CN116069879A - Method, device, equipment and storage medium for predicting pedestrian track - Google Patents

Method, device, equipment and storage medium for predicting pedestrian track Download PDF

Info

Publication number
CN116069879A
CN116069879A CN202211420061.4A CN202211420061A CN116069879A CN 116069879 A CN116069879 A CN 116069879A CN 202211420061 A CN202211420061 A CN 202211420061A CN 116069879 A CN116069879 A CN 116069879A
Authority
CN
China
Prior art keywords
pedestrian
current
motion
information
track
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
Application number
CN202211420061.4A
Other languages
Chinese (zh)
Other versions
CN116069879B (en
Inventor
罗德宁
张葛祥
李嘉迪
全雪峰
马忠丽
杨强
刘启虞
王嘉伟
朱明�
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu University of Information Technology
Original Assignee
Chengdu University of Information Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chengdu University of Information Technology filed Critical Chengdu University of Information Technology
Priority to CN202211420061.4A priority Critical patent/CN116069879B/en
Publication of CN116069879A publication Critical patent/CN116069879A/en
Application granted granted Critical
Publication of CN116069879B publication Critical patent/CN116069879B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method, a device, equipment and a storage medium for predicting pedestrian tracks, wherein the method comprises the following steps: acquiring past historical track information of a current pedestrian, and obtaining pedestrian motion characteristics of the current pedestrian by encoding the historical track information; processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics; and predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.

