CN104657746B - A kind of method for detecting abnormality based on track of vehicle similitude - Google Patents
A kind of method for detecting abnormality based on track of vehicle similitude Download PDFInfo
- Publication number
- CN104657746B CN104657746B CN201510046984.1A CN201510046984A CN104657746B CN 104657746 B CN104657746 B CN 104657746B CN 201510046984 A CN201510046984 A CN 201510046984A CN 104657746 B CN104657746 B CN 104657746B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- vehicle
- track
- mtd
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Traffic Control Systems (AREA)
- Train Traffic Observation, Control, And Security (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of method for detecting abnormality based on track similitude, the deflection sequence of track of vehicle and track of vehicle is extracted first, then by calculating track between similarity measurement clustered, according to deflection sequential extraction procedures exemplary trajectory, exemplary trajectory is calculated again and the similarity measurement of such track of vehicle sets up deviation statistics model, the confidential interval of similarity measurement is obtained, the similarity measurement of track to be measured and exemplary trajectory is finally calculated, exception is judged whether according to confidential interval.This method realization approach is more directly perceived, is easily integrated into traffic abnormity warning system;The present invention make track of vehicle just, abnormal discrimination it is more obvious, make abnormality detection accuracy rate higher, effective foundation provided for further abnormality processing.
Description
Technical field
The invention belongs to abnormal behaviour mode identification technology, more specifically, it is related to a kind of based on track of vehicle
The method for detecting abnormality of similitude.
Background technology
With the development of society and becoming increasingly conspicuous for traffic safety problem, the concept of intelligent transportation is produced immediately, and
In the case where the fast development of video monitoring system is with application, people need to recognize the people of many special scenes or object (such as
The vehicle moved in convict in old man, prison and traffic in home for destitute) behavior it is whether correct, therefore, abnormality detection is
Through becoming a highly important application, and the movable information for obtaining moving object is the basis of unusual checking.In recent years
Come, unusual checking turns into the study hotspot of machine vision and area of pattern recognition, many scholar's research are applied to difference
Scene and the anomaly detection method for meeting different demands.
Anomaly detection method can substantially be divided into two classes:Method based on local motion feature and the side based on track
Method.The former is to extract each pixel or the feature of certain fixed size block of pixels, then to the feature modeling of extraction;The latter is
Moving target in actual scene is tracked by technological means, then obtained track analyzed.Side based on track
Method is more applicable under many circumstances with its advantage in terms of macroscopic view.The movement locus of moving target is further to analyze it
Behavior, abnormality detection and cloud, which are surveyed etc., provides foundation, such as track of vehicle in traffic video, and it contains abundant characteristic information,
The live structure of these information illustrations can extract the direction of motion and speed there is provided the generation clue of time, and can be right
The correlation of moving target makes inferences, and effectively monitors unlawful practice violating the regulations, prevents traffic accident.
Trajectory clustering is the main contents of trajectory analysis research, and the starting point of trajectory clustering is according to the phase between track
The division to track is automatically realized like degree.Track similarity is the foundation place of trajectory clustering, conventional track similarity side
Method mainly includes Euclidean distances, principal component analysis (PCA), Hausdorff distances, (most long common sequence) LCCS, dynamic
Time alignment (DTW), hidden Markov model (HMM) etc..Dynamic time warping (Dynamic Time Wraping, DTW) is
It is a kind of method for the similarity for weighing the different time series of two length based on Dynamic Programming (DP) thought, it is main to use
In template matches, such as speech recognition, gesture identification, data mining and information retrieval etc., DTW distances are between measuring and calculating track
Apart from when be also well used.In the case of many traffic abnormities, such as an once big fluctuation, its elsewhere belongs to normal row
It is considered that being normal behaviour during situation about sailing, and the situation of snakelike traveling belongs to abnormal behaviour always, under this background, classical
DTW rangings be difficult to be able to distinguish, it is necessary to be improved to this method when calculating two trajectory distances, in both of these case,
And applied in actual scene, watch effect.
The content of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of abnormal inspection based on track of vehicle similitude
Survey method, it is possible to increase abnormal accuracy rate in the snakelike traveling of identification.
