20-04-2012, 03:10 PM
Easy Image Processing: Camera Interfacing for Robotics
[attachment=20455]
I’ve been interested in robotics since I was a little boy.
Back when I was in junior high school, I built a mobile
robot platform out of the drive portion of a child’s
motorized car and a Commodore VIC-20. Over the
years, advances in technology have made experimenting
with robotics more enjoyable. The Game Boy
Camera is an ingenious addition to the Game Boy
Color game unit that came out a couple of years ago.
It’s a black-and-white digital camera with a resolution
of 123 x 128 pixels, operating at a rate of one to 30
frames per second.
The camera’s original price was between $40 and
$50, making it somewhat cost-prohibitive for hobbyists.
However, because the product was recently discontinued,
I found some on eBay selling for between
$10 and $20. The reduced price makes the camera an
attractive solution if you’re interested in robot navigation.
It’s even less costly than a single Polaroid sonar
module ($30 to $50) and in the same ballpark as
reflective infrared sensors ($5 to $15).
The sensor inside the camera is a highly integrated
CMOS array with built-in edge enhancement and
extraction. Built-in image processing enables a microcontroller
to perform object detection and tracking,
provided certain assumptions about the background of
the image are valid.
Atmel’s AT90S8515 microcontroller has an external
memory interface bus that allows you to easily connect
an SRAM IC. The on-chip hardware UART makes it
possible to output processed data without consuming
precious processing resources, and the timers enable
it to control hobby servo motors without much work. In
addition, the AVR series of microcontrollers has a highspeed
RISC architecture (e.g., most instructions take
one or two clock cycles) that makes timing calculations
simple. In short, the flexibility of the AVR 8-bit
microcontrollers makes attaching special-purpose
peripherals like the Game Boy Camera a breeze.
Obstructed Views
Ian Horswill’s 1993 dissertation, “Specialization of
Perceptual Processes,” details some of his research
concerning image processing for robotics. Horswill outlines
various assumptions that may be made by a
robotic image-processing system for detecting obstacles
in a given environment.
Developing
The software for this project consists of two parts. The
first section consists of the assembly code in the AVR
microcontroller that talks to the camera, RAM, and serial
interface. The second part includes the C program for
a Linux-based PC that reads and writes camera registers.
In addition, this portion captures images and
obtains depth information, nearest object information,
or object-tracking locations.
I assembled the microcontroller code with the tavrasm
AVR assembler and programmed the microcontroller
with sp12. I wrote the C program for the host PC using
the Simple DirectMedia Layer (SDL) library, which is a
public cross-platform graphics and multimedia-programming
library.
[attachment=20455]
I’ve been interested in robotics since I was a little boy.
Back when I was in junior high school, I built a mobile
robot platform out of the drive portion of a child’s
motorized car and a Commodore VIC-20. Over the
years, advances in technology have made experimenting
with robotics more enjoyable. The Game Boy
Camera is an ingenious addition to the Game Boy
Color game unit that came out a couple of years ago.
It’s a black-and-white digital camera with a resolution
of 123 x 128 pixels, operating at a rate of one to 30
frames per second.
The camera’s original price was between $40 and
$50, making it somewhat cost-prohibitive for hobbyists.
However, because the product was recently discontinued,
I found some on eBay selling for between
$10 and $20. The reduced price makes the camera an
attractive solution if you’re interested in robot navigation.
It’s even less costly than a single Polaroid sonar
module ($30 to $50) and in the same ballpark as
reflective infrared sensors ($5 to $15).
The sensor inside the camera is a highly integrated
CMOS array with built-in edge enhancement and
extraction. Built-in image processing enables a microcontroller
to perform object detection and tracking,
provided certain assumptions about the background of
the image are valid.
Atmel’s AT90S8515 microcontroller has an external
memory interface bus that allows you to easily connect
an SRAM IC. The on-chip hardware UART makes it
possible to output processed data without consuming
precious processing resources, and the timers enable
it to control hobby servo motors without much work. In
addition, the AVR series of microcontrollers has a highspeed
RISC architecture (e.g., most instructions take
one or two clock cycles) that makes timing calculations
simple. In short, the flexibility of the AVR 8-bit
microcontrollers makes attaching special-purpose
peripherals like the Game Boy Camera a breeze.
Obstructed Views
Ian Horswill’s 1993 dissertation, “Specialization of
Perceptual Processes,” details some of his research
concerning image processing for robotics. Horswill outlines
various assumptions that may be made by a
robotic image-processing system for detecting obstacles
in a given environment.
Developing
The software for this project consists of two parts. The
first section consists of the assembly code in the AVR
microcontroller that talks to the camera, RAM, and serial
interface. The second part includes the C program for
a Linux-based PC that reads and writes camera registers.
In addition, this portion captures images and
obtains depth information, nearest object information,
or object-tracking locations.
I assembled the microcontroller code with the tavrasm
AVR assembler and programmed the microcontroller
with sp12. I wrote the C program for the host PC using
the Simple DirectMedia Layer (SDL) library, which is a
public cross-platform graphics and multimedia-programming
library.