Untitled Document

Moving People

Title: Moving people

Name: Theparit Peerasathien

Class: Computer Vision (4005-757)

Professor: Roger S. Gaborski

website: http://hicomvision.tripod.com/

Introduction
The motion analysis system usually consists of a video camera and a PC with a certain software to acquire people locations for each image. Program uses automatic tracking algorithms. The task of these algorithms is automatic recognition of certain people within the image, the determination of their locations, and the tracking of those objects along the complete sequence.

Algorithm
This application uses the technique of extraction the object location for each individual image frame. The methods can typically be separated into offline tracking (after record). The methods use the difference between each frames of a video sequence to detect the moving people. To detect moving pixels in the image, it models the background as an image without people in it. The program uses this image background for subtracting it with the video frames and result with specific threshold value. Then the program regions the final result with colored rectangles and output with the original video file

(source: http://www.cvg.rdg.ac.uk/projects/advisor/index.html)

Region Tracker Algorithm tracks the moving object all the time. This includes region splitting and merging using predictions from the previous frame.
Main features:

region splitting and merging using predictions
adjust bounding box from Active Shape Tracker results
identify static regions for background integration

(source: http://www.cvg.rdg.ac.uk/projects/advisor/index.html)

Objectives

Research motion tracking and image processing
Implement basic computer vision recognision

Result

The program will automatic detect the people in the video file with position
Bounding the moving people with rectangle.

Software
This application uses MATLAB for all program code and design GUI with MATLAB GUI.

Project Timeline
            Week1: Study about tracking algorithms and related materials
            Week2: Design program classes interfaces and methods
          Week3:: Writing Code
            Week4: Testing and writing final project

Update #3 April, 27 with Change in Update#2

I extract the object from each image by using the original image without any people substract with each frame in video. The result from substraction is the objects not appearing in the original image will be in black color.

The original image (Extracted from video file)