For achieving the above object, a kind of method for detecting abnormality based on track of vehicle similitude of the present invention, its feature
It is, comprises the following steps:
(1) track of vehicle, is extracted
The positional information of target vehicle is gathered, the positional information to collection carries out gaussian filtering, then the position to being filtered
Information carries out feature extraction, and the position feature of extraction is constituted 2n*m matrix M, and wherein 2n is matrix M line number, and m is matrix M's
Columns;
In matrix M, the first row that every two row is represented in a track of vehicle, two rows represents abscissa, and the second row is represented
Ordinate, then include in n bar track of vehicle numbers, every track of vehicle in matrix M comprising m sampled point;
The matrix M ' that matrix M every two row is constitutedi, i=1,2 ..., n, matrix M 'iI-th vehicle is obtained after transposition again
The vector sequence L of tracki={ qj=(xj,yj), j=1,2 ..., m }, wherein, qjRepresent horizontal stroke, the ordinate of j-th of sampled point
(xj,yj);
(2) the deflection sequence of track of vehicle, is obtained
(2.1) track of vehicle deflection β, is extractedj
(2.2) the deflection sequence of track of vehicle, is determined
If the deflection of last sampled point is identical with the deflection of previous sampled point in every track of vehicle, i.e. βm
=βm-1;
The then vector sequence L of every track of vehicleiCorresponding deflection sequence BiIt is represented by:Bi=(β1,β2,...,
βm);
(3), track of vehicle is clustered
(3.1) similarity measurement between track of vehicle, is calculated
Utilize the phase between any two tracks of vehicle in the DTW trajectory distance algorithm calculating matrix M based on weight after improvement
Like property measurement
Wherein, teIt is e-th of value outside the threshold value on required regular path T;tfIt is within the threshold value on required regular path T
F-th value;we、wfThe respectively weight of the similarity measurement of outer and in threshold value the point of threshold value, and meeting
M1, n1 are respectively the number put outside T upper threshold values, within threshold value;Lk、LlFor two tracks of vehicle arbitrarily extracted from matrix M
Vector sequence, k, l=1,2 ..., n;
(3.2) the similarity measurement matrix K of track of vehicle, is obtained
The similarity measurement of all tracks of vehicle is calculated by step (3.1), n*n similarity measurement matrix K is constituted:
Wherein, σ is scale parameter, for controlling KklWith LkAnd LlThe speed that distance increases and decayed, KklFor in matrix K
The element of k rows l row;
(3.3), track of vehicle is clustered
The cluster number of spectral clustering is first determined, that is, gathers for ζ classes, regard n*n similarity measurement matrix K as spectral clustering
Algorithm is inputted, and all tracks of vehicle are gathered for ζ classes;
(4) exemplary trajectory and the deviation statistics modeling of track of vehicle, are extracted
τ classes track of vehicle after cluster is subjected to interval division according to ordinate, τ=1,2 ..., ζ are initial interval long
Degree is set to sx, the variance of such track of vehicle deflection in each interval is calculated, and set variance threshold values η and interval merging threshold
Value ε;
(4.1) if, in some interval, the variance of any one track of vehicle is more than η, then is fallen into a trap in step (4.2)
When calculating deflection average, give up the deflection of this track of vehicle in the interval;
(4.2), according to deflection sequence BiCalculate each interval deflection average αi,T is merging proparea
Between number;
If the difference of the deflection average of two adjacent intervals is more than or equal to ε, two adjacent intervals keep constant;
If the difference of the deflection average of two adjacent intervals is less than ε, merge the two intervals, and calculate the area after merging
Between deflection averageS is merges number between back zone, and union operation is up to being finally unable to combine interval
Untill;
Then calculate respectively each interval horizontal stroke, ordinate a little be worth to the interval intermediate value c a little, then by
Deflection average calculates slope
Obtaining slope isAnd point c straight line is crossed, as such track of vehicle is in the interval exemplary trajectoryAgain will
Each interval exemplary trajectory connects the exemplary trajectory L for obtaining such track of vehiclecτ;
(4.3) trajector deviation statistical model, is set up
According to track of vehicle similarity measurement calculation formula in step (3.1), all tracks and typical case in τ classes are calculated
Track LcτSimilarity measurement, obtain the similarity measurement sequence D of τ class tracks of vehicleτr, r represents τ class tracks of vehicle
Bar number;
In τ class tracks of vehicle, to each sampled point of every track of vehicle according to exemplary trajectory LcτAsk correspondence is vertical to sit
Abscissa x under markpq, p=1,2 ..., r;Q=1,2 ..., m, xpqRepresent pth bar track of vehicle q-th of sampled point it is vertical
Abscissa of the coordinate in exemplary trajectory;The abscissa of each sampled point is subtracted into corresponding xpq, then every vehicle is counted respectively
Track abscissa it is positive and negative, the similarity measurement that positive number is at most calculated for just, the similarity measurement that negative is at most calculated be it is negative, then
DτrAs D'τr;
To D'τrIn all similarity measurements carry out Gauss modelings, and obey N (uτ,στ 2), i.e. Dτ~N (uτ,στ 2)
Wherein Dτr' (p) is represented in τ class tracks of vehicle, pth bar track of vehicle and exemplary trajectory LcτSimilarity measurements
Amount;
Setting confidence level is 1- α, then confidential interval is
Wherein, nτFor trace bar number in τ class tracks of vehicle,
(5), the abnormality detection of track of vehicle
According to track of vehicle similarity measurement calculation formula in step (3.1), track of vehicle to be measured and every quasi-representative are calculated
The similarity measurement of track, if any one similarity measurement is in confidential interval, the track of vehicle is normal, otherwise the vehicle
Track is abnormal.
What the goal of the invention of the present invention was realized in:
Method for detecting abnormality of the invention based on track similitude, extracts the deflection of track of vehicle and track of vehicle first
Sequence, then by calculating track between similarity measurement clustered, according to deflection sequential extraction procedures exemplary trajectory, then calculate allusion quotation
Type track and the similarity measurement of such track of vehicle set up deviation statistics model, obtain the confidential interval of similarity measurement, most
The similarity measurement of track to be measured and exemplary trajectory is calculated afterwards, and exception is judged whether according to confidential interval.This method, which is realized, to be thought
Road is more directly perceived, is easily integrated into traffic abnormity warning system;The present invention make track of vehicle just, abnormal discrimination it is more obvious,
Make abnormality detection accuracy rate higher, effective foundation is provided for further abnormality processing.
