Semantic Trajectories - PowerPoint Presentation

Slide1

Semantic TrajectoriesModeling and Analysis

By: Shahab HelmiSlide2

OutlineLarger datasets are becoming available from GPS, GSM, RFID, and other sensors.Interest in movement has shifted from

raw movement data analysis

to more

application-oriented ways

of analyzing segments of movement.

Hence,

semantically rich trajectories

has have been promoted.

We will review:

Constructing trajectories from movement tracks

Enriching trajectories with semantic information

Using data mining techniques to analyze semantic trajectoriesSlide3

Preliminaries & DefinitionsTrajectory: a

trajectory is

the path that a moving object follows

through space as a

function

of time. Thus, it can be captured as a time-stamped series of location points, denoted as {x1, y1, t1, x2, y2, t2, ..., xN, yN, tN} where xi, yi represent geographic coordinates of the moving object at time ti and N is the total number of elements in the series.Semantic Trajectory: a trajectory that has been enhanced with annotations and/or one or several complementary segmentations (e.g. segmented by “start” and “stop” points). Slide4

From Raw To Sound TrajectoriesRaw movement data: data points collected from tracking devices, such as GPS.We would like to turn imperfect raw movement data into a trajectory dataset that is

correct and manageable

from the viewpoint of the targeted application.

Steps:

Trajectory data cleaning

Trajectory map-matchingCompression of trajectory dataSlide5

Trajectory Data CleaningMost query

processing and indexing techniques are built upon this assumption

that

spatio

-temporal

positions of moving objects can be precisely provided.Real-life trajectory data is far from being reliable enough for applications:NoisesPoor GPS SignalBattery Outage…GPS errors can be divided into to main categories:Systematic errors: happens due to low number of available satellites and invalidates the GPS position. Can be solved by automatic filtering methods -> for example using the maximum speed of the object.Random errors: happens duo to external reasons. Can be reduced by smoothing methods which are based on statistical analysis -> Kernel based methods, regression based methods …We have to find noisy points (outliers) replace them with a better value. For example, we could use the Maximum speed of an object to find outliers.Slide6

Trajectory Map-MatchingFor objects that moves in a network, such as road network.There are two reconstruction levels:

Replacing the point with a point inside the network:

Geometric

map-matching: point-to point, point-to-curve, curve-to-curve

Topological map matching: adjacency and connectivity of the graph are important.

Probabilistic map-matching: consider speed, direction, heading …HybridTransforming the raw trajectory into a semantic map-matched trajectory, for example a. sequence of road segments.Slide7

Compression TechniquesTrajectory data in applications grow progressively and intensively as the tracking time

goes

by.

Such enormous

amounts of

data can sooner or later lead to storage, transmission, computation, and display challenges.Objectives of compression algorithms:Reducing the size of dataset.Reducing the computation complexity.Supporting low devastation (reduced and original trajectories are not too different)Algorithms can be categorized to 3 groups:Top-down algorithms recursively split trajectory and keep the key points (DP).Bottom-up algorithms in each step add a point that has the lowest cost.Windowing algorithms (online reduction).Slide8

Semantic Enrichment of TrajectoriesThere are three general steps:

Trajectory segmentation into Episodes

Episodes annotation

Trajectory annotation Slide9

Trajectory Segmentation Into Episodes

Trajectory segmentation is driven by application-dependent criteria

.

The most common one: “

stop”

s and “move”s periods.The challenge is to find the stop points:No movement at all during some length of time -> probably stopping at a place of interest (POI) such as a “Restaurant”, the “Eifel tower”, a “Museum” and etc.A five minute gap that car is not moving or low speed (Krumm and Horvitz [2006]).Moving with a almost constant speed and direction for a fishing boat.…Slide10

Episodes annotationAfter finding

stop

points, we could annotate episodes with

activities

or

POIs:At home, at work, …In bus, driving, walkingShopping, walking, ……Slide11

Trajectory annotation Synthesizing all the information in the trajectory into a singe label that characterizes the whole trajectory. For example, considering annotations of episodes we could come up with the “

Tourist

” label for the trajectory.Slide12

Extracting Behavioral Knowledge From TrajectoriesKnowledge discovery from trajectories aims at identifying

behaviors

, either

among individual

trajectories or groups of trajectories

.Spatial and Spatiotemporal PatternsGranularity of Trajectory PatternsGlobal Vs. Partial PatternsIndividual Vs. Group PatternsConstrained Trajectory PatternsCommon techniques:Clustering trajectories sharing similar characteristics such as shape, speed, direction…Classifying trajectories in predefined classes.Discovering common sequences of movements (from A to B to C)Identifying objects that their movements are related to each other (leadership, flock, …)Slide13

Semantic-Based Knowledge DiscoverySemantics-based behavior discovery techniques can be divided in two main groups:

approaches

searching for common behaviors that are previously

unknown

.

approaches looking for specific behaviors.Slide14

Discovering Unknown BehaviorsMeet: A group of trajectories end at the same region C.

If we

use semantic trajectories with stops annotated with

POIs

(

school A, school B, . . . , cinema Lux). we can discover the frequent semantic behavior “going from school to cinema on Wednesday afternoon”, which corresponds to the fact that this cinema offers special price tickets for students on Wednesday afternoons.The Semantic Trajectory Data Mining Query Language (ST-DMQL) [Bogorny et al. 2009] allows users to specify semantic enrichment of trajectories with contextual domain information,The language is implemented in Weka-STPM [Bogorny et al. 2011], an extended version of the Weka data

mining toolkit

and the first toolkit for multilevel mining of semantic trajectories. It is a

free and

open-source

tool that also

provides spatial

visualization of the semantic

trajectories and

behaviors.

Another tool that analyzes semantic trajectories to infer behavior is

M-ATLAS

[

Giannotti

et al. 2011]. This system provides support for both raw and

semantic trajectories

, and is organized in a plug-and-play architecture that allows the

easy integration

of different mining algorithms, from clustering to classification techniques.

Spatio

-temporal vs.

semantic behaviorSlide15

Discovering Unknown BehaviorsChasing behavior:

provide an algorithm that evaluates if an

individual

(a person or

animal) called

the stalker intentionally follows another individual called the target. The stalker must follow the target for a certain time period, and during this period, the movement of the two individuals must remain with similar speed and direction. Moreover, the target must always be in front of the stalker. Siqueira and Bogorny [2011].Avoidance behavior: present an algorithm for identifying the trajectories that avoids a static object. For example, when analyzing human trajectories an avoidance of street cameras may reveal a suspect behavior; A trajectory shows an Avoidance behavior when it moves towards a target geographic object, turns around this object without intersecting it, and after avoiding the target object, the trajectory returns to its original path. Alvares et al. [2011

].Slide16

ReferencesSemantic Trajectories Modeling and Analysis, ACM Computing Surveys, Vol. 45, No. 4, Article 42, Publication date: August 2013

.

Computing with

Spatial Trajectories,

Yu

Zheng, Xiaofang Zhou, SpringerMobility Data Management and Exploration, Nikos Pelekis, Yannis Theodoridis, Spronger