Description

Method, device, equipment and storage medium for predicting pedestrian track
Technical Field
The present invention relates to the field of traffic environment sensing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for predicting a pedestrian track.
Background
At present, researchers at home and abroad extract interaction characteristics of people and vehicles more singly, and interaction information in the interaction characteristics cannot be fully utilized. People start to strengthen the study on the safety of the unmanned system, and the unmanned decision system can make the next planning in advance by predicting the track route of pedestrians, so that traffic accidents are avoided. However, it is difficult to predict the trajectory of the pedestrian, and first, the trajectory prediction of the pedestrian is affected by many factors, such as the motion trajectory of other pedestrians around, the distribution of surrounding obstacles, the ground situation, etc., and the trajectory of the pedestrian is diversified. Secondly, the motion trail between pedestrians is mutually influenced, for example, in order to avoid other people nearby, the pedestrians subconsciously change the motion trail before themselves to protect themselves from collision and the like. Traffic environment perception is an extremely important part of unmanned technology, and pedestrian track prediction is a difficulty to be solved and optimized. The traditional method can only simply predict the linear sequence, and easily neglect the motion information characteristics of people, thereby resulting in poor accuracy. With the use of the neural network in the aspect of pedestrian track prediction, pedestrian motion characteristics and intention collection and analysis are greatly enhanced, although the accuracy of prediction is increased, excessive load and excessive characteristic information of the network are caused, and the accuracy cannot be expected.
Disclosure of Invention
The technical problem solved by the scheme provided by the embodiment of the invention is that the accuracy is low due to the fact that the network collection characteristic information of the pedestrian track is too much in a traffic scene prediction pedestrian of pedestrian crossing a road when a vehicle is in a static state.
The method for predicting the pedestrian track provided by the embodiment of the invention comprises the following steps:
acquiring past historical track information of a current pedestrian, and obtaining pedestrian motion characteristics of the current pedestrian by encoding the historical track information;
processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics;
and predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.
According to an embodiment of the present invention, a device for predicting a pedestrian track includes:
the processing module is used for acquiring past historical track information of the current pedestrian and obtaining the pedestrian motion characteristics of the current pedestrian by encoding the historical track information; processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics;
and the prediction module is used for predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.
An electronic device provided in an embodiment of the present application includes: a memory; a processor; a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement a method of predicting a pedestrian trajectory.
A computer-readable storage medium provided in an embodiment of the present application has a computer program stored thereon; the computer program is executed by a processor to implement a method of predicting a pedestrian trajectory.
According to the scheme provided by the embodiment of the invention, the extraction accuracy of the hidden characteristics of the pedestrian movement is improved, a plurality of good prediction tracks are possible on the premise of conforming to the social rule, and the result is more diversified and selectable.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention. In the drawings:
FIG. 1 is a flowchart of a method for predicting a pedestrian trajectory according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an apparatus for predicting a pedestrian trajectory according to an embodiment of the present invention;
fig. 3 is a flowchart of a method for extracting hidden characteristics of pedestrian motion based on social behavior according to an embodiment of the present invention.
Detailed Description
The following detailed description of the preferred embodiments of the present invention is provided in conjunction with the accompanying drawings, and it is to be understood that the preferred embodiments described below are merely illustrative and explanatory of the invention, and are not restrictive of the invention.
The embodiment of the invention is suitable for the situation that the network of pedestrian trails predicted by pedestrian trails of pedestrians passing through roads in a static state of the vehicle collects characteristic information and passes through the environment.
Fig. 1 is a flowchart of a method for predicting a pedestrian track according to an embodiment of the present invention, where, as shown in fig. 1, the method includes:
step S101: acquiring past historical track information of a current pedestrian, and obtaining pedestrian motion characteristics of the current pedestrian by encoding the historical track information;
step S102: processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics;
step S103: and predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.
Specifically, the step of obtaining the pedestrian motion characteristics of the current pedestrian by encoding the historical track information includes: converting the past history track information of the current pedestrian from a coordinate space to a feature space through a fully connected network; and coding the characteristic space and the pedestrian motion characteristic of the current pedestrian at the previous moment to obtain the pedestrian motion characteristic of the current pedestrian.
The step of obtaining the pedestrian motion characteristics of the current pedestrian by carrying out coding processing on the characteristic space and the pedestrian motion characteristics of the current pedestrian at the previous moment comprises the following steps:
Figure 455914DEST_PATH_IMAGE001
wherein ,
Figure 227561DEST_PATH_IMAGE002
refers to pedestrian motion characteristics;
Figure 709358DEST_PATH_IMAGE003
refers to a function;
Figure 591863DEST_PATH_IMAGE004
refers to a feature space;
Figure 960659DEST_PATH_IMAGE005
refers to a function
Figure 903207DEST_PATH_IMAGE003
Is used for the weight parameters of the (c),
Figure 872300DEST_PATH_IMAGE006
is the weight parameter of the encoder.
Further, the processing the pedestrian motion characteristics by using the social attention mechanism to obtain the weight information of the current pedestrian includes: according to the pedestrian movement characteristics of the current pedestrian, calculating the relative movement information of the current pedestrian and each adjacent pedestrian; calculating the attention weight of the current pedestrian and each adjacent pedestrian by using the relative motion information; and obtaining the weight information of the current pedestrian by using the relative motion information and the attention weight.
Wherein, the obtaining the pedestrian motion hiding feature of the current pedestrian by using the weight information and the pedestrian motion feature includes:
Figure 292917DEST_PATH_IMAGE007
wherein ,
Figure 93383DEST_PATH_IMAGE008
is a pedestrian motion hiding feature;
Figure 206832DEST_PATH_IMAGE009
is the motion state information of the adjacent pedestrian at the last moment,
Figure 132063DEST_PATH_IMAGE010
is the surrounding pedestrians at the previous moment
Figure 638262DEST_PATH_IMAGE011
Is to pedestrians of (a)
Figure 699759DEST_PATH_IMAGE012
The influence of the future trajectory is such that, weight information
Figure 249689DEST_PATH_IMAGE013
Is weight information;
Figure 662216DEST_PATH_IMAGE014
is noise.
Specifically, the predicting the pedestrian track of the current pedestrian by using the pedestrian motion hiding feature of the current pedestrian includes: acquiring initial motion state information of a current pedestrian, and updating the initial motion state information by utilizing pedestrian motion hiding characteristics of the current pedestrian to obtain updated motion state information; and converting the updated current motion state into a coordinate space to obtain the predicted pedestrian track of the current pedestrian.
Wherein, the step of obtaining the predicted pedestrian track of the current pedestrian by converting the updated current motion state into a coordinate space comprises the following steps:
Figure 18111DEST_PATH_IMAGE015
wherein ,
Figure 199693DEST_PATH_IMAGE016
is to predict the pedestrian trajectory:
Figure 671257DEST_PATH_IMAGE017
is the current state of motion after the update,
Figure 571080DEST_PATH_IMAGE005
refers to a function
Figure 137190DEST_PATH_IMAGE003
Weight parameters of (c).
Fig. 2 is a schematic diagram of an apparatus for predicting a pedestrian track according to an embodiment of the present invention, as shown in fig. 2, including: the processing module is used for acquiring past historical track information of the current pedestrian and obtaining the pedestrian motion characteristics of the current pedestrian by encoding the historical track information; processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics; and the prediction module is used for predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.
An electronic device provided in an embodiment of the present application includes: a memory; a processor; a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement a method of predicting a pedestrian trajectory.
A computer-readable storage medium provided in an embodiment of the present application has a computer program stored thereon; the computer program is executed by a processor to implement a method of predicting a pedestrian trajectory.
Fig. 3 is a flowchart of a method for extracting hidden characteristics of pedestrian motion based on social behavior, which is provided by an embodiment of the invention, and includes the following steps:
s1, a past track is subjected to long and short time sequence network in an encoder
Figure 704438DEST_PATH_IMAGE018
Coding to obtain pedestrians
Figure 65012DEST_PATH_IMAGE019
Motion characteristics of (a)
Figure 983290DEST_PATH_IMAGE020
。
Through a fully-connected network
Figure 353091DEST_PATH_IMAGE021
Sequence of trajectories of current pedestrians
Figure 322315DEST_PATH_IMAGE022
Conversion from coordinate space to feature space
Figure 119370DEST_PATH_IMAGE023
；
Figure 259364DEST_PATH_IMAGE024
Figure 432857DEST_PATH_IMAGE025
Is a fully connected network in the encoder
Figure 443538DEST_PATH_IMAGE021
Weight parameters of (c).
By a function of
Figure 677073DEST_PATH_IMAGE026
Will be
Figure 773205DEST_PATH_IMAGE027
State after linear embedding and before pedestrian
Figure 15968DEST_PATH_IMAGE028
Inputting the information into an LSTM module of an encoder to encode until the observation sequence is finished, encoding all the information, and encoding the motion characteristics of the pedestrian u
Figure 881155DEST_PATH_IMAGE020
Updating;
Figure 98641DEST_PATH_IMAGE001
in the formula ：
Figure 682069DEST_PATH_IMAGE005
is a function of
Figure 931785DEST_PATH_IMAGE029
Is used for the weight parameters of the (c),
Figure 182638DEST_PATH_IMAGE030
is the weight parameter of the encoder, initialized by pre-training fine-tuning.
The long and short time sequence network internal structure circulation unit can establish a time sequence dependency relationship with a longer distance. Firstly, the sequence at the previous moment is read, coding is carried out, a hidden state and a simple sequence transfer rule are obtained, and the sequence is transmitted through a first door: forgetting gates, inputting a value between 0 and 1 to each memory cell, 1 indicating complete retention, and 0 indicating complete rejection. Via a second gate: the input gate will determine what information is stored in the memory unit. Via a third gate: and updating the gate, removing unnecessary features, leaving important features, updating the state of the memory element, and obtaining the sequence transfer rule at the next moment. And (3) a last door: and the output gate is used for transmitting the final internal screening result to the external hidden state, so that the hidden state at the next moment is obtained, and the sequence at the next moment is further obtained.
S2, the motion characteristics of the pedestrians obtained in the step S1 are obtained
Figure 492396DEST_PATH_IMAGE031
The pedestrian to be tested is subjected to social attention mechanism
Figure 94279DEST_PATH_IMAGE019
Generating weight information
Figure 147686DEST_PATH_IMAGE032
To evaluate the impact of other pedestrians on the pedestrians to be tested.
S21, calculating pedestrians
Figure 66094DEST_PATH_IMAGE019
Adjacent pedestrians around it
Figure 546754DEST_PATH_IMAGE033
Relative motion information of (a)
Figure 370354DEST_PATH_IMAGE034
；
Pedestrian
Figure 227451DEST_PATH_IMAGE035
Pedestrian closely interacting with his surroundings
Figure 921738DEST_PATH_IMAGE033
The relative position information is composed of
Figure 838878DEST_PATH_IMAGE036
Performing calculation, and then
Figure 884195DEST_PATH_IMAGE036
Through a fully-connected network
Figure 561295DEST_PATH_IMAGE037
Mapping to
Figure 110088DEST_PATH_IMAGE034
Obtaining pedestrians
Figure 526026DEST_PATH_IMAGE035
Pedestrian closely interacting with his surroundings
Figure 793059DEST_PATH_IMAGE033
Information of relative motion between them
Figure 991959DEST_PATH_IMAGE034
，
Figure 926417DEST_PATH_IMAGE038
Is composed of (1) pedestrians
Figure 185360DEST_PATH_IMAGE019
And
Figure 967720DEST_PATH_IMAGE033
combining Euclidean distances between (2) pedestrians
Figure 704732DEST_PATH_IMAGE033
With pedestrians
Figure 556013DEST_PATH_IMAGE019
Azimuth of (i.e
Figure 985857DEST_PATH_IMAGE019
Velocity vector sum of (2)
Figure 961904DEST_PATH_IMAGE019
And
Figure 33765DEST_PATH_IMAGE033
included angle between connection vectors of (c) and (3) nearest approachingDistance (minimum distance they will reach in the future if both objects maintain the current speed) is made up of three parts, where
Figure 677236DEST_PATH_IMAGE039
For the weight parameter of the full connection layer, the calculation formula is as follows:
Figure 28714DEST_PATH_IMAGE040
Figure 757635DEST_PATH_IMAGE041
in the formula ,
Figure 367608DEST_PATH_IMAGE042
the present invention uses the first eight sequence tracks as the past tracks.
S22, calculating the attention weight of each adjacent pedestrian.
Pedestrian
Figure 865586DEST_PATH_IMAGE019
And
Figure 371653DEST_PATH_IMAGE033
between (a) and (b)
Figure 915767DEST_PATH_IMAGE034
Through the whole connecting layer
Figure 798273DEST_PATH_IMAGE043
Will be
Figure 432647DEST_PATH_IMAGE038
Embedding in
Figure 375196DEST_PATH_IMAGE044
In (a) the number of the components,
Figure 78709DEST_PATH_IMAGE045
is an adjacent pedestrian
Figure 499326DEST_PATH_IMAGE033
Is provided.
Figure 299792DEST_PATH_IMAGE046
in the formula ,
Figure 147663DEST_PATH_IMAGE034
is the interactive motion information of pedestrian a and the adjacent pedestrians around him, N is the total number of pedestrians,
Figure 72893DEST_PATH_IMAGE047
is that
Figure 579092DEST_PATH_IMAGE034
A common rank of linear mapping weights applied to the motion profile information,
Figure 906168DEST_PATH_IMAGE048
is a full connection layer
Figure 456098DEST_PATH_IMAGE049
Weight parameters of (c).
S23, will
Figure 868625DEST_PATH_IMAGE034
And
Figure 162203DEST_PATH_IMAGE050
the attention weight of each adjacent pedestrian is obtained through scalar product and softmax operation
Figure 343786DEST_PATH_IMAGE051
，
Figure 799038DEST_PATH_IMAGE050
Is the motion characteristic information of all pedestrians,
Figure 777489DEST_PATH_IMAGE052
indicating the number of pedestrians that are present,
Figure 343600DEST_PATH_IMAGE035
、
Figure 910847DEST_PATH_IMAGE053
。
Figure 537001DEST_PATH_IMAGE054
Figure 189699DEST_PATH_IMAGE055
,
Figure 293921DEST_PATH_IMAGE056
s3, weight information obtained according to S2 in a decoder
Figure 450096DEST_PATH_IMAGE051
In combination with pedestrians
Figure 778309DEST_PATH_IMAGE019
Motion state of (2)
Figure 387145DEST_PATH_IMAGE031
And adjacent pedestrians
Figure 639266DEST_PATH_IMAGE057
Motion state of (2)
Figure 649947DEST_PATH_IMAGE058
Obtaining useful pedestrian motion hiding features
Figure 883483DEST_PATH_IMAGE059
。
Figure 979615DEST_PATH_IMAGE060
in the formula ,
Figure 425639DEST_PATH_IMAGE061
is an adjacent pedestrian
Figure 87565DEST_PATH_IMAGE057
The motion state information of the last moment,
Figure 226422DEST_PATH_IMAGE062
is the surrounding pedestrians at the previous moment
Figure 888479DEST_PATH_IMAGE057
Is to pedestrians of (a)
Figure 138195DEST_PATH_IMAGE019
Future trajectory is used for the control of the (c),
Figure 654627DEST_PATH_IMAGE063
is noise.
S4, hiding the features according to the motion obtained in the S3
Figure 964385DEST_PATH_IMAGE064
And the current motion state of the pedestrian
Figure 300689DEST_PATH_IMAGE065
Predicting pedestrian trajectories
Figure 354095DEST_PATH_IMAGE066
。
Pedestrians received by long-short time sequence network in decoder
Figure 193875DEST_PATH_IMAGE019
Is the initial current motion state information of
Figure 956426DEST_PATH_IMAGE065
,
Figure 248867DEST_PATH_IMAGE065
Is an encoder
Figure 433861DEST_PATH_IMAGE067
State of (2)
Figure 128147DEST_PATH_IMAGE065
Cascade high stage noise
Figure 310867DEST_PATH_IMAGE063
Obtained.
Figure 356183DEST_PATH_IMAGE068
Subsequent updating
Figure 751393DEST_PATH_IMAGE065
It is necessary to send the motion state information of the previous moment
Figure 316497DEST_PATH_IMAGE069
And the attention mechanism module at the last moment
Figure 404539DEST_PATH_IMAGE070
The screened useful pedestrian motion hiding features are combined into a long-short time sequence network.
Figure 999468DEST_PATH_IMAGE071
in the formula ,
Figure 932789DEST_PATH_IMAGE072
is a decoding unit function of a long and short time sequence network,
Figure 132827DEST_PATH_IMAGE073
is the weight of the long and short timing network in the decoder.
Then by passing through
Figure 391770DEST_PATH_IMAGE074
The function updates the current motion state
Figure 880520DEST_PATH_IMAGE075
Conversion to coordinate space, obtaining predicted future track
Figure 899422DEST_PATH_IMAGE076
：
Figure 422808DEST_PATH_IMAGE077
in the formula
Figure 118231DEST_PATH_IMAGE078
Is a function of
Figure 359857DEST_PATH_IMAGE079
Is a weight of (2).
Although the present invention has been described in detail hereinabove, the present invention is not limited thereto and various modifications may be made by those skilled in the art in accordance with the principles of the present invention. Therefore, all modifications made in accordance with the principles of the present invention should be understood as falling within the scope of the present invention.

Claims (10)

1. A method of predicting a pedestrian trajectory, comprising:
acquiring past historical track information of a current pedestrian, and obtaining pedestrian motion characteristics of the current pedestrian by encoding the historical track information;
processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics;
and predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.
2. The method of claim 1, wherein the obtaining the pedestrian motion characteristic of the current pedestrian by encoding the historical track information comprises:
converting the past history track information of the current pedestrian from a coordinate space to a feature space through a fully connected network;
and coding the characteristic space and the pedestrian motion characteristic of the current pedestrian at the previous moment to obtain the pedestrian motion characteristic of the current pedestrian.
3. The method according to claim 2, wherein the obtaining the pedestrian motion characteristic of the current pedestrian by performing encoding processing on the characteristic space and the pedestrian motion characteristic of the current pedestrian at a previous time comprises:
Figure 270672DEST_PATH_IMAGE002
wherein ,
Figure 718971DEST_PATH_IMAGE003
refers to pedestrian motion characteristics;
Figure 498708DEST_PATH_IMAGE004
refers to a function;
Figure 893917DEST_PATH_IMAGE005
refers to a feature space;
Figure 255760DEST_PATH_IMAGE006
refers to a function
Figure 78222DEST_PATH_IMAGE004
Is used for the weight parameters of the (c),
Figure 141993DEST_PATH_IMAGE007
is the weight parameter of the encoder.
4. The method of claim 2, wherein the processing the pedestrian motion profile using a social-awareness mechanism to obtain the weight information of the current pedestrian comprises:
according to the pedestrian movement characteristics of the current pedestrian, calculating the relative movement information of the current pedestrian and each adjacent pedestrian;
calculating the attention weight of the current pedestrian and each adjacent pedestrian by using the relative motion information;
and obtaining the weight information of the current pedestrian by using the relative motion information and the attention weight.
5. The method of claim 4, wherein using the weight information and the pedestrian motion characteristics to obtain the pedestrian motion concealment characteristic for the current pedestrian comprises:
Figure 340893DEST_PATH_IMAGE008
wherein ,
Figure 9772DEST_PATH_IMAGE009
is a pedestrian motion hiding feature;
Figure 534294DEST_PATH_IMAGE010
is the motion state information of the adjacent pedestrian at the last moment,
Figure 288624DEST_PATH_IMAGE011
is the surrounding pedestrians at the previous moment
Figure 573106DEST_PATH_IMAGE012
Is to pedestrians of (a)
Figure 96491DEST_PATH_IMAGE013
Influence of future trajectory, weight information
Figure 57494DEST_PATH_IMAGE014
Is weight information;
Figure 299119DEST_PATH_IMAGE015
is noise.
6. The method of claim 5, wherein predicting the pedestrian trajectory of the current pedestrian using the pedestrian motion hiding feature of the current pedestrian comprises:
acquiring initial motion state information of a current pedestrian, and updating the initial motion state information by utilizing pedestrian motion hiding characteristics of the current pedestrian to obtain updated motion state information;
and converting the updated current motion state into a coordinate space to obtain the predicted pedestrian track of the current pedestrian.
7. The method of claim 6, wherein the obtaining the predicted pedestrian trajectory for the current pedestrian by translating the updated current motion state into a coordinate space comprises:
Figure 574243DEST_PATH_IMAGE016
wherein ,
Figure 14451DEST_PATH_IMAGE017
is to predict the pedestrian trajectory:
Figure 349618DEST_PATH_IMAGE018
is the current state of motion after the update,
Figure 360430DEST_PATH_IMAGE006
refers to a function
Figure 439245DEST_PATH_IMAGE004
Weight parameters of (c).
8. An apparatus for predicting a pedestrian trajectory, comprising:
the processing module is used for acquiring past historical track information of the current pedestrian and obtaining the pedestrian motion characteristics of the current pedestrian by encoding the historical track information; processing the pedestrian motion characteristics by using a social attention mechanism to obtain weight information of the current pedestrian, and obtaining pedestrian motion hidden characteristics of the current pedestrian by using the weight information and the pedestrian motion characteristics;
and the prediction module is used for predicting the pedestrian track of the current pedestrian by utilizing the pedestrian motion hiding characteristics of the current pedestrian to obtain the predicted pedestrian track of the current pedestrian.
9. An electronic device, comprising: a memory; a processor; a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of any of claims 1-7.
10. A computer-readable storage medium, characterized in that a computer program is stored thereon; the computer program being executed by a processor to implement the method of any of claims 1-7.
CN202211420061.4A 2022-11-14 2022-11-14 Method, device, equipment and storage medium for predicting pedestrian track Active CN116069879B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211420061.4A CN116069879B (en) 2022-11-14 2022-11-14 Method, device, equipment and storage medium for predicting pedestrian track