Brief description of the drawings
Fig. 1 is the flow chart of the method for detecting abnormality of the invention based on track of vehicle similitude;
Fig. 2 is track of vehicle cluster schematic diagram.
Embodiment
The embodiment to the present invention is described below in conjunction with the accompanying drawings, so as to those skilled in the art preferably
Understand the present invention.Requiring particular attention is that, in the following description, when known function and design detailed description perhaps
When can desalinate the main contents of the present invention, these descriptions will be ignored herein.
Embodiment
Fig. 1 is the flow chart of the method for detecting abnormality of the invention based on track of vehicle similitude.
In the present embodiment, as shown in figure 1, a kind of method for detecting abnormality based on track of vehicle similitude of the present invention, main
To include three parts:Trajectory extraction stage, clustering phase and abnormality detection stage.
It is described in detail below for three phases, it is specific as follows:
S1, trajectory extraction stage
S1.1, acquisition vehicle position information
The positional information of target vehicle is gathered, the positional information to collection carries out gaussian filtering, then the position to being filtered
Information carries out feature extraction, and the position feature of extraction is constituted 2n*m matrix M, and wherein 2n is matrix M line number, and m is matrix M's
Columns;In the present embodiment, as shown in table 1, n=6, m=16;
Table 1 is matrix M;
3 | 9 | 12 | 19 | 24 | 30 | 36 | 44 | 51 | 54 | 64 | 69 | 78 | 80 | 84 | 88 |
126 | 118 | 110 | 105 | 98 | 90 | 82 | 70 | 62 | 58 | 45 | 38 | 26 | 20 | 18 | 13 |
40 | 44 | 48 | 51 | 54 | 58 | 61 | 66 | 63 | 70 | 76 | 77 | 79 | 82 | 85 | 90 |
135 | 127 | 117 | 110 | 103 | 95 | 82 | 90 | 81 | 75 | 61 | 55 | 50 | 43 | 35 | 25 |
56 | 57 | 62 | 64 | 67 | 71 | 73 | 75 | 76 | 77 | 78 | 79 | 80 | 84 | 86 | 92 |
125 | 119 | 108 | 101 | 95 | 90 | 85 | 72 | 65 | 61 | 59 | 55 | 50 | 43 | 34 | 27 |
153 | 159 | 162 | 169 | 174 | 180 | 186 | 194 | 201 | 204 | 214 | 219 | 228 | 230 | 234 | 238 |
126 | 117 | 110 | 105 | 98 | 90 | 82 | 70 | 62 | 58 | 45 | 38 | 26 | 22 | 17 | 12 |
175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 | 175 |
126 | 118 | 110 | 105 | 98 | 90 | 82 | 70 | 62 | 58 | 45 | 38 | 26 | 22 | 17 | 12 |
178 | 177 | 176 | 175 | 174 | 173 | 172 | 171 | 170 | 169 | 168 | 167 | 166 | 165 | 162 | 160 |
126 | 118 | 110 | 105 | 98 | 90 | 82 | 70 | 62 | 58 | 45 | 38 | 26 | 22 | 17 | 12 |
Table 1
In matrix M, the first row that every two row is represented in a track of vehicle, two rows represents abscissa, and the second row is represented
Ordinate, then include in 6 track of vehicle numbers, every track of vehicle in matrix M comprising m sampled point;
The matrix M that matrix M every two row is constitutedi', i=1,2 ..., 6, matrix Mi' i-th vehicle is obtained after transposition again
The vector sequence L of tracki={ qj=(xj,yj), j=1,2 ..., 16 }, wherein, qjRepresent horizontal stroke, the ordinate of j-th of sampled point
(xj,yj);
S1.2, the deflection sequence for obtaining track of vehicle
1) track of vehicle deflection β, is extractedj
2) the deflection sequence of track of vehicle, is determined
If the deflection of last sampled point is identical with the deflection of previous sampled point in every track of vehicle, i.e. βm
=βm-1;In the present embodiment, by taking first track of vehicle in matrix M as an example, the 15th sampled point (84,18) and the 16th adopt
The deflection of sampling point (88,13) is identical, i.e. β16=β15;
The then vector sequence L of every track of vehicleiCorresponding deflection sequence BiIt is represented by:Bi=(β1,β2,...,
βm);
S2, clustering phase
Similarity measurement between S2.1, calculating track of vehicle
In the present embodiment, first DTW distance algorithms are briefly described, it is specific as follows:
DTW is a kind of method for mode matching proposed based on the thoery of dynamic programming (DP), the algorithm by Time alignment and away from
It is combined from Likelihood Computation, some vectors is compressed, expands or changes by searching the similar features between 2 tracks, with
Just its characteristic quantity is corresponding with mode standard.Its principle is as follows:
If T (t × N) and R (r × N) is 2 polynary tracks, test lot and reference batch are represented respectively, wherein t and r are
Sampling number, N is variable number.DTW uses dynamic programming principle, and non-linearly misplace 2 tracks, arranges similar case, makes
Wherein one track it is each vectorial corresponding with another track each vector, to obtain the beeline between two tracks.If
itAnd jtRespectively LkAnd LlThe coordinate on the time on track, 1≤it≤ t, 1≤jt≤r.DTW sets up P in t × r grids
The F* sequences of individual point:
F*={ c (1), c (2) ..., c (p) ..., c (P) },
Wherein max (t, r)≤P≤t+r, c (p)=[i (p) j (p)].
