ECE 8893 Project Three: Pedestrian Traffic Flow 1.3 Spring 2008
1
Summary: This project explores pedestrian traffic surveillance and analysis. The task is to
identify people who are moving in a scene (e.g., walking, running, or riding bikes along a
sidewalk, indoor corridor, crosswalk, etc.), track them, and compute statistics of pedestrian
traffic flow.
P3-1 Collect Input Sequences: Use your web-cam to collect an input image sequence of
pedestrian traffic. This sequence and those collected by the other teams in the class will be used
as input sequences to test the program you write in part P3-2. The image sequence should satisfy
the following specification:
1. The pedestrian flow must be primarily bi-directional and orthogonal to the camera’s
shooting direction (i.e., people move left/right across the camera’s field of view).
2. The pedestrian traffic must be within a walkway/corridor that is 3-4 meters wide.
3. The camera must be located approximately 10 meters from the center of the pedestrian
flow corridor, at a height of approximately 1.5 meters.
4. The pedestrian flow rate must be approximately 15-30 people/minute, crossing a vertical
tripwire bisecting the image.
5. The images must be collected (using the Collect program described below) at a rate of 2-
5 frames/second. The sequence should contain a total of 500-750 frames.
6. The image sequence must be temporally contiguous; do not edit or splice image
sequences.
Collect program:
In order to collect images on your computer, you can download and extract files from
Collect.zip. Then make a directory on your C or D drive named ‘C:\Images\Seqs\’. Load
‘Collect’py’ into Idle and run it. At the prompt, type:
>>> Collect(750, 0.0)
Your program should begin collecting 750 image files in the sequence directory as fast as your
camera and hard disk allows. You can adjust the second argument to add additional delay (in
seconds between captured images. You can adjust the first argument to collect more or less
frames.
This program uses a neat Python package called VideoCapture by Markus Gritsch (from
http://videocapture.sourceforge.net/).
Reminder about proper conduct:
In all activities associated with this class, particularly in gathering videos for test input
data, do nothing to harm yourself or others (including risking personal injury, causing
embarrassment, or invading privacy).
For this part, you must create a directory that contains your collected image files plus an info file
that contains important data on the image sequence. To create this file, modify the
readme.txt in the provided example sequence 20070403-09.zip shown below.
