Wednesday, June 2

Institute for Practical Bicycle Video Detection Research

ENOUGH of all this Bike School stuff for the moment. Let's do something pretentious. Hence the post title. Any of you are welcome to purchase the rights to it from me if we agree to suitable financial terms.

In English, this post is about how to get traffic signals to change for you when you are on a bike, and not those namby pamby induction ones, but the big video cameras. This is a tough one. HERE, I talked about induction loop signals. Amusingly, Rantwick, HERE, tested that tired old theory that magnets will help and found they don't.

Today, I delve further into a more mysterious area of traffic signals, the Video Imaging Vehicle Detection System (VIVDS in traffic speak). If you search around on the internet, you'll find two classes of information about VIVDS and cyclists. The first, epitomized HERE (regrettably, since it is usually FULL of good stuff), HERE, HERE, and HERE (among dozens of others), basically give no useful information at all about what the traveling cyclist needs to know to trigger such a system. All they say is "these are great for cyclists." Well, it ain't necessarily so. What these articles don't tell you is that cameras only work well when properly installed and calibrated, and presuming that the windstorm that hit last night didn't swing the camera around 79 degrees (which I have seen happen). Fortunately, there is also a second kind of information, which completely ignores bicycles, but which DOES give solid information. As it turns out, those "these are great for bikes" articles didn't LOOK at the second articles. Which is why you are fortunate to be reading this. Consider yourself "cutting edge."

FUNDAMENTALS
If you aren't the type of person who simply is highly trusting of the traffic types that do stuff like create door zone bike lanes, or stop signs just for the heck of it, read on. The first two photos are from a Texas A&M study on VIVDS system effectiveness, and the third is from a Purdue study. You can read these for yourself HERE and HERE. In truth, I forgot to copy the link when I pinched the pictures, but you'll see them in various forms in the links, along with all kinds of trivia it didn't occur to me might apply to cyclists.

Anyway, a VIVDS system works by detecting pixel changes in a "zone" that is programmed into the system when the camera is installed. You may notice that the cameras are usually installed fairly high in the air. This is so they can get a good view of the area leading up to the intersection stop line without also capturing clutter such as the horizon. As you can see in the first photo, the camera is aimed at the stop line and the roadway leading up to the stop line.  The horizon is purposely out of view. IF you happen to favor riding shoulders in such a situation, you may also notice that the shoulder disappears out of the view of the camera. Thus, we learn our first lesson. WHEN YOU WANT TO TRIGGER A CAMERA SIGNAL TO CHANGE, MOVE INTO A TRAFFIC LANE WELL BEFORE COMING UPON THE STOP LINE. This way you will be detected. Few cameras will be set to detect things on the shoulder, sidewalk, or median. More often than not, they are puposely set to AVOID detecting such areas. These things create false signals. Don't be a false signal unless you like waiting forever.


As you may see in the second photo, the camera is not simply set to just read a big blob waiting patiently for the light. It is broken into zones, so the camera can detect traffic approaching the intersection as well as traffic waiting for the light. This is another reason you want to move into the traffic lane well in advance of the intersection. However, take a good look at the Stop-Line Detection zones and imagine the situation of a cyclist that stops where the nose of the gray car stopped. That cyclist has given up at least half his/her pixels to the zone boundary. What's more, if the vagaries of weather have tilted that camera up just a smidge, the cyclist will be outside the detection zone altogether. This leads us to the second lesson. WHEN STOPPING FOR A CAMERA, STOP A BIT BACK FROM THE STOP LINE SO THE CAMERA SEES YOU. This also brings up the third lesson. WHEN DECIDING WHERE TO STOP, LOOK AT WHERE THE CAMERA POINTS. I have seen cameras spun by the wind around North Texas. The zones will move the way the camera has shifted. You want to be IN that detection zone. If the camera is pointed towards Oklahoma, it won't see you unless you are on your way south FROM Okalahoma. God LOVE those OKIES, except for the ones that drive red pickups with OU stickers on them.
You may also encounter one of these cameras at night. Well, they do not work by magic and they don't have IR or some special night vision. They get triggered by what they see. One more reason you don't want to depend only on a reflector at night. I do not know definitely if it helps to point your light at the camera, but it seems to me that pointing a blinky headlight at a camera at night ought to catch its attention. I've tried it and the light triggered, but it might have changed anyway. Still, I often check to see how far I can lift my front wheel when I am stopped at these types of light in the dark. Which brings up the final rule, AT NIGHT, NINJA CAN WAIT FOREVER! You KNEW there was some reason you needed to ride with lights at night. Now you know at least one reason why that you never read in the bike books. For all you know, there MIGHT be a ninja in that middle lane.

PS: As a duly designated remote tester, I EXPECT that you all will journey to some video signal camera you've never been able to trigger and give all this high tech stuff a go. I hope to hear of success, but these can be tricky devils. Unlike the induction units, there are not cuts in the pavement to clue you in. AND, if you can get me in touch with Eli Mowbray, of the Santa Cruz Public Works Department, you get double points. I suspect that guy can fill in the gaps a casual internet search leaves behind.

11 comments:

Khal said...

Good stuff, Steve. Thanks. Your site definitely gets added to the Los Alamos List of Infamy.

RANTWICK said...

I haven't noticed any cams here, but I'm familiar with setting up zones on security cams at work designed to trigger (or not trigger) on motion... nice post. I continue to appreciate your engineering ways and useful links to "real" information.

John Romeo Alpha said...

There are so many cameras mounted at intersections In These Troubled Times I'm not sure which ones are VIVDS. Is there a key identifying feature of a VIVDS cam?

cycler said...

Of course the average Ninja just runs the light (cross traffic, what cross traffic?)

Most of the places I ride just have timed lights- they change all night and whether anyone is there or not. They have different timing at different times of day, and a few of them downtown actually blink after 2AM, but if I'm at work that late, I normally get myself a zipcar to get home...
If I see a camera controlled light, I'll do my best to set it off though!

cycler said...

Oh, and just out of curiosity, when you can't get a signal to change, I presume that there are no other cars going your direction to trigger the signal. In that case, is there generally cross traffic? If not, even stick in the mud law abiding me will run a persistent red. If a bicyclist runs a light in the forest.....

Steve A said...

JRA, I'm hoping I snookered CycleDog into doing a post about the various types of detectors. If not, maybe Rantwick will scoop us all. For now, the ones that change signals look a lot like your garden variety security camera and look down and toward you from the other side of the intersection as high as they can mount it. I'll take the shots when I get a chance. I'm in LCI training starting tomorrow. Our session on Saturday is 12 hours long!

Cycler. Darn, the notion that Ninja are mostly unaffected by details such as traffic signals never occurred to me. And, "out of curiosity," I select from a variety of alternatives, because I have no wish to die from old age due to a non functioning traffic signal. Some options are more dramatic than others, depending on the audience. The fundamental rule of all those approaches is I call the traffic department so they know they've got a bum signal.

If police are coming from the other way, I lie the bike down where I am, run over to the side of the road to trigger the pedestrian signal, SPRINT back to the bike, and then wave to the policeman with a "thumb's up." I may be an engineer, but that doesn't mean I can't be a "Drama King" when the situation warrants. That is fun even if I know the signal would trigger without the "dash for the button."

One option that is very cautious and safe around my parts is to make a free right turn, followed by a U turn, and then followed by another free right turn. Mostly, however, my long commute allows me to just make a free right, followed by an alternate route that loses me little time. Commuters DO have an advantage!

Steve A said...

PS: When I lay the bike down, it is ALWAYS with the drivetrain side facing up. Don't want to bung up any "C" stuff!

danc said...

To all interested readers:
Advanced traffic cameras make intersections safer for cyclists Boise, Idaho

Application of video detection is a little troubling. The reporter and traffic engineer state the cyclist will be detected only if in the bike lane. This not good if you are positioning to making a left turn, avoiding debris out side the bike lane. The video demonstrate the system working and shows a potential right hook at 1:17 min:sec caused by staying the bike lane when traveling straight thru.

Thanks for trying to be bicycle friendly Boise!

Steve A said...

Danc, kudos for a very interesting report! Be cautious about concluding that the cyclist will only be detected if in the bike lane. I watched the video twice and what it really said is that the cyclist would now be detected IN the bike lane.

In reality, if "Figure 5" above had a bike lane instead of being a wide outside lane, no cyclist in that bike lane would be detected. What I THINK Ada did was to add a zone so a cyclist over on the right would get picked up. That IS a good thing unless they turned down the sensitivity in the other zones, which is unlikely unless they want to irritate motorcyclists.

Any cyclist in a traffic lane would get picked up almost as easily as a motorcyclist if he/she rides through the zone and the sensitivity isn't set too low.

It was very interesting to hear them say that the zones were set to not trigger until enough traffic stacked up. That suggests you might wait a LONG time if you were by yourself, even if you were driving your Land Rover instead of your Cannondale.

I shall keep that link in my favorites as I continue to plumb the finer points of making lights change for me.

In Traffic 101 yesterday, I was surprised to find that my students didn't know how induction sensors work, and neither did the instructors. It gave everyone something relevant to hear while we waited for the light to change. Yes, my wheels were both precisely over the sensor wire.

danc said...

Steve A wrote "what it really said is that the cyclist would now be detected IN the bike lane."

Correct, however the devil is in the details, did the traffic engineer understand the need to detect cyclists in other lanes or only the bike lane? Hmmm.

Steve A wrote "Any cyclist in a traffic lane would get picked up almost as easily as a motorcyclist if he/she rides through the zone and the sensitivity isn't set too low."

Hopefully!

Steve A wrote "It was very interesting to hear them say that the zones were set to not trigger until enough traffic stacked up. That suggests you might wait a LONG time if you were by yourself, even if you were driving your Land Rover instead of your Cannondale."

In a general sense if a automatic devise do not detect a cyclist or another lone vehicle, drivers after a reasonable period, may consider the device defective, treat it the stop light as stop light and proceed with caution. Careful programming should cover those lulls, 2-4 AM?

Steve A wrote "In Traffic 101 yesterday, I was surprised to find that my students didn't know how induction sensors work, and neither did the instructors. It gave everyone something relevant to hear while we waited for the light to change. Yes, my wheels were both precisely over the sensor
wire."

Cyclistview has excellent introduction and overview of the induction loops: Traffic Signal Detection .

A 20" BMX rim (no axle or frame) on 2x4 wood block is sufficient to check sensitivity of the loop. No fretting over carbon frame or leaning the bike over. Bob Shanteau, PhD estimated 2 or 3 wraps of armature wire would fix carbon rims.

Steve A said...

I suspect that the report was misleading. I expect that the program is set to change the light quickly if a bunch of vehicles show up, or to take a lot longer if a single one is waiting.

There is one item I'd note in that Cyclistview set of charts and one item I'd like to know that isn't addressed. The one item I'd note is that the strategy for finding an invisible sensor won't work well if it is a type Q. Most of ours in Texas are type A or type Q. They seem to favor the type Q in turn lanes.

The one item I'd like to see addressed is the common practice when a sensor is put in to replace an existing one. I see quite a few where there is a type A loop surrounding another type A loop. I suspect the outside loop is the active one but really don't know. Sometimes the cuts are close and it might matter, but not always.

The IMPORTANT thing - if you pass through a light as being nonfunctional, you owe it to your fellow travellers to alert the authorities. ALWAYS. When I call, I sometimes mention my vehicle won't trip the light. Sometimes I tell a white lie and say my motorcycle won't do it. Actually, a bicycle IS a motorcycle that uses a very low power output organic engine, so it's just misleading. If I was in the middle lane and failed to trigger the camera, usually I don't want to bother to try to explain why I was there on my bike in the first place. I don't think I earn points in heaven from trying to educate dispatchers about cycling over the phone. I figure if the work order notes motorcycle failed to trip the signal, they'll check sensitivity and maybe turn it up.

Post a Comment

No Need for Non-Robot proof here!