In DTW algorithms, F* sequences, which can be regarded as, makes most short one of gauged distance between two tracks be in t × r grids
Optimal path.If d (it,jt) it is to represent T (t × N) and R (r × N) respectively in itAnd jtThe Euclidean distance value at moment:
Wherein, wcFor the weights of each variable, reflect the relative importance of each measured variable.Construct a similarity matrix
DA, for describing the diversity factor between two tracks.Can be by d (i with the algorithm of Dynamic Programmingt,jt) obtain DA(i, j), application
The local restriction that Itakura is proposed can obtain following stepping type:
D in formulaA(1,1)=d (1,1).
The DTW range formulas for defining track on this basis are as follows:
Wherein, φ is the number put on regular path,It is on required regular path TIndividual value.
In the present invention, DTW distance algorithms are improved further, after being improved the DTW tracks based on weight away from
From algorithm, i.e.,:
Wherein, teIt is e-th of value outside the threshold value on required regular path T;tfBe threshold value on required regular path T it
F-th interior of value;we、wfThe respectively weight of the similarity measurement of outer and in threshold value the point of threshold value, and meeting
M1, n1 are respectively the number put outside T upper threshold values, within threshold value;Lk、LlFor two tracks of vehicle arbitrarily extracted from matrix M
Vector sequence, k, l=1,2 ..., n;
Recycle in the DTW trajectory distance algorithms based on weight after improving, calculating matrix M between any two tracks of vehicle
Similarity measurement;
S2.2, the similarity measurement matrix K for obtaining track of vehicle
By the DTW trajectory distance algorithms based on weight in step S2.1, the similarity measurements of all tracks of vehicle are calculated
Amount, constitutes n*n similarity measurement matrix K:
Wherein, σ is scale parameter, for controlling KklWith LkAnd LlThe speed that distance increases and decayed, KklFor in matrix K
The element of k rows l row;
S2.3, track of vehicle is clustered
It is so-called cluster be exactly data point is divided into several classes or cluster so that between the data point in same class have compared with
High similarity, and there is higher distinctiveness ratio between inhomogeneity.The algorithm idea of spectral clustering originates from spectrogram Partition Theory, it
Using input feature vector vector as the fixed point in figure, using the similarity between characteristic vector as the side on connection summit, schemed by finding
Optimization realize the classification of input feature vector vector.
The cluster number of spectral clustering is first determined, that is, is gathered for ζ classes, then it is poly- using n*n similarity measurement matrix K as spectrum
Class algorithm is inputted, and all tracks of vehicle are gathered for ζ classes;
In the present embodiment, spectral clustering uses k mean clusters;As shown in Fig. 26 tracks of vehicle gather for 2 classes, i.e.,
TR1, TR2, TR3 are poly- for a class, and TR4, TR5, TR6 are poly- in order to another kind of.
S3, abnormality detection stage
τ classes track of vehicle after cluster is subjected to interval division according to ordinate, τ=1,2 ..., ζ are initial interval long
Degree is set to sx, the variance of such track of vehicle deflection in each interval is calculated, and set variance threshold values η and interval merging threshold
Value ε;
In the present embodiment, by taking three tracks of vehicle of this class of TR1, TR2, TR3 as an example, initial siding-to-siding block length sx=
20, variance threshold values η=0.5, interval merges threshold epsilon=0.1;
If S3.1, in some interval, the variance of any one track of vehicle is more than η, then is calculated in step S3.2
During deflection average, give up the deflection of this track of vehicle in the interval;
Table 2 is the variance statistic table of track of vehicle deflection in each interval;
Table 2
In table 2, variances of the track of vehicle TR2 in interval [60,80] and interval [80,100] is more than η=0.5, so house
Abandon the deflection of this track of vehicle in the interval;
S3.2, according to deflection sequence BiCalculate each interval deflection average αi,T is merging proparea
Between number;
If the difference of the deflection average of two adjacent intervals is more than or equal to ε, two adjacent intervals keep constant;
If the difference of the deflection average of two adjacent intervals is less than ε, merge the two intervals, and calculate the area after merging
Between deflection averageS is merges number between back zone, and union operation is up to being finally unable to combine interval
Untill;
Table 3 is each interval course bearing angle average statistical form
It is interval | [0,20] | [20,40] | [40,60] | [60,80] | [80,100] | [100,120] | [120,140] |
Average | 1.919 | 1.996 | 2.001 | 2.332 | 1.936 | 2.123 | 2.055 |
