Hi Lars, 

Great prompt. 

One case study is Hwy 3 in the East Kootenay region of BC. in 2016 they set up a video alert system on a few high-collision sections of the highway. I can't confirm the exact method of detection but suspect based on hardware and the instal that they are just using motion and basic filtering to trigger an alert drivers can see. By anecdote the local drivers believe it falses positive very often and as a result traffic takes no notice. 

I suspect that to have a flashing alert impact driving behaviour it needs to be very accurate. But I also wonder if traffic will care at all, even if it is accurate. We know that AI video can detect animals, even in busy backgrounds with training. I can see a problem where cameras accurately detect wildlife, but if the wildlife is not visible to traffic, drivers learn to ignore the warning signs. Do you know of any studies of efficacy on driver behaviour where these systems have been tested?

There is also the enforcement aspect. Speed limits work because their are consequences to speeding. But these warning signs are rarely connected to locally reduced speeds - kinda like the warning signs on corner speed on some roads. Maybe drivers just perceive this as a guideline rather than a rule. 

Another thing is the practicality of deployments. As the physical camera hardware, assuming IR for nightime, means any deployment is going to be over $10,000 and transportation authority would be the responsible party, this pushes purchases into public tenders - at least in richer countries. Getting through public tenders is a miserable experience and adds huge costs. The requirements in these tenders of course will also vary and favour specific solutions or suppliers. All this creates a product ecosystem that relies on commercial offers with large spending on marketing and sales.

For the pro side:

Strangely enough wildlife crossings have in some jurisdiction become political such as in Island Park Idaho. Perhaps cameras are a solution to this.