E Engineering project proposal

1

Name: Class:

Final Project PROJECT TOPIC:

Object Tracking using Image Registration and Kalman Filter

PAPER REFERENCE:

Paper: Naidu, VPS. and Raol, J.R. “Object Tracking using Image Registration and Kalman Filter.” Multi Sensor Data Fusion Lab, Flight Mechanics and Control Division, National Aerospace Laboratories, Bangalore-17, India

OBJECTIVE:

1. Generate simulated data following steps listed below: a. Design a registration algorithm called “Sum Absolute Difference” for detecting

target location (M x N) from frame to frame. b. Using the target location, compute the center of the target mass. Center of the

target mass serves as a point for the synthetic data. c. Repeat steps ‘a’ and ‘b’ to calculate center of the target mass on each frame.

2. Employing the created data in step one, apply Kalman filter to the data and make predictions for the location of the mass in the next frame.

3. Use the real-world human subject data to test and analyze Kalman filter results

OVERALL APPROACH:

1. Human operator select a target of interest to be tracked from the reference frame 2. Image registration algorithm locates the selected target location in the current frame with

respect to the reference frame a. Sum Absolute Difference (SAD) b. Normalized Cross Correlation (NCC)

3. Target locations from the image registration algorithm are passed to a target state estimator (Kalman Filer)

4. The estimator continuously estimates the position of a moving target in a sequence of video frames.

2

DATA:

Human subject is moving at about constant speed, but the center of the mass of the subject changes as subject moves from frame to frame. Therefore, the motion is not linear. Three types of data sets will be analyzed for target tracking with the subject moving: 1) along the image plane, 2) orthogonal to the image plane, and 3) diagonal to the image plane.

For target moving along the plan:

Real-world human subject data collected @ 30 Hz

3

For target moving diagonal to the plan:

For target moving orthogonal to the plan: