Senior Project

1
Senior Project Electrical Engineering -
2006Motion Detection Using MATLABLaura DeMar,
Jin HanAdvisor Prof. Rudko

• Objectives
• Model a lizards visual perception of movements
  in natural scenes
• Use an elementary motion detector of the
  correlation type

In the correlation-type movement detector two
mirror subunits have a delay and multiplication.
The spatially filtered input is correlated at a
given distance with the imprint of the preceding
frames given by the temporal filtered output.

• Maximum Response Experimentation
• Calculated the size of the maximum response for
  various object sizes, speeds, and memory time
  constants.

• Input Simulation
• Movie with several frames, in which an object
  moves with constant velocity n0 from frame to
  frame.

• Results
• There is a shift in the desired peak of the
  maximum response for large object sizes with
  different tau. Thus, the peak of the maximum
  response does not occur when its velocity is
  equal to the correlation distance
• Large objects are not tuned for different memory
  time constants
• For a small correlation distance and a small tau,
  the maximum response is broader
• Tuning of the motion detector depends on tau
  because there were more maximum response peaks
  for a small tau
• Gain is inversely proportional to the time
  constant
• Program can be tested further for varying speeds
  between frames
• Results need to be correlated with lizard
  physiological data

Filtering

• Temporal Filter
• Lowpass filter
• Time constant tau determines memory of the filter
• The larger tau, the longer the memory of the
  filter
• Filtered signals mimic those that would appear in
  lizards retinal image

• Spatial Filter
• Symmetric Gaussian filter
• Enhances the edges of the object