Implementation of Vision Systems
for Mobile Robots


WORK IN PROGRESS


 

Vision systems for mobile robots involve image processing and interpretation, in order to reconstruct the world seen from the camera or for object recognition.
Image processing require a representation of the image in RAM memory, so it requires digitizing the analog NTSC or PAL signal coming from the camera and transfering the image to the RAM.

To this end a set of hardware and software components are required:

  1. Video Camera
  2. Frame grabber (digitization device)
  3. Device Drivers (OS driver for FG management)
  4. Image acquisition library (high-level functions for image acquisition)
  5. Image processing tools

 

I had some experience with both Matrox Meteor and BT848-based frame grabbers. I found both of them suitable for image processing applications. Moreover, some BT848-based FG is very cheap.

As for the OS, I am very happy working with my Linux system. It is not difficult, and very interesting, to write a device driver for Linux.

There at least two different BT848 drivers for Linux, but I consider bt848 driver to be more appropriate for image processing.

Image acquisition libraries are usually included in the driver package.

Many image processing tools are available (see for example Khoros).

 


Luca Iocchi. July 24, 1998.