CN115239767B - Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment - Google Patents

Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment Download PDF

Info

Publication number
CN115239767B
CN115239767B CN202211154363.1A CN202211154363A CN115239767B CN 115239767 B CN115239767 B CN 115239767B CN 202211154363 A CN202211154363 A CN 202211154363A CN 115239767 B CN115239767 B CN 115239767B
Authority
CN
China
Prior art keywords
probability
pedestrian
behavior
hidden
walking state
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202211154363.1A
Other languages
Chinese (zh)
Other versions
CN115239767A (en
Inventor
孙立山
邵娟
张亚军
王欣桐
王鑫伟
魏有军
李宇杰
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.)
Beijing University of Technology
Original Assignee
Beijing University of 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 Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN202211154363.1A priority Critical patent/CN115239767B/en
Publication of CN115239767A publication Critical patent/CN115239767A/en
Application granted granted Critical
Publication of CN115239767B publication Critical patent/CN115239767B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/72—Data preparation, e.g. statistical preprocessing of image or video features
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30241—Trajectory
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Human Computer Interaction (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention relates to the technical field of pedestrian behavior and state prediction, and provides a dynamic passenger flow behavior situation prediction method, a system, a storage medium and equipment, wherein the method comprises the following steps: acquiring passenger flow track characteristic data, and identifying and obtaining the walking state and hidden behavior of each pedestrian at each historical moment; counting initial probability, mapping probability and transition probability according to the walking states and hidden behaviors of all pedestrians at all historical moments; according to the statistical initial probability, mapping probability and transition probability, the maximum walking state probability of each pedestrian at each to-be-predicted time is calculated, the walking state sequence and the hidden behavior sequence of each pedestrian are obtained through prediction, abnormal passenger flow can be accurately and rapidly identified, and the evolution situation of high-density passenger flow can be accurately reflected.

Description

Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment
Technical Field
The invention belongs to the technical field of pedestrian behavior and state prediction, and particularly relates to a dynamic passenger flow behavior situation prediction method, a dynamic passenger flow behavior situation prediction system, a storage medium and a device.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
At present, the phenomenon of passenger flow gathering inside a rail transit junction is frequent, the comfort of passengers and the transfer efficiency are seriously influenced, and even potential safety hazards are formed. The existing management and control for high-density passenger flow is mostly carried out by depending on the working experience of rail transit operators, the subjective randomness is strong, and the effect is not good enough. The walking state of the passenger flow can effectively reflect the potential complex behaviors of the crowd, and is directly related to the safety evaluation of public transport service and the occurrence of abnormal events. Therefore, how to accurately grasp passenger behavior characteristics and predict pedestrian hidden behavior evolution situation is helpful for timely discovering and early warning public safety events, and has important significance for assisting management and control of passenger flow of operators.
At the present stage, aiming at the identification and prediction of the pedestrian behaviors and states, the method is mainly realized by original video image data and a passenger flow simulation model:
(1) The analysis method based on video image processing comprises the following steps: the behavior state of the crowd is analyzed from the global external expression of the crowd, the pedestrian behavior is identified by extracting the characteristics of the crowd, for example, the states of stillness, walking, running and the like are defined by the head and shoulder coordinate information of the pedestrian in each frame of image of the video, the posture of the human body and other image textural characteristics;
(2) The passenger flow simulation model analysis method comprises the following steps: for example, a cellular automata model is based on a grid principle, pedestrian movement is mainly simulated by constructing pedestrian movement rules, passenger flow characteristics are described through speed and direction changes, but the individual selection differences of the model are only considered, and most of model algorithms are heuristic, so that results are unpredictable; the social force model mainly describes the motion characteristics of individual pedestrians through interaction force among passenger flows, can effectively simulate and predict behaviors of following, avoiding, exceeding and the like of the pedestrians, but is slow in operation speed when large-scale crowd motion is simulated due to the complex mathematical algorithm.
From the current state of research:
(1) The video image processing method is easily influenced by the illumination background of a monitoring place and the erection angle of a monitoring camera, so that the shape size, position and speed of a person in an image have deviation with the reality, and particularly in an area with insufficient illumination in a closed space of a subway, the state change of the person cannot be accurately analyzed.
(2) The existing passenger flow simulation model prediction method is mainly completed based on variable parameters such as individual speed, direction and walking rules of pedestrians, the analysis and state quantization of common behaviors of the pedestrians are rarely involved, the problems of complex modeling of various passenger groups, low accuracy of high-density simulation prediction and the like exist, and effective guidance from individual pedestrians to passenger flow behavior state recognition prediction research is difficult to achieve.
Disclosure of Invention
In order to solve the technical problems in the background art, the invention provides a dynamic passenger flow behavior situation prediction method, a dynamic passenger flow behavior situation prediction system, a storage medium and a dynamic passenger flow behavior situation prediction device, which can accurately and quickly identify abnormal passenger flow and accurately reflect the high-density passenger flow evolution situation.
In order to achieve the purpose, the invention adopts the following technical scheme:
a first aspect of the present invention provides a dynamic passenger flow behavior situation prediction method, including:
acquiring passenger flow track characteristic data, and identifying and obtaining the walking state and hidden behavior of each pedestrian at each historical moment;
counting initial probability, mapping probability and transition probability according to the walking states and hidden behaviors of all pedestrians at all historical moments;
and according to the statistical initial probability, mapping probability and transition probability, predicting to obtain a walking state sequence and a hidden behavior sequence of each pedestrian by calculating the maximum walking state probability of each pedestrian at each moment to be predicted.
Further, the method for identifying the walking state comprises the following steps:
according to the actual track of the pedestrian in front, obtaining state track templates of the pedestrian in different walking states by moving the spatial position distance on the vertical axis and the moving time length on the horizontal axis;
calculating error probability between the trajectory of the pedestrian and the trajectory templates in different states;
and taking the walking state corresponding to the minimum error probability as the final walking state of the pedestrian.
Further, the method for identifying the hidden behavior comprises the following steps: according totThe distance of the pedestrian from the pedestrian ahead at the moment, antTime of day andt-1 a change in distance at a time instant, identifying hidden behavior of the pedestrian.
Further, the initial probability is: at the initial time in all historical times, the probability of various hidden behaviors;
alternatively, the first and second electrodes may be,
the mapping probability is: the probability of different observable walking states expressed under various hidden behaviors;
alternatively, the first and second electrodes may be,
the transition probability is: the probability of transition between classes of hidden behavior at two consecutive times.
Further, the maximum walking state probability is expressed as:
P(O t )=max{P(O t-1 )·P(H t |H t-1 )·P(O t |H t )}
wherein the content of the first and second substances,P(O t ) Represents the maximum probability of a walking state at time t,P(H t |H t-1 ) Is represented bytHidden behavior at time-1H t-1 TotHidden behavior of a time of dayH t The transition probability of (a) is,P(O t |H t ) Is shown intTemporal hidden behaviorH t And the state of walkingO t The probability of mapping between.
Further, the upper hidden behavior that achieves the maximum walking state probability at the last predicted time is represented as:
Figure DEST_PATH_IMAGE001
wherein the content of the first and second substances,O T indicating the walking state at time T.
