CS7322: Computer Vision II: Final Project Proposal


 

"Traffic Flow Monitor"

by James Mossman


 

The Idea:

Monitoring scenes remotely through the use of cameras has been common place for years. This project will use a camera to monitor the traffic flow on "TBD" Street here in Atlanta.   Four goals have been set for this project:  

1) count the number of vehicles that pass through a common plane of the movie images.

2) remove the moving vehicles from the movie images, resulting in a fairly static background-only image (i.e. movie of road w/ moving vehicles removed).

3) taking movie sequence created from Goal#2 and adding "X" symbols to represent the locations of the moving vehicles that were removed

4) attempt construction of a panorampic (SP?) background-only image of the road by taking images from a camera that is rotating  

Approach:

How do you plan to do this?  What methods?  

For goal #1:

Optical flow will be computed for each frame. Each frame will then be converted into an intensity image (i.e. xvel^2 + yvel^2). Edge detection will be performed on the resulting intensity image. Morphological operators will be applied to image to improve connectivity of motion "blobs". The number of "blobs" that pass through a given line will indicate the number of cars.

For goal #2:

Motion "blobs" identified during goal#1 indicate regions of motion that need to be replaced in order to achieve a static background image. Scanning each frame for "static" data to replace the motion "blob"s should allow construction of a fairly static background-only movie sequence.

For goal #3:

For each frame from goal#1, identify the centroid of each motion blob. Place an "X" symbol in each frame from goal#2 centered at the location calculated from the goal#1 centroid information.

For goal #4:

Technique 'TBD'.  

Proposed Time-line:

Week 1:
Obtain camcorder footage of Street. Digitize the footage to allow analysis of traffic flow.
Week 2:
Verify that required computational tools are available and running as expected (i.e. convolution, optical flow, RAW I/O, etc.). Perform preliminary analysis of data to indicate whether simple image differencing or optical flow analysis will allow traffic volume to be estimated (i.e. counting the number of cars).
Week 3:
Finish system so that the number of cars passing through the image plane can be counted. Write Progress Report
Week 4-6:
"TBD"