Table 3
In the present embodiment, it is interval to merge threshold epsilon=0.1, then understand to close interval [0,20], [20,40], [40,60]
And, [100,120], [120,140] merge, specific as shown in table 4;
Table 4 is the course bearing angle average statistical form after interval merge
It is interval | [0,60] | [60,80] | [80,100] | [100,140] |
Average | 1.972 | 2.332 | 1.936 | 2.089 |
Table 4
Then calculate respectively each interval horizontal stroke, ordinate a little be worth to the interval intermediate value c a little, then by
Deflection average calculates slope
Obtaining slope isAnd point c straight line is crossed, as such track of vehicle is in the interval exemplary trajectoryEach interval exemplary trajectory is connected to the exemplary trajectory L for obtaining such track of vehicle againcτ;
S3.3, set up trajector deviation statistical model
According to the DTW trajectory distance algorithms based on weight in step S2.1, all tracks and typical rail in τ classes are calculated
MarkSimilarity measurement, obtain the similarity measurement sequence D of τ class tracks of vehicleτr, r represents article of τ class tracks of vehicle
Number;
In τ class tracks of vehicle, to each sampled point of every track of vehicle according to exemplary trajectory LcτAsk correspondence is vertical to sit
Abscissa x under markpq, p=1,2 ..., r;Q=1,2 ..., m, xpqRepresent pth bar track of vehicle q-th of sampled point it is vertical
Abscissa of the coordinate in exemplary trajectory;The abscissa of each sampled point is subtracted into corresponding xpq, then every vehicle is counted respectively
Track abscissa it is positive and negative, the similarity measurement that positive number is at most calculated for just, the similarity measurement that negative is at most calculated be it is negative, then
DτrAs D'τr;
In the present embodiment, if qq=(xq,yq) it is track LqPoint is up-sampled, then xpqFor yqSubstitute into exemplary trajectory LcτFall into a trap
The abscissa of calculation.Because similarity measure values are not negative in itself, it is assumed that Dτr=(d1,d2,d3),d1,d2,d3>=0, that is, represent
L in τ classes trackcτ1、Lcτ2、Lcτ33 tracks respectively with exemplary trajectory LcτThe similarity measurement of calculating is 1,2,1, if now Lcτ1
All sampled point abscissa x on trackqSubtract xpq, it is that loser is in the majority, and Lcτ2、Lcτ3It is many for positive person, then D'τr=(- d1,d2,
d3)。
To D'τrIn all similarity measurements carry out Gauss modelings, and obey N (uτ,στ 2), i.e. Dτ~N (uτ,στ 2)
Wherein Dτr' (p) is represented in τ class tracks of vehicle, pth bar track of vehicle and exemplary trajectory LcτSimilarity measurements
Amount;
Setting confidence level is 1- α, then confidential interval is
Wherein, nτFor trace bar number in τ class tracks of vehicle,
In the present embodiment, confidence level 1- α>90%.
S3.4, track of vehicle abnormality detection
According to the DTW trajectory distance algorithms based on weight in step S2.1, track of vehicle to be measured and every quasi-representative rail are calculated
The similarity measurement of mark, if any one similarity measurement is in confidential interval, the track of vehicle is normal, otherwise the vehicle rail
Mark is abnormal.
Example
In the unusual checking for applying the present invention to track of vehicle, the data of collection are 42 normal traces, every
Track gathers 25 location point information, and the track data taken is 84*25, and every two row represents a track, every track it is every
Two row represent a positional information.After 42 normal traces are carried out with similarity measurement and spectral clustering, it has been divided into two classes, then
Exemplary trajectory is extracted as reference template, the abnormal tracks of 10 normal traces and 10 are finally taken, this 20 tracks to be measured are allowed
The method before improving and after improvement is respectively applied to, as shown in table 5, the DTW algorithms after improvement can significantly improve inspection to its result
The precision of survey.
Table 5 is abnormality detection experimental result contrast table before and after improving
Track classification | Actual bar number | Detection bar number before improving | Bar number is detected after improvement |
Normally | 10 | 7 | 9 |
It is abnormal | 10 | 13 | 11 |
Table 5
Although illustrative embodiment of the invention is described above, in order to the technology of the art
Personnel understand the present invention, it should be apparent that the invention is not restricted to the scope of embodiment, to the common skill of the art
For art personnel, as long as various change is in the spirit and scope of the present invention that appended claim is limited and is determined, these
Change is it will be apparent that all utilize the innovation and creation of present inventive concept in the row of protection.
