CN114239902A - User track prediction and abnormal track detection method and system - Google Patents
User track prediction and abnormal track detection method and system Download PDFInfo
- Publication number
- CN114239902A CN114239902A CN202111247387.7A CN202111247387A CN114239902A CN 114239902 A CN114239902 A CN 114239902A CN 202111247387 A CN202111247387 A CN 202111247387A CN 114239902 A CN114239902 A CN 114239902A
- Authority
- CN
- China
- Prior art keywords
- point
- track
- feature vector
- data
- foot
- 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.)
- Pending
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 187
- 238000000034 method Methods 0.000 claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 238000004364 calculation method Methods 0.000 claims abstract description 16
- 230000004927 fusion Effects 0.000 claims description 25
- 206010034701 Peroneal nerve palsy Diseases 0.000 claims description 20
- 238000005259 measurement Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 238000013145 classification model Methods 0.000 claims description 3
- 238000007477 logistic regression Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 230000005856 abnormality Effects 0.000 description 5
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for user trajectory prediction and abnormal trajectory detection, wherein the method comprises the following steps: s100, acquiring road network data, geographic information data and high-risk spatiotemporal information data; s200, preprocessing original track data based on a road network calculation engine and road network data, wherein the preprocessing comprises outlier removal, missing item filling, road network matching and period division; s300, identifying all foot points/stagnation points of the user track as a foot point/stagnation point candidate set based on the preprocessed track data; s400, extracting road network structure feature vectors, track structure feature vectors and foot-falling points/standing points semantic feature vectors in the track data based on the road network data, the geographic information data and the foot-falling points/standing points candidate set. The method and the device can give consideration to the prediction of the short-term track trend and the track stagnation point of the user, and have higher accuracy and better prediction effect.
Description
Technical Field
The invention relates to the field of spatio-temporal data analysis, in particular to a user trajectory prediction and abnormal trajectory detection method and system.
Background
The existing user trajectory prediction technology is generally limited to predicting one of the short-term trajectory movement or the trajectory stagnation point of a user, and cannot simultaneously perform prediction analysis on the short-term trajectory movement or the trajectory stagnation point. The existing abnormal track detection method mainly measures the abnormality according to the deviation degree of a local track compared with a normal track, and does not consider the relation between the local track and a stationary point or a space-time risk point. Therefore, a technical solution that can consider both the short-term trajectory movement and trajectory stagnation point prediction of the user and can also consider the current trajectory abnormality is needed.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a user track prediction and abnormal track detection method and system, which can give consideration to the prediction of the short-term track movement and track stagnation point of a user, and have higher prediction accuracy and better prediction effect.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a user trajectory prediction and abnormal trajectory detection method comprises the following steps:
s100, acquiring road network data, geographic information data and high-risk spatiotemporal information data;
s200, preprocessing original track data based on a road network calculation engine and the road network data, wherein the preprocessing comprises abnormal value removal, missing item filling, road network matching and period division;
s300, identifying all foot points/stagnation points of the user track as a foot point/stagnation point candidate set based on the preprocessed track data;
s400, extracting road network structure feature vectors, track structure feature vectors and foot-falling points/standing points semantic feature vectors in the track data based on the road network data, the geographic information data and the foot-falling points/standing points candidate set;
s500, performing feature fusion on the road network structure feature vector, the track structure feature vector and the foot-falling point/stationing point semantic feature vector in the track data to obtain a fused current track feature vector, and performing historical feature fusion on the fused current track feature vector and a historical track feature vector in the same period to obtain a historical feature vector;
s600, generating a corresponding track sequence based on the fused current track characteristic vector and the historical characteristic vector, wherein the track sequence comprises a predicted road section characteristic vector and a predicted foot-landing point/standing point characteristic vector;
s700, regularizing a candidate road section set at a next time point of a current road section and the feature vector of the predicted road section to obtain a predicted road section, and regularizing the candidate road section set based on the foot-drop point/stationing point and the feature vector of the predicted foot-drop point/stationing point to obtain a predicted foot-drop point/stationing point;