Further, after the maximum walking state probability of the last prediction time is obtained, the maximum walking state probability corresponds to the optimal walking state of which the walking state is the last prediction time, and the upper layer hiding behavior for realizing the maximum walking state probability is used as the optimal hiding behavior of the last prediction time; and obtaining an optimal walking state sequence and a hidden behavior sequence according to the maximum walking state probability of each predicted moment and the recorded optimal walking behavior of the predecessor by optimal path backtracking.
A second aspect of the invention provides a dynamic passenger flow behavior situation prediction system comprising:
a data acquisition module configured to: acquiring passenger flow track characteristic data, and identifying and obtaining the walking state and the hidden behavior of each pedestrian at each historical moment;
a statistics module configured to: counting initial probability, mapping probability and transition probability according to the walking states and hidden behaviors of all pedestrians at all historical moments;
a prediction module configured to: and according to the statistical initial probability, mapping probability and transition probability, predicting to obtain a walking state sequence and a hidden behavior sequence of each pedestrian by calculating the maximum walking state probability of each pedestrian at each moment to be predicted.
A third aspect of the invention provides a computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for dynamic passenger flow behavior situation prediction as described above.
A fourth aspect of the invention provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the dynamic passenger flow behavior situation prediction method as described above when executing the program.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a dynamic passenger flow behavior situation prediction method, which takes common behaviors in the walking process of pedestrians as entry points, provides a pedestrian walking state recognition method and a pedestrian hidden behavior recognition method, captures the pedestrian state in real time through track matching, can accurately and quickly recognize abnormal passenger flow, can effectively avoid problems of background shielding, illumination and angle compared with an analysis method based on image processing, and reduces the difficulty of feature extraction.
The invention provides a dynamic passenger flow behavior situation prediction method which is based on a hidden Markov prediction model, effectively quantifies the mapping relation between behaviors and states in complex passenger flow and the hidden incidence relation between different behaviors, and realizes the prediction of hidden behaviors and states of pedestrians by a numerical simulation method.
The invention provides a dynamic passenger flow behavior situation prediction method which is beneficial to timely discovering and early warning public safety events and has important significance for assisting rail transit operators in carrying out passenger flow management and control.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a block diagram of a two-level hierarchical hidden Markov model according to a first embodiment of the present invention;
FIG. 2 is a schematic diagram of a pedestrian real track and state track template according to a first embodiment of the present invention;
fig. 3 is a schematic diagram illustrating a method for recognizing a hidden behavior of a pedestrian according to a first embodiment of the present invention;
FIG. 4 is a diagram illustrating a pedestrian walking decision process according to a first embodiment of the present invention;
FIG. 5 is a diagram illustrating hidden behavior transition change relationships according to a first embodiment of the present invention;
FIG. 6 (a) is a diagram illustrating a mapping and transition relationship in a hidden Markov model according to an embodiment of the present invention;
fig. 6 (b) is a schematic diagram of the maximum likelihood path for pedestrian hidden behavior transition according to the first embodiment of the present invention;
fig. 7 is a diagram illustrating a prediction result according to a first embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example one
The embodiment provides a dynamic passenger flow behavior situation prediction method, which specifically comprises the following steps:
step 1, obtaining passenger flow track characteristic data to obtain the walking state and hidden behavior of each pedestrian at each historical moment.
Step 101, collecting passenger flow track characteristic data.
The passenger flow trajectory characteristic data comprises positions of a plurality of pedestrians at a plurality of historical moments, and characteristic data such as pedestrian walking distance and pedestrian reflection time can be obtained according to position change in unit time.
And 102, identifying the walking state of each pedestrian by adopting a pedestrian walking state identification method based on the passenger flow track characteristic data.
As shown in fig. 1, the walking characteristics in the passenger flow are divided into two levels, namely a low level and a high level; the first level is a pedestrian walking state, and the pedestrian walking state is divided into three types, namely Conservative (Conservative), normal (Normal) and Aggressive (Aggressive); the second level is pedestrian hiding behavior, and the pedestrian hiding behavior is divided into four types of following (following), relaxing (relaxing), surpassing (Changing) and adjusting (animation).
If a pedestrian is considered to be a following pedestrian, the pedestrian located in front of and closest to the following pedestrian is referred to as a leading pedestrian (or a leading pedestrian).
The first layer of pedestrian walking state division is to perform clustering based on instant reaction time between pedestrians in front and following pedestrians, the pedestrian reaction time is mainly divided into three types according to a clustering result, and numerical values are mainly concentrated on 0.4s, 0.7s and 1.0s, so that the pedestrian walking state is divided into three types of aggressive type, normal type and conservative type.
Under the state of radical excitation, the following pedestrians can be more concentrated on the information of the pedestrians in front, the walking speed is higher, the behavior change is more sensitive, and the reaction time is shorterτAnd a spatial distancedSmall, reaction timeτ=0.4; in the normal state of the process, the temperature of the reaction vessel,τ=0.7; under the state of conservation, the walking speed can be reduced, and the reaction time is shortenedτAnd a spatial distancedThe size of the composite material is larger,τ=1.0; the pedestrian walking state result can be directly observed according to the fact that the trajectory data of the pedestrian under each time step is matched with the state model, and the pedestrian walking state result is defined as a display state.
The second layer is a hidden behavior of the pedestrian at a higher level, the hidden behavior is a subjective consciousness of the pedestrian, cannot be directly obtained quantitatively according to real-time monitoring, and needs to be identified by combining the spatial position change relationship between the target passenger individual and the adjacent passenger, so that the hidden behavior is called as the hidden behavior.
