When we think of cars and computer vision, the obvious applications are all on the car. Lane sensing, obstacle sensing, and seeing other cars enables driver assistance, and eventually autonomous vehicles. But what if someone did the opposite, and enabled the road to see the cars? That’s exactly what Volkswagen and the municipal government in Carmel, Indiana, are doing.
“The future of transportation will require more data connections and sophisticated analysis than ever before,” said Johan de Nysschen, chief operating officer at Volkswagen Group of America. “We see our tool as an opportunity to provide cities with more usable data on mobility to help shape their future transportation needs, from pedestrians and vehicle drivers through public transit. We’re grateful to the City of Carmel for partnering in this shared research effort.”
I’ve known some people who work in street planning, and the current method for measuring street utilization is to put the little pneumatic tubes over the street that click a counter every time a vehicle passes over. One particularly innovative program started using modified game cameras (cameras that are usually hung on the trees near a trail to get pictures of wildlife). The Volkswagen program seems to catch much richer data that governments can do a lot more with.
The cameras automatically blur out people’s faces and their license plates, so nobody’s individual movements can be tracked using the system. Instead, they simply count how many parking spaces are being taken up, how many people drive in each lane, and how many pedestrians there are and where they walk. This all helps with routine planning needs, so that new building goes where it’s needed most.
For everything from small accidents to major disasters, the camera data can generate important information. By sending the right resources for what’s actually happening on the ground, lives can be saved.
“With enhanced software from Volkswagen, Carmel can more easily expand our view of traffic flow throughout the city,” said Jim Brainard, Mayor of the City of Carmel, Indiana. “As we plan for public transportation needs in our future, our ability to capture more sophisticated information will help us make better decisions.”
This better data will be available through a dashboard that the city can access to get both long-term data as well as instant snapshots.
The software originated from a factory project by Volkswagen. They used computer vision to analyze factory production to make it more efficient, as well as for quality checks. After using it for complex tasks in a complex environment, they realized that the software could perform other tasks in cities. They’re in talks with other cities to use the software for not only street and walking path analysis, but also for multi-modal transportation.
The Good & The Bad
There’s a lot of good that can come out of such a system.
The biggest thing is that scarce resources can be targeted for where they’ll have maximum effect on improving life in the places being analyzed. This can also help free up funds for more impactful projects, so everyone can come out a winner on this, at least in theory. It may also save lives by helping with not only incident response, but also other phases of emergency management like mitigation and planning.
To really get the full benefit, the data need to be collected from more points around a city or town to get more information and not focus too much on some small areas.
There are possible downsides to this, though.
For one, it’s all about where the data is being collected. If fine-grained data is being collected only in the nicer parts of a town, then the nicer parts of the town will be the squeaky wheel that gets the grease. The other parts of a town, where more racial minorities and “rednecks” live, could be further neglected and continue the cycle of poverty for people who are already stuck in that rut.
In other words, some serious analysis needs to go into making sure these analysis programs don’t amplify disparities rather than help solve them.
Anonymization Doesn’t Prevent Disparity
Another problem is that when used for public safety, these sorts of systems have created problems (some deadly) in the past. While data is going to be anonymized, that doesn’t alleviate all potential civil rights concerns. The experience with “shot spotter” microphones in some cities is particularly instructive, as they’ve led to the wrong person being pursued, and to the denial of civil rights to innocent people through false positives.
Because these gunshot detection microphones were primarily installed in the “bad neighborhoods,” they led to an increase in aggressive police responses to places with more people of color. When police already have problems with racial bias, this has been a recipe for disaster.
I don’t know whether these sorts of issues could crop up with Volkswagen’s cameras when feeding information into dispatchers, but great care must be taken to make sure they solve rather than amplify problems or transform them into other problems.
Privacy Is Still An Issue
Privacy concerns are also not always solved by anonymization. As researchers have found, it’s almost always possible to dig into the data and identify individuals again.
Here’s a potential scenario: You stayed up late working from home and decide to take a 2:00 AM stroll in the night air to de-stress before bed. You happen to be the only person walking through a neighborhood while a crime occurred, and then pass by a convenience store on the way back home. When police investigate the crime, they see that only one person walked through the neighborhood at that time, and that they passed by the convenience store. They then request footage from the store, which has your face.
You then go from being an innocent person on a nighttime stroll to a person of interest in some crime that you’ll need to lawyer up to defend yourself from being convicted for. This may sound outlandish, but cameras have led to false arrests in the past, and the mere anonymization of data doesn’t preclude this.
We need to be very cautious with these innovative technologies.