s800, based on the predicted road section, the predicted foothold/stationing point, the high-risk spatiotemporal information data and a candidate route set between the current road section and the predicted foothold/foothold, obtaining two indexes of a route deviation index and a potential risk index as abnormal state measurement indexes, and performing state detection to obtain a detection result.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, in S200, performing road network matching on the original trajectory data includes:
and matching the user track in the track data with the road network data through a road network calculation engine, and acquiring the road section corresponding to each time point of the user track from the road network data.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, S300 includes:
and after acquiring the road sections corresponding to each time point of the user track from the road network data, comprehensively customizing the track data in the period, and acquiring the foot-falling points/standing points in the period through a foot-falling point/standing point clustering device to be used as a foot-falling point/standing point candidate set.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, in S400, extracting a feature vector of a road network structure in the trajectory data includes:
taking a road section obtained by road network matching of the track data as a starting point, taking a time circle of walking for 30 minutes and the like as a receptive field, obtaining all road sections in the receptive field through a receptive field local filter, and extracting a road network structure feature vector as a road network structure feature vector in the track data;
extracting a track structure feature vector in the track data, including:
acquiring all road sections of user tracks in the road network data in the track data, carrying out graph embedding on each road section in a anonymous walking mode in a graph embedding algorithm, and acquiring a track structure feature vector as a track structure feature vector in the track data;
extracting the semantic feature vectors of the footfall/standing points in the track data, which comprises the following steps:
and acquiring information associated with the foot-falling points/the standing points in the candidate set of foot-falling points/standing points from the geographic information data, and screening the attribute of the information as a semantic feature vector of the foot-falling points/the standing points in the trajectory data through feature engineering.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, in S500, feature fusion is performed on a road network structure feature vector, a trajectory structure feature vector, and a footer/anchor semantic feature vector in the trajectory data, including:
unifying the road network structure feature vector, the track structure feature vector and the foot-falling point/standing point semantic feature vector in the track data into the same dimension in a vector embedding mode;
and performing feature fusion on the road network structure feature vector, the track structure feature vector and the foot-landing point/resident point semantic feature vector in the track data through an Attention operator, a full-connection operator or a summation operator.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, S600 includes:
and inputting the fused current track characteristic vector and the historical characteristic vector, the motion state of the current track and whether the current track is at the historical footfall point/stagnation point into a seq2seq model based on a neural network, and outputting the predicted road section characteristic vector and the predicted footfall point/stagnation point characteristic vector.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, S700 includes:
acquiring a candidate road section of a current road section at a next time point by the road network calculation engine to serve as the candidate road section set;
performing dot product operation on the structural feature vector of the candidate road section set and the predicted road section feature vector, and taking the road section with the largest dot product value as the predicted road section;
and performing dot product operation on the semantic feature vector of the pin point/stationing point candidate set and the predicted pin point/stationing point feature vector, and taking the pin point/stationing point with the largest dot product value as the predicted pin point/stationing point.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, in S800, a route deviation index and a potential risk index are obtained based on the predicted road segment, the predicted foothold/stationing point, the high-risk spatiotemporal information data, and a candidate route set between the current road segment and the predicted foothold/foothold, including:
acquiring all potential shortest route sections from the current road section to the predicted footfall point/stagnation point through the road network calculation engine as the candidate route set;
calculating a probability of occurrence of the predicted section in the candidate route set as the route deviation index;
calculating the coverage rate of the candidate route set on the high-risk spatio-temporal information data as the potential risk indicator.
Further, in the method for predicting a user trajectory and detecting an abnormal trajectory as described above, in S800, the abnormal state metric is subjected to state detection by using a detection classification model based on logistic regression.
The embodiment of the invention also provides a system for predicting the user track and detecting the abnormal track, which comprises the following steps:
the data acquisition module is used for acquiring road network data, geographic information data and high-risk spatiotemporal information data;
the preprocessing module is used for preprocessing original track data based on a road network computing engine and the road network data, and comprises abnormal value removal, missing item filling, road network matching and period division;
the foot point/standing point identification module is used for identifying all foot points/standing points of the user track as a foot point/standing point candidate set based on the track data after preprocessing;
the feature extraction module is used for extracting a road network structure feature vector, a track structure feature vector and a foot-falling point/standing point semantic feature vector from the track data based on the road network data, the geographic information data and the foot-falling point/standing point candidate set;
the feature fusion module is used for performing feature fusion on a road network structure feature vector, a track structure feature vector and a foot point/stationing point semantic feature vector in the track data to obtain a fused current track feature vector, and performing historical feature fusion on the fused current track feature vector and a historical track feature vector in the same period to obtain a historical feature vector;
a track sequence generating module, configured to generate a corresponding track sequence based on the fused current track feature vector and the historical feature vector, where the track sequence includes a predicted road section feature vector and a predicted landing point/stagnation point feature vector;
the prediction result regularization module is used for regularizing a candidate road section set at a time point under a current road section and the feature vector of the prediction road section to obtain a prediction road section, and regularizing the candidate road section set based on the foot-drop point/the feature vector of the prediction foot-drop point/the feature vector of the stagnation point to obtain a prediction foot-drop point/the stagnation point;
and the state detection module is used for acquiring two indexes of a route deviation index and a potential risk index as abnormal state measurement indexes based on the predicted road section, the predicted foothold/stationing point, the high-risk spatio-temporal information data and a candidate route set between the current road section and the predicted foothold/foothold, and performing state detection to obtain a detection result.
The invention has the beneficial effects that: the method and the system provided by the invention can give consideration to the prediction of the short-term track movement direction and the track stagnation point of the user, and can comprehensively analyze the abnormality of the current track by combining the local track movement direction, the future potential stagnation point and the space-time risk points defined under different scenes, thereby ensuring higher prediction accuracy and better prediction effect.
Drawings
Fig. 1 is a schematic flowchart of a user trajectory prediction and abnormal trajectory detection method provided in an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a user trajectory prediction and abnormal trajectory detection system according to an embodiment of the present invention;
fig. 3 is a flowchart of a user trajectory prediction and abnormal trajectory detection system according to an embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and the detailed description.
As shown in fig. 1, a method for predicting a user trajectory and detecting an abnormal trajectory includes:
s100, acquiring road network data, geographic information data and high-risk spatiotemporal information data;
in this embodiment, the data most necessary for trajectory prediction is road network data, and the road network data may be stored in different ways according to different selected road network calculation engines. In order to improve the prediction accuracy, the semantic characteristics of the track need to be acquired, public POI (Point of Information) data is selected to assist in representing the semantic characteristics of the track, and the semantic characteristics of the track can be stored in a geographic database or stored in a database after being matched with a road network. The high-risk spatiotemporal information data are data for assisting in representing track abnormality, are acquired and defined according to business requirements, for example, the fire fighting field is a key place, the public security field is a large-scale activity collection and distribution place, a key place and the like, and adopt the same storage mode as the POI data.
S200, preprocessing original track data based on a road network calculation engine and road network data, wherein the preprocessing comprises outlier removal, missing item filling, road network matching and period division;
in this embodiment, the original trajectory data has many missing information or abnormal values, such as a large-scale deviation in short-term positioning, missing positioning information, and the like. Firstly, exception processing operations such as elimination, smoothing and the like are required for an abnormal value; then filling missing values, filling missing items in modes of interpolation, resampling and the like for small-range missing, and selecting a proper interpolation scheme for processing large-range missing according to scenes; the track data needs to be matched with a road network, and the track points are matched with the road network by combining the road network data through a road network calculation engine to obtain a road section corresponding to each track; finally, the trajectory data needs to be divided into at least one fixed period, such as hours, days, weeks, months, etc., so as to predict the relevant information in the integrated period.
S300, identifying all foot points/stagnation points of the user track as a foot point/stagnation point candidate set based on the preprocessed track data;
in this embodiment, the foothold point is a place where the user stays, and the stationing point is a place where the user stays for a long time. After the road section corresponding to each time point of the track is obtained, track records in the self-defined period are comprehensively defined, and the foot-falling points/stagnation points in the period are obtained by utilizing a space-time clustering algorithm so as to determine potential destinations of the user when going out, and form a foot-falling point/stagnation point candidate set.
S400, extracting road network structure feature vectors, track structure feature vectors and foot-falling points/standing points semantic feature vectors in the track data based on the road network data, the geographic information data and the foot-falling points/standing points candidate set;
for the road network structure feature vector, the feature of the road section relative to the road section in the receptive field needs to be calculated, the receptive field is defined according to the scene (default is 30 minutes walking and other time circles), and a road section set in the receptive field corresponding to each road section as a starting point in the preset road network data is obtained; the method comprises the steps of inputting a road segment set obtained by a receptive field, extracting topological features of a road network, outputting feature vectors of the road network, selecting different indexes for different fields of feature extraction modes, such as integration degree and selectivity in space syntax, proximity degree and intermediate degree in space design network analysis, and obtaining embedded feature vectors of all road segments in the receptive field by adopting a graph neural network mode.
For the track structure feature vector, a part of data in user track data can be selected according to business needs to calculate (one week is selected by default), a road section sub-graph of a user track in a road network is obtained, the frequency of skip among the user road sections is set as the weight among the road sections, and each road section is subjected to graph embedding in an anonymous walking mode in a graph embedding algorithm by taking a foot point as a starting point, so that the track structure feature is obtained. For the road sections outside the road section sub-graph, the track characteristics can be initialized to zero vectors or random vectors according to requirements.
And for the semantic feature vectors of the foot points/the stationing points, aiming at each foot point/stationing point, screening the attributes of the geographic semantic information data through feature engineering to be used as the semantic features of the foot points/the stationing points. For the road section where the non-footfall point/stationing point belongs to, the semantic features of the road section are defaulted to zero vectors.
S500, performing feature fusion on a road network structure feature vector, a track structure feature vector and a foot-down point/resident point semantic feature vector in track data to obtain a fused current track feature vector, and performing historical feature fusion on the fused current track feature vector and a historical track feature vector in the same period to obtain a historical feature vector;
in this embodiment, the three features need to be fused in the feature fusion stage, the fusion mode adopts a vector embedding mode to unify the three features into the same dimension, and then the three feature vectors are fused by using an Attention operator, a full-join operator or a summation operator. The information recorded by a single track cannot reflect the intention of a user, only the track containing the context can reflect the real intention of the user, the fused current track characteristic vector and the historical track characteristic vector in the same period are input into a historical characteristic fusion device together, and the historical characteristic vector is represented by a summation operator or a GRU (neural network unit) and other neural network models.
S600, generating a corresponding track sequence based on the fused current track characteristic vector and the fused historical characteristic vector, wherein the track sequence comprises a predicted road section characteristic vector and a predicted foot-landing point/standing point characteristic vector;
in this embodiment, in order to synthesize the short-term motion prediction and the medium-term and long-term stagnation point prediction, in the track sequence generation stage, the motion state (whether to stay) of the current track, the comprehensive characteristics (the current track characteristic vector and the historical characteristic vector fused in S500) of the current track, and whether the current track is in the historical footfall point/stagnation point are input into the seq2seq model of the neural network, and a vector composed of two parts is output, where one part is a predicted road section characteristic vector (i.e., a road section characteristic vector at the next time point) and the other part is a predicted footfall point/stagnation point characteristic vector (i.e., a next footfall point/stagnation point characteristic vector).
S700, regularizing a candidate road section set and a predicted road section feature vector at a next time point of a current road section to obtain a predicted road section, and regularizing the candidate road section set and the predicted road section feature vector to obtain a predicted road section, and the predicted road section feature vector is obtained based on a foot drop point/parking point candidate set and a predicted foot drop point/parking point feature vector;
in this embodiment, the candidate road segment set is composed of candidate road segments at the next time point obtained by the road network calculation engine based on the current road segment. And performing dot product operation on the structural feature vector of the candidate road section set and the predicted road section feature vector output in the step S600, and taking the candidate road section with the largest dot product value as a final predicted road section. Similarly, the semantic feature vector of the candidate set of the foot-falling point/standing point obtained in S300 and the predicted foot-falling point/standing point feature vector output in S600 are subjected to dot product operation, and the foot-falling point/standing point with the largest dot product value is taken as the final predicted foot-falling point/standing point.
