,k=v
L,k-1+A*t,v
R,k=V,a
L,k=-A,a
R,k=0,(3)
Left foot is in the preceding semiperiod of feet:
force=F,v
L,k=V,v
R,k=v
R,k-1+A*t?a
L,k=0,a
R,k=A,(4)
Left foot is in the later half cycle of feet:
force=-F,v
L,k=V,v
R,k=v
R,k-1+A*t?a
L,k=0,a
R,k=-A,(5)
Force representes physiological action power in the formula, and F representes the physiological action power of the actual generation of pedestrian, and v representes movement velocity; V representes the speed of generation under the pedestrian F acting force, and a representes acceleration, and A representes the acceleration that the pedestrian produces under the F acting force; L representes left foot, and R representes right crus of diaphragm, and k representes k constantly;
When V=0, the pedestrian promptly is under the walking state, and when V>0, the pedestrian promptly is under the state of jogging; As shown in Figure 2, the acceleration scalar value has obvious periodic to change consistently in the motion process, the direction of acceleration then all the time by the swing toe to feet, the speed of feet then is at the uniform velocity to increase progressively or at the uniform velocity successively decrease.
2) Kalman filtering is followed the tracks of
Be used as state variable to pedestrian's position P and speed V, acceleration A is used as the input quantity of system, and the state equation that then obtains system is:
Wherein S representes state variable, and i representes pedestrian's mark, and Ω representes the state model error;
Represent state-transition matrix with Φ, Ψ representes the transition matrix of acceleration to speed and displacement conversion, and when the radar scanning frequency was 37.5Hz, T was 0.026 second; Then have:
The systematic observation equation is following:
Wherein M is an observation vector, and H is the transition matrix of S to M, and ζ is an observational error;
Equation (10) and equation (11) are the state equation and the observation equation of Kalman filtering; Employing standard Kalman Filter Technology comes state vector is carried out the continuous real-time optimal estimation, obtains the estimated position and the variance of each pedestrian's target under each sampling instant;
3) renewal of speed and acceleration
To convert feet into be one-period to definition swing pin from beginning to swing to; In one-period; The speed of swing pin and the speed of feet are upgraded by Kalman filtering automatically; The then every one-period of crossing of the acceleration of swing pin upgrades once; Upgrade with the average acceleration value in the last cycle, as the acceleration input quantity in next cycle, direction is all the time by swinging toe to feet; The distance of swing pin motion is s in the assumption period, and the time is t, and then more new formula is following for acceleration:
As shown in Figure 3; After each frame data is through fusion and cluster; Just begin existing tracking target (being recorded in the variable) is carried out model prediction; At first with the predicted position of existing tracking target be search center present frame with the step candidate point in certain this frame of threshold range search as observation station, successfully promptly found the observation station of tracking target in this frame if search for, then carry out filtering; Obtain the current location of filtered, simultaneously renewal speed, acceleration (if one-period just finishes) as tracking target; Promptly do not find the step candidate point as observation station if follow the tracks of failure; Then with the current location of model prediction result as tracking target; And this type of tracking target is labeled as lose follows the tracks of, count, when losing when following the tracks of frame number and reaching 20 frames; We think that this tracking target follows the tracks of failure or left target place, stop to follow the tracks of.Step candidate point for the observation station that is not identified as existing tracking target; We think that it possibly be new pedestrian or noise; We adopt simple means of tracking under the situation that can not confirm its authenticity at once, and these points are carried out the tracking in certain frame number.Because the radar scanning frequency is high, the move distance between the consecutive point is less, is that setting each similar point constantly of appropriate threshold search in center just can find new candidate point so need only with the candidate point; If do not find, then this candidate point is a noise probably, follows the tracks of if found then continued; Carry out the Kalman filtering tracking after determining that it is real pedestrian; So repeat above step, all leave target place up to all tracking targets, this experiment finishes.