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211420061.4A CN116069879B (en) 2022-11-14 2022-11-14 Method, device, equipment and storage medium for predicting pedestrian track

Publications (2)

Publication Number Publication Date
CN116069879A true CN116069879A (en) 2023-05-05
CN116069879B CN116069879B (en) 2023-06-20

Family

ID=86173865

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211420061.4A Active CN116069879B (en) 2022-11-14 2022-11-14 Method, device, equipment and storage medium for predicting pedestrian track

Country Status (1)

Country Link
CN (1) CN116069879B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111044045A (en) * 2019-12-09 2020-04-21 中国科学院深圳先进技术研究院 Navigation method and device based on neural network and terminal equipment
CN111311009A (en) * 2020-02-24 2020-06-19 广东工业大学 Pedestrian trajectory prediction method based on long-term and short-term memory
CN111459168A (en) * 2020-04-23 2020-07-28 上海交通大学 Fused automatic-driving automobile pedestrian crossing track prediction method and system
CN112766561A (en) * 2021-01-15 2021-05-07 东南大学 Generating type confrontation track prediction method based on attention mechanism
WO2022222034A1 (en) * 2021-04-20 2022-10-27 深圳市锐明技术股份有限公司 Pedestrian trajectory prediction method and apparatus, electronic device, and storage medium

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111044045A (en) * 2019-12-09 2020-04-21 中国科学院深圳先进技术研究院 Navigation method and device based on neural network and terminal equipment
CN111311009A (en) * 2020-02-24 2020-06-19 广东工业大学 Pedestrian trajectory prediction method based on long-term and short-term memory
CN111459168A (en) * 2020-04-23 2020-07-28 上海交通大学 Fused automatic-driving automobile pedestrian crossing track prediction method and system
CN112766561A (en) * 2021-01-15 2021-05-07 东南大学 Generating type confrontation track prediction method based on attention mechanism
WO2022222034A1 (en) * 2021-04-20 2022-10-27 深圳市锐明技术股份有限公司 Pedestrian trajectory prediction method and apparatus, electronic device, and storage medium

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
ALEXANDRE ALAHI等: "Social LSTM:Human Trajectory Prediction in Crowded Spaces", 2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), pages 961 - 971 *
李文礼等: "基于视野域机制的行人轨迹预测", 计算机应用研究, vol. 40, no. 1, pages 80 - 85 *
欧阳俊等: "基于GAN和注意力机制的行人轨迹预测", 激光与光电子学进展, vol. 57, no. 14, pages 161 - 172 *

Also Published As

Publication number Publication date
CN116069879B (en) 2023-06-20

Similar Documents

Publication Publication Date Title
CN111400620B (en) User trajectory position prediction method based on space-time embedded Self-orientation
Messaoud et al. Non-local social pooling for vehicle trajectory prediction
CN114299723B (en) Traffic flow prediction method
CN114372116A (en) Vehicle track prediction method based on LSTM and space-time attention mechanism
Ren et al. Short‐Term Traffic Flow Prediction: A Method of Combined Deep Learnings
CN115438856A (en) Pedestrian trajectory prediction method based on space-time interaction characteristics and end point information
CN115510174A (en) Road network pixelation-based Wasserstein generation countermeasure flow data interpolation method
Meng et al. Forecasting human trajectory from scene history
CN116203971A (en) Unmanned obstacle avoidance method for generating countering network collaborative prediction
Zhao et al. End‐to‐end autonomous driving decision model joined by attention mechanism and spatiotemporal features
Yang et al. Multi-granularity scenarios understanding network for trajectory prediction
CN113408786B (en) Traffic characteristic prediction method and system
Zhang et al. Step attention: Sequential pedestrian trajectory prediction
Chauhan et al. Traffic flow forecasting using attention enabled Bi-LSTM and GRU hybrid model
CN117429461A (en) Automatic driving automobile passing decision-making method for urban no-signal road section
CN116069879B (en) Method, device, equipment and storage medium for predicting pedestrian track
CN116501820A (en) Vehicle track prediction method, device, equipment and storage medium
Zhang et al. Obstacle‐transformer: A trajectory prediction network based on surrounding trajectories
CN113191539B (en) High-density composite scene track prediction method based on heterogeneous graph aggregation network
US20230394284A1 (en) Method and system for predicting trajectories for maneuver planning based on a neural network
Chen et al. Traffic flow prediction based on cooperative vehicle infrastructure for cloud control platform
CN110489671B (en) Road charging pile recommendation method based on encoder-decoder model
CN116702836A (en) Traffic participant trajectory prediction method and device and related equipment
Chu et al. Deep Encoder Cross Network for Estimated Time of Arrival
Chan et al. Short Term Traffic Flow Prediction with Neighbor Selecting Gated Recurrent Unit

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