S800, acquiring two indexes of a route deviation index and a potential risk index as abnormal state measurement indexes based on the predicted road section, the predicted foothold/stationing point, the high-risk spatio-temporal information data and a candidate route set between the current road section and the predicted foothold/foothold, and carrying out state detection to obtain a detection result.
In this embodiment, a road network calculation engine obtains all potential shortest route segments of a current route segment to a predicted footfall point/stationing point output in S700 as a candidate route set, calculates a probability of occurrence of the predicted route segment output in S700 in the candidate route set as a route deviation index, calculates a coverage rate of the candidate route set on high-risk spatio-temporal information data as a potential risk index, uses the two indexes as abnormal state measurement indexes, performs state detection on the abnormal state measurement indexes through a detection classification model based on logistic regression, and detects whether a current trajectory state is normal.
As shown in fig. 2, a system for predicting a user trajectory and detecting an abnormal trajectory includes:
a data acquisition module 100, configured to acquire road network data, geographic information data, and high-risk spatiotemporal information data;
the preprocessing module 200 is configured to perform preprocessing on the original trajectory data based on the road network calculation engine 900 and the road network data, including outlier removal, missing item filling, road network matching, and cycle division;
a foot point/standing point identification module 300, configured to identify all foot points/standing points of the user trajectory as a foot point/standing point candidate set based on the preprocessed trajectory data;
the feature extraction module 400 is configured to extract a road network structure feature vector, a track structure feature vector, and a foot-drop point/standing point semantic feature vector in the track data based on the road network data, the geographic information data, and the foot-drop point/standing point candidate set;
the feature fusion module 500 is configured to perform feature fusion on a road network structure feature vector, a track structure feature vector, and a foot-down point/stationary point semantic feature vector in the track data to obtain a fused current track feature vector, and perform historical feature fusion on the fused current track feature vector and a historical track feature vector in the same period to obtain a historical feature vector;
a track sequence generating module 600, configured to generate a corresponding track sequence based on the fused current track feature vector and the historical feature vector, where the track sequence includes a predicted road section feature vector and a predicted landing point/stagnation point feature vector;
the prediction result regularization module 700 is configured to regularize the candidate road segment set and the predicted road segment feature vector at a time point of the current road segment to obtain a predicted road segment, and regularize the candidate road segment set and the predicted road segment set and predicted road segment set;
and the state detection module 800 is configured to obtain two indexes, namely a route deviation index and a potential risk index, as abnormal state measurement indexes based on the predicted road section, the predicted foothold/stationing point, the high-risk spatio-temporal information data, and a candidate route set between the current road section and the predicted foothold/foothold, and perform state detection to obtain a detection result.
In this embodiment, as shown in fig. 3, in the preprocessing module 200, the abnormal value processing is performed on the trajectory data first, and then the missing term interpolation is performed; the interpolated track data is subjected to road network matching, and a road network calculation engine 900 is called to calculate pre-stored road network data to acquire relevant information; inputting the data in a certain period of integration into a foot-drop point/standing point clustering device for foot-drop point/standing point identification, and adding candidate standing points/foot-drop point sets for newly-appeared foot-drop points/standing points; for the matched track data, extracting the track structure characteristics of the data in a certain period; extracting semantic features according to information which is in space-time correlation with the footfall/stationing points in the geographic information base through a semantic feature extractor, and defaulting the non-footfall/stationing points into zero vectors or random vectors; all road sections in the receptive field of the road sections in the road network are obtained through the receptive field local screener and road network characteristics are extracted; performing feature fusion on the track structure feature vector, the road network structure feature vector and the track semantic feature vector; the fused track characteristic vector and the historical track characteristic vector in the same period enter a historical characteristic fusion device to be subjected to historical characteristic fusion; track characteristics, historical characteristics, the state of the current track, and whether the current track is in a historical foothold/stationing point or not are input into the track sequence generation module 600 for track prediction, and a future stationing point/foothold characteristic vector and a road section characteristic vector are finally output; regularizing the prediction result by a prediction result regularization module 700 to obtain a prediction stagnation point/footfall point and a prediction road section; the state detection module 800 combines the comprehensive predicted road section, the predicted stopping point/foot point, and the candidate route set between the current road section and the predicted stopping point/foot point with the customized high-risk spatio-temporal information base to output whether the current trajectory is in an abnormal state.