The observation of the first layer walking state is the reaction timeτAnd distance traveleddOn the basis, a pedestrian following track microscopic discrete state standard template under three state types of aggressive, conventional and conservative is constructed on the basis of the front pedestrian track, the following pedestrian track is matched with the state standard template, the maximum likelihood sequence of the pedestrian state is searched, and the discrete state of the following pedestrian track under each time step is identified.
As shown in fig. 2, a specific method of identifying the walking state of the pedestrian is as follows:
step 1021, according to the actual track of the pedestrian ahead, moving the spatial position distance (walking distance) on the vertical axisdAnd length of time of movement on the horizontal axis (reaction time)τTo obtain the state of following the pedestrian in the state of being excited,State trajectory templates in normal and conservative states:
Figure 684277DEST_PATH_IMAGE002
in the formula (I), the compound is shown in the specification,a、nandcrepresents aggressive, normal and conservative states, respectively;x i (t) Is composed ofiIn the state track template in the statetThe location of the time of day;τ i is composed ofiReaction time under conditions;
Figure DEST_PATH_IMAGE003
for the front pedestrian in the actual trackt-τiThe location of the time of day;d i is composed ofiThe travel distance in the state.
Step 1022, calculating the track line of the following pedestrian and the three-state track template, and calculating the time point of each tracktError distance of (2):
△d i (t)=|x F (t)-x i (t)|，i∈(a,n,c)
in the formula (I), the compound is shown in the specification,x F (t) Is shown intA locus point of a pedestrian is followed at a momentd i (t) Is composed oftConstantly following pedestriansiVertical distance between the track templates in the state.
Step 1023, the error distance is converted into an error cost, and the conversion formula is as follows:
C i (t)=△d i (t)，i∈(a,n,c)
step 1024, calculating error probability based on the simple logit model:
Figure 86570DEST_PATH_IMAGE004
the smaller the error probability, the better the fitting degree of the track points of the following pedestrians and the track points of the state template.
Step 1025, calculating the total error probability of the track sequence:
Figure DEST_PATH_IMAGE005
and the sequence with the minimum total error probability is the observed discrete state maximum likelihood sequence of the track of the following pedestrian under each time step.
And step 1026, taking the walking state corresponding to the minimum error probability as the final walking state of the pedestrian.
As shown in fig. 2, the real pedestrian trajectory is obtained by identifying the walking state by using a pedestrian walking state identification method, and the final state observation result is shown in table 1.
TABLE 1 matching results of real trajectory and three state trajectory templates
Figure 725362DEST_PATH_IMAGE006
103, for the identification of the hidden behavior, firstly, four behavior characteristics of following, relaxing, exceeding and adjusting need to be analyzed; according to the analysis result, giving out the numerical threshold definition of four behaviors based on the track data of the front pedestrian and the rear following pedestrian; and (4) combining with numerical definition, and identifying four behaviors according to the spatial position change relationship between the target passenger individual and the near-adjacent passenger.
As shown in fig. 3, the hidden behavior in the second layer is identified according to the spatial position variation relationship between the target passenger individual and the near-adjacent passenger, specifically: according totThe distance of the pedestrian from the pedestrian ahead at the moment, antTime andt-1 a change in distance at a time instant, identifying hidden behavior of the pedestrian. The method comprises the following specific steps:
following behavior: the space distance between the pedestrian following behind and the pedestrian ahead is always kept in a basically stable state, and the pedestrian following behind and the pedestrian ahead are in relative restState, as in the region of FIG. 3
Figure DEST_PATH_IMAGE007
(leading pedestrian and following pedestrian 1);
relaxation behavior: the space distance between the rear following pedestrian and the front pedestrian is gradually increased, as shown in the area of fig. 3
Figure 277828DEST_PATH_IMAGE008
(leading pedestrian and following pedestrian 1);
and (4) overtaking behavior: the following pedestrian behind overtakes the pedestrian 1 ahead, and the spatial distance between the two changes to a negative value, as shown in the region of fig. 3
Figure DEST_PATH_IMAGE009
Shown as (follower pedestrian 2 and follower pedestrian 1);
and (3) adjusting the behavior: in order to restore oneself to a stable state, the space distance between the rear following pedestrian and the front pedestrian is gradually reduced, as shown in the area of fig. 3
Figure 888938DEST_PATH_IMAGE010
(follower pedestrian 2 versus follower pedestrian 1).
The next time, the following pedestrian (following pedestrian 1) takes a stable following action again, as shown by the region in fig. 3
Figure DEST_PATH_IMAGE011
As shown.
The numerical threshold of the four behavior characteristics is specifically defined as follows:
d t =x Lt -x Ft ，△d t =d t -d t-1 ，t∈(1,2,3,…,T)
in the formula (I), the compound is shown in the specification,x Lt for the pedestrian in front oftThe position of the moment of time is,x Ft to follow the pedestriantThe position of the moment of time is,d t is composed oftDistance between the pedestrian ahead and the following pedestrian behind at the moment, Δd t Is composed oftThe distance of the moment from the previous moment varies.
(1) If it is notd t-1 ＞0，d t > 0, and-0.01m≤△d t ≤0.01mTo follow the pedestriantThe moment is a following behavior;
(2) If it is notd t-1 ＞0，d t > 0, and Δd t ≥0.01mTo follow the pedestriantThe moment is relaxation behavior;
(3) If it is usedd t-1 ＞0，d t Is less than 0, follows the pedestriantThe moment is an overtaking behavior;
(4) If it is notd t-1 ＞0，d t > 0, and Δd t ≤-0.01mFollowing the pedestriantThe moment is the adjustment action.
Step 2: counting initial probability, mapping probability and transition probability; by calculation oftTemporal hidden behavior andt+1 transition probability between hidden behaviors, obtainingt+Probability of behavior at time 1 until recursion to timeTAcquiring the probability of the hidden behavior sequence; under the condition that the hidden behavior sequence probability is determined, calculating the probability of generating the walking state sequence under the hidden behavior sequence through the mapping probability between the hidden behavior and the walking state.
