Proposing a vision camera as a traffic awareness tool for aircraft strikes me a roughly as pointless as hydraulic power distribution around an office, driving electric generators at each desk. All that’s needed in Europe or anywhere else is to deploy this:
greg_mp wrote:
UAV uses camera for obstacle avoidance, it can be used, but the speed and range are very far from th eone used in GA. but I think with some good cameras (2 at least) and nice algos, it can work.
Vision AI is very fashion at the present time. However there are many potential issues with vision systems:
- they are limited by the resolution of the lens+sensor,
- you are trading off resolution if you want a larger field of view,
- the range is limited by what the system “sees” i.e. see the previous points,
- it does not work at night or in degraded meteorological conditions;
- besides, sensors are expensive,
- vision AI require a lot of processing power, i.e. cost and electrical power
As always, the sensor needs to be adapted to what you want to “see”. That’s why, in automotive for instance, several different sensors are used concurrently: mainly vision, radar, and lidar. None is perfect but a combination provides reliable results.
I agree that vision can work for obstacle avoidance in UAV. However IMHO in this particular case lidar, or even better radar, would achieve the same results with greater accuracy, increased efficiency and speed (less processing needed), longer battery life, and at a much lower cost.
In full disclosure: this is my daily business and is something we regularly pitch to our customers.
Silvaire wrote:
Proposing a vision camera as a traffic awareness tool
Not exactly a camera, but this is what we have been doing in VFR for over 100 years :)
Apologies for the pun :)
Edit: PS. Agree with you it’s completely pointless for GA traffic awareness, as it neither has the required range nor the 360° field of view. Some sort of radio-based system with ADSB and/or Flarm is much better.
For UAV and pure obstacle/terrain avoidance, it’s possibly another story. But then a radar is better, faster and cheaper than a camera system.
greg_mp wrote:
I think with some good cameras (2 at least) and nice algos, it can work.
Correct – this will happen. It’s going to take a few more years, but it will come. The tech already exists and is rapidly maturing, but not exactly available to the general public. I worked on a primitive vision system for non-civilian use in the 80s, and the algorithms were solid, but the compute power was 1000x too slow. If you do the math, we have plenty of power now. Multiple cameras and multiple CPUs, each focused on a small section of the sky, it will be better than the human eye.
Much of the technology we use today would have been considered impossible 30 years ago.
We are talking about collision avoidance, not traffic awarness. Nothing can beat a TCAS on that point, but just for collision avoidance, you just need to see around you in a 1nm, that can be seen with a good camera system, and even at night.
Not exactly a camera, but this is what we have been doing in VFR for over 100 years :)
Unsuccessfully…. 
The goal should be not to see other traffic unless it’s necessary – with traffic info it’s usually practical instead to change course slightly and avoid being that close, even when there’s a lot of traffic. You spot some of them by eye anyway but I no longer prioritize visual contact outside of the immediate airport traffic area in which everybody is maneuvering to or from the same spot, and thereby close.
ADS-B works, and it is interesting to me that after only three years of operating in terminal area environments where 99.9% of everybody has it and I can see them clearly on an iPad, this thread feels to me like I’m reading something from the Stone Age. How quickly we adjust and reset our expectations.
Really the best thing for me is not having to hassle with ATC voice communication to avoid other traffic in dense traffic areas, at least for the most part. If I’m approaching or departing and talking to ATC in Class D or whatever I almost always ‘see’ the traffic and am already avoiding it before they call it for me, if they call it for me. The exception is really close in when my head needs to be 90% out of the cockpit for other reasons, e.g. I’m on a base leg entry maneuvering to land and there’s another plane head on, doing the same for a parallel runway. In that circumstance you need to see-and-avoid and aren’t looking at a traffic display. Otherwise and especially en route cumbersome back-and-forth ATC voice communication doesn’t have a lot to offer, and see-and-avoid is no better than it ever was.
Much of the technology we use today would have been considered impossible 30 years ago.
If not impossible then impractical, agreed. Especially at low cost which along with everybody using the same basic technology is the key issue.
Really the best thing for me is not having to hassle with ATC voice communication to avoid other traffic in dense traffic areas, at least for the most part.
This I don’t get. Visual is OK but audible is not? Or in other words, to sit in the cockpit with your eyes glued to a pad is “a good thing”, but to have a system where you don’t need to have your eyes glued to a pad is “a bad thing” ?
Where this even is an issue is at super crowded airsport sites with lots of background traffic to start with, gliders, skydivers and so on. In controlled airspaces these are controlled anyway.
Silvaire wrote:
Unsuccessfully….The goal should be not to see other traffic unless it’s necessary – with traffic info it’s usually practical instead to change course slightly and avoid being that close, even when there’s a lot of traffic. You spot some of them by eye anyway but I no longer prioritize visual contact outside of the immediate airport traffic area in which everybody is maneuvering to or from the same spot, and thereby close.