The user track prediction and abnormal track detection method and system provided by the invention can give consideration to the prediction of the short-term track movement direction and the track stagnation point of the user, and can comprehensively analyze the abnormality of the current track by combining the local track movement direction, the future potential stagnation point and the space-time risk points defined under different scenes, so that the prediction accuracy is higher and the prediction effect is better.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is intended to include such modifications and variations.
Claims (10)
1. A user trajectory prediction and abnormal trajectory detection method is characterized by comprising the following steps:
s100, acquiring road network data, geographic information data and high-risk spatiotemporal information data;
s200, preprocessing original track data based on a road network calculation engine and the road network data, wherein the preprocessing comprises abnormal value removal, missing item filling, road network matching and period division;
s300, identifying all foot points/stagnation points of the user track as a foot point/stagnation point candidate set based on the preprocessed track data;
s400, extracting road network structure feature vectors, track structure feature vectors and foot-falling points/standing points semantic feature vectors in the track data based on the road network data, the geographic information data and the foot-falling points/standing points candidate set;
s500, performing feature fusion on the road network structure feature vector, the track structure feature vector and the foot-falling point/stationing point semantic feature vector in the track data to obtain a fused current track feature vector, and performing historical feature fusion on the fused current track feature vector and a historical track feature vector in the same period to obtain a historical feature vector;
s600, generating a corresponding track sequence based on the fused current track characteristic vector and the historical characteristic vector, wherein the track sequence comprises a predicted road section characteristic vector and a predicted foot-landing point/standing point characteristic vector;
s700, regularizing a candidate road section set at a next time point of a current road section and the feature vector of the predicted road section to obtain a predicted road section, and regularizing the candidate road section set based on the foot-drop point/stationing point and the feature vector of the predicted foot-drop point/stationing point to obtain a predicted foot-drop point/stationing point;
s800, based on the predicted road section, the predicted foothold/stationing point, the high-risk spatiotemporal information data and a candidate route set between the current road section and the predicted foothold/foothold, obtaining two indexes of a route deviation index and a potential risk index as abnormal state measurement indexes, and performing state detection to obtain a detection result.
2. The method for predicting user trajectories and detecting abnormal trajectories according to claim 1, wherein the step S200 of performing road network matching on the original trajectory data comprises:
and matching the user track in the track data with the road network data through a road network calculation engine, and acquiring the road section corresponding to each time point of the user track from the road network data.
3. The method for predicting user trajectories and detecting abnormal trajectories according to claim 2, wherein the step S300 comprises:
and after acquiring the road sections corresponding to each time point of the user track from the road network data, comprehensively customizing the track data in the period, and acquiring the foot-falling points/standing points in the period through a foot-falling point/standing point clustering device to be used as a foot-falling point/standing point candidate set.
4. The method according to claim 2, wherein the step of extracting the feature vector of the road network structure from the trajectory data in S400 comprises:
taking a road section obtained by road network matching of the track data as a starting point, taking a time circle of walking for 30 minutes and the like as a receptive field, obtaining all road sections in the receptive field through a receptive field local filter, and extracting a road network structure feature vector as a road network structure feature vector in the track data;
extracting a track structure feature vector in the track data, including:
acquiring all road sections of user tracks in the road network data in the track data, carrying out graph embedding on each road section in a anonymous walking mode in a graph embedding algorithm, and acquiring a track structure feature vector as a track structure feature vector in the track data;
extracting the semantic feature vectors of the footfall/standing points in the track data, which comprises the following steps:
and acquiring information associated with the foot-falling points/the standing points in the candidate set of foot-falling points/standing points from the geographic information data, and screening the attribute of the information as a semantic feature vector of the foot-falling points/the standing points in the trajectory data through feature engineering.
5. The method according to claim 1, wherein in S500, feature fusion is performed on the road network structure feature vector, the track structure feature vector, and the footer/anchor semantic feature vector in the track data, and the method includes:
unifying the road network structure feature vector, the track structure feature vector and the foot-falling point/standing point semantic feature vector in the track data into the same dimension in a vector embedding mode;
and performing feature fusion on the road network structure feature vector, the track structure feature vector and the foot-landing point/resident point semantic feature vector in the track data through an Attention operator, a full-connection operator or a summation operator.
6. The method for predicting user trajectories and detecting abnormal trajectories according to claim 1, wherein the step S600 comprises:
and inputting the fused current track characteristic vector and the historical characteristic vector, the motion state of the current track and whether the current track is at the historical footfall point/stagnation point into a seq2seq model based on a neural network, and outputting the predicted road section characteristic vector and the predicted footfall point/stagnation point characteristic vector.
7. The method for predicting user trajectories and detecting abnormal trajectories according to claim 1, wherein the step S700 comprises:
acquiring a candidate road section of a current road section at a next time point by the road network calculation engine to serve as the candidate road section set;
performing dot product operation on the structural feature vector of the candidate road section set and the predicted road section feature vector, and taking the road section with the largest dot product value as the predicted road section;
and performing dot product operation on the semantic feature vector of the pin point/stationing point candidate set and the predicted pin point/stationing point feature vector, and taking the pin point/stationing point with the largest dot product value as the predicted pin point/stationing point.
8. The method for predicting user trajectories and detecting abnormal trajectories according to claim 1, wherein in step S800, obtaining a trajectory deviation index and a potential risk index based on the predicted road segments, the predicted foothold/stationing points, the high-risk spatiotemporal information data, and a candidate trajectory set between a current road segment and the predicted foothold/foothold, comprises:
acquiring all potential shortest route sections from the current road section to the predicted footfall point/stagnation point through the road network calculation engine as the candidate route set;
calculating a probability of occurrence of the predicted section in the candidate route set as the route deviation index;
calculating the coverage rate of the candidate route set on the high-risk spatio-temporal information data as the potential risk indicator.
9. The method according to claim 1, wherein in step S800, the abnormal state metric is detected by a detection classification model based on logistic regression.
10. A user trajectory prediction and abnormal trajectory detection system is characterized by comprising:
the data acquisition module is used for acquiring road network data, geographic information data and high-risk spatiotemporal information data;
the preprocessing module is used for preprocessing original track data based on a road network computing engine and the road network data, and comprises abnormal value removal, missing item filling, road network matching and period division;
the foot point/standing point identification module is used for identifying all foot points/standing points of the user track as a foot point/standing point candidate set based on the track data after preprocessing;
the feature extraction module is used for extracting a road network structure feature vector, a track structure feature vector and a foot-falling point/standing point semantic feature vector from the track data based on the road network data, the geographic information data and the foot-falling point/standing point candidate set;
the feature fusion module is used for performing feature fusion on a road network structure feature vector, a track structure feature vector and a foot point/stationing point semantic feature vector in the track data to obtain a fused current track feature vector, and performing historical feature fusion on the fused current track feature vector and a historical track feature vector in the same period to obtain a historical feature vector;
a track sequence generating module, configured to generate a corresponding track sequence based on the fused current track feature vector and the historical feature vector, where the track sequence includes a predicted road section feature vector and a predicted landing point/stagnation point feature vector;
the prediction result regularization module is used for regularizing a candidate road section set at a time point under a current road section and the feature vector of the prediction road section to obtain a prediction road section, and regularizing the candidate road section set based on the foot-drop point/the feature vector of the prediction foot-drop point/the feature vector of the stagnation point to obtain a prediction foot-drop point/the stagnation point;
and the state detection module is used for acquiring two indexes of a route deviation index and a potential risk index as abnormal state measurement indexes based on the predicted road section, the predicted foothold/stationing point, the high-risk spatio-temporal information data and a candidate route set between the current road section and the predicted foothold/foothold, and performing state detection to obtain a detection result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111247387.7A CN114239902A (en) | 2021-10-26 | 2021-10-26 | User track prediction and abnormal track detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111247387.7A CN114239902A (en) | 2021-10-26 | 2021-10-26 | User track prediction and abnormal track detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114239902A true CN114239902A (en) | 2022-03-25 |
Family
ID=80743241
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111247387.7A Pending CN114239902A (en) | 2021-10-26 | 2021-10-26 | User track prediction and abnormal track detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114239902A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115002679A (en) * | 2022-07-18 | 2022-09-02 | 北京航天泰坦科技股份有限公司 | Trajectory deviation rectifying processing method and device |
| WO2024098956A1 (en) * | 2022-11-10 | 2024-05-16 | 中国测绘科学研究院 | Method for fusing social media data and moving track data |
-
2021
- 2021-10-26 CN CN202111247387.7A patent/CN114239902A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115002679A (en) * | 2022-07-18 | 2022-09-02 | 北京航天泰坦科技股份有限公司 | Trajectory deviation rectifying processing method and device |
| CN115002679B (en) * | 2022-07-18 | 2022-11-15 | 北京航天泰坦科技股份有限公司 | Trajectory deviation rectifying processing method and device |
| WO2024098956A1 (en) * | 2022-11-10 | 2024-05-16 | 中国测绘科学研究院 | Method for fusing social media data and moving track data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105467975B (en) | A kind of equipment fault diagnosis method | |
| KR102274389B1 (en) | Method for building anomaly pattern detection model using sensor data, apparatus and method for detecting anomaly using the same | |
| CN114239902A (en) | User track prediction and abnormal track detection method and system | |
| CN105548764B (en) | A kind of Fault Diagnosis for Electrical Equipment method | |
| CN110232082B (en) | Anomaly detection method for continuous space-time refueling data | |
| US8682824B2 (en) | Method and device for monitoring the state of a facility | |
| Chang et al. | Forecasting of manufacturing cost in mobile phone products by case-based reasoning and artificial neural network models | |
| CN110390262A (en) | Video analysis method, apparatus, server and storage medium | |
| CN104778245A (en) | Similar trajectory mining method and device on basis of massive license plate identification data | |
| CN105631596A (en) | Equipment fault diagnosis method based on multidimensional segmentation fitting | |
| KR102361423B1 (en) | Artificial intelligence system and method for predicting maintenance demand | |
| Sun et al. | Time series anomaly detection based on GAN | |
| US20040064450A1 (en) | Method for preparing data to be analyzed, data analysis method, data analysis device, data preparation program, data analysis program, data prediction device, data prediction method, data prediction program and computer | |
| CN109960232A (en) | The method that the selection method of leading auxiliary parameter and plant maintenance diagnose in advance | |
| CN109784253A (en) | A kind of detection method of bicycle user abnormal behaviour | |
| CN115170027A (en) | Data analysis method, device, equipment and storage medium | |
| Liu et al. | Detecting loaded trajectories for hazardous chemicals transportation | |
| CN109492065B (en) | Extraction method of indoor semantic map space-time relationship | |
| CN112767438B (en) | Multi-target tracking method combining space-time motion | |
| CN112650949B (en) | Regional POI (point of interest) demand identification method based on multi-source feature fusion collaborative filtering | |
| CN116956089A (en) | Training method and detection method for temperature anomaly detection model of electrical equipment | |
| CN114550291A (en) | Gait feature extraction method, device and equipment | |
| CN115132368A (en) | Infectious disease prevention and control monitoring method and system based on big data platform | |
| Asgari et al. | Transport mode detection when fine-grained and coarse-grained data meet | |
| CN112434648A (en) | Wall shape change detection method and system |
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 |