Claims (2)
1. a kind of method for detecting abnormality based on track of vehicle similitude, it is characterised in that comprise the following steps:
(1) track of vehicle, is extracted
The positional information of target vehicle is gathered, the positional information to collection carries out gaussian filtering, then the positional information to being filtered
Feature extraction is carried out, the position feature of extraction is constituted 2n*m matrix M, wherein 2n is matrix M line number, and m is matrix M row
Number;
In matrix M, the first row that every two row is represented in a track of vehicle, two rows represents abscissa, and the second row represents vertical seat
Mark, then include in n bar track of vehicle numbers, every track of vehicle in matrix M comprising m sampled point;
The matrix M of the two rows composition of a track of vehicle will be represented in matrix Mi', i=1,2 ..., n, matrix Mi' again after transposition
Obtain the vector sequence L of i-th track of vehiclei={ qj=(xj,yj), j=1,2 ..., m }, wherein, qjRepresent j-th of sampling
Horizontal stroke, the ordinate (x of pointj,yj);
(2) the deflection sequence of track of vehicle, is obtained
(2.1) track of vehicle deflection β, is extractedj
<mrow>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>arctan</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>></mo>
<mn>0</mn>
<mo>,</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>&NotEqual;</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>arctan</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>&pi;</mi>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo><</mo>
<mn>0</mn>
<mo>,</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo><</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>arctan</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>&pi;</mi>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo><</mo>
<mn>0</mn>
<mo>,</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>></mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<mfrac>
<mi>&pi;</mi>
<mn>2</mn>
</mfrac>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mn>0</mn>
<mo>,</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo><</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mfrac>
<mi>&pi;</mi>
<mn>2</mn>
</mfrac>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mn>0</mn>
<mo>,</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>></mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
(2.2) the deflection sequence of track of vehicle, is determined
If the deflection of last sampled point is identical with the deflection of previous sampled point in every track of vehicle, i.e. βm=
βm-1;
The then vector sequence L of every track of vehicleiCorresponding deflection sequence BiIt is represented by:Bi=(β1,β2,...,βm);
(3), track of vehicle is clustered
(3.1) similarity measurement between track of vehicle, is calculated
Utilize the similitude between any two tracks of vehicle in the DTW trajectory distance algorithm calculating matrix M based on weight after improvement
Measurement
<mrow>
<msub>
<mi>D</mi>
<mrow>
<mi>i</mi>
<mi>m</mi>
<mi>p</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>L</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>L</mi>
<mi>l</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>e</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mrow>
<mi>m</mi>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>w</mi>
<mi>e</mi>
</msub>
<msub>
<mi>t</mi>
<mi>e</mi>
</msub>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>f</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mrow>
<mi>n</mi>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>w</mi>
<mi>f</mi>
</msub>
<msub>
<mi>t</mi>
<mi>f</mi>
</msub>
</mrow>
Wherein, teIt is e-th of value outside the threshold value on required regular path T;tfIt is within the threshold value on required regular path T
F-th value;we、wfThe respectively weight of the similarity measurement of outer and in threshold value the point of threshold value, and meeting
M1, n1 are respectively the number put outside T upper threshold values, within threshold value;Lk、LlFor two tracks of vehicle arbitrarily extracted from matrix M
Vector sequence, k, l=1,2 ..., n;
(3.2) the similarity measurement matrix K of track of vehicle, is obtained
The similarity measurement of all tracks of vehicle is calculated by step (3.1), n*n similarity measurement matrix K is constituted:
<mrow>
<msub>
<mi>K</mi>
<mrow>
<mi>k</mi>
<mi>l</mi>
</mrow>
</msub>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mi>l</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<msub>
<mi>D</mi>
<mrow>
<mi>i</mi>
<mi>m</mi>
<mi>p</mi>
</mrow>
</msub>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>L</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>L</mi>
<mi>l</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>/</mo>
<mo>(</mo>
<mrow>
<mn>2</mn>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>k</mi>
<mo>&NotEqual;</mo>
<mi>l</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, σ is scale parameter, for controlling KklWith LkAnd LlThe speed that distance increases and decayed, KklFor row k in matrix K
The element of l row;
(3.3), track of vehicle is clustered
The cluster number of spectral clustering is first determined, that is, gathers for ζ classes, regard n*n similarity measurement matrix K as spectral clustering
Input, all tracks of vehicle are gathered for ζ classes;
(4) exemplary trajectory and the deviation statistics modeling of track of vehicle, are extracted
τ classes track of vehicle after cluster is subjected to interval division according to ordinate, τ=1,2 ..., ζ, initial siding-to-siding block length are set
For sx, the variance of such track of vehicle deflection in each interval is calculated, and set variance threshold values η and interval merging threshold epsilon;
(4.1) if, in some interval, the variance of any one track of vehicle is more than η, then the calculating side in step (4.2)
During to angle average, give up the deflection of this track of vehicle in the interval;
(4.2), according to deflection sequence BiCalculate each interval deflection average T is between merging proparea
