A more general version asks for the smallest ball that contains points of at least p colors, where p < k.
Consider both optimal and approximately optimal algorithms.
This differs from the art gallery theorem in that we want the optimal answer for a specific polygon. What can you prove about the quality of your algorithm's solution?
For some recent work and motivation, you can read these papers: "On Finding a Guard that Sees Most and a Shop that Sells Most" and "Adaptive Beacon Placement".