الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 105 |  |  |
| | | from | 105 | | |
Abstract
There has been a tremendous growth in movement data due to availability
of devices that could be used to track movement of objects.
Tracking an object gives rise to a sequence of points in time and space,
called a trajectory. One of the main functions is evaluating similarity
between moving objects’ trajectories and it has gained much attention
in many application domains.
There exist similarity measures in the literature that propose evaluating
similarity between trajectories in the form of time stamped
values, de nes some meaning of similarity and propose algorithms for
computing it. The user is restricted to that meaning of similarity
while it should be application dependent and only determined by the
user. Therefore, there is a lack of genericness where there is a need for
a generic approach where users can de ne the meaning of similarity.
In this thesis, a new parametrized similarity operator, TWEDistance,
is proposed. This operator is based on one of the discrete similarity
measures, time warp edit distance, where the meaning of similarity is
generic and left for user to de ne and it is implemented in Secondo.
The similarity measures in the literature that are based on the
discrete form of a trajectory is also a ected by the sampling rate differences
as it is de ned over sequences of points. Therefore,this thesis
propose to deal with the nature of trajectory’s data as a continuous
function. Continuous based similarity is evaluated using interpolation,
regression and curve barcoding.
In the experimental evaluation, rst, the accuracy of the TWEDis tance operator is evaluated based on di erent meaning of similarity
and the results were as settled in the experiments settings with an
accuracy up to 100%. Second, a comparative study is made between
TWEDistance operator, regression, interpolation and curve barcoding
while considering sampling rate di erences. The results showed that
interpolation gives higher average accuracy up to 90%. Experiments
were made over both real and synthetic datasets.
2.