See Also: New page on Object Tracking 

A two-pass edge detection process is applied to the live video image (left). A horizontal pass reveals vertical edges (red) and a vertical pass reveals horizontal edges (red). This image pair pre-dates a targeting process that searched for high concentrations of edge transitions and then estimates the center (in the horizontal direction) of the object. The deviation of of object center from view center creates an error term which is used to steer the robot toward the object.

Below is a mock-up of the Trinity College Fire Fighting Course used by the Dallas robot club (DPRG) for an annual contest.

Using Delphi, an Object Pascal Visual Programming Tool together with Silicon Graphics' OpenGL, a  library containing a robust set of functions for performing 3D Rendering, I produced a real-time simulation of my robot.

The main purpose of the simulation is to provide a training tool for my visual maze navigation software. Views are rendered from the robot's point of view (2nd 3D graphics image) then will be analyzed by the robot's vision processing software.

I have not completed that interface which requires rendering images to a bitmap. This is not a complex task, but will require a little research.

Calculations to move the simulated robot are very straight forward, but I am lacking a near exact mimic of robot driving dynamics to be added in the future. Also, this graphical representation can be used to display the estimated position of the real robot by reading motor encoder counts (wheel odometry.) Visual information from the robot can help to make this estimate very accurate, i.e. correcting for errors in wheel position due to slippage on floor, floor irregularities, or errors in wheel size to name a few sources of error.

Simulated Robot Explores the Maze

Robot's point of view