CN111126438B - Driving behavior recognition method and system - Google Patents

Driving behavior recognition method and system Download PDF

Info

Publication number
CN111126438B
CN111126438B CN201911156134.1A CN201911156134A CN111126438B CN 111126438 B CN111126438 B CN 111126438B CN 201911156134 A CN201911156134 A CN 201911156134A CN 111126438 B CN111126438 B CN 111126438B
Authority
CN
China
Prior art keywords
driving behavior
sequence
speed
minimum
confusion
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
CN201911156134.1A
Other languages
Chinese (zh)
Other versions
CN111126438A (en
Inventor
王震坡
刘鹏
崔丁松
张照生
武烨
龙超华
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 Bitnei Corp ltd
Beijing Institute of Technology BIT
Original Assignee
Beijing Bitnei Corp ltd
Beijing Institute of Technology BIT
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 Bitnei Corp ltd, Beijing Institute of Technology BIT filed Critical Beijing Bitnei Corp ltd
Priority to CN201911156134.1A priority Critical patent/CN111126438B/en
Publication of CN111126438A publication Critical patent/CN111126438A/en
Application granted granted Critical
Publication of CN111126438B publication Critical patent/CN111126438B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G06F18/24155—Bayesian classification

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention discloses a driving behavior recognition method and system, and relates to the technical field of automobile driving. The method comprises the following steps: dividing the driving behavior data by using a driving behavior sequence variable point detection algorithm to obtain divided driving behavior fragments; calculating the confusion degree of LDA models with different clustering numbers; comparing to obtain minimum confusion degree; and inputting the driving behavior segment characteristic frequency statistical matrix into an LDA model corresponding to the minimum confusion degree to obtain the driving behavior category. According to the method, the driving behavior data is segmented by adopting a driving behavior sequence variable point detection algorithm, so that the driving behavior data can be segmented more accurately, and the formed driving behavior segments are more complete; and measuring the LDA model by using the confusion degree, and obtaining the driving behavior category according to the LDA model with the optimal clustering number, wherein the obtained driving behavior category is more complete and comprehensive.

