3D model based vehicle tracking

Hi,

I am doing a project on a submitted paper which requires the tracking of vehicles, the procedure is as follows

The video captured by a calibrated camera is transformed into an image sequence. For each image, |*motion detection is conducted.|* Regions of interest (ROIs) that contain moving vehicles are detected. Each ROI is handled independently. If a new moving target is found, the tracking procedure is initialized by recognizing its vehicle type and assigning an initial value to its pose. After the initialization, the tracking takes place using pose prediction and pose refinement. Pose prediction is the estimation of the position of the vehicle in the current frame This position is the initial value for pose refinement. Pose refinement involves a search for the optimal mapping between the projected 3D model and the 2D image data.

HOW IS THE MOTION DETECTION DONE?? HOW ARE THE ROIs OF MOVING VEHICLES DETECTED?? HOW CAN THIS ENTIRE THING BE DONE??? PLEASE HELP ME..

THANK YOU

[Andreas Goser]

From your text, I am not sure how much of the project is already done, what constraints exist, etc. I will create an answer based on the sub-question "how can this entire thing be done" without knowing constraints

Answer 1

[Andreas Goser]

In case you can use the Video and Image Processing Blockset, you can have a look at multiple implementations of tracking here:

http://www.mathworks.com/products/viprocessing/demos.html

Depending on your exact need of motion detection, you can also use the blockset.