CN108734008B - Method for removing anonymity of moving track data anonymized to vehicle based on parking record - Google Patents
Method for removing anonymity of moving track data anonymized to vehicle based on parking record Download PDFInfo
- Publication number
- CN108734008B CN108734008B CN201810385918.0A CN201810385918A CN108734008B CN 108734008 B CN108734008 B CN 108734008B CN 201810385918 A CN201810385918 A CN 201810385918A CN 108734008 B CN108734008 B CN 108734008B
- Authority
- CN
- China
- Prior art keywords
- parking
- points
- anonymous
- tracks
- track
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/50—Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems
- G06F21/55—Detecting local intrusion or implementing counter-measures
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention relates to a method for removing anonymity of vehicle anonymous movement track data based on parking records, which comprises the following steps: extracting all parking points in the track, removing parking caused by passengers getting on and off, and dividing parking into manned parking and no-load parking; removing abnormal points of manned parking, and modeling the manned parking; removing traffic lights and traffic jam parking in the no-load parking points by using the obtained model, and reserving interesting parking points; and constructing feature vectors based on TF-IDF for the interested parking points, comparing the feature vectors of the known tracks with the feature vectors of the anonymous tracks, and finding the anonymous tracks matched with the known tracks in the anonymous track set. The invention realizes that the undeniated original track can be matched from the anonymous moving track set by depending on the parking spot record, verifies the stability and uniqueness of the characteristic of the parking spot interested by a single driver and reveals the security risk of the anonymous track data.
Description
Technical Field
The invention relates to the technical field of privacy security of anonymous movement tracks, in particular to a method for de-anonymizing anonymous movement track data of a vehicle based on parking records.
Background
With the development of mobile terminals and positioning technologies, the movement trajectory data of vehicles can be easily acquired, and the movement trajectories generally contain rich spatiotemporal information, and valuable information can be obtained by reasonably mining and analyzing the data. Meanwhile, the moving tracks contain personal privacy information, so that malicious attackers can deduce various privacy events interested by the malicious attackers through the moving tracks, thereby causing the privacy security problem.
In order to protect the privacy of the track data, the track data is generally preprocessed by using a related privacy protection technology before being released. At present, two general categories are common, one is to modify the original track and reduce the precision of the track in space-time, for example, reduce the resolution of the recorded track or insert noise into the track to protect privacy. But will result in serious data distortion and low usability; another approach to track anonymization is to substitute pseudonyms (unique random identifiers) for the participant's true identity, so that the participant's true identity cannot be associated with the pseudonyms in any way. The method is easy to implement, protects privacy under the condition of keeping low expenditure and unchanged original data, keeps the maximum availability of data, and is widely adopted.
The attacker can identify the trajectory of the attacked from the anonymous trajectory by the bypass information. It is assumed that an attacker can access some anonymous set of traces, including the traces of his attack targets. In the attack method, a plurality of time-space 'snapshots' (namely position information of a vehicle at a certain time) of an attack target vehicle in an anonymous track occurrence time period need to be acquired, so that tracks which are consistent with the acquired time-space 'snapshots' are identified from the anonymous track set. The attack has strong space-time limitation on 'snapshot', thereby limiting the implementation of de-anonymization attack. However, whether a more general attack mode exists or not can be determined, and the anonymous attack can be successfully implemented without acquiring the spatio-temporal information overlapped with the anonymous trajectory data storage time.
Disclosure of Invention
The invention aims to solve the technical problem of providing a method for de-anonymizing vehicle anonymous movement track data based on parking records, analyzing interesting parking point characteristics caused by personal preference of a driver, verifying the stability and uniqueness of the interesting parking point characteristics of a single driver, and disclosing the security risk of the anonymous track data.
The technical scheme adopted by the invention for solving the technical problems is as follows: the method for removing anonymity of the anonymous moving track data of the vehicle based on the parking record is provided, a plurality of moving track fragments of a taxi in any time period are obtained, and the track of the taxi is identified by comparing the holding track with the concentrated track of the anonymous track, and comprises the following steps:
(1) extracting all parking points in the track, removing parking caused by passengers getting on and off, and dividing parking into manned parking and idle parking, wherein the manned parking comprises parking caused by traffic lights and parking caused by traffic jam;
(2) removing abnormal points of manned parking, and modeling the manned parking;
(3) removing traffic lights and traffic jam parking in the no-load parking point by using the obtained model, and reserving the interesting parking point;
(4) and constructing feature vectors based on TF-IDF for the interested parking points, comparing the feature vectors of the known tracks with the feature vectors of the anonymous tracks, and finding the anonymous tracks matched with the known tracks in the anonymous track set.
The step (1) is specifically as follows: and (4) screening out points with positions not changing along with time from the track set, and judging whether the parking is carried out by passengers or not according to the manned states before and after the parking in the track set of the taxi.
In the step (2), dura is used for representing parking time, dist is used for representing the distance between a parking point and an intersection, inter represents the interval time between two parking points, and a characteristic vector S is constructedtpThe mahalanobis distance is used to remove outliers of manned parking.
The method for removing the abnormal points of manned parking by using the Mahalanobis distance specifically comprises the following steps: setting the Mahalanobis distance as
Wherein p is the degree of freedom, and the parking points exceeding the threshold are regarded as abnormal points; the mahalanobis distance is calculated as:
wherein StpiRepresenting the feature vector of the ith parking spot, and mu represents the mean value of the feature vectors of a group of parking spots; v represents the covariance matrix of the feature vector of the parking point set.
In the step (3), a manned parking point training One-classSVM classifier is used after the abnormal parking point is removed, so as to distinguish whether the parking point is a traffic jam or a traffic light waiting parking point; and identifying and removing the parking spots of traffic lights and the like in the no-load parking spots by using a classifier, and reserving the parking spots as interesting parking spots.
The step (4) is specifically as follows: the taxi running track is regarded as a text, the road section number is regarded as each word, and the number of the parking spots of the road section is the number of times of the word appearing in the text; weight tf-idfi，jThe calculation formula of (2) is as follows: tf-idfi，j＝tfi，j×idfi，jWherein, in the step (A),
Nifor vehicles v in training dataiTotal number of road segments traversed; m isjIncluding road segments r for training datajThe number of tracks of (a); ciFor vehicles v in training dataiThe class to which the locus belongs; m isi，jIs CiIncluding road section rjThe number of tracks of (a); t is ti，jIs rjAt CiThe number of occurrences in (a); tf isi，jIs rjAt viFrequency of occurrence in the trace of (a); the extracted parking points in the training data form track characteristics, the parking points in the test data form test track characteristics, the cosine similarity values are used for matching, and the similarity between the two vector inner product spaces is measured by measuring the cosine value of the included angle of the two vector inner product spaces; it is determined from the cosine value of the angle between the two vectors whether the two vectors point in approximately the same direction.
Advantageous effects
Due to the adoption of the technical scheme, compared with the prior art, the invention has the following advantages and positive effects: the invention realizes that the undeniated original track can be matched from the anonymous moving track in a centralized way by depending on the parking spot record, analyzes the interesting parking spot characteristics caused by personal preference of the driver, verifies the stability and uniqueness of the interesting parking spot characteristics of a single driver and reveals the security risk of anonymous track data.
Drawings
FIG. 1 is a flow chart of a method for de-anonymizing taxi tracks based on a stopping point;
FIG. 2 is a detailed classification diagram of taxi stops;
FIG. 3 is a parking spot feature vector distribution plot for a Shanghai taxi;
FIG. 4 is a parking point feature vector distribution diagram of Shenzhen taxi;
FIG. 5 is a diagram of the result of classification of empty taxi parking spots by SVM;
FIG. 6 is a result diagram of empty taxi parking spots of Shenzhen taxi classified by SVM;
FIG. 7 is a histogram of taxi track matching accuracy;
FIG. 8 is a line graph of the effect of test trace length change on matching accuracy;
FIG. 9 is a graph of the results of testing the effect of trace length changes on matching accuracy.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
The embodiment of the invention relates to a method for removing anonymity of an anonymous moving track of a taxi according to a taxi parking record, which comprises the steps of removing passenger parking and dividing passenger parking and no-load parking as shown in figure 1; removing abnormal points of manned parking, and modeling the manned parking; removing parking points of traffic lights and traffic jams in no-load parking according to the model and reserving interesting parking points; constructing a feature vector of the parking point of interest and performing track matching, wherein:
except for parking of passengers and passengers, the parking of passengers and the parking of no-load are divided, the type of parking spots is distinguished, and the method is a preparation work for classifying the parking spots of the taxi.
And removing abnormal points of manned parking, acquiring equal traffic lights and parking points of traffic jam in the manned parking points, and training an One-class SVM classifier by using the abnormal points to distinguish whether the parking points are the parking points of traffic jam and equal traffic lights.
According to the trained SVM classifier, parking points of traffic lights and traffic jams in no-load parking are removed, and interesting parking points are reserved, so that the interesting parking points reflecting the intention of a taxi driver are obtained.
And constructing a track characteristic vector, carrying out track matching, identifying the corresponding moving track of the vehicle, and verifying the matching accuracy.
The present embodiment adopts taxi GPS report data of shanghai (906 cars) and shenzhen (1945 cars) as an original movement trajectory data set. Each track data contains specific information: the number ID of the vehicle, the current position, the longitude and latitude of the current position, the timestamp, the driving speed, the driving angle and the motion state of the vehicle. The motion state can display whether the taxi carries people or not. And each road section and each intersection in the road network of Shanghai and Shenzhen have unique numbers in the respective road network. Before processing the trajectory data, the trajectory is preprocessed. Because there is an error in the positioning by the GPS, the position is corrected by the road network information in the electronic map, the trajectories are all restored to the roads, and the position of the vehicle relative to the map is determined. The GPS points of the track are mapped onto the corresponding road segments by using the ST-Matching map Matching algorithm. The movement track of a taxi throughout the day is recorded in each GPS log file in the data set. The complete track of the vehicle is divided into a plurality of sub-tracks by day. Data of 28 days and 31 days are recorded for Shanghai and Shenzhen respectively. Shanghai and Shenzhen use 20 and 22 days of continuous traces to form a set of 906 and 1945 anonymous traces, respectively, and the remaining 8 and 9 days to separately form a test trace fragment corresponding to the set of anonymous traces using the daily traces.
It should be noted that, since the parking spots of the taxi can be classified into three categories, i.e., boarding and disembarking parking, mid-way parking with passengers, and empty parking, in order to obtain useful information, the boarding and disembarking parking spots need to be filtered out first, and the boarding and disembarking parking spots can be easily removed according to the information contained in the data. Fig. 2 is a detailed classification of parking spots.
In the present embodiment, the parking caused by traffic jam and waiting for traffic lights is eliminatedThree characteristics of the parking time, the distance from the parking point to the intersection and the interval time between two parking points are used. Because the vehicle stops when walking due to traffic jam, the parking time and the interval time of two parking points are directly related, and the distance between the traffic light and the parking point to the intersection and the parking time are also directly related, so that the three attributes are used as the characteristic attributes of the parking points to form the characteristic vector, and the purpose of distinguishing the parking points can be achieved. Using dura to represent parking time, dist to represent the distance between a parking point and an intersection, and inter to represent the interval time between two parking points, and constructing a characteristic vector S of the parking pointtpDura, dist, inter). Fig. 3 and 4 show the statistical results of the parking spots of the vehicle, and some parking spots form a bar graph, which indicates that the vehicle often stops at the position for one month. Parking points which are located at the same position for several consecutive days are the features of the tracks, and some feature points like edges are also the features that some tracks are distinguished from other tracks.
In the present embodiment, abnormal parking points among manned parking points are filtered out using mahalanobis distance, and the threshold value of mahalanobis distance is set to
(p is a degree of freedom), and a parking point exceeding a threshold is regarded as an abnormal point. The mahalanobis distance is calculated as:
wherein StpiA feature vector representing the ith parking spot,
and v represents the covariance matrix of the characteristic vector S of the parking point set.