Description

Driving behavior recognition method and system
Technical Field
The invention relates to the technical field of automobile driving, in particular to a driving behavior identification method and system.
Background
With the rapid development of intelligent technologies of automobiles, the identification and prediction of driver behaviors have become key technologies for establishing advanced driving assistance systems (Advanced DrivingAssistant System, ADAS), which can realize the full cooperative driving of drivers and vehicles. Driving behavior recognition has been the subject of investigation from many different perspectives since the 50 s of the 20 th century. The research shows that the driving behavior recognition can be applied to the fields of traffic flow models, road infrastructure optimization, traffic jam prediction, establishment of advanced personalized driving assistance systems, design of automatic driving vehicles and the like.
The existing driving behavior recognition method of the vehicle is mainly researched aiming at specific driving scenes, and driving behavior states are mainly divided into four types of acceleration, deceleration, uniform speed and idle speed, so that the recognition research of detailed driving behavior features is not deep enough, and the driving behavior segments are poor in segmentation precision and ambiguous in clustering features. Therefore, there is a problem in that the classification of driving behavior states is small.
Disclosure of Invention
The invention aims to provide a driving behavior recognition method and system, which solve the problem of less classification of driving behavior states.
In order to achieve the above object, the present invention provides the following solutions:
a driving behavior recognition method, comprising:
acquiring driving behavior data;
carrying out symbolization processing on the driving behavior data to obtain a symbolized speed time sequence;
dividing the driving behavior data by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain divided driving behavior fragments;
preprocessing the driving behavior data to obtain a driving behavior fragment characteristic frequency statistical matrix;
obtaining an LDA model, and initializing the clustering number of the LDA model;
inputting the characteristic frequency statistical matrix of the driving behavior segment into the LDA model to obtain a first driving behavior class;
Calculating the confusion degree of the LDA model according to the first driving behavior category, and recording the confusion degree and the LDA model corresponding to the confusion degree;
judging whether the clustering number of the LDA model is equal to a preset number or not to obtain a first judging result;
if the first judgment result is negative, the clustering number of the LDA models is increased, and the step of inputting the driving behavior segment characteristic frequency statistical matrix into the LDA models to obtain a first driving behavior class is returned, and the confusion degree and the LDA models corresponding to the confusion degree are updated and recorded;
the first judgment result is that all recorded confusion degrees are compared, and the minimum confusion degree is obtained;
obtaining an LDA model corresponding to the minimum confusion according to the minimum confusion;
and inputting the driving behavior segment characteristic frequency statistical matrix into an LDA model corresponding to the minimum confusion degree to obtain a driving behavior class.
Optionally, the performing symbolization processing on the driving behavior data to obtain a symbolized speed time sequence specifically includes:
acquiring a speed time sequence S in the driving behavior data velocity ＝{v 1 ,v 2 ,…,v n }, v is 1 ,v 2 ,…,v n Representing sequence points, n representing the total number of the sequence points;
Obtaining a symbolization speed time sequence S according to a formula (1), wherein the formula (1):
in the formula (1), v j Representing the velocity time series S velocity J represents the sequence point number, j E [2, n-1 ]]。
Optionally, the dividing the driving behavior data by using a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain a divided driving behavior segment specifically includes:
calculating the information entropy of speed sequence points in the symbolized speed time sequence according to the formula (2):
H(S)＝-∑p(s i )log(p(s i ))(i＝1,2,...,n) (2)；
in the formula (2), H (S) represents the velocity sequenceInformation entropy of column points, S represents the symbolization speed time sequence, s= { S 1 ,s 2 ,…,s n }，s 1 ,s 2 ,…,s n Symbolized values representing sequence points, i.e., speed sequence points; p(s) i ) Representing the ith velocity sequence point s i Is a probability of occurrence of (2); s is(s) i =1 represents acceleration, s i =0 denotes constant velocity, s i -1 represents deceleration;
calculating the ith velocity sequence point s according to equation (3) i Local minimum entropy E of (2) j ：
E j ＝H(S Forward )+H(S Backward ) (3)；
In the formula (3), H (S Forward ) Representing the velocity sequence points s i Entropy of information of previous symbolization speed time sequence, S Forward ＝{s 1 ,s 2 ,…,s i }；H(S Backward ) Representing the velocity sequence points s i Information entropy of the subsequent symbolization speed time sequence, S Backward ＝{s i+1 ,s i+2 ,…,s n }；
According to the symbolized speed time sequence, driving state dividing points of the symbolized speed time sequence are obtained by adopting a variable step sliding window method;
And acquiring all driving state dividing points, wherein driving behavior data between any two adjacent driving state dividing points are divided driving behavior fragments.
Optionally, the driving state dividing point is obtained by adopting a variable step sliding window method according to the symbolization speed time sequence, which specifically includes:
acquiring a preset minimum length of a window, a preset maximum length of the window, a preset window offset step length and a preset window length;
initializing a sequence number j and a window length l of a speed sequence point;
acquiring a sequence to be detected according to the sequence number of the initialized speed sequence point and the initialized window length;
acquiring a first region from the sequence to be detected according to the window offset step length and the window length;
calculating the minimum local entropy of all the speed sequence points in the first area according to a formula (3), and obtaining a minimum local entropy set;
calculating the smallest minimum local entropy in the smallest local entropy set according to formula (4):
E s ＝min{E j+f ,E j+f+1 ,...,E j+l-f-1 ,E j+l-f } (4)
e in formula (4) s Representing the smallest minimum local entropy;
judging whether the minimum local entropy is unique in the first area, and obtaining a second judging result;
if the second judgment result is yes, carrying out third judgment;
If the second judgment result is negative, fourth judgment is carried out;
the third judgment is to judge whether the reduction times of all the minimum local entropies before the speed sequence point corresponding to the minimum local entropies and the increase times of all the minimum local entropies after the speed sequence point corresponding to the minimum local entropies are simultaneously larger than the preset times or not, so as to obtain a third judgment result; if the third judgment result is negative, carrying out the fourth judgment;
the third judgment result is that a speed sequence point corresponding to the minimum local entropy is the driving state dividing point, and the fourth judgment is carried out;
the fourth judgment is to judge whether the length of the window is larger than or equal to the maximum length of the window, so as to obtain a fourth judgment result;
the fourth judgment result is that the sequence number j of the speed sequence point is updated, 1 is added to j, the window length l is equal to the minimum length of the window, and the fifth judgment is carried out;
if the fourth judgment result is negative, updating the window length, adding 1 to the window length l, and carrying out the fifth judgment;
the fifth judgment is to judge whether j+l is greater than or equal to n, so as to obtain a fifth judgment result;
The fifth judgment result is that the driving state dividing points are obtained;
and if the fifth judgment result is negative, updating the sequence to be detected according to the updated window length and the updated sequence number of the speed sequence point, and returning to the step of acquiring a first area from the sequence to be detected according to the window offset step length and the window length.
Optionally, the preprocessing the driving behavior data to obtain a driving behavior segment feature frequency statistical matrix specifically includes:
normalizing the driving behavior characteristics in the driving behavior data to obtain normalized driving behavior characteristics; the driving behavior features include a speed feature, a current feature, and an acceleration feature;
dividing each driving behavior feature into 20 feature intervals according to the difference value between the maximum value and the minimum value of the driving behavior feature;
calculating the frequency of the speed, the current and the acceleration in each driving behavior segment in each characteristic interval by a counting statistical method to obtain a frequency matrix;
and counting all the driving behavior fragments to obtain frequency matrixes of all the driving behavior fragments, and combining the frequency matrixes of all the driving behavior fragments into a characteristic frequency statistic matrix of the driving behavior fragments.
Optionally, the calculating the confusion degree of the LDA model according to the first driving behavior category, and recording the confusion degree and the LDA model corresponding to the confusion degree specifically includes:
calculating the confusion according to equation (5):
in the formula (5), p (D) represents the degree of confusion; d represents the number of the driving behavior segments, D represents the sequence number of the driving behavior segments，d∈D；p(ω d ) Representing the occurrence probability of each driving behavior feature in each driving behavior segment; p (omega) d ) =p (z|d) ×p (ω|z), p (z|d) represents the probability of occurrence of each of the driving behavior categories in each of the driving behavior segments, and p (ω|z) represents the probability of occurrence of each of the driving behavior features in each of the driving behavior categories; r is R d Representing the total length of the driving behavior segment.
A driving behavior recognition system, comprising:
the acquisition module is used for acquiring driving behavior data;
the symbolizing speed time sequence module is used for symbolizing the driving behavior data to obtain a symbolizing speed time sequence;
the driving behavior segment module is used for dividing the driving behavior data by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain divided driving behavior segments;
The driving behavior segment characteristic frequency statistical matrix module is used for preprocessing the driving behavior data to obtain a driving behavior segment characteristic frequency statistical matrix;
the initialization module is used for acquiring an LDA model and initializing the clustering number of the LDA model;
the first driving behavior category module is used for inputting the driving behavior segment characteristic frequency statistical matrix into the LDA model to obtain a first driving behavior category;
a confusion degree module, configured to calculate a confusion degree of the LDA model according to the first driving behavior category, and record the confusion degree and an LDA model corresponding to the confusion degree;
the first judging module is used for judging whether the clustering number of the LDA model is equal to the preset number or not to obtain a first judging result; if the first judgment result is negative, executing an updating module; the first judgment result is yes, and a minimum confusion degree module is executed;
the updating module is used for increasing the clustering number of the LDA models, executing a first driving behavior category module, updating and recording the confusion degree and the LDA models corresponding to the confusion degree;
the minimum confusion degree module is used for comparing the confusion degrees of all records to obtain the minimum confusion degree;
The LDA model module is used for obtaining an LDA model corresponding to the minimum confusion according to the minimum confusion;
and the driving behavior category module is used for inputting the driving behavior segment characteristic frequency statistical matrix into the LDA model corresponding to the minimum confusion degree to obtain the driving behavior category.
Optionally, the symbolizing speed time sequence module specifically includes:
an acquisition unit for acquiring a speed time series S in the driving behavior data velocity ＝{v 1 ,v 2 ,…,v n }, v is 1 ,v 2 ,…,v n Representing sequence points, n representing the total number of the sequence points;
a calculating unit, configured to obtain a symbolized speed time sequence S according to formula (1), where formula (1):
in the formula (1), v j Representing the velocity time series S velocity J represents the sequence point number, j E [2, n-1 ]]。
Optionally, the driving behavior segment module specifically includes:
an information entropy unit, configured to calculate information entropy of speed sequence points in the symbolized speed time sequence according to formula (2):
H(S)＝-∑p(s i )log(p(s i ))(i＝1,2,...,n) (2)；
in the formula (2), H (S) represents the information entropy of the speed sequence point, S represents the symbolized speed time sequence, and s= { S 1 ,s 2 ,…,s n }，s 1 ,s 2 ,…,s n Symbolizing representing sequence pointsNumerical value, i.e. speed sequence point; p(s) i ) Representing the ith velocity sequence point s i Is a probability of occurrence of (2); s is(s) i =1 represents acceleration, s i =0 denotes constant velocity, s i -1 represents deceleration;
a local minimum entropy unit for calculating the ith speed sequence point s according to formula (3) i Local minimum entropy E of (2) j ：
E j ＝H(S Forward )+H(S Backward ) (3)；
In the formula (3), H (S Forward ) Representing the velocity sequence points s i Entropy of information of previous symbolization speed time sequence, S Forward ＝{s 1 ,s 2 ,…,s i }；H(S Backward ) Representing the velocity sequence points s i Information entropy of the subsequent symbolization speed time sequence, S Backward ＝{s i+1 ,s i+2 ,…,s n }；
The driving state dividing point unit is used for obtaining driving state dividing points of the symbolized speed time sequence by adopting a variable step sliding window method according to the symbolized speed time sequence;
and the driving behavior segment unit is used for acquiring all driving state division points, and driving behavior data between any two adjacent driving state division points are divided driving behavior segments.
Optionally, the confusion degree module specifically includes:
a confusion unit for calculating the confusion according to formula (5):
in the formula (5), p (D) represents the degree of confusion; d represents the number of the driving behavior fragments, D represents the sequence number of the driving behavior fragments, and D epsilon D; p (omega) d ) Representing the occurrence probability of each driving behavior feature in each driving behavior segment; p (omega) d ) =p (z|d) ×p (ω|z), p (z|d) representing each of the driving behaviorsThe occurrence probability of each driving behavior category in a segment, p (omega|z) represents the occurrence probability of each driving behavior feature in each driving behavior category; r is R d Representing the total length of the driving behavior segment.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a driving behavior recognition method and system. The method comprises the following steps: acquiring driving behavior data; carrying out symbolization processing on driving behavior data to obtain a symbolized speed time sequence; dividing driving behavior data by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain divided driving behavior fragments; preprocessing driving behavior data to obtain a driving behavior fragment characteristic frequency statistical matrix; obtaining an LDA model, and initializing the clustering number of the LDA model; inputting the characteristic frequency statistical matrix of the driving behavior segment into an LDA model to obtain a first driving behavior class; calculating the confusion degree of the LDA model according to the first driving behavior category, and recording the confusion degree and the LDA model corresponding to the confusion degree; judging whether the clustering number of the LDA models is equal to the preset number or not, and obtaining a first judging result; if the first judgment result is negative, the clustering number of the LDA models is increased, and the step of inputting the driving behavior segment characteristic frequency statistical matrix into the LDA models to obtain a first driving behavior class is returned, and the confusion degree and the LDA models corresponding to the confusion degree are updated and recorded; if the first judgment result is yes, comparing all recorded confusion degrees to obtain the minimum confusion degree; obtaining an LDA model corresponding to the minimum confusion according to the minimum confusion; and inputting the driving behavior segment characteristic frequency statistical matrix into an LDA model corresponding to the minimum confusion degree to obtain the driving behavior category. According to the method, the driving behavior data is segmented by adopting a driving behavior sequence variable point detection algorithm to obtain segmented driving behavior fragments, the driving behavior data can be segmented more accurately, the segmentation points of the driving behavior state can be identified more accurately, the misjudgment condition of the driving behavior state is reduced, and the formed driving behavior fragments are more complete; and measuring the LDA model by using the confusion degree to obtain the optimal clustering number, and obtaining the driving behavior category according to the LDA model with the optimal clustering number, wherein the obtained driving behavior category is more complete and comprehensive.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a driving behavior recognition method according to an embodiment of the present invention;
FIG. 2 is a flowchart of a variable step sliding window method according to an embodiment of the present invention;
FIG. 3 is a system configuration diagram of a driving behavior recognition system according to an embodiment of the present invention;
fig. 4 is a block diagram of an LDA model according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
The invention provides a driving behavior recognition method, and fig. 1 is a flowchart of the driving behavior recognition method provided by the embodiment of the invention. Referring to fig. 1, the driving behavior recognition method includes:
step 101, driving behavior data is acquired. And acquiring speed data, current data and acceleration data of 50 pure electric passenger cars from the national monitoring and management platform of the new energy automobile as driving behavior data.
And 102, performing symbolization processing on the driving behavior data to obtain a symbolized speed time sequence.
Step 102, specifically includes:
acquiring a speed time series S in driving behavior data velocity ＝{v 1 ,v 2 ,…,v n }, v is 1 ,v 2 ,…,v n Representing sequence points, n representing the total number of sequence points. The driving behavior recognition method of the invention uses the speed data to divide the driving behavior data, and the speed data in the driving behavior data can not be directly applied to the driving behavior sequence variable point detection algorithm to sign the speed time sequence.
Obtaining a symbolization speed time sequence S according to a formula (1), wherein the formula (1):
In the formula (1), v j Representing a velocity time series S velocity J represents the sequence point number, j E [2, n-1 ]]In this embodiment, the first sequence point and the last sequence point of the speed time sequence are not subjected to symbolization. Step 102 represents the original speed time series as a symbolized speed time series including-1, 0 and 1 through symbolization processing, wherein the symbolized speed time series can be used for a driving behavior sequence change point detection algorithm based on minimum local entropy.
In order to improve the information accuracy of the symbolization speed time sequence S and eliminate the influence of certain short-time operation or uniform operation on the driving behavior data segmentation, the following rule is added on the basis of the formula (1): if the symbolic value of any sequence point is 0, but the symbolic values of the previous sequence point and the next sequence point of the sequence point are consistent and are not 0, the symbolic values of the sequence point are consistent with the state of the symbolic values of the previous sequence point and the next sequence point.
And 103, dividing the driving behavior data by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain divided driving behavior fragments.
Step 103, specifically includes:
calculating the information entropy of speed sequence points in the symbolized speed time sequence according to the formula (2):
H(S)＝-∑p(s i )log(p(s i ))(i＝1,2,...,n) (2)；
in the formula (2), H (S) represents information entropy of a velocity sequence point, S represents a symbolized velocity time sequence, and s= { S 1 ,s 2 ,…,s n }，s 1 ,s 2 ,…,s n Symbolized values representing sequence points, i.e., speed sequence points; p(s) i ) Representing the ith velocity sequence point s i Is a probability of occurrence of (2); s is(s) i =1 represents acceleration, s i =0 denotes constant velocity, s i -1 represents deceleration; i represents the sequence number of the velocity sequence point.
Calculating an ith velocity sequence point s according to formula (3) i Local minimum entropy E of (2) j ：
E j ＝H(S Forward )+H(S Backward ) (3)；
In the formula (3), H (S Forward ) Representing a velocity sequence point s i Entropy of information of previous symbolization speed time sequence, S Forward ＝{s 1 ,s 2 ,…,s i }；H(S Backward ) Representing a velocity sequence point s i Information entropy of the subsequent symbolization speed time sequence, S Backward ＝{s i+1 ,s i+2 ,…,s n }。
And obtaining all driving state division points in the symbolized speed time sequence by adopting a variable step sliding window method according to the symbolized speed time sequence. As can be seen from the formula (3), any one speed sequence point in the symbolic speed time sequence may be a driving state division point, and only when the minimum local entropy of the speed sequence point is the minimum value of the minimum local entropy of all the speed sequence points, the speed sequence point is a real driving state division point, and the boundary reliability determined by the driving state division point is maximum; if the driving state division point is not the real driving state division point, the minimum local entropy of the driving state division point is necessarily greater than the minimum local entropy of the real driving state division point. Since the driving behavior is a short period process, the driving behavior class of the driving behavior data is related only to the information entropy of the adjacent speed sequence points in the driving behavior data. Therefore, the invention adopts a variable step length sliding window method to calculate the minimum local entropy of the symbolized speed time sequence and the real driving state dividing point. If the driving state division points are not contained in the window, the change of the minimum local entropy of all possible driving state division points in the window is irregular. Although the minimum local entropy corresponding to different window lengths and different types of driving behavior states is different, as long as the driving state dividing points exist in the window, the value of the minimum local entropy of the possible driving state dividing points is a process of changing from large to small from the first possible driving state dividing point, then the value of the minimum local entropy reaches the minimum at the real driving state dividing point, and then the value gradually changes from small to large.
Fig. 2 is a flowchart of a variable step sliding window method according to an embodiment of the present invention, referring to fig. 2, the variable step sliding window method includes:
step 1031, setting a minimum length of the window, a maximum length of the window, a window offset step length and a window length, wherein the setting range of the minimum length of the window is within 5 seconds(s) -10s, and the setting range of the maximum length of the window is within 10s-20s. The shorter the window length setting is, the higher the sensitivity to the driving behavior state change is, and the longer the window length setting is, the lower the sensitivity to the driving behavior state change is. Since the driving operation behavior is a process quantity, too short window length may cause too sensitive recognition of the driving behavior state, too long window length may cause too sparse recognition of the driving behavior state, and in this embodiment, the minimum window length Minl is 7 seconds, the maximum window length Maxl is 15 seconds, and the window offset step f is 2.
Step 1032, obtaining a preset minimum length Minl of the window, a preset maximum length Maxl of the window, a window offset step f and a window length l.
Step 1033, initializing a sequence number j and a window length l of the velocity sequence point.
Step 1034, obtaining the sequence number of the initialized speed sequence point and the initialized window length, and the speed sequence point corresponding to the sequence number of the initialized speed sequence point according to the sequence number of the initialized speed sequence point and the initialized window length.
Step 1035, according to the initialized speed sequence point and the initialized window length, acquiring a sequence to be detected, specifically: intercepting a symbolized speed time sequence segment with a window length according to the initialized speed sequence point and the window length, namely marking a sequence to be detected as S t ＝{s j ,s j+1 ,…,s j+l ' l represents window length, s j ,s j+1 ,…,s j+l A speed sequence point representing the sequence to be detected.
Step 1036, obtaining a first region from the sequence to be detected according to the window offset step length and the window length, wherein the first region is S e ＝{s j+f ,s j+f+1 ,…,s j+l-f }，s j+f ,s j+f+1 ,…,s j+l-f Representing a velocity sequence point in the first region. In order to prevent too little of the time series of symbolizing speeds before any speed sequence point or too little of the time series of symbolizing speeds after any speed sequence point, which results in the minimum local entropy being unreliable, a first area is acquired from the sequence to be detected according to the window offset step size and the window length. If only one sequence point or no sequence exists in the sequence to be detected, the entropy value of the minimum local entropy cannot be calculated, so that the first area is acquired from the sequence to be detected according to the window offset step length and the window length.
Step 1037, calculating the minimum local entropy of all the velocity sequence points in the first region according to the formula (3), and obtaining a minimum local entropy set, wherein the minimum local entropy set is E= { E j+f ,E j+f+1 ,...,E j+l-f-1 ,E j+l-f }，E j+f ,E j+f+1 ,…,E j+l-f-1 ,E j+l-f Representing the minimum of all velocity sequence points in the first regionLocal entropy.
Step 1038, calculating the smallest local entropy of the smallest local entropy set according to formula (4).
E s ＝min{E j+f ,E j+f+1 ,...,E j+l-f-1 ,E j+l-f } (4)
E in formula (4) s Representing the smallest minimum local entropy.
Step 1039, determining whether the minimum local entropy is unique in the first region, and obtaining a second determination result.
If the second determination result is yes, a third determination is made in step 10310.
And if not, performing a fourth determination in step 10312.
Step 10310, determining whether the decreasing times of all the minimum local entropies before the speed sequence point corresponding to the minimum local entropies and the increasing times of all the minimum local entropies after the speed sequence point corresponding to the minimum local entropies are larger than the preset times p at the same time for the third time, and obtaining a third determination result. In this embodiment, the preset number p is 2. If the third determination result is yes, go to step 10311; and if not, performing a fourth determination in step 10312.
In step 10311, the speed sequence point corresponding to the smallest local entropy is the driving state dividing point, and the fourth determination step 10312 is performed.
Step 10312, determining whether the window length is greater than or equal to the maximum length of the window for the fourth time, to obtain a fourth determination result. If the fourth determination result is yes, go to step 10313; and if not, step 10314 is performed.
Step 10313, update the sequence number j of the velocity sequence point, let j add 1, let the window length l equal to the minimum length of the window, and make the fifth judgment in step 10315.
Step 10314, updating the window length, adding 1 to the window length l, and performing the fifth determination in step 10315.
Step 10315, determining whether j+l is greater than or equal to n for the fifth time to obtain a fifth determination result. If yes, go to step 10316; and if not, step 10317 is performed.
Step 10316, a driving state division point is acquired.
Step 10317, updating the sequence to be detected according to the updated window length and the sequence number of the updated speed sequence point, and returning to step 1036 "acquire the first region from the sequence to be detected according to the window offset step length and the window length".
The driving behavior sequence variable point detection algorithm can divide a long driving behavior data into a plurality of tiny driving behavior fragments, and the driving state dividing point of each driving behavior fragment can represent the change of the driving state of a driver. Since the LDA model cannot directly process the driving behavior data, the driving behavior data is segmented by adopting a driving behavior sequence variable point detection algorithm to obtain fine driving behavior segments, and the driving behavior segments can be applied to the LDA model.
And acquiring all driving state dividing points, wherein driving behavior data between any two adjacent driving state dividing points are divided driving behavior fragments.
And 104, preprocessing the driving behavior data to obtain a driving behavior segment characteristic frequency statistical matrix. The LDA (Latent Dirichlet Allocation) model is a document topic generation model, also called a three-layer Bayesian probability model, comprises word, topic and document three-layer structures, is also a hidden dirichlet allocation model, and is a generation type unsupervised machine learning algorithm.
Step 104 specifically includes:
and carrying out normalization processing on the driving behavior characteristics in the driving behavior data to obtain normalized driving behavior characteristics. The driving behavior features include a speed feature, a current feature, and an acceleration feature.
Dividing each driving behavior feature into 20 feature intervals according to the difference value between the maximum value and the minimum value of the driving behavior feature, and particularly dividing the normalized speed feature into 20 feature intervals according to the difference value between the maximum value and the minimum value of the speed feature; dividing the normalized current characteristic into 20 characteristic intervals according to the difference value between the maximum value and the minimum value of the current characteristic; and equally dividing the normalized acceleration characteristic into 20 characteristic sections according to the difference value between the maximum value and the minimum value of the acceleration characteristic.
And calculating the frequency of the speed, the current and the acceleration in each driving behavior segment in each characteristic interval by a counting statistical method to obtain a frequency matrix. The frequency matrix is a 1×60 matrix including 3 driving behavior features and 20 feature intervals of each driving behavior feature.
And counting all the driving behavior fragments to obtain frequency matrixes of all the driving behavior fragments, and combining the frequency matrixes of all the driving behavior fragments into a characteristic frequency statistic matrix of the driving behavior fragments. The driving behavior segment characteristic frequency statistical matrix is a matrix of D multiplied by 60, and D represents the number of driving behavior segments.
Step 105, obtaining an LDA model, and initializing the clustering number of the LDA model. FIG. 4 is a block diagram of an LDA model provided by an embodiment of the present invention, in FIG. 4, α represents a hyper-parameter of Dirichlet (Dirichlet) distribution of a topic under each driving behavior segment; θ d Representing the distribution of driving behavior categories under the d-th driving behavior segment; phi (phi) k Representing a distribution of driving behavior characteristics under a kth subject; z d,m Representing the driving behavior category of the j-th speed sequence point in the d-th driving behavior segment; omega d,m Representing the driving behavior category corresponding to the j sequence point in the finally generated d driving behavior segment; m is M d A number representing 3 features of speed, current and acceleration; d represents the number of driving behavior segments, D represents the sequence number of the driving behavior segments, D epsilon D; k represents the number of potential driving behavior categories-topics, K represents the topic number, K epsilon K; beta represents the hyper-parameters of the Dirichlet distribution of the physical features under each topic. In this embodiment, the document of the LDA model corresponds to the driving behavior segment, the topic corresponds to the driving behavior class, and the word corresponds to the driving behavior feature, so as to obtain the distribution situation of the driving behavior class in different driving behavior states, where the driving behavior class is composed of various driving behavior features. In the LDA model, each driving behavior segment is constructed as a combination of K topics, assuming K is a known numberAnd does not change.
