6 ms·

> By eye, yes. By digital camera? That would have to be pretty high. At least on clear days. Have you ever seen a hobby drone flown? I challenge you to see a t

> By eye, yes. By digital camera? That would have to be pretty high. At least on clear days. Have you ever seen a hobby drone flown? I challenge you to see a typical model flying at 250ft feet with anything not using a telephoto lens, even on a clear day, unless you already knew where it was (aka saw it from a lower altitude and followed it up).

kbenson 8y ago
> I challenge you to see a typical model flying at 250ft feet with anything not using a telephoto lens, even on a clear day, unless you already knew where it was Do you think you could identity a drone moving at 50 feet using a video camera that recorded as 640x480? Because approximately the same number of pixels would by used on a 4k capable video camera at 250 feet, and that's before any optical zoom at all. As long as you can identify something moving, it can be zoomed in on and classified (either by the same camera or a secondary one dedicated to the task), or just give the camera more pixels and use digital zoom. A few of those in a cluster would cover an area quite well I think. For another way to think of this, an object that is 8" (assuming a 12" drone at an angle) should appear to be a quarter the size of the moon to the naked eye at 300 feet.[1] I don't think that's something a good quality camera should have trouble picking up. 1: http://www.1728.org/angsize.htm http://www.1728.org/angsize.htm, input angle in minutes 7.5, size .666, using their example which says the moon appears to be about 40 minutes in angle size.
- jsjohnst 8y ago
  > For another way to think of this, an object that is 8" (assuming a 12" drone at an angle) should appear to be a quarter the size of the moon to the naked eye at 300 feet. If a drone was a solid spherical mass like the moon, your guesstimate might have some merit, but it’s not. In reality, your 8” is already way overshooting it. Furthermore, the amount of light reflecting by the moon is significantly higher providing much higher contrast for edge detection. The matte white plastic of many common drones however easily blends into the sky. Am I saying no camera would do it? No, that’s obviously wrong. But the post I replied to used a very vague wording which implies a majority of cameras would do it, which is definitely not the case. Happy to provide real photos to inspect this weekend (assuming it’s not still raining) at 100, 200, 300, and 350ft at whatever megapixel you’d like (up to 22mp) and focal length (from 8-1600mm, 35mm equivalent) so you can see what I mean.
  - kbenson 8y ago
    > In reality, your 8” is already way overshooting it. Given the body area of most these drones, I'm not so sure. I'd best the visible cross-section is pretty close to 64 sq. inches from most angles, if you include the small amount of area from the arms, but ignoring rotors. > But the post I replied to used a very vague wording which implies a majority of cameras would do it I don't read what I said that way. I said a digital camera, as in there exists a camera or type of camera that would work. I mean, there's some pretty crappy digital cameras out there, if you want to go really cheap. > Happy to provide real photos to inspect this weekend (assuming it’s not still raining) at 100, 200, 300, and 350ft at whatever megapixel you’d like (up to 22mp) and focal length (from 8-1600mm, 35mm equivalent) so you can see what I mean. Sure, I would be interested in see how it looks in practice. I purposefully limited myself to a little over 8 megapixels (4k) as I assumed video, and that's fairly easy to get with consumer grade hardware, which sets a nice precedent. If cheap consumer grade hardware can do it, a government solution with possibly special purpose hardware would have no problem. Knowing how a decent still camera views it at full res with little or no zoom would be a good way to look at is, as I think that would approximate a high-end video system, or a mid-range one in a couple years. I get emails on replies, so I shouldn't miss it even if it's a few days later.
- learc83 8y ago
  >Should appear to be a quarter the size of the moon to the naked eye at 300 The moon is illuminated by the sun, so not really a fair comparison. Take a drone and paint it the color of the sky and it's going to be a lot more difficult than “a quarter the size of the moon” would imply. A drone can also fly much higher than 300 feet, and it's not a solid object. But I've written a fair bit of object detection code though, and I'm certain a decent camera could do it fairly accurately. False positives are the problem though. Birds, airplanes (farther away), insects (closer to the camera), clouds. All of those things are much more common than drones and so even a very small false positive percentage is going to cause this system to be completely unreliable. That’s all without even mentioning licensed drones. How will the system handle not sounding the alarm when it detects a licensed drone? All drones will have transponders I assume? So now the camera needs a radio, and access to a national database of transponder codes. Assuming drones become more popular how does the camera know which transponder code belongs to which drone? Signal strength? This will be different depending on environment. Signal direction? You need multiple receivers for this. How do you stop a drone from emulating another drone’s transponder code? A challenge response system with a public/private key pair? And the most important question yet. What do you do when you spot an unauthorized drone? If a cop spots a car he can chase it. Does each camera have an interceptor drone? How does the interceptor drone bring down the unauthorized drone without causing any collateral damage? What happens if there are multiple unauthorized drones in an area? How do you trace the unauthorized drone back to the owner? Unlike a car, the operator isn't in the vehicle. Do you just allow the cameras to hand off monitoring to other nearby cameras until it lands and send cops to deal with it? How do you build a camera network with enough cameras to provide continuous monitoring, taking into account trees and buildings. All of that just to say yeah it’s possible, but it’s a lot harder and a lot more expensive than it seems. Much too hard and much too expensive to feasible anytime soon. Unless drones prove to be a lot more popular and effective than truck, car, and suitcase bombs, than I think they will, no one is going to pay for all of this. That’s not to say that we won’t have more licensing and registration for drones, but I don’t think it’s going to prove very enforceable or very effective at anything more than reducing the number of nuisance complaints. Maybe it’s worth it just to do that, but I’d need to see real data first. That's also not to say that some kind of automated drone defense won't almost certainly become more popular in very specific high security areas like the White House.

kbenson 8y ago

> I challenge you to see a typical model flying at 250ft feet with anything not using a telephoto lens, even on a clear day, unless you already knew where it was Do you think you could identity a drone moving at 50 feet using a video camera that recorded as 640x480? Because approximately the same number of pixels would by used on a 4k capable video camera at 250 feet, and that's before any optical zoom at all. As long as you can identify something moving, it can be zoomed in on and classified (either by the same camera or a secondary one dedicated to the task), or just give the camera more pixels and use digital zoom. A few of those in a cluster would cover an area quite well I think. For another way to think of this, an object that is 8" (assuming a 12" drone at an angle) should appear to be a quarter the size of the moon to the naked eye at 300 feet.[1] I don't think that's something a good quality camera should have trouble picking up. 1: http://www.1728.org/angsize.htm http://www.1728.org/angsize.htm, input angle in minutes 7.5, size .666, using their example which says the moon appears to be about 40 minutes in angle size.

jsjohnst 8y ago
> For another way to think of this, an object that is 8" (assuming a 12" drone at an angle) should appear to be a quarter the size of the moon to the naked eye at 300 feet. If a drone was a solid spherical mass like the moon, your guesstimate might have some merit, but it’s not. In reality, your 8” is already way overshooting it. Furthermore, the amount of light reflecting by the moon is significantly higher providing much higher contrast for edge detection. The matte white plastic of many common drