NTU/Intel M2M Project: Wireless Sensor Networks Content Analysis

advertisement
NTU/Intel M2M Project: Wireless Sensor Networks
Content Analysis and Management Special Interest Group
Data Analysis Team
Monthly Report
1. Team Organization
Principal Investigator: Shou-De Lin
Co-Principal Investigator: Mi-Yen Yeh
Team Members: Chih-Hung Hsieh (postdoc), Yi-Chen Lo (PhD student), Perng-Hwa Kung
(Graduate student), Ruei-Bin Wang (Graduate student), Yu-Chen Lu (Undergraduate student),
Kuan-Ting Chou (Undergraduate student), Chin-en Wang (Graduate student)
2. Discussion with Champions
a. Number of meetings with champion in current month: 1
b. Major comments/conclusion from the discussion: working on paper submission
3. Progress between last month and this month
a. Topic1: Video Summarization using MSWave.
-
Apply our previous work Mswave to video summarization, and submit IEEE
International Conference on Image Processing (ICIP) 2014.
b. Topic2: Distributed Nearest Neighbor Search of Time Series Using Dynamic Time
Warping
-
Finish the paper submission.
c. Topic 3: Intelligent Transportation System (ITS) Machine Learning: Predict whether
driver will stop at intersection or not without using video data.
2) Building Datasets of Driving Events with Minimum Human Effort.
-
Motivation: We treat the collected trajectories as multiple time-series data
and try to design a computer-aiding way to efficiently build datasets of
multiple driving events with minimum human interactions. The resulted
datasets will be provided for data analysis of further ITS applications.
-
Challenge: The most important issues to extract the sub-sequences as
samples of interesting driving behaviors with computer aiding are the
followings: 1) the events will occur as fragments starting from any positions of
the whole trajectory; 2) the same events derived by even the same drivers or
not often vary in length.
-
Method:
i.
A modified Dynamic Time Warping (DTW) based method: To cope with
the problem of non-deterministic correspondence, matching or
extracting algorithms without needing the “perfect-matching” are
useful.
ii.
To describe the corresponding motions of different driving behaviors,
we developed a new version of description which incorporating
speed/orientation measurements jointly and describing the motions in
vertical and horizontal way, respectively. The modified DTW-based
method will be applied to both the vertical and horizontal time-series
data.
-
Progress: Currently, we have finished the implementation of algorithms. And
there are 2,996 resulted candidates for left-turn dataset being generated. The
informations for each candidate including: corresponding driver-ID,
trajectory-ID, distances from the left-turn query sequence, the start-time and
end-time for this candidate.
3) To-do list.
-
Now, we are implementing user-interface to efficiently evaluate the precision
the resulted dataset.
-
Further applications for this datasets will be carefully formulated. Not only to
build a prediction model of multiple driving behaviors for real-time driving
safety, but also off-line and long-term driving profile analysis will be
concerned. For example, we can use DTW to distinguish aggressive left-turn
from regular ones, such that we can evaluate the risk index of a driver’s
current status (by calculating the ratio of aggressive left-turn cases for one
driver) or evaluate the risk index of an intersection (by calculating the ratio of
aggressive left-turn cases happening on one intersection).
4. Brief plan for the next month
a. We will continuous paper survey and refine our proposed approaches.
b. To implement our proposed approaches and evaluate their performance.
5. Research Byproducts
a. Paper: N/A
b. Served on the Editorial Board of International Journals: N/A
c. Invited Lectures: N/A
d. Significant Honors / Awards: N/A
Download