You can actually see cellular automata as various ways to generalize the automata studied by automata theory. While automata studied in automata theory typically have only one reading (and possibly writing) head, have access to a single cell of memory storage at a time (per storage), cellular automata may use multiple reading/writing heads and read multiple cells of the memory storage per operation. Also, unlike other automata, cellular automata may output more than one symbol per operation.
But traditional texbooks on automata theory don't treat cellular automata as a member of the automata family. Cellular automata didn't see much practical use and, maybe because of that, it doesn't have a kind of established hierarchy and a host of theorems covering each aspect as does the automata of automata theory.
There's also a "personal" aspect of this relation. Stephen Wolfram, the person who invested a lot of effort into popularizing the subject has done so with a lot of controversy. He might have been overly "enthusiastic" about his research, while the rest of scientific community responded with patronizing and diminishing commentary (something you might be familiar with from visiting popular social Q&A sites). And so much so that today, for "serious" figures in the said scientific community embracing cellular automata into the family of automata would be too embarrassing (and nobody likes that, yikes!)