a b s

Testing was done at every step of developing our system, from basic camera communication to the final implementation of the vision network. The majority of this initial testing dealt with verifying the color based particle filters, therefore it was tested for slight occlusion, rotation, scale change and illumination changes. Once it was proven to be effective under those conditions we tested its robustness, by tracking targets making erratic and swift motions. Because the tracking method relies on causal data if the target were ever lost it would not be able to recover. As the image below shows it was able to track the target throughout the entire frame and because the final location of its closest match is still in the targets head we can rest assured that our target was never lost.

Figure 1: Target Tracking With Path Shown As A Gradient

All of the testing on fixed frames was extended to 350 degree pan, 90 degree tilt, and 40x zoom cameras for better coverage of our testing space. In order to measure the proficiency of tracking, a live graph like the one shown below was used for demonstrating how well the camera was able to keep the target centered in frame. Due to the long period to process successive commands in the camera, the tracking will focus the target to the center, but it will have a slight under-damped effect.

Figure 2: Target Displacement from Center of Frame