Claims (1)
1. pedestrian's method for tracking target based on multilasered optical radar may further comprise the steps:
(1) setting of radar site and system calibrating;
(2) data acquisition and fusion;
(3) pairing of the cluster in motor point and step point;
(4) based on the target following of manikin and Kalman filtering;
It is characterized in that:
The setting and the system calibrating of said step (1) radar site: be about to the diverse location that a plurality of radars are arranged on scene, the position choice principle is set is: (a) make any position of target in scene all be not easy to be blocked; (b) evenly arrange laser radar at the scene edge, with the possibility that the raising target is detected, radar scanning plan range floor level 23~25cm, this guarantees that highly radar in most of the cases can both detect and obtain the range data of two pin of pedestrian; Demarcation to a plurality of radars realizes through following method: some bar-shaped objects are set in scene, and the coordinate of establishing i the bar-shaped object that from N radar, obtains is:
And suppose that the coordinate with the 1st radar is a reference coordinates, on two dimensional surface, establish the transition matrix R of 1 radar fix of a n radar fix to the
nWith translational movement T
nFor:
Then obtain i o'clock transfer equation from 1 radar fix of a n radar fix to the:
Transition matrix and translational movement are unique when i>3 confirms, through demarcating data-switching that different radar fixs system is obtained down under same radar fix, in the formula: x representes abscissa value a little, and y representes ordinate value a little;
(2) data acquisition of said step and fusion: promptly through on data transmission to a main frame of hub with a plurality of radars collections, the time of while record acquisition; Data acquisition between a plurality of radars is by unified outer synchronous triggering; Extract earlier the background data of each radar automatically through the data of gathering the above frame number of 50 frames, the method for distilling of background data has adopted histogram method, promptly adds up in institute's acquisition frame the range information of data point on each direction; And do statistics with histogram; The frequency of occurrences is the highest is the background data on this direction, preserves this background data, and range data on all directions that then the same time collected and background data are relatively; If then think background dot less than the threshold quantity that sets and remove apart from difference, the foreground point that remaining promptly is in the scene is the motor point; Coordinate conversion is carried out through calibrating parameters in the foreground point that each radar of synchronization obtains, promptly obtain the exercise data point after this fusion that collects constantly under the unified global coordinate system;
The cluster in said step (3) motor point and the pairing of step point: the exercise data point that promptly fusion is obtained; The phase mutual edge distance is carried out cluster less than the point of first distance threshold; And with the position of type center of interior point as candidate's step; Then the step candidate point is carried out the coupling on time domain and the spatial domain, to confirm whether certain two candidate's step point belongs to same pedestrian, and two kinds of situation are arranged: near (a) search matched two step point prediction positions of the existing target that Kalman filtering obtains; If two candidate's step points in this frame respectively with two predicted positions less than the second distance threshold value, judge that then two candidate's step points are the observation station that has target; (b) if current candidate's step point fails in existing target, to find coupling, then belong to fresh target or noise point; The step candidate point that in present frame, does not mate with the search of the 3rd distance threshold then is if find then with match point associating becoming together two the step points that belong to same fresh target that find;
Said step (4) is based on the target following of modelling of human body motion and Kalman filtering: promptly the pedestrian has been set up motion model, and adopted Kalman filtering to carry out target following according to this model:
5.1) foundation of pedestrian dummy
Research shows that pedestrian's state is divided into: stand, walking, jog, the pedestrian always has a feet in the walking and the motion process of jogging; A swing pin; And both periodically exchange, and feet was always not liftoff when different was walking, and remained static; And feet can be liftoff and move distance can be arranged in the cycle when jogging, based on above mention stand, the modeling of walking, the three kinds of pedestrian's states of jogging respectively as follows:
The pattern of standing: do not have the driving of physiological action power, the speed of left and right sides pin is zero, and acceleration also is zero;
force=0,v
L,k=0,v
R,k=0,a
L,k=0,a
R,k=0,(1)
Walking and jogging pattern: feet has an invariable speed; The swing pin has an invariable physiological action power; Produce the constant physiological action power of constant magnitude of a forward and negative sense in preceding semiperiod and later half cycle respectively, thereby the motion pin is formed acceleration;
With T is the time interval between two frames, and right crus of diaphragm is in the preceding semiperiod of feet:
force=F,v
L,k=v
L,k-1+A*t,v
R,k=V,a
L,k=A,a
R,k=0,(2)
Right crus of diaphragm is in the later half cycle of feet:
force=-F,v
L,k=v
L,k-1-A*t,v
R,k=V,a
L,k=-A,a
R,k=0,(3)
Left foot is in the preceding semiperiod of feet:
force=F,v
L,k=V,v
R,k=v
R,k-1+A*t a
L,k=0,a
R,k=A,(4)
Left foot is in the later half cycle of feet:
force=-F,v
L,k=V,v
R,k=v
R,k-1-A*t,a
L,k=0,a
R,k=-A,(5)
Force representes physiological action power in the formula, and F representes the physiological action power of the actual generation of pedestrian, and v representes movement velocity; V representes the speed of generation under the pedestrian F acting force, and a representes acceleration, and A representes the acceleration that the pedestrian produces under the F acting force; L representes left foot, and R representes right crus of diaphragm, and k representes k constantly;
When V=0, the pedestrian promptly is under the walking state, and when V>0, the pedestrian promptly is under the state of jogging;
5.2) the Kalman filtering tracking
Be used as state variable to pedestrian's position P and speed V, acceleration A is used as the input quantity of system, and the state equation that then obtains system is:
Wherein S representes state variable, and i representes pedestrian's mark, and Ω representes the state model error;
Represent state-transition matrix with Φ, Ψ representes the transition matrix of acceleration to speed and displacement conversion, and when the radar scanning frequency was 37.5Hz, T was 0.026 second; Then have:
The systematic observation equation is following:
Wherein M is an observation vector, and H is the transition matrix of S to M, and ζ is an observational error;
Equation (10) and equation (11) are the state equation and the observation equation of Kalman filtering; Employing standard Kalman Filter Technology comes state vector is carried out the continuous real-time optimal estimation, obtains the estimated position and the variance of each pedestrian's target under each sampling instant;
5.3) renewal of speed and acceleration
To convert feet into be one-period to definition swing pin from beginning to swing to; In one-period; The speed of swing pin and the speed of feet are upgraded by Kalman filtering automatically; The then every one-period of crossing of the acceleration of swing pin upgrades once; Upgrade with the average acceleration value in the last cycle, as the acceleration input quantity in next cycle, direction is all the time by swinging toe to feet; The distance of swing pin motion is s in the assumption period, and the time is t, and then more new formula is following for acceleration:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011101071373A CN102253391B (en) | 2011-04-19 | 2011-04-19 | Multi-laser-radar-based pedestrian target tracking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011101071373A CN102253391B (en) | 2011-04-19 | 2011-04-19 | Multi-laser-radar-based pedestrian target tracking method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102253391A CN102253391A (en) | 2011-11-23 |
CN102253391B true CN102253391B (en) | 2012-11-28 |
Family
ID=44980756
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2011101071373A Expired - Fee Related CN102253391B (en) | 2011-04-19 | 2011-04-19 | Multi-laser-radar-based pedestrian target tracking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102253391B (en) |
Families Citing this family (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104063740B (en) * | 2013-03-21 | 2017-11-17 | 日电(中国)有限公司 | Office's group of entities identifying system, method and device |
WO2014172708A1 (en) | 2013-04-19 | 2014-10-23 | Polaris Sensor Technologies, Inc. | Pedestrian right of way monitoring and reporting system and method |
CN103325115B (en) * | 2013-06-14 | 2016-08-10 | 上海交通大学 | A kind of method of monitoring people counting based on overhead camera head |
CN104915628B (en) * | 2014-03-14 | 2018-09-25 | 株式会社理光 | The method and apparatus that scene modeling based on in-vehicle camera carries out movement pedestrian's prediction |
CN105652895A (en) * | 2014-11-12 | 2016-06-08 | 沈阳新松机器人自动化股份有限公司 | Mobile robot human body tracking system and tracking method based on laser sensor |
CN105182283B (en) * | 2015-08-17 | 2017-08-25 | 周口师范学院 | A kind of passive radar fixes target time domain direction-finding method |
CN105866782B (en) * | 2016-04-04 | 2018-08-17 | 上海大学 | A kind of moving object detection system and method based on laser radar |
JP6929028B2 (en) * | 2016-07-29 | 2021-09-01 | キヤノン株式会社 | Devices, methods and programs that detect people using range-finding sensors |
CN106468772B (en) * | 2016-09-23 | 2018-10-23 | 南京特艺科技有限公司 | A kind of multistation radar human body tracing method measured based on distance-Doppler |
CN109343049B (en) * | 2017-11-10 | 2022-04-26 | 毫末智行科技有限公司 | Method and apparatus for tracking movable target |
CN108345004A (en) * | 2018-02-09 | 2018-07-31 | 弗徕威智能机器人科技(上海)有限公司 | A kind of human body follower method of mobile robot |
CN108596117B (en) * | 2018-04-28 | 2021-03-02 | 河北工业大学 | Scene monitoring method based on two-dimensional laser range finder array |
WO2020015748A1 (en) * | 2018-07-20 | 2020-01-23 | Suteng Innovation Technology Co., Ltd. | Systems and methods for lidar detection |
CN109581302B (en) * | 2018-12-12 | 2021-04-16 | 北京润科通用技术有限公司 | Vehicle-mounted radar data tracking method and system |
CN109870692B (en) * | 2019-04-16 | 2023-10-20 | 浙江力邦合信智能制动***股份有限公司 | Radar looking around system and data processing method |
CN111048208A (en) * | 2019-12-28 | 2020-04-21 | 哈尔滨工业大学(威海) | Indoor solitary old man walking health detection method based on laser radar |
CN111476822B (en) * | 2020-04-08 | 2023-04-18 | 浙江大学 | Laser radar target detection and motion tracking method based on scene flow |
CN112099620A (en) * | 2020-08-11 | 2020-12-18 | 中国人民解放军军事科学院国防科技创新研究院 | Combat collaboration system and method for soldier and team combat |
CN112084372B (en) * | 2020-09-14 | 2024-01-26 | 北京数衍科技有限公司 | Pedestrian track updating method and device |
CN112684430A (en) * | 2020-12-23 | 2021-04-20 | 哈尔滨工业大学(威海) | Indoor old person walking health detection method and system, storage medium and terminal |
CN112926514A (en) * | 2021-03-26 | 2021-06-08 | 哈尔滨工业大学(威海) | Multi-target detection and tracking method, system, storage medium and application |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8605947B2 (en) * | 2008-04-24 | 2013-12-10 | GM Global Technology Operations LLC | Method for detecting a clear path of travel for a vehicle enhanced by object detection |
US7924146B2 (en) * | 2009-04-02 | 2011-04-12 | GM Global Technology Operations LLC | Daytime pedestrian detection on full-windscreen head-up display |
-
2011
- 2011-04-19 CN CN2011101071373A patent/CN102253391B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN102253391A (en) | 2011-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102253391B (en) | Multi-laser-radar-based pedestrian target tracking method | |
US11900536B2 (en) | Visual-inertial positional awareness for autonomous and non-autonomous tracking | |
Xiao et al. | Dynamic-SLAM: Semantic monocular visual localization and mapping based on deep learning in dynamic environment | |
US10366508B1 (en) | Visual-inertial positional awareness for autonomous and non-autonomous device | |
Yan et al. | EU long-term dataset with multiple sensors for autonomous driving | |
US10410328B1 (en) | Visual-inertial positional awareness for autonomous and non-autonomous device | |
US20240202938A1 (en) | Fault-tolerance to provide robust tracking for autonomous and non-autonomous positional awareness | |
Held et al. | Precision tracking with sparse 3d and dense color 2d data | |
US10390003B1 (en) | Visual-inertial positional awareness for autonomous and non-autonomous device | |
US20210201509A1 (en) | Monitoring method and device for mobile target, monitoring system and mobile robot | |
Barnes et al. | Find your own way: Weakly-supervised segmentation of path proposals for urban autonomy | |
KR102053802B1 (en) | Method of locating a sensor and related apparatus | |
EP3516582A1 (en) | Autonomous route determination | |
CN102034355A (en) | Feature point matching-based vehicle detecting and tracking method | |
CN105760846B (en) | Target detection and localization method and system based on depth data | |
KR20130030208A (en) | Egomotion estimation system and method | |
EP3710985A1 (en) | Detecting static parts of a scene | |
CN102074018A (en) | Depth information-based contour tracing method | |
JP2009198445A (en) | Device and method for object detection | |
CN115479598A (en) | Positioning and mapping method based on multi-sensor fusion and tight coupling system | |
EP2610783A2 (en) | Object recognition method and descriptor for object recognition | |
CN106228570A (en) | A kind of Truth data determines method and apparatus | |
CN112833892A (en) | Semantic mapping method based on track alignment | |
Li et al. | Multi-sensor fusion for robust localization with moving object segmentation in complex dynamic 3D scenes | |
CN112432653B (en) | Monocular vision inertial odometer method based on dotted line characteristics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121128 Termination date: 20180419 |