Advanced Topics in Computer Vision for Video Surveillance

Instructor: Tony X. Han

Email: hantx-AT-missouri.edu

Phone: (573)-882-6630

Meeting Time: Tuesday 3:30PM - 6:00PM

Meeting Place: MIDDLEBUSH HALL 10

Office Hour: 30min after class or by appointment

Course Webpage: http://web.missouri.edu/~hantx/ECE8001/

Content:

This course will emphasize the advanced topics for video surveillance.

(Note: This is a tentative schedule. The topics or the time allocated to each topic might be modified based on interests of the class and the research progress.)

Week	Lecture Topics
1: 01/20	Introduction, Motivation, Overview, Software, tools.
2: 01/29, 01/31	Background modeling, color space,
3: 02/05, 02/07	Optical flow, Motion segmentation, Lucas-Kanade Algorithm
4: 02/12, 02/14	Template Matching, Distance Transform,
5: 02/19, 02/21	Meanshift algorithm, gradient descent method for tracking, Mini project announced
6: 02/26, 02/28	Dynamic models 1: Linear Model, Kalman filter
7: 03/01, 03/06	Dynamic models 2: Condensation, Sequential Monte Carlo simulation
8: 03/11, 03/13	Object detection 1: features: SIFT, HOG, Kadir-Brady interest point detector.
9: 03/18, 03/20	HOG, Maximal Stable Extremal Region, Appearance updating
Spring Break	Mini project and progress report due after spring break.
11: 04/01, 04/03	Object detection 2: unsupervised learning: K-means, EM, Affinity propagation.
12: 04/08, 04/10	Object detection 3:supervised learning: SVM, Ada-boosting, Decision tree
13: 04/15, 04/17	Modern Segmentation: Graph Cut, Normalized Cut.
14: 04/22, 04/24	HMM, Markov Random Field, Belief propagation, Conditional Random Field.
15: 04/29, 05/01	Guest Lecture, Optional Topics
16: 05/06, 05/07	Final project presentation by each group and final reports due.

Announcement:

· Mini-project: background modeling using (GMM, eigen-background, wall-flower, W4, or…)
Download the data here. The first 18 frames can be used for background learning.

· Please prepare your paper presentation according to the presentation schedule.

Prerequisites:

· ECE 7850 (same as CS 4650/7650): Image Processing; or consent from the instructor.

Optional Texts, Papers and Resource:

· Proceedings of ICCV, CVPR, ECCV from year 1998 to now.

· CMU Computer Vision Resource

· USC Annotated IP and Vision Bibliography

· IEEE Transactions on Pattern Analysis and Machine Intelligence

· IEEE Transactions on Circuits and Systems for Video Technology

General Notices:

· There is no paper based exam. However, there will be one mini-projects and one final project.

· Discussion of assignment among classmates is encouraged. However, if you turn in an obviously copied assignment, you will receive a final grade of F for the course

· The students can do the final projects in a team of 2~3 people.

Grading:

· Final project: 70% (Based on Final presentations and reports)

· Mini-project: 15% (Based reports, program and experimental results submitted)

· Paper Presentation: 10% (Each of you will present one paper in class relevant to the topics. The paper will be picked up by you from a list of papers on video surveillance topics)

· Attendance: 5% (This is a discussion oriented graduate level course, so I do require you to attend the class)

Acknowledgment (by the order of last name):

· Thanks to Professor Michael J. Black at CS dept of Brown to allow me to use many of his computer vision lecture slides.

· Thanks to Professor William T. Freeman at EECS dept of MIT to allow me to use many of his computer vision lecture slides.

· Thanks to Professor Andrew W. Moore at CS dept of CMU to allow me to use many of his statistical data mining tutorial slides.