Image Processing

1
Guroo Vision for a Humanoid Robot
Student Daniel J. Stonier Supervisor Gordon
Wyeth
Developing a vision system to assist in position
and pose estimation for the Guroo a humanoid
robot.
Image Processing Analysis is carried out in YUV
(luminance or brightness, red chrominance, blue
chrominance) colour space due to the
insensitivity of UV to brightness. Colour
Segmentation UV tables or pie slicing. Morphologi
cal Erosion Reduces noise effects. Run-Length
Encoding Identifies coloured 'runs'. Blobbing
Groups connected 'runs'. Blob Analysis Identify
objects from the blob list.
Software for the SH4 - VBSH4 The SH4 is
responsible for all aspects of image processing.
This ensures data transfer from the vision system
to the rest of the Guroo's integrated systems
causes little network delay.
Guroo Hardware CCD Camera Configurable image
captures. SH4 Vision Board Image Processing. PC
Runs the vision debugging software.

• Software for the PC - VDebug
• Platform Windows/Linux using OpenGL.
• Retrieves and displays images from the SH4.
• Performs image processing.
• Displays and processes bitmaps.
• The ability to render and process bitmaps allows
  image processing to be tested and polished before
  final placement on the SH4. New algorithms can be
  tested using VDebug and then quickly converted
  over to the vision board.

Programming for the SH4 Host Platform Windows
(Mingw)/ Linux. Target Platform SH4
processor. Compiler GNU SH4 cross-compilers. Memo
ry 512k SRAM.