Now, what is interesting is that it doesn't involve the actual picture of a ball landing on the court. Rather, it is more complicated than that. It takes a series of video frames of the trajectory of the ball and then calculates where it will land on the court. So it isn't a direct picture of where there ball lands, but rather, using mechanics, calculate where it would have landed. Of course, the factors involved in producing such calculation is a bit more complicated since a number of parameters must be considered.
Tests have always been conducted outdoors, encompassing situations that take the following factors into consideration:
* Wind (and therefore camera wobble);
* Bright sunlight at different times of the day;
* Shadows covering part or the majority of the court;
* Dark or overcast conditions;
* Artifical floodlights.
Still, considering how winds can be swirling in different directions, and how many of these tennis balls are hit with a variety of spins, one would tend to wonder if some of the very, very close calls (ball landing 1/4 inch on the line, etc.) are valid. The problem with this is that there isn't any kind of documentation of the degree of accuracy of its call. Where are the "error bars"?
Still, I suppose it is better than the human judgment call, especially on the very close one. At the very least, it stopped the players from arguing with the umpire. Would have been interesting to see how this would have changed how John McEnroe behaved on the courts if it had been around back then. :)