How would you go about computing the bucket?

You could take the centroid and then round it down to the nearest bucket interval increment, then divide by the interval. That should yield the index into the buckets.