According to the method, the mapping relation between the commonalities and behaviors of the pedestrians is quantified according to the decision making process of the walking of the pedestrians; establishing a hidden Markov model structure with two levels based on a mapping relation between a walking state and a hidden behavior and a transfer relation between the hidden behavior and the hidden behavior; and calculating the mapping probability between the hidden behaviors of the pedestrians and the walking state and the transition probability between the hidden behaviors of the pedestrians in the hidden Markov model structure through a large amount of pedestrian track data.
As shown in fig. 4, the decision process for the pedestrian to walk is as follows: when the pedestrian arrives at the destination, a corresponding walking route is formulated by combining the environment of the pedestrian and the condition attribute of the pedestrian. In order to realize the route target, the pedestrian can schedule the body function, adopt appropriate subjective behaviors (such as following behavior, relaxation behavior, exceeding behavior and adjusting behavior) and realize the optimization and the most comfort of the self walking route. These behaviors can influence the direction of pedestrian operation, i.e., take an aggressive state, a normal state, or a conservative state. Therefore, the high-level subjective behavior selection of the pedestrian directly results in the difference of the walking state of the pedestrian.
The structural division of the hidden markov model in two levels according to the relation between hidden behavior of a pedestrian and walking state is shown in fig. 6 (a): the upper layer is a hidden behavior, the lower layer is a walking state influenced by the hidden behavior, and the three states of the lower layer are influenced by the high-level behavior. Due to the influence of subjective consciousness, the hidden behaviors of the pedestrians can be randomly changed according to the surrounding environment and other factors, and the behavior conversion change relationship is shown in fig. 5.
The embodiment constructs a hidden Markov model-based pedestrian hidden behavior and state dynamic prediction model based on a mapping and transfer relationship in a hidden Markov model as shown in FIG. 6 (a); providing a hidden behavior sequence probability and a corresponding walking state sequence probability calculation method; and predicting a hidden behavior sequence which enables the probability of the walking state sequence to be maximum through a Viterbi algorithm.
The pedestrian hiding behavior and state dynamic prediction model is assumed as follows:
(1) The observation values output by the hidden Markov model are strictly independent;
(2) During the hidden behavior transition, the pedestrian's next transitional behavior is only related to the current behavior, and not to the previous behavior.
Thus, the behavior transfer process only needs to consider paths that maximize the likelihood function over the entire set of paths. For multiple traces from start to end, there is only one optimal behavior transition (possibly the minimum state transition cost, the safest and/or most comfortable transition line, etc.). The goal is therefore to find the maximum likelihood path for the pedestrian hidden behavior transition with the highest joint mapping and transition probability on the markov chain, as shown in fig. 6 (b).
And step 201, counting initial probability, mapping probability and transition probability according to the walking states and behaviors of all pedestrians at all historical moments.
Initial probability (behavior probability at initial time of pedestrian trajectory): collecting the common behaviors of the initial moments of the pedestrians in the trajectory data, recording the occurrence frequency of various behaviors, and calculating the occurrence probability of the initial moments of various common behaviors. That is, the initial probability is a probability that various types of hidden behaviors appear at the initial time among all the history times.
Mapping probability (mapping probability between hidden behavior and walking state): acquiring a walking state and a hidden behavior discrete sequence according to the trajectory data, and calculating the number of simultaneous mapping times of the walking state and the hidden behavior discrete sequence; thereby preliminarily obtaining the mapping probability between the walking state and the hidden behavior. That is, the mapping probability is the probability of different observable walking states expressed under various hidden behaviors.
Transition probability (behavior transition probability between hidden behaviors): and acquiring the hidden behaviors and the state sequence of the pedestrians according to the trajectory data, calculating the change times between the hidden behaviors of the pedestrians at adjacent moments, and preliminarily obtaining the transition probability between the hidden behaviors. That is, the transition probability is the transition probability between various types of hidden behavior at two consecutive times. Examples of statistical initial, transition and mapping probabilities are given in tables 2, 3 and 4, respectively.
TABLE 2 probability of initial hidden behavior
Figure 845220DEST_PATH_IMAGE012
TABLE 3 transition probability of hidden behavior
Figure DEST_PATH_IMAGE013
TABLE 4 mapping probability between hidden behavior and Walking state
Figure 394145DEST_PATH_IMAGE014
Step 202, according to the statistical initial probability, mapping probability and transition probability, the walking state sequence and the hidden behavior sequence of each pedestrian are obtained through prediction by calculating the maximum walking state probability of each pedestrian at each moment to be predicted. Specifically, after the maximum walking state probability at the last prediction time is obtained, the walking state corresponding to the maximum walking state probability is the optimal walking state at the last prediction time, and the upper layer hiding behavior for realizing the maximum walking state probability is used as the optimal hiding behavior at the last prediction time; and obtaining an optimal walking state sequence and a hidden behavior sequence according to the maximum walking state probability of each predicted moment and the recorded optimal hidden behavior of the previous moment by optimal path backtracking.
Given a hidden Markov model, defineO=(O t |t=1,…,T) For walking observable state sequences, defineH=(H t |t=1,…,T) A sequence of behaviors is hidden for the pedestrian, wherein,H=(H t |t=1,…,T) In (1)TThe individual behaviors are not observed, and the behavior at each momentH t There are many options, and the behavior sequence probability calculation relationship is as follows:
Figure DEST_PATH_IMAGE015
in the formula, P: (H 1 ) The initial probability of the behavior is represented,P(H t |H t-1 ) Is represented bytHidden behavior at time-1H t-1 TotHidden behavior of a momentH t The transition probability of (2).
Generating a sequence of walking states on condition that a sequence of hidden behaviors is foundO=(O t |t=1,…,T) Is calculated as follows：
Figure 248837DEST_PATH_IMAGE016
In the formula (I), the compound is shown in the specification,P(O 1 ,O 2 ,…,O t |H 1 ,H 2 ,…,H t ) Representing the mapping probability between the hidden behavior and the sequence of walking states,P(O t |H t ) Is shown intTemporal hidden behaviorH t And the state of walkingO t The mapping probability between them.
