I'm no expert on this, so please ignore this if its not relevant. I had this idea and then went hunting for it on the internet and found this (which is pretty close to what I had in mind)
Would this be relevant for KBO detection?
"This includes extending detection ranges to fainter magnitudes at the noise limit of the imagery and operating in dense cluttered star fields as encountered at low Galactic latitudes."
The basic idea as I understand it is to stack (add) images in all possible combinations based on expected movement of the targets and then look for them as round blobs. i.e. imagine if you already knew where and how the KBO was moving, you could then get the telescope to track it and do a long exposure of just the KBO (enhance its SNR?). Stars, which would be streaks would not add. I'm guessing they could also be suppressed by subtraction. i.e. move and stack 1-0,2-1,3-2,4-3 etc
I'd imagine you would need a lot of short exposures adding up to the same result as a long exposure but filtered on the expected movement of the target.
The paper mentions this is computationally intense, but they where talking about using a 2 Ghz Pentium 4. I'd imagine something a bit more powerful (GPGPU based?) could be throw at the problem today.