The generation process of the LDA model comprises the following steps:
(1)α→θ d →z d,m i.e. document-topic procedure. Document-topic procedure means that a topic class is generated in the distribution of "driving behavior segment-driving behavior class" in the d-th driving behavior segment, the document-topic procedure obeys the following distribution:
the theme distribution theta of the d-th driving behavior segment is generated by sampling from Dirichlet (Dirichlet) distribution of the super parameter alpha d ：
θ d ～Dir(α) (6)
From the topic distribution θ d The subject z of the mth driving behavior category in the mth driving behavior segment is generated by sampling d,m ：
z d,m ～Mult(θ d ) (7)
(2)β→φ k →ω d,m |k＝z d,m I.e. a topic-word process. The topic-word process is represented at k=z d,m Subject z under the limitation of (2) d,m In the distribution of the driving behavior category-driving behavior characteristics, a driving behavior category is generated, and the theme-word process obeys the following distribution:
sampling from a Dirichlet distribution of a superparameter beta to generate a topic z d,m Is a distribution phi of driving behavior characteristics k ：
φ k ～Dir(β) (8)
From the distribution phi of driving behaviour characteristics k Mid-sampling to generate driving behavior class omega d,m ：
ω d,m ～Mult(φ k ) (9)
(3) And (3) repeating the steps (1) - (2) until the frequency statistical matrix of all driving behavior segments is traversed, and obtaining the distribution situation of the speed characteristic, the acceleration characteristic and the current characteristic of each driving behavior type.
The LDA model needs to have two parameters of driving behavior fragment-driving behavior category distribution and driving behavior category-driving behavior characteristic distribution, and the two parameters need to be deduced through training and learning; the LDA model comprises a variation inference algorithm and an Expectation-maximization (EM) algorithm, and training and learning are carried out on the LDA model based on the EM algorithm.
And 106, inputting the characteristic frequency statistical matrix of the driving behavior segment into an LDA model to obtain a first driving behavior class.
Step 107, calculating the confusion degree of the LDA model according to the first driving behavior category, and recording the confusion degree and the LDA model corresponding to the confusion degree. Confusion (Preplexity) is a measure of the LDA model, and is also a measure of information theory, commonly used for comparison of probabilistic models. The number of optimal clustering categories of the LDA model and the number of driving behavior categories are determined through the minimum confusion value.
Step 107 specifically includes:
calculating the degree of confusion according to equation (5):
in the formula (5), p (D) represents the degree of confusion; d represents the number of driving behavior segments, D represents the sequence number of the driving behavior segments, D epsilon D; p (omega) d ) Representing the occurrence probability of each driving behavior feature in each driving behavior segment; p (omega) d ) =p (z|d) ×p (ω|z), p (z|d) represents the occurrence probability of each driving behavior class in each driving behavior segment, that is, θ d The method comprises the steps of carrying out a first treatment on the surface of the p (ωz) represents the probability of occurrence of each driving behavior feature in each driving behavior class, i.e., φ k ；R d Representing the total length of the driving behavior segment.
And step 108, judging whether the clustering number of the LDA models is equal to the preset number, and obtaining a first judging result.
Step 109, if the first judgment result is no, increasing the clustering number of the LDA models, returning to step 106, "inputting the driving behavior segment feature frequency statistical matrix into the LDA models to obtain a first driving behavior category", and updating and recording the confusion degree and the LDA models corresponding to the confusion degree ".
And step 110, if the first judgment result is yes, comparing the confusion degrees of all records to obtain the minimum confusion degree.
And step 111, obtaining an LDA model corresponding to the minimum confusion degree according to the minimum confusion degree. In this embodiment, the number of clusters of the LDA model corresponding to the smallest confusion degree is 13.
And step 112, inputting the driving behavior segment characteristic frequency statistical matrix into an LDA model corresponding to the minimum confusion degree to obtain the driving behavior category. Step 112 further includes inputting each driving behavior segment into the LDA model corresponding to the smallest degree of confusion, resulting in a probability that each driving behavior segment corresponds to each driving behavior class. 4 driving behavior states can be obtained by the present embodiment: acceleration behavior, deceleration behavior, uniform behavior, and idle behavior, 13 driving behavior categories: low speed slow acceleration, medium speed rapid acceleration, medium and low speed slow acceleration, medium and low speed-medium speed rapid acceleration, medium speed braking deceleration, low speed deceleration, medium and high speed uniform velocity, medium and low speed uniform velocity, and idle speed. The driving behavior categories in this embodiment are shown in table 1:
table 1 driving behavior classification of electric-only passenger car
The driving behavior sequence change point detection algorithm can more accurately divide driving behavior data and more accurately identify change points of driving behavior states, namely driving state division points. On the aspect of the recognition result, the driving behavior recognition method can better recognize the change position of the driving behavior state, the misjudgment condition of the driving behavior state is very few, and the formed driving behavior fragment is more complete.
Compared with the traditional method, the method can correlate the abstract linear segmentation with the specific driving behavior characteristics by utilizing the LDA model; for the cluster analysis result of the LDA model, the driving behavior recognition method can represent the driving behavior category in each driving behavior segment in a probability form with higher recognition reliability; the LDA model is a generated model, and only new driving behavior data or detection behavior quantity is required to be added, namely, only the joint distribution of new driving behavior categories is required to be calculated.
The invention takes mass, real-time and dynamic vehicle operation data of the national supervision and management platform of the new energy automobile as a main data source of driving behavior data, covers a large number of pure electric vehicles, and can cover various vehicle driving scenes and a large number of driving behavior characteristics. The method has the advantages of high identification accuracy, relatively high calculation speed and certain application capability in the practical application level.
The invention provides a driving behavior recognition system, and fig. 3 is a system structure diagram of the driving behavior recognition system provided by the embodiment of the invention. Referring to fig. 3, the driving behavior recognition system includes:
an acquisition module 201, configured to acquire driving behavior data. And acquiring speed data, current data and acceleration data of 50 pure electric passenger cars from the national monitoring and management platform of the new energy automobile as driving behavior data.
The symbolizing speed time sequence module 202 is configured to symbolize driving behavior data to obtain a symbolizing speed time sequence.
The symbolizing speed time sequence module 202 specifically includes:
an acquisition unit for acquiring a speed time series S in the driving behavior data velocity ＝{v 1 ,v 2 ,…,v n }, v is 1 ,v 2 ,…,v n Representing sequence points, n representing the total number of sequence points.
A calculating unit, configured to obtain a symbolized speed time sequence S according to formula (1), where formula (1):
in the formula (1), v j Representing a velocity time series S velocity J represents the sequence point number, j E [2, n-1 ]]The first sequence point and last of the velocity time sequence in this embodimentOne sequence point is not symbolized. If the symbolic value of any sequence point is 0, but the symbolic values of the previous sequence point and the next sequence point of the sequence point are consistent and are not 0, the symbolic values of the sequence point are consistent with the state of the symbolic values of the previous sequence point and the next sequence point.
The driving behavior segment module 203 is configured to segment driving behavior data according to the symbolized speed time sequence by using a driving behavior sequence variable point detection algorithm, so as to obtain a segmented driving behavior segment.
The driving behavior segment module 203 specifically includes:
an information entropy unit for calculating information entropy of speed sequence points in the symbolized speed time sequence according to formula (2):
H(S)＝-∑p(s i )log(p(s i ))(i＝1,2,...,n) (2)；
in the formula (2), H (S) represents information entropy of a velocity sequence point, S represents a symbolized velocity time sequence, and s= { S 1 ,s 2 ,…,s n }，s 1 ,s 2 ,…,s n Symbolized values representing sequence points, i.e., speed sequence points; p(s) i ) Representing the ith velocity sequence point s i Is a probability of occurrence of (2); s is(s) i =1 represents acceleration, s i =0 denotes constant velocity, s i -1 represents deceleration; i represents the sequence number of the velocity sequence point.
A local minimum entropy unit for calculating an ith speed sequence point s according to formula (3) i Local minimum entropy E of (2) j ：
E j ＝H(S Forward )+H(S Backward ) (3)；
In the formula (3), H (S Forward ) Representing a velocity sequence point s i Entropy of information of previous symbolization speed time sequence, S Forward ＝{s 1 ,s 2 ,…,s i }；H(S Backward ) Representing a velocity sequence point s i Information entropy of the subsequent symbolization speed time sequence, S Backward ＝{s i+1 ,s i+2 ,…,s n }。
And the driving state dividing point unit is used for obtaining all driving state dividing points in the symbolized speed time sequence by adopting a variable step sliding window method according to the symbolized speed time sequence.
The driving state dividing point unit specifically includes:
a first subunit, configured to set a minimum length of a window, a maximum length of the window, a window offset step size, and a window length; the minimum window length Minl is 7 seconds, the maximum window length Maxl is 15 seconds, and the window offset step f is 2.
The second subunit is configured to obtain a preset minimum length Minl of the window, a preset maximum length Maxl of the window, a window offset step f, and a window length l.
And a third subunit, configured to initialize a sequence number j and a window length l of the speed sequence point.
And the fourth subunit is used for obtaining the sequence number of the initialized speed sequence point, the initialized window length and the speed sequence point corresponding to the sequence number of the initialized speed sequence point according to the sequence number of the initialized speed sequence point and the initialized window length.
A fifth subunit, configured to obtain a sequence to be detected according to the initialized speed sequence point and the initialized window length, where the sequence to be detected specifically is: intercepting a symbolized speed time sequence segment with a window length according to the initialized speed sequence point and the window length, namely marking a sequence to be detected as S t ＝{s j ,s j+1 ,…,s j+l ' l represents window length, s j ,s j+1 ,…,s j+l A speed sequence point representing the sequence to be detected.
A sixth subunit, configured to obtain a first area from the sequence to be detected according to the window offset step length and the window length, where the first area is S e ＝{s j+f ,s j+f+1 ,…,s j+l-f }，s j+f ,s j+f+1 ,…,s j+l-f Representing a velocity sequence point in the first region.
A seventh subunit, configured to calculate a minimum local entropy of all the speed sequence points in the first area according to formula (3), and obtain a minimum local entropy set, and a minimum local entropy set Combined into E= { E j+f ,E j+f+1 ,...,E j+l-f-1 ,E j+l-f }，E j+f ,E j+f+1 ,…,E j+l-f-1 ,E j+l-f Representing the minimum local entropy of all velocity sequence points in the first region.
And an eighth subunit for calculating a smallest local entropy of the smallest local entropy set according to formula (4).
E s ＝min{E j+f ,E j+f+1 ,...,E j+l-f-1 ,E j+l-f } (4)
E in formula (4) s Representing the smallest minimum local entropy.
And a ninth subunit, configured to determine whether the minimum local entropy is unique in the first area, and obtain a second determination result. If the second judgment result is yes, executing a tenth subunit; and if the second judgment result is negative, executing the twelfth subunit.
And a tenth subunit, configured to determine whether the number of decreases of all the minimum local entropies before the speed sequence point corresponding to the minimum local entropy and the number of increases of all the minimum local entropies after the speed sequence point corresponding to the minimum local entropy are simultaneously greater than a preset number of times p, to obtain a third determination result. In this embodiment, the preset number of times p=2. If the third judgment result is yes, executing an eleventh subunit; and if the third judgment result is negative, executing the twelfth subunit.
An eleventh subunit for taking the speed sequence point corresponding to the smallest minimum partial entropy as a driving state division point, and executing the twelfth subunit.
And the twelfth subunit is configured to determine whether the window length is greater than or equal to the maximum length of the window, so as to obtain a fourth determination result. The fourth judging result is that the thirteenth subunit is executed; and if the fourth judgment result is negative, executing the fourteenth subunit.
Thirteenth subunit, for updating the sequence number j of the speed sequence point, making j be added with 1, making the window length l equal to the minimum length of the window, and executing fifteenth subunit.
A fourteenth subunit for updating the window length, adding 1 to the window length l, and executing the fifteenth subunit.
And the fifteenth subunit is configured to determine whether j+l is greater than or equal to n, to obtain a fifth determination result. The fifth judgment result is yes, executing a sixteenth subunit; and if the fifth judgment result is negative, executing the seventeenth subunit.
Sixteenth subunit for acquiring driving state division points.
Seventeenth subunit, configured to update the sequence to be detected according to the updated window length and the sequence number of the updated speed sequence point, and execute the sixth subunit.
And the driving behavior segment unit is used for acquiring all driving state division points, and driving behavior data between any two adjacent driving state division points is divided driving behavior segments.
The driving behavior segment characteristic frequency statistics matrix module 204 is configured to preprocess driving behavior data to obtain a driving behavior segment characteristic frequency statistics matrix.
The driving behavior segment characteristic frequency statistics matrix module 204 specifically includes:
And the normalization unit is used for normalizing the driving behavior characteristics in the driving behavior data to obtain normalized driving behavior characteristics. The driving behavior features include a speed feature, a current feature, and an acceleration feature.
The characteristic interval unit is used for equally dividing each driving behavior characteristic into 20 characteristic intervals according to the difference value between the maximum value and the minimum value of the driving behavior characteristic, and particularly equally dividing the normalized speed characteristic into 20 characteristic intervals according to the difference value between the maximum value and the minimum value of the speed characteristic; dividing the normalized current characteristic into 20 characteristic intervals according to the difference value between the maximum value and the minimum value of the current characteristic; and equally dividing the normalized acceleration characteristic into 20 characteristic sections according to the difference value between the maximum value and the minimum value of the acceleration characteristic.
And the matrix unit is used for calculating the frequency of the speed, the current and the acceleration in each driving behavior segment in each characteristic interval by a counting statistical method to obtain a frequency matrix. The frequency matrix is a 1×60 matrix including 3 driving behavior features and 20 feature intervals of each driving behavior feature.
And the driving behavior segment characteristic frequency statistics matrix unit is used for counting all driving behavior segments to obtain frequency matrixes of all driving behavior segments, and the frequency matrixes of all driving behavior segments are combined into the driving behavior segment characteristic frequency statistics matrix. The driving behavior segment characteristic frequency statistical matrix is a matrix of D multiplied by 60, and D represents the number of driving behavior segments.
The initialization module 205 is configured to obtain an LDA model, and initialize the number of clusters of the LDA model. FIG. 4 is a block diagram of an LDA model provided by an embodiment of the present invention, in FIG. 4, α represents a hyper-parameter of Dirichlet (Dirichlet) distribution of a topic under each driving behavior segment; θ d Representing the distribution of driving behavior categories under the d-th driving behavior segment; phi (phi) k Representing a distribution of driving behavior characteristics under a kth subject; z d,m Representing the driving behavior category of the j-th speed sequence point in the d-th driving behavior segment; omega d,m Representing the driving behavior category corresponding to the j sequence point in the finally generated d driving behavior segment; m is M d A number representing 3 driving behavior features of speed, current and acceleration; d represents the number of driving behavior segments, D represents the sequence number of the driving behavior segments, D epsilon D; k represents the number of potential driving behavior categories-topics, K represents the topic number, K epsilon K; beta represents the hyper-parameters of the Dirichlet distribution of the physical features under each topic. In the LDA model, each driving behavior segment is constructed as a combination of K topics, assuming that K is a known number and does not change. The LDA model needs to have two parameters of driving behavior fragment-driving behavior category distribution and driving behavior category-driving behavior characteristic distribution, and the two parameters need to be deduced through training and learning; in the embodiment, an EM algorithm is adopted to train and learn an LDA model.
The first driving behavior class module 206 is configured to input the driving behavior segment feature frequency statistics matrix into the LDA model to obtain a first driving behavior class.
A confusion degree module 207, configured to calculate a confusion degree of the LDA model according to the first driving behavior category, and record the confusion degree and the LDA model corresponding to the confusion degree.
The confusion degree module 207 specifically includes:
a confusion unit for calculating a confusion according to formula (5):
in the formula (5), p (D) represents the degree of confusion; d represents the number of driving behavior segments, D represents the sequence number of the driving behavior segments, D epsilon D; p (omega) d ) Representing the occurrence probability of each driving behavior feature in each driving behavior segment; p (omega) d ) =p (z|d) ×p (ω|z), p (z|d) represents the occurrence probability of each driving behavior class in each driving behavior segment, that is, θ d The method comprises the steps of carrying out a first treatment on the surface of the p (ωz) represents the probability of occurrence of each driving behavior feature in each driving behavior class, i.e., φ k ；R d Representing the total length of the driving behavior segment.
A first judging module 208, configured to judge whether the number of clusters of the LDA model is equal to a preset number, to obtain a first judging result; if the first judgment result is negative, executing an updating module; and if the first judgment result is yes, executing a minimum confusion degree module.
The updating module 209 is configured to increase the number of clusters of the LDA models, and execute the first driving behavior classification module to update and record the confusion degree and the LDA model corresponding to the confusion degree.
The minimum confusion module 210 is configured to compare the confusion of all records, and obtain the minimum confusion.
The LDA model module 211 is configured to obtain an LDA model corresponding to the minimum confusion according to the minimum confusion. In this embodiment, the number of clusters of the LDA model corresponding to the smallest confusion degree is 13.
The driving behavior category module 212 is configured to input the driving behavior segment feature frequency statistics matrix into an LDA model corresponding to the minimum confusion degree, to obtain a driving behavior category; the LDA model further used to input each driving behavior segment into the LDA model corresponding to the smallest degree of confusion may result in a probability that each driving behavior segment corresponds to each driving behavior class.
4 driving behavior states can be obtained by the driving behavior recognition system of the present embodiment: acceleration behavior, deceleration behavior, uniform behavior, and idle behavior, 13 driving behavior categories: low speed slow acceleration, medium speed rapid acceleration, medium and low speed slow acceleration, medium and low speed-medium speed rapid acceleration, medium speed braking deceleration, low speed deceleration, medium and high speed uniform velocity, medium and low speed uniform velocity, and idle speed.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.

Claims (6)

1. A driving behavior recognition method, characterized by comprising:
acquiring driving behavior data;
carrying out symbolization processing on the driving behavior data to obtain a symbolized speed time sequence;
dividing the driving behavior data by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain divided driving behavior fragments;
Preprocessing the driving behavior data to obtain a driving behavior fragment characteristic frequency statistical matrix;
obtaining an LDA model, and initializing the clustering number of the LDA model;
inputting the characteristic frequency statistical matrix of the driving behavior segment into the LDA model to obtain a first driving behavior class;
calculating the confusion degree of the LDA model according to the first driving behavior category, and recording the confusion degree and the LDA model corresponding to the confusion degree;
judging whether the clustering number of the LDA model is equal to a preset number or not to obtain a first judging result;
if the first judgment result is negative, the clustering number of the LDA models is increased, and the step of inputting the driving behavior segment characteristic frequency statistical matrix into the LDA models to obtain a first driving behavior class is returned, and the confusion degree and the LDA models corresponding to the confusion degree are updated and recorded;
the first judgment result is that all recorded confusion degrees are compared, and the minimum confusion degree is obtained;
obtaining an LDA model corresponding to the minimum confusion according to the minimum confusion;
inputting the driving behavior segment characteristic frequency statistical matrix into an LDA model corresponding to the minimum confusion degree to obtain a driving behavior class;
The step of carrying out symbolization processing on the driving behavior data to obtain a symbolized speed time sequence specifically comprises the following steps:
acquiring a speed time sequence S in the driving behavior data velocity ＝{v 1 ,v 2 ,…,v n }, v is 1 ,v 2 ,…,v n Representing sequence points, n representing the total number of the sequence points;
obtaining a symbolization speed time sequence S according to a formula (1), wherein the formula (1):
in the formula (1), v j Representing the velocity time series S velocity J represents the sequence point number, j E [2, n-1 ]]；
The driving behavior data is segmented by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain segmented driving behavior segments, and the method specifically comprises the following steps:
calculating the information entropy of speed sequence points in the symbolized speed time sequence according to the formula (2):
H(S)＝-∑p(s i )log(p(s i ))(i＝1,2,...,n) (2)；
in the formula (2), H (S) represents the information entropy of the speed sequence point, S represents the symbolized speed time sequence, and s= { S 1 ,s 2 ,…,s n }，s 1 ,s 2 ,…,s n Symbolized values representing sequence points, i.e., speed sequence points; p(s) i ) Representing the ith velocity sequence point s i Is a probability of occurrence of (2); s is(s) i =1 represents acceleration, s i =0 denotes constant velocity, s i -1 represents deceleration;
calculating the ith velocity sequence point s according to equation (3) i Local minimum entropy E of (2) j ：
E j ＝H(S Forward )+H(S Backward ) (3)；
In the formula (3), H (S Forward ) Representing the velocity sequence points s i Entropy of information of previous symbolization speed time sequence, S Forward ＝{s 1 ,s 2 ,…,s i }；H(S Backward ) Representing the velocity sequence points s i Information entropy of the subsequent symbolization speed time sequence, S Backward ＝{s i+1 ,s i+2 ,…,s n }；
According to the symbolized speed time sequence, driving state dividing points of the symbolized speed time sequence are obtained by adopting a variable step sliding window method;
and acquiring all driving state dividing points, wherein driving behavior data between any two adjacent driving state dividing points are divided driving behavior fragments.
2. The driving behavior recognition method according to claim 1, wherein the driving state division points are obtained by using a variable step sliding window method according to the symbolized speed time sequence, and the driving state division points specifically include:
acquiring a preset minimum length of a window, a preset maximum length of the window, a preset window offset step length and a preset window length;
initializing a sequence number j and a window length l of a speed sequence point;
acquiring a sequence to be detected according to the sequence number of the initialized speed sequence point and the initialized window length;
acquiring a first region from the sequence to be detected according to the window offset step length and the window length;
calculating the minimum local entropy of all the speed sequence points in the first area according to a formula (3), and obtaining a minimum local entropy set;
Calculating the smallest minimum local entropy in the smallest local entropy set according to formula (4):
E s ＝min{E j+f ,E j+f+1 ,...,E j+l-f-1 ,E j+l-f } (4)
e in formula (4) s Representing the minimum local entropy, f representing the window offset step;
judging whether the minimum local entropy is unique in the first area, and obtaining a second judging result;
if the second judgment result is yes, carrying out third judgment;
if the second judgment result is negative, fourth judgment is carried out;
the third judgment is to judge whether the reduction times of all the minimum local entropies before the speed sequence point corresponding to the minimum local entropies and the increase times of all the minimum local entropies after the speed sequence point corresponding to the minimum local entropies are simultaneously larger than the preset times or not, so as to obtain a third judgment result; if the third judgment result is negative, carrying out the fourth judgment;
the third judgment result is that a speed sequence point corresponding to the minimum local entropy is the driving state dividing point, and the fourth judgment is carried out;
the fourth judgment is to judge whether the length of the window is larger than or equal to the maximum length of the window, so as to obtain a fourth judgment result;
the fourth judgment result is that the sequence number j of the speed sequence point is updated, 1 is added to j, the window length l is equal to the minimum length of the window, and the fifth judgment is carried out;
If the fourth judgment result is negative, updating the window length, adding 1 to the window length l, and carrying out the fifth judgment;
the fifth judgment is to judge whether j+l is greater than or equal to n, so as to obtain a fifth judgment result;
the fifth judgment result is that the driving state dividing points are obtained;
and if the fifth judgment result is negative, updating the sequence to be detected according to the updated window length and the updated sequence number of the speed sequence point, and returning to the step of acquiring a first area from the sequence to be detected according to the window offset step length and the window length.
3. The driving behavior recognition method according to claim 1, wherein the preprocessing the driving behavior data to obtain a driving behavior segment feature frequency statistical matrix specifically includes:
normalizing the driving behavior characteristics in the driving behavior data to obtain normalized driving behavior characteristics; the driving behavior features include a speed feature, a current feature, and an acceleration feature;
dividing each driving behavior feature into 20 feature intervals according to the difference value between the maximum value and the minimum value of the driving behavior feature;
Calculating the frequency of the speed, the current and the acceleration in each driving behavior segment in each characteristic interval by a counting statistical method to obtain a frequency matrix;
and counting all the driving behavior fragments to obtain frequency matrixes of all the driving behavior fragments, and combining the frequency matrixes of all the driving behavior fragments into a characteristic frequency statistic matrix of the driving behavior fragments.
4. A driving behavior recognition method according to claim 3, wherein the calculating the confusion degree of the LDA model according to the first driving behavior class and recording the confusion degree and the LDA model corresponding to the confusion degree specifically includes:
calculating the confusion according to equation (5):
in the formula (5), p (D) represents the degree of confusion; d represents the number of the driving behavior fragments, D represents the sequence number of the driving behavior fragments, and D epsilon D; p (omega) d ) Representing the occurrence probability of each driving behavior feature in each driving behavior segment; p (omega) d ) =p (z|d) ×p (ω|z), p (z|d) represents the probability of occurrence of each of the driving behavior categories in each of the driving behavior segments, and p (ω|z) represents the probability of occurrence of each of the driving behavior features in each of the driving behavior categories; r is R d Representing the total length of the driving behavior segment.
5. A driving behavior recognition system, characterized by comprising:
the acquisition module is used for acquiring driving behavior data;
the symbolizing speed time sequence module is used for symbolizing the driving behavior data to obtain a symbolizing speed time sequence;
the driving behavior segment module is used for dividing the driving behavior data by adopting a driving behavior sequence variable point detection algorithm according to the symbolized speed time sequence to obtain divided driving behavior segments;
the driving behavior segment characteristic frequency statistical matrix module is used for preprocessing the driving behavior data to obtain a driving behavior segment characteristic frequency statistical matrix;
the initialization module is used for acquiring an LDA model and initializing the clustering number of the LDA model;
the first driving behavior category module is used for inputting the driving behavior segment characteristic frequency statistical matrix into the LDA model to obtain a first driving behavior category;
a confusion degree module, configured to calculate a confusion degree of the LDA model according to the first driving behavior category, and record the confusion degree and an LDA model corresponding to the confusion degree;
The first judging module is used for judging whether the clustering number of the LDA model is equal to the preset number or not to obtain a first judging result; if the first judgment result is negative, executing an updating module; the first judgment result is yes, and a minimum confusion degree module is executed;
the updating module is used for increasing the clustering number of the LDA models, executing a first driving behavior category module, updating and recording the confusion degree and the LDA models corresponding to the confusion degree;
the minimum confusion degree module is used for comparing the confusion degrees of all records to obtain the minimum confusion degree;
the LDA model module is used for obtaining an LDA model corresponding to the minimum confusion according to the minimum confusion;
the driving behavior category module is used for inputting the driving behavior segment characteristic frequency statistical matrix into an LDA model corresponding to the minimum confusion degree to obtain a driving behavior category;
the symbolizing speed time sequence module specifically comprises:
an acquisition unit for acquiring a speed time series S in the driving behavior data velocity ＝{v 1 ,v 2 ,…,v n }, v is 1 ,v 2 ,…,v n Representing sequence points, n representing the total number of the sequence points;
a calculating unit, configured to obtain a symbolized speed time sequence S according to formula (1), where formula (1):
In the formula (1), v j Representing the velocity time series S velocity J represents the sequence point number, j E [2, n-1 ]]；
The driving behavior segment module specifically comprises:
an information entropy unit, configured to calculate information entropy of speed sequence points in the symbolized speed time sequence according to formula (2):
H(S)＝-∑p(s i )log(p(s i ))(i＝1,2,...,n) (2)；
in the formula (2), H (S) represents the information entropy of the speed sequence point, S represents the symbolized speed time sequence, and s= { S 1 ,s 2 ,…,s n }，s 1 ,s 2 ,…,s n Symbolized values representing sequence points, i.e., speed sequence points; p(s) i ) Representing the ith velocity sequence point s i Is a probability of occurrence of (2); s is(s) i =1 represents acceleration, s i =0 denotes constant velocity, s i -1 represents deceleration;
a local minimum entropy unit for calculating the ith speed sequence point s according to formula (3) i Local minimum entropy E of (2) j ：
E j ＝H(S Forward )+H(S Backward ) (3)；
In the formula (3), H (S Forward ) Representing the velocity sequence points s i Entropy of information of previous symbolization speed time sequence, S Forward ＝{s 1 ,s 2 ,…,s i }；H(S Backward ) Representing the velocity sequence points s i Information entropy of the subsequent symbolization speed time sequence, S Backward ＝{s i+1 ,s i+2 ,…,s n }；
The driving state dividing point unit is used for obtaining driving state dividing points of the symbolized speed time sequence by adopting a variable step sliding window method according to the symbolized speed time sequence;
and the driving behavior segment unit is used for acquiring all driving state division points, and driving behavior data between any two adjacent driving state division points are divided driving behavior segments.
6. The driving behavior recognition system of claim 5, wherein the confusion module specifically comprises:
a confusion unit for calculating the confusion according to formula (5):
in the formula (5), p (D) represents the degree of confusion; d represents the number of the driving behavior fragments, D represents the sequence number of the driving behavior fragments, and D epsilon D; p (omega) d ) Representing the occurrence probability of each driving behavior feature in each driving behavior segment; p (omega) d ) =p (z|d) ×p (ω|z), p (z|d) represents the probability of occurrence of each of the driving behavior categories in each of the driving behavior segments, and p (ω|z) represents the probability of occurrence of each of the driving behavior features in each of the driving behavior categories; r is R d Representing the total length of the driving behavior segment.
CN201911156134.1A 2019-11-22 2019-11-22 Driving behavior recognition method and system Active CN111126438B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911156134.1A CN111126438B (en) 2019-11-22 2019-11-22 Driving behavior recognition method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911156134.1A CN111126438B (en) 2019-11-22 2019-11-22 Driving behavior recognition method and system

Publications (2)

Publication Number Publication Date
CN111126438A CN111126438A (en) 2020-05-08
CN111126438B true CN111126438B (en) 2023-11-14

Family

ID=70498092

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911156134.1A Active CN111126438B (en) 2019-11-22 2019-11-22 Driving behavior recognition method and system

Country Status (1)

Country Link
CN (1) CN111126438B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113159105B (en) * 2021-02-26 2023-08-08 北京科技大学 Driving behavior unsupervised mode identification method and data acquisition monitoring system
CN113312707B (en) * 2021-06-18 2023-09-08 深圳市神驼科技有限公司 Truck state self-adaptive real-time detection method and device
CN113715834B (en) * 2021-08-24 2023-05-16 广州五舟软件技术有限公司 Vehicle driving behavior recognition method, device, electronic equipment and storage medium
CN114241429A (en) * 2021-12-15 2022-03-25 平安科技（深圳）有限公司 Driving risk level determination method and device, computer equipment and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1199847A (en) * 1997-09-30 1999-04-13 Nissan Motor Co Ltd Driving pattern recognizing device
JP2014235605A (en) * 2013-06-03 2014-12-15 株式会社デンソー Driving scene label estimation device
JP2015056058A (en) * 2013-09-12 2015-03-23 株式会社デンソー Driving support device
CN108694407A (en) * 2017-04-11 2018-10-23 西安邮电大学 A kind of driving behavior recognition methods based on mobile terminal
CN109086794A (en) * 2018-06-27 2018-12-25 武汉理工大学 A kind of driving behavior mode knowledge method based on T-LDA topic model

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10300922B2 (en) * 2017-09-29 2019-05-28 Denso International America, Inc. Risk assessment system for assessing current driver behavior relative to past behavior and behaviors of other drivers

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1199847A (en) * 1997-09-30 1999-04-13 Nissan Motor Co Ltd Driving pattern recognizing device
JP2014235605A (en) * 2013-06-03 2014-12-15 株式会社デンソー Driving scene label estimation device
JP2015056058A (en) * 2013-09-12 2015-03-23 株式会社デンソー Driving support device
CN108694407A (en) * 2017-04-11 2018-10-23 西安邮电大学 A kind of driving behavior recognition methods based on mobile terminal
CN109086794A (en) * 2018-06-27 2018-12-25 武汉理工大学 A kind of driving behavior mode knowledge method based on T-LDA topic model

Also Published As

Publication number Publication date
CN111126438A (en) 2020-05-08

Similar Documents

Publication Publication Date Title
CN111126438B (en) Driving behavior recognition method and system
Xue et al. Rapid Driving Style Recognition in Car‐Following Using Machine Learning and Vehicle Trajectory Data
US9527384B2 (en) Driving context generation system for generating driving behavior description information
Deng et al. Driving style recognition method using braking characteristics based on hidden Markov model
Tang et al. Driver lane change intention recognition of intelligent vehicle based on long short-term memory network
JP7398383B2 (en) Vehicle classification based on telematics data
Li et al. Driving style classification based on driving operational pictures
CN106956680B (en) Electric automobile driving behavior recognition analysis method
JP2014234113A (en) Vehicular reference value generation apparatus
WO2021093012A1 (en) Method and apparatus for extracting time series data formation mode and terminal device
JP2017071333A (en) Drive assisting device
Halim et al. Deep neural network-based identification of driving risk utilizing driver dependent vehicle driving features: A scheme for critical infrastructure protection
Moosavi et al. Driving style representation in convolutional recurrent neural network model of driver identification
CN114882069A (en) Taxi track abnormity detection method based on LSTM network and attention mechanism
Xing et al. Recognizing driver braking intention with vehicle data using unsupervised learning methods
Marciniuk et al. Machine learning applied to acoustic-based road traffic monitoring
Anil et al. Driver behavior analysis using K-means algorithm
Schwarz Time series categorization of driving maneuvers using acceleration signals
Safont et al. Comparison of dimensionality reduction methods for road surface identification system
Wang et al. A two-layer risky driver recognition model with context awareness
Gao et al. Driver identification based on stop-and-go events using naturalistic driving data
Wen et al. Analysis of vehicle driving styles at freeway merging areas using trajectory data
Wang et al. Driving Style Identification Strategy Based on DS Evidence Theory
Zhang et al. A recognition model for acceleration intention of automobile drivers based on fuzzy clustering
CN116340767B (en) Electric automobile travel energy consumption probability distribution prediction method, system and product

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