In the embodiment, the trained SVM classifier is used for identifying and filtering out traffic jams in manned parking spots and parking spots with traffic lights and the like, so that the required interesting parking spots are obtained. A One-class SVM classifier is trained by using manned parking points with abnormal parking points removed, and the classifier can judge whether the parking points are the parking points of traffic lights and the like. Because only interesting parking spots and non-interesting parking spots exist in the empty parking spots, after all traffic jams and the non-interesting parking spots of the traffic lights are removed, the reserved parking spots are the interesting parking spots. The results of classification using the SVM classifier are shown in fig. 5 and 6. The parts with darker colors are parking points caused by traffic jam and traffic lights, the parts with lighter colors are required interesting parking points, and track feature vectors formed by the interesting parking points are used for calculating matching accuracy.
In the present embodiment, the extracted parking points of interest are used to construct a trajectory feature vector. Constructing feature vectors uses a modified TF-IDF method in which the trajectories are analyzed in a manner that processes text. The taxi driving track is regarded as a text, the road section number is regarded as each word, and the number of the parking spots of the road section is the number of times of the word appearing in the text. Weight tf-idfi，jThe calculation formula of (2) is as follows:
tf-idfi，j＝tfi，j×idfi，j.
wherein the content of the first and second substances,
wherein N isiFor vehicles v in training dataiTotal number of road segments traversed; m isjIncluding road segments r for training datajThe number of tracks of (a); ciFor vehicles v in training dataiThe class to which the locus belongs; m isi，jIs CiIncluding road section rjThe number of tracks of (a); t is ti，jIs rjAt CiThe number of occurrences in (a); tf isi，jIs rjAt viThe frequency of occurrence in the trace of (a).
Obtaining the weight value tf-idfi，jThen, the following trajectory feature vector can be formed:
Wi＝(ωi，1，ωi，2，...，ωi，n).
wherein ω isi，jIs represented by rjOf weight value, i.e. ωi，j＝tf-idfi，j。
The parking points extracted from the training data constitute trajectory features, and the parking points in the test data constitute test trajectory features, which are matched by applying the cosine similarity values. The similarity between two vector inner product spaces is measured by measuring the cosine of the angle between them. It is determined from the cosine value of the angle between the two vectors whether the two vectors point in approximately the same direction. In the comparison process, the size of the vector is not considered, and only the pointing direction of the vector is considered. The cosine similarity between two vectors is expressed in terms of dot product, and the formula is as follows:
wherein A, B represent two different trajectory feature vectors, respectively.
Firstly, the cosine similarity degree similarity of the test track and each vehicle track in the anonymous track set is calculated, the maximum cosine similarity value of the test track and each vehicle track is respectively calculated, and finally, the maximum value is found and is regarded as belonging to the same vehicle as the test track.
In the present embodiment, when the feature vector of each test trajectory is obtained, the vehicle to which it belongs is already known. The final accuracy can be known only by comparing the anonymous track information obtained by comparison with the real track information. The effect degree of the method for extracting the features can be embodied by analyzing the accuracy. The specific accuracy formula is as follows:
wherein n iscorrcetRepresents the classification result after the test, nallRepresenting the true category.
The accuracy of the parking spot based de-anonymization method is shown in fig. 7. The accuracy rates are 48.58% and 39.66% respectively. Namely, 444 tracks can be distinguished by the method in the Shanghai 906 vehicle track, and 720 tracks can be distinguished by the method in the 1945 tracks of Shenzhen. Furthermore, de-anonymization accuracy will rise significantly as the length of the trace (i.e., number of days) obtained by the attacker increases (meaning more parking spot information of interest). As shown in FIG. 8 and FIG. 9, Shanghai data and Shenzhen data rise from 48.58% and 39.66% to 85.22% and 66.67%, respectively. Therefore, as long as an attacker obtains enough real data of the user, the original track of the victim can be accurately matched, and the privacy of the user can be stolen.
The invention can realize that the undeniated original track can be matched from the anonymous moving track in a centralized way by depending on the parking spot record, analyzes the interesting parking spot characteristics caused by personal preference of the driver, verifies the stability and uniqueness of the interesting parking spot characteristics of a single driver and reveals the security risk of anonymous track data.
Claims (3)
1. A method for removing anonymity of moving track data anonymized to a vehicle based on parking records is characterized in that a plurality of moving track fragments of a taxi in any time period are obtained, and tracks of the taxi are identified by comparing known tracks with tracks concentrated in anonymous tracks, and comprises the following steps:
(1) extracting all parking points in the track, removing parking caused by passengers getting on and off, and dividing parking into manned parking and idle parking, wherein the manned parking comprises parking caused by traffic lights and parking caused by traffic jam;
(2) removing abnormal points of manned parking, and modeling the manned parking; wherein, dura is used for representing the parking time, dist is used for representing the distance between a parking point and an intersection, inter is used for representing the interval time between two parking points, and a characteristic vector S of the parking point is constructedtpRemoving abnormal points of manned parking by using mahalanobis distance; the method for removing the abnormal points of manned parking by using the Mahalanobis distance specifically comprises the following steps: setting the Mahalanobis distance as
Wherein p is the degree of freedom, and the parking points exceeding the threshold are regarded as abnormal points; the mahalanobis distance is calculated as:
wherein StpiRepresenting the feature vector of the ith parking spot, and mu represents the mean value of the feature vectors of a group of parking spots; v represents a covariance matrix of the characteristic vector of the parking point set;
(3) removing traffic lights and traffic jam parking in the no-load parking point by using the obtained model, and reserving the interesting parking point;
(4) and constructing feature vectors based on TF-IDF for the interested parking points, comparing the feature vectors of the known tracks with the feature vectors of the anonymous tracks, and finding the anonymous tracks matched with the known tracks in the anonymous track set.
2. The method for de-anonymizing the vehicle-anonymous movement locus data based on the parking record according to claim 1, wherein the step (1) is specifically as follows: and (4) screening out points with positions not changing along with time from the track set, and judging whether the parking is carried out by passengers or not according to the manned states before and after the parking in the track set of the taxi.
3. The method for de-anonymizing the vehicle anonymous movement locus data based on the parking record according to claim 1, wherein in the step (3), the person-carrying parking point after the abnormal parking point is removed is used for training an One-class SVM classifier, so that the One-class SVM classifier can distinguish whether the parking point is a parking point of a traffic jam and a traffic light; and identifying and removing the parking points of traffic lights and the like in the no-load parking points by using the One-class SVM classifier, wherein the reserved parking points are the parking points of interest.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810385918.0A CN108734008B (en) | 2018-04-26 | 2018-04-26 | Method for removing anonymity of moving track data anonymized to vehicle based on parking record |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810385918.0A CN108734008B (en) | 2018-04-26 | 2018-04-26 | Method for removing anonymity of moving track data anonymized to vehicle based on parking record |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108734008A CN108734008A (en) | 2018-11-02 |
| CN108734008B true CN108734008B (en) | 2021-12-07 |
Family
ID=63939306
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810385918.0A Active CN108734008B (en) | 2018-04-26 | 2018-04-26 | Method for removing anonymity of moving track data anonymized to vehicle based on parking record |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108734008B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110766937B (en) * | 2019-05-22 | 2020-10-20 | 滴图(北京)科技有限公司 | Parking spot identification method and device, electronic equipment and readable storage medium |
| CN110428621B (en) * | 2019-07-30 | 2022-07-15 | 山东交通学院 | Track data-based monitoring and early warning method for dangerous driving behavior of floating car |
| CN110428604B (en) * | 2019-07-30 | 2022-04-22 | 山东交通学院 | Taxi illegal parking monitoring and early warning method based on track and map data |
| CN111224940B (en) * | 2019-11-15 | 2021-03-09 | 中国科学院信息工程研究所 | Anonymous service traffic correlation identification method and system nested in encrypted tunnel |
| DE102020204045A1 (en) | 2020-03-27 | 2021-09-30 | Volkswagen Aktiengesellschaft | Vehicle, method, device and computer program for a vehicle for determining a traffic density from at least one movement profile of a vehicle |
| CN113641887A (en) * | 2021-08-26 | 2021-11-12 | 河南工业大学 | Mobile track de-anonymization method and system based on semantic track mode |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105354857A (en) * | 2015-12-07 | 2016-02-24 | 北京航空航天大学 | Matching method for vehicle track shielded by overpass |
| CN105608919A (en) * | 2014-11-21 | 2016-05-25 | 杭州海康威视数字技术股份有限公司 | Method and device for determining station position |
| CN105810006A (en) * | 2016-03-29 | 2016-07-27 | 福建工程学院 | Method and system for recognizing roadside parking place |
| CN107945037A (en) * | 2017-11-27 | 2018-04-20 | 北京工商大学 | A kind of social networks based on node structure feature goes de-identification method |
-
2018
- 2018-04-26 CN CN201810385918.0A patent/CN108734008B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105608919A (en) * | 2014-11-21 | 2016-05-25 | 杭州海康威视数字技术股份有限公司 | Method and device for determining station position |
| CN105354857A (en) * | 2015-12-07 | 2016-02-24 | 北京航空航天大学 | Matching method for vehicle track shielded by overpass |
| CN105810006A (en) * | 2016-03-29 | 2016-07-27 | 福建工程学院 | Method and system for recognizing roadside parking place |
| CN107945037A (en) * | 2017-11-27 | 2018-04-20 | 北京工商大学 | A kind of social networks based on node structure feature goes de-identification method |
Non-Patent Citations (1)
| Title |
|---|
| 移动轨迹数据去匿名化攻击方法;钟建友等;《计算机工程》;20161215;第133-138页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108734008A (en) | 2018-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108734008B (en) | Method for removing anonymity of moving track data anonymized to vehicle based on parking record | |
| CN106846801B (en) | Method for detecting abnormal behavior of regional loitering based on vehicle track | |
| US9501690B2 (en) | Passive driver identification | |
| CN105654730B (en) | A kind of fake-licensed car identification for crossing vehicle big data analysis based on bayonet | |
| CN109214345B (en) | Method for searching driving track of card-changing vehicle based on similarity comparison | |
| CN109242024B (en) | Vehicle behavior similarity calculation method based on checkpoint data | |
| Nakashima et al. | Passenger counter based on random forest regressor using drive recorder and sensors in buses | |
| CN106875679B (en) | Method and device for identifying accompanying vehicle | |
| US20070282519A1 (en) | System and method for analyzing traffic disturbances reported by vehicles | |
| CN106297304A (en) | A kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data | |
| CN108665699B (en) | Method and device for predicting vehicle appearance place | |
| CN114999181B (en) | Highway vehicle speed abnormality identification method based on ETC system data | |
| CN115035491A (en) | Driving behavior road condition early warning method based on federal learning | |
| Wu et al. | Clustering of several typical behavioral characteristics of commercial vehicle drivers based on GPS data mining: Case study of highways in China | |
| CN114926824A (en) | Method for judging bad driving behavior | |
| CN114997777A (en) | Vehicle movement feature identification method based on track information | |
| CN114724122A (en) | Target tracking method and device, electronic equipment and storage medium | |
| CN114821490A (en) | Airport departure layer illegal vehicle automatic identification method based on deep learning | |
| JP2003256980A (en) | Device, system and program for supporting traffic safety | |
| CN111780620B (en) | Unmanned aerial vehicle potential threat determination method | |
| Li et al. | Driving performances assessment based on speed variation using dedicated route truck GPS data | |
| CN110097074B (en) | Vehicle track compression method based on sequence similarity | |
| Van Hinsbergh et al. | Vehicle point of interest detection using in-car data | |
| Padarthy et al. | Investigation on identifying road anomalies using in-vehicle sensors for cooperative applications and road asset management | |
| Kumar et al. | Effect of vehicle size on crash risk in a heterogeneous traffic scenario: a bivariate extreme value approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |