Trajectory Matching Algorithm Based on Clustering and GPR in Video Retrieval - Intelligent Computing for Sustainable Energy and Environment - page 374

R. Hu et al and L. Le Thi introduce retrieval methods using free-hand sketched
queries as target ones [8] [9]. C. Williams et al adopt relevance feedback strategies
to build a more user-centered retrieval system [10].
In the following parts, these methods are adopted: main segments abstraction,
reformed k-means cluster and Gauss Process Regression model. Advantages of
this processing framework are: (a) By extracting main features of trajectories,
nonsignificant features like scale or length take no opposite effect on similar
trajectories clustering and have less requirement on time synchronization. (b)
Gaussian Process Regression model has no request on the integrity of an motion
trajectory. In other words, even the similar trajectories of a partial one can be
found out.
The rest of this paper is organized as follows: In section 2, algorithm frame-
work is described in details and each part of this method is introduced succes-
sively. Simulation results are presented and explained in Section 3. Then the
conclusion is shown in Section 4.

2 Framework for Trajectory Matching

The whole algorithm can be divided into three parts: trajectory disposing, tra-
jectories clustering and particular trajectory matching.

2.1 Trajectory Disposing

Sampling. Moving objects are essential temporal-spatial features of videos.
This paper starts from sampling points from available trajectories and these
points contain features like position, velocity and acceleration. Sample points as
below

Point ( x,y,t,v x ,v y ,a x ,a y ) (1)

( x,y )isthe2 D position of the point. t is the instantaneous time when moving
object goes through ( x,y ). v x and v y , a x and a y are corresponding instantaneous
velocities and accelerations in each direction respectively.
Interval distance between Point i +1 and Point i is constant denoted by l .And
then, trajectories are divided into segments as below

Segment ( PointS,PointE,θ,a ) (2)

PointS and PointE are the starting point and ending point. θ is the angle that
expresses the segment's motion direction. a expresses the average velocity change
of moving object from PointS to PointE .