Yep. That is why I said that having one module per space that integrates a sensor and an indicator "has some obvious attractions to it". I'm merely pointing out that there are lots of options that can/should be considered.I had in mind a single module that contains the electronics (CPU, radio, etc), lamp, power supply (from mains), and sensor (probably cheap CCD video), and lots and lots of software. One module could be bolted to the ceiling of the parking garage for each space, or mounted on poles sans ceiling. All the modules would mesh and report their status to a master PC somewhere on the internet. I bet I could manufacture something like this for somewhere between $20 and $50 ea. in high volume.
For the TS's purposes, it probably doesn't matter too much how low cost something like this could be manufactured in high volume unless there is someone out there actually manufacturing it in high volume (and, who knows, there might be). If they end up actually having to design and build the system (or have someone do so for them), then a very different set of tradeoffs come into play -- in particular the NRE on that "lots and lots of software". Of course, even here putting everything into a self-contained module has potential NRE advantages over separate sensor/indicator solutions.
One option, instead of image analysis (though there are an increasing number of open-source packages for that available) is to use a radar module. It shouldn't take as much processing to decide if the space is available and it should be able to manage interference from other units pretty easily given the time frames involved.