Thus, given the initial parameters, the probability that an observable behavior sequence is generated is calculated as follows:
Figure DEST_PATH_IMAGE017
step 2021, firstly, two variables are introduced, and the calculation is carried out from the 1 st moment to be predicted to the 1 st momenttThe probability value of the optimal path of the walking state at the moment to be predicted is obtained, and the optimal path of the walking state at the next moment is recurred according to the probability value, namely
Figure 950339DEST_PATH_IMAGE018
Wherein the content of the first and second substances,P(O t ) To representtThe maximum probability of a walking state at a moment,P(H t |H t-1 ) Is represented bytHidden behavior at time-1H t-1 TotHidden behavior of a momentH t The transition probability of (a) is,P(O t |H t ) Is shown intTemporal hidden behaviorH t And the state of walkingO t The probability of mapping between.
To store memory rollback paths, define
Figure DEST_PATH_IMAGE019
Precursor behaviors of the final layer hidden behaviors of the local optimal path in the pedestrian state are as follows:
Figure 333916DEST_PATH_IMAGE020
for the examples given in tables 2, 3 and 4, it is assumed thatT=3, then:
①ttime =1, the calculation process is as follows:
Figure DEST_PATH_IMAGE021
Figure 166874DEST_PATH_IMAGE022
Figure DEST_PATH_IMAGE023
tthe maximum walking state probability at time =1 is:
Figure 264405DEST_PATH_IMAGE024
(2) recording the previous behavior of the upper layer hiding behavior for realizing the optimal walking state:
Figure DEST_PATH_IMAGE025
(3) continuing to predicttThe probability value of the optimal path of the walking state at the moment of =2 is calculated as follows:
Figure 115686DEST_PATH_IMAGE026
Figure DEST_PATH_IMAGE027
/>
tthe maximum walking state probability at time =2 is:
Figure 906050DEST_PATH_IMAGE028
(4) recording the previous behavior of the upper layer hiding behavior for realizing the optimal walking state:
Figure DEST_PATH_IMAGE029
(5) then, continue to predicttThe probability value of the optimal path of the walking state at the moment of =3 is calculated as follows:
Figure 226304DEST_PATH_IMAGE030
Figure DEST_PATH_IMAGE031
tthe maximum walking state probability at time =3 is:
Figure 767007DEST_PATH_IMAGE032
(6) recording the previous behavior of the upper hidden behavior for realizing the optimal walking state:
Figure DEST_PATH_IMAGE033
step 2022, whenTWhen the time prediction process is terminated, a walking state occursO T The maximum probability of (c) is:
Figure 223527DEST_PATH_IMAGE034
for example, whenTWhen the prediction process is terminated at time =3, a walking state occursO 3 The maximum probability of (c) is:
Figure DEST_PATH_IMAGE035
step 2023,TThe optimal final hidden behavior corresponding to the walking state with the maximum probability at the moment is as follows:
Figure 715951DEST_PATH_IMAGE036
for example,Tthe optimal final hidden behavior corresponding to the walking state with the maximum probability at time =3 is as follows:
Figure DEST_PATH_IMAGE037
step 2024, backtracking process is based onTTime-optimal hidden behavior
Figure 976031DEST_PATH_IMAGE038
FromT-1, starting recurrence and optimal path backtracking, and calculating as follows:
Figure DEST_PATH_IMAGE039
wherein the content of the first and second substances,
Figure 212102DEST_PATH_IMAGE040
to representtAn optimal hiding behavior of the moment in time>
Figure DEST_PATH_IMAGE041
To representtPredecessor behavior of optimal hidden behavior at time + 1. />
Up tot=1 hour obtainedObtaining a hidden behavior sequence which maximizes the probability of walking state sequence
Figure 257550DEST_PATH_IMAGE042
。
For example, by optimal path backtracking, the last step is first computedt= hidden behavior of maximum path reached at moment 3, i.e. found above
Figure DEST_PATH_IMAGE043
。
Figure 888251DEST_PATH_IMAGE044
And then, tracing one step forward through an optimal path backtracking calculation formula, namely that the current behavior with the maximum probability is transferred from the previous step which hidden behavior is transferred from, namely:
Figure DEST_PATH_IMAGE045
the last-but-one step is reached, tracing from which starting behavior the optimal path is transferred, i.e.:
Figure 668251DEST_PATH_IMAGE046
the backtracking process is based onTOptimal hiding behavior at time =3
Figure DEST_PATH_IMAGE047
FromT-1 starts recursion untiltAcquiring optimal hidden behavior sequence when =1H * =(R,A,A) And state sequenceO * =(c,n,n)，tFig. 7 shows the results of the prediction at time points =1 to 3.
The embodiment takes the common behavior in the walking process of the pedestrian as an entry point, provides a pedestrian walking state identification method and a pedestrian hidden behavior identification method, captures the pedestrian state in real time through track matching, can accurately and quickly identify abnormal passenger flow, can effectively avoid the problems of background shielding, illumination and angle compared with an analysis method based on image processing, and reduces the difficulty of feature extraction.
Compared with the existing simulation modeling method, the hidden Markov prediction model-based pedestrian behavior prediction method based on the hidden Markov model can effectively quantify the mapping relation between behaviors and states in complex passenger flow and the hidden incidence relation between different behaviors, and realizes the prediction of hidden behaviors and states of pedestrians through a numerical simulation method. Meanwhile, the hidden behavior characteristics of pedestrians can be effectively depicted, the hidden behavior characteristics are not limited by grids, the calculation efficiency is high, and the high-density passenger flow evolution situation can be accurately reflected.
Example two
The embodiment provides a dynamic passenger flow behavior situation prediction system, which specifically comprises the following modules:
a data acquisition module configured to: acquiring passenger flow track characteristic data, and identifying and obtaining the walking state and hidden behavior of each pedestrian at each historical moment;
a statistics module configured to: counting initial probability, mapping probability and transition probability according to the walking states and hidden behaviors of all pedestrians at all historical moments;
a prediction module configured to: and according to the statistical initial probability, mapping probability and transition probability, predicting to obtain a walking state sequence and a hidden behavior sequence of each pedestrian by calculating the maximum walking state probability of each pedestrian at each moment to be predicted.
It should be noted that, each module in the present embodiment corresponds to each step in the first embodiment one to one, and the specific implementation process is the same, which is not described herein again.
EXAMPLE III
The present embodiment provides a computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the steps of the dynamic passenger flow behavior situation prediction method as described in the first embodiment above.
Example four
The embodiment provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps in the dynamic passenger flow behavior prediction method according to the first embodiment are implemented.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (7)

1. The dynamic passenger flow behavior situation prediction method is characterized by comprising the following steps:
acquiring passenger flow track characteristic data, and identifying and obtaining the walking state and hidden behavior of each pedestrian at each historical moment;
according to the walking states and hidden behaviors of all pedestrians at all historical moments, counting initial probability, mapping probability and transition probability;
according to the statistical initial probability, mapping probability and transition probability, predicting to obtain a walking state sequence and a hidden behavior sequence of each pedestrian by calculating the maximum walking state probability of each pedestrian at each moment to be predicted; the identification method of the walking state comprises the following steps: according to the actual track of the pedestrian in front, obtaining state track templates of the pedestrian in different walking states by moving the spatial position distance on the vertical axis and the moving time length on the horizontal axis; calculating error probability between the trajectory of the pedestrian and the trajectory templates in different states; taking the walking state corresponding to the minimum error probability as the final walking state of the pedestrian;
the identification method of the hidden behaviors comprises the following steps: recognizing the hidden behavior of the pedestrian according to the distance between the pedestrian and the pedestrian in front at the time t and the distance change between the time t and the time t-1;
the maximum walking state probability is expressed as:
P(O t )＝max{P(O t-1 )·P(H t |H t-1 )·P(O t |H t )}
wherein, P (O) t ) Represents the maximum probability of walking state at time t, P (H) t |H t-1 ) Representing hidden behavior H from time t-1 t-1 Hidden behavior H to time t t Transition probability of (A), P (O) t |H t ) Representing a hidden behavior H at time t t And a running state O t The probability of mapping between.
2. The dynamic passenger flow behavior situation prediction method according to claim 1, characterized in that the initial probability is: the probability of various hidden behaviors occurring at the initial time in all historical times;
or, the mapping probability is: the probability of different observable walking states shown under various hidden behaviors; or, the transition probability is: the probability of transition between classes of hidden behavior at two consecutive times.
3. The dynamic passenger flow behavior situation prediction method according to claim 1, characterized in that the upper hidden behavior that achieves the maximum walking state probability at the last prediction moment is expressed as:
Figure FDA0003928514990000011
wherein, O T Indicating the walking state at time T.
4. The dynamic passenger flow behavior situation prediction method according to claim 1, characterized in that after obtaining the maximum walking state probability at the last prediction time, the walking state at the last prediction time corresponding to the maximum walking state probability is the optimal walking state at the last prediction time, and the upper layer hidden behavior realizing the maximum walking state probability is taken as the optimal hidden behavior at the last prediction time; and obtaining the optimal walking state and the optimal hidden behavior at the corresponding moment according to the maximum walking state probability at each predicted moment by backtracking the optimal path, and obtaining an optimal walking state sequence and a hidden behavior sequence.
5. Dynamic passenger flow behavior situation prediction system, characterized by, includes:
a data acquisition module configured to: acquiring passenger flow track characteristic data, and identifying and obtaining the walking state and hidden behavior of each pedestrian at each historical moment;
a statistics module configured to: counting initial probability, mapping probability and transition probability according to the walking states and hidden behaviors of all pedestrians at all historical moments;
a prediction module configured to: according to the statistical initial probability, mapping probability and transition probability, predicting to obtain a walking state sequence and a hidden behavior sequence of each pedestrian by calculating the maximum walking state probability of each pedestrian at each moment to be predicted;
the identification method of the walking state comprises the following steps: according to the actual track of the pedestrian in front, obtaining state track templates of the pedestrian in different walking states by moving the space position distance on the vertical axis and the moving time length on the horizontal axis; calculating error probability between the trajectory of the pedestrian and the trajectory templates in different states; taking the walking state corresponding to the minimum error probability as the final walking state of the pedestrian;
the identification method of the hidden behavior comprises the following steps: recognizing the hidden behavior of the pedestrian according to the distance between the pedestrian and the pedestrian in front at the time t and the distance change between the time t and the time t-1;
the maximum walking state probability is expressed as:
P(O t )＝max{P(O t-1 )·P(H t |H t-1 )·P(O t |H t )}
wherein, P (O) t ) Represents the maximum probability of walking state at time t, P (H) t |H t-1 ) Representing hidden behavior H from time t-1 t-1 Hidden behavior H to time t t Transition probability of (A), P (O) t |H t ) Representing a hidden behavior H at time t t And a running state O t The probability of mapping between.
6. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for dynamic passenger flow behavior prediction according to any one of claims 1-4.
7. Computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor realizes the steps in the method for dynamic passenger flow behavior prediction according to any of the claims 1-4 when executing the program.
CN202211154363.1A 2022-09-22 2022-09-22 Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment Active CN115239767B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211154363.1A CN115239767B (en) 2022-09-22 2022-09-22 Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211154363.1A CN115239767B (en) 2022-09-22 2022-09-22 Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment

Publications (2)

Publication Number Publication Date
CN115239767A CN115239767A (en) 2022-10-25
CN115239767B true CN115239767B (en) 2023-03-24

Family

ID=83666986

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211154363.1A Active CN115239767B (en) 2022-09-22 2022-09-22 Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment

Country Status (1)

Country Link
CN (1) CN115239767B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109272168A (en) * 2018-10-09 2019-01-25 南京地铁集团有限公司 A kind of urban track traffic for passenger flow trend method

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150347918A1 (en) * 2014-06-02 2015-12-03 Disney Enterprises, Inc. Future event prediction using augmented conditional random field
EP3706034A1 (en) * 2019-03-06 2020-09-09 Robert Bosch GmbH Movement prediction of pedestrians useful for autonomous driving
CN111329485A (en) * 2020-02-25 2020-06-26 北京机械设备研究所 Gait recognition method and device based on IMU
CN114511999B (en) * 2020-11-17 2023-09-01 宇通客车股份有限公司 Pedestrian behavior prediction method and device
CN114898293A (en) * 2022-05-20 2022-08-12 南京理工大学 Pedestrian crossing group multi-mode trajectory prediction method for automatically driving automobile

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109272168A (en) * 2018-10-09 2019-01-25 南京地铁集团有限公司 A kind of urban track traffic for passenger flow trend method

Also Published As

Publication number Publication date
CN115239767A (en) 2022-10-25

Similar Documents

Publication Publication Date Title
Behrmann et al. Unified fully and timestamp supervised temporal action segmentation via sequence to sequence translation
Bhattacharyya et al. Accurate and diverse sampling of sequences based on a “best of many” sample objective
Wu et al. Learning to anticipate egocentric actions by imagination
CN109146921B (en) Pedestrian target tracking method based on deep learning
Zhang et al. Artificial intelligence and its applications
Chen et al. Pointgpt: Auto-regressively generative pre-training from point clouds
CN109671102A (en) A kind of composite type method for tracking target based on depth characteristic fusion convolutional neural networks
Ren et al. A novel squeeze YOLO-based real-time people counting approach
CN105654139A (en) Real-time online multi-target tracking method adopting temporal dynamic appearance model
Zhang et al. Social-IWSTCNN: A social interaction-weighted spatio-temporal convolutional neural network for pedestrian trajectory prediction in urban traffic scenarios
CN111291098A (en) Track mode mining method and device
Li et al. Pedestrian trajectory prediction with learning-based approaches: A comparative study
CN115512152A (en) Ship track classification method and system combining CNN (CNN) neural network and LSTM neural network
CN112861758A (en) Behavior identification method based on weak supervised learning video segmentation
Anderson et al. Stochastic sampling simulation for pedestrian trajectory prediction
Shao et al. Failure detection for motion prediction of autonomous driving: An uncertainty perspective
CN115272712A (en) Pedestrian trajectory prediction method fusing moving target analysis
Cong et al. DACR-AMTP: adaptive multi-modal vehicle trajectory prediction for dynamic drivable areas based on collision risk
Zhu et al. Tri-HGNN: Learning triple policies fused hierarchical graph neural networks for pedestrian trajectory prediction
Wang et al. Multi-agent trajectory prediction with spatio-temporal sequence fusion
CN115239767B (en) Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment
Zhou et al. Sa-sgan: A vehicle trajectory prediction model based on generative adversarial networks
Zhu et al. Fast Adaptive Character Animation Synthesis Based on Greedy Algorithm
CN112487920B (en) Convolution neural network-based crossing behavior identification method
Wu et al. SMART: Scalable Multi-agent Real-time Simulation via Next-token Prediction

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