ADS-B works, and it is interesting to me that after only three years of operating in terminal area environments where 99.9% of everybody has it and I can see them clearly on an iPad, this thread feels to me like I’m reading something from the Stone Age. How quickly we adjust and reset our expectations.
Agree with you @Silvaire. I personally am not yet to the point where I can say that “I no longer prioritize visual contact”, as ADS-B is not mandatory in Europe. However, looking at the FLARM display is definitely part of my scan in busy areas. I purposely placed it on top of the panel when rebuilding it, so as to keep the eyes outside as much as I can.
One of my biggest fear is drones, possibly operated “wildly” by uneducated / relentless pilots, outside of the prescribed zones and altitude.
Agree with you @Silvaire. I personally am not yet to the point where I can say that “I no longer prioritize visual contact”, as ADS-B is not mandatory in Europe. However, looking at the FLARM display is definitely part of my scan in busy areas. I purposely placed it on top of the panel when rebuilding it, so as to keep the eyes outside as much as I can.
I didn’t mean to downplay the importance of looking outside the plane, and understand that Europe is very different from the environment I’m describing. Even in what is now a US ADS-B OUT ‘mandatory’ environment I was myself the exempted guy without a transponder of any kind for years, in a different plane. Also, ADS-B is only ‘mandatory’ in the US in areas where traffic is very dense. However within those areas exempted traffic has got to be less than one in a thousand planes nowadays, and the reality is that I’ve never since 2020 had any traffic surprise me. That is very different than it was before, and I suspect the mid-airs that we had locally every couple of years are going to have decreased in number dramatically when the record can be examined after say ten years.
ADS-B in the US wasn’t primarily intended for on-board traffic awareness, it was for ATC’s use in ‘surveillance’. That’s because when the signal happens to be coming from an FAA ground station and not the other plane directly there is a delay and the location of the plane is not shown accurately. However in my experience in my environment that has not been a factor, I find it transforms your awareness of the traffic around you, and in practice it appears to me you can ‘see’ virtually everybody nearby in any dense traffic area where I fly. Also, in e.g. Foreflight their position can be shown on the VFR sectional in relation to terrain and other geographical features. It’s a completely transformational situation.
@LeSving, for traffic avoidance, listening to ATC, understanding what was said to you specifically out of an almost non-stop stream of words, forming a phrase, waiting for a gap and responding appropriately while simultaneously looking for the traffic visually (and often failing to see it) feels very, very clunky in comparison with glancing at a display every so often and simply acting on the much clearer info it provides. Also, a critical phase for traffic awareness is when say 10-25 miles from an urban area that may have several busy airports nearby, with converging traffic. In those US areas you are likely not talking to ATC anyway when VFR – there is much too much traffic for everybody aloft to be talking to ATC.
FAA forecast for 2023 is a total GA fleet of 209,140 aircraft. Not all of these are required to have ADS-B Out in rule airspace because those without an electrical system are exempted in much of the rule airspace. Total number of equipped with ADS-B Out are 165,351 of which 155,906 are good installs, the rest are NPE (Non performing emitter). Some of the NPE are SIL=0, but not many. Most are due to a bad install or with missing required data. My Bonanza was NPE for a while because the pressure altitude was missing. The serial port I was using to send the pressure altitude from the transponder to the UAT ADS-B Out system had failed, fortunately there was a spare serial port available to resolve the issue.
So 79% are installed with ADS-B Out with 75% good installs. Not every issue, probably almost all that make an aircraft NPE, are unrelated to SIL=0, so the target is still likely to be displayed on my certified panel display. Pretty much all with a position and either a pressure or geometric altitude will show on my ForeFlight display, as it does not care about SIL or NPE status. The targets good or bad are received by my certified receiver, but passed on to ForeFlight via BT. What NPE status does is to prevent ATC to use the ADS-B Out position. The rule prevents non equipped aircraft to operate in the designated rule airspace without prior advanced permission via an automated system. Fortunately this does not apply to aircraft that are NPE, because they are equipped, just that they are inop or not meeting requirements. Such NPE aircraft can request permission at any time including in flight. As long as one has a working transponder with mode C, permission is generally granted by ATC. NPE aircraft also won’t get the ground station service of TISB, which generates an ADS-B like broadcast for aircraft that are not equipped,
I also agree with the sentiment that the ADS-B Out is the more valuable in that at least others with inexpensive receivers can see your location. It still takes two to tango. Most mid air collisions involve a faster airplane overcoming a slower one, often where the slower one is in the pattern and the faster one descends onto the other aircraft. Often only the faster aircraft is going to be able to gain a visual before the collision, but even that is sometimes blocked by fusalage or other structure..