Number;
If the difference of the deflection average of two adjacent intervals is more than or equal to ε, two adjacent intervals keep constant;
If the difference of the deflection average of two adjacent intervals is less than ε, merge the two intervals, and calculate interval after merging
Deflection average S is merges number between back zone, and union operation is untill finally combine interval is unable to;
Then calculate respectively each interval horizontal stroke, ordinate a little be worth to the interval intermediate value c a little, then by direction
Angle average calculates slope
<mrow>
<msub>
<mi>k</mi>
<msup>
<mi>i</mi>
<mo>*</mo>
</msup>
</msub>
<mo>=</mo>
<mi>t</mi>
<mi>a</mi>
<mi>n</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msub>
<mover>
<mi>&alpha;</mi>
<mo>~</mo>
</mover>
<msup>
<mi>i</mi>
<mo>*</mo>
</msup>
</msub>
<mo>&CenterDot;</mo>
<mi>&pi;</mi>
</mrow>
<mn>180</mn>
</mfrac>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Obtaining slope isAnd point c straight line is crossed, as such track of vehicle is in the interval exemplary trajectoryAgain by each
Interval exemplary trajectory connects the exemplary trajectory L for obtaining such track of vehiclecτ;
(4.3) trajector deviation statistical model, is set up
According to track of vehicle similarity measurement calculation formula in step (3.1), all tracks and exemplary trajectory in τ classes are calculated
LcτSimilarity measurement, obtain the similarity measurement sequence D of τ class tracks of vehicleτr, r represents article number of τ class tracks of vehicle;
In τ class tracks of vehicle, to each sampled point of every track of vehicle according to exemplary trajectory LcτAsk under correspondence ordinate
Abscissa xpq, p=1,2 ..., r;Q=1,2 ..., m, xpqRepresent the ordinate of q-th of sampled point of pth bar track of vehicle
Abscissa in exemplary trajectory;The abscissa of each sampled point is subtracted into corresponding xpq, then every track of vehicle is counted respectively
Abscissa it is positive and negative, the similarity measurement that positive number is at most calculated is just, and the similarity measurement that negative is at most calculated is negative, then DτrInto
For D'τr;
To D'τrIn all similarity measurements carry out Gauss modelings, and obey N (uτ,στ 2), i.e. Dτ~N (uτ,στ 2)
<mrow>
<msub>
<mi>u</mi>
<mi>&tau;</mi>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>r</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>p</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>r</mi>
</munderover>
<msup>
<msub>
<mi>D</mi>
<mrow>
<mi>&tau;</mi>
<mi>r</mi>
</mrow>
</msub>
<mo>&prime;</mo>
</msup>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
<msup>
<msub>
<mi>&sigma;</mi>
<mi>&tau;</mi>
</msub>
<mn>2</mn>
</msup>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>r</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>p</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>r</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msup>
<msub>
<mi>D</mi>
<mrow>
<mi>&tau;</mi>
<mi>r</mi>
</mrow>
</msub>
<mo>&prime;</mo>
</msup>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>u</mi>
<mi>&tau;</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
Wherein Dτr' (p) is represented in τ class tracks of vehicle, pth bar track of vehicle and exemplary trajectory LcτSimilarity measurement;
Setting confidence level is 1- α, then confidential interval is
<mrow>
<mo>&lsqb;</mo>
<msub>
<mover>
<mi>D</mi>
<mo>&OverBar;</mo>
</mover>
<mi>&tau;</mi>
</msub>
<mo>-</mo>
<mfrac>
<msub>
<mi>&sigma;</mi>
<mi>&tau;</mi>
</msub>
<msqrt>
<msub>
<mi>n</mi>
<mi>&tau;</mi>
</msub>
</msqrt>
</mfrac>
<msub>
<mi>z</mi>
<mrow>
<mi>&alpha;</mi>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msub>
<mo>,</mo>
<msub>
<mover>
<mi>D</mi>
<mo>&OverBar;</mo>
</mover>
<mi>&tau;</mi>
</msub>
<mo>+</mo>
<mfrac>
<msub>
<mi>&sigma;</mi>
<mi>&tau;</mi>
</msub>
<msqrt>
<msub>
<mi>n</mi>
<mi>&tau;</mi>
</msub>
</msqrt>
</mfrac>
<msub>
<mi>z</mi>
<mrow>
<mi>&alpha;</mi>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msub>
<mo>&rsqb;</mo>
</mrow>
Wherein, nτFor trace bar number in τ class tracks of vehicle,
(5), the abnormality detection of track of vehicle
According to track of vehicle similarity measurement calculation formula in step (3.1), track of vehicle to be measured and every quasi-representative track are calculated
Similarity measurement, if any one similarity measurement is in confidential interval, the track of vehicle is normal, otherwise the track of vehicle
It is abnormal.
2. the method for detecting abnormality according to claim 1 based on track of vehicle similitude, it is characterised in that described step
Suddenly in (3.3), spectral clustering uses k mean clusters.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510046984.1A CN104657746B (en) | 2015-01-29 | 2015-01-29 | A kind of method for detecting abnormality based on track of vehicle similitude |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510046984.1A CN104657746B (en) | 2015-01-29 | 2015-01-29 | A kind of method for detecting abnormality based on track of vehicle similitude |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104657746A CN104657746A (en) | 2015-05-27 |
CN104657746B true CN104657746B (en) | 2017-09-12 |
Family
ID=53248845
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510046984.1A Expired - Fee Related CN104657746B (en) | 2015-01-29 | 2015-01-29 | A kind of method for detecting abnormality based on track of vehicle similitude |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104657746B (en) |
Families Citing this family (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106055665B (en) * | 2016-06-02 | 2020-03-17 | 兰州大学 | Context-aware Web service recommendation method and system based on abnormal value elimination |
CN109923595B (en) * | 2016-12-30 | 2021-07-13 | 同济大学 | Urban road traffic abnormity detection method based on floating car data |
EP3729397A1 (en) * | 2018-01-23 | 2020-10-28 | Siemens Mobility GmbH | System, device and method for detecting abnormal traffic events in a geographical location |
CN108387932B (en) * | 2018-01-29 | 2019-10-18 | 成都承义科技有限公司 | A kind of recognition methods of seismic signal and device |
CN108680174B (en) * | 2018-05-10 | 2019-05-10 | 长安大学 | A method of map match abnormal point is improved based on machine learning algorithm |
CN110688435B (en) * | 2018-07-04 | 2022-04-05 | 北京嘀嘀无限科技发展有限公司 | Similar track searching method and system |
EP3853789A1 (en) | 2018-10-16 | 2021-07-28 | Huawei Technologies Co., Ltd. | Improved trajectory matching based on use of quality indicators empowered by weighted confidence values |
CN111079776A (en) * | 2018-10-22 | 2020-04-28 | 百度在线网络技术(北京)有限公司 | Method and device for judging track abnormity of obstacle vehicle |
CN109726737B (en) * | 2018-11-27 | 2020-11-10 | 武汉极意网络科技有限公司 | Track-based abnormal behavior detection method and device |
CN109858517B (en) * | 2018-12-25 | 2021-03-30 | 中国石油大学(华东) | Track similarity measurement method taking motion direction as dominant |
CN110443288B (en) * | 2019-07-19 | 2021-09-28 | 浙江大学城市学院 | Trajectory similarity calculation method based on sequencing learning |
US11443184B2 (en) | 2019-08-19 | 2022-09-13 | Toyota Research Institute, Inc. | Methods and systems for predicting a trajectory of a road agent based on an intermediate space |
CN110990678B (en) * | 2019-10-29 | 2022-05-10 | 浙江大学城市学院 | Trajectory similarity calculation method based on enhanced recurrent neural network |
CN111046942A (en) * | 2019-12-09 | 2020-04-21 | 交控科技股份有限公司 | Turnout fault judgment method and device |
CN111159254B (en) * | 2019-12-30 | 2023-07-25 | 武汉长江通信产业集团股份有限公司 | Vehicle and personnel association method based on big data processing |
CN111768430B (en) * | 2020-06-23 | 2023-08-11 | 重庆大学 | Expressway outfield vehicle tracking method based on multi-feature cascade matching |
CN112230253B (en) * | 2020-10-13 | 2021-07-09 | 电子科技大学 | Track characteristic anomaly detection method based on public slice subsequence |
CN113158415B (en) * | 2021-02-23 | 2023-09-08 | 电子科技大学长三角研究院(衢州) | Vehicle track similarity evaluation method based on error analysis |
CN113204914B (en) * | 2021-04-12 | 2022-06-17 | 哈尔滨工业大学 | Flight data abnormity interpretation method based on multi-flight data characterization modeling |
CN113992399A (en) * | 2021-10-22 | 2022-01-28 | 安徽江淮汽车集团股份有限公司 | Vehicle networking Sybil tool detection method based on mobile behavior |
CN114372382B (en) * | 2022-03-22 | 2022-06-10 | 交通运输部公路科学研究所 | Method, device and storage medium for evaluating reliability of vehicle track |
CN115859129B (en) * | 2023-02-27 | 2023-07-14 | 南昌工程学院 | Vehicle driving track similarity measurement method and system based on sparse satellite positioning |
CN117523382A (en) * | 2023-07-19 | 2024-02-06 | 石河子大学 | Abnormal track detection method based on improved GRU neural network |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102855638A (en) * | 2012-08-13 | 2013-01-02 | 苏州大学 | Detection method for abnormal behavior of vehicle based on spectrum clustering |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI430212B (en) * | 2010-06-08 | 2014-03-11 | Gorilla Technology Inc | Abnormal behavior detection system and method using automatic classification of multiple features |
-
2015
- 2015-01-29 CN CN201510046984.1A patent/CN104657746B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102855638A (en) * | 2012-08-13 | 2013-01-02 | 苏州大学 | Detection method for abnormal behavior of vehicle based on spectrum clustering |
Non-Patent Citations (2)
Title |
---|
A Novel Approach for Clustering Similarity Trajectories Using Averaging Method in Video Surveillance;Hui Qing See等;《IEEE》;20111231;第29-33页 * |
基于Hausdorff距离的视觉监控轨迹分类算法;曲琳 等;《吉林大学学报》;20091130;第39卷(第6期);第1618-1624页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104657746A (en) | 2015-05-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104657746B (en) | A kind of method for detecting abnormality based on track of vehicle similitude | |
CN103235933B (en) | A kind of vehicle abnormality behavioral value method based on HMM | |
Chen et al. | A just-in-time-learning-aided canonical correlation analysis method for multimode process monitoring and fault detection | |
CN105279365B (en) | For the method for the sample for learning abnormality detection | |
CN103164706B (en) | Object counting method and device based on video signal analysis | |
US8682824B2 (en) | Method and device for monitoring the state of a facility | |
CN110084292A (en) | Object detection method based on DenseNet and multi-scale feature fusion | |
CN106874894A (en) | A kind of human body target detection method based on the full convolutional neural networks in region | |
CN103473540B (en) | The modeling of intelligent transportation system track of vehicle increment type and online method for detecting abnormality | |
CN102609720B (en) | Pedestrian detection method based on position correction model | |
US20130318011A1 (en) | Method for Detecting Anomalies in Multivariate Time Series Data | |
CN106226050A (en) | A kind of TFDS fault automatic identifying method | |
KR101538843B1 (en) | Yield management system and method for root cause analysis using manufacturing sensor data | |
CN103761748A (en) | Method and device for detecting abnormal behaviors | |
CN106384092A (en) | Online low-rank abnormal video event detection method for monitoring scene | |
CN105354542B (en) | A kind of video accident detection method under crowd scene | |
CN103034860A (en) | Scale-invariant feature transform (SIFT) based illegal building detection method | |
KR20190072652A (en) | Information processing apparatus and information processing method | |
CN105425150A (en) | Motor fault diagnosis method based on RBF and PCA-SVDD | |
CN106682573A (en) | Pedestrian tracking method of single camera | |
CN112766301B (en) | Oil extraction machine indicator diagram similarity judging method | |
CN112070043A (en) | Safety helmet wearing convolutional network based on feature fusion, training and detecting method | |
CN106023499A (en) | Fiber security signal dual identification method and system | |
CN102254183A (en) | Face detection method based on AdaBoost algorithm | |
CN105469380A (en) | Method and device for detecting shielding against object |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170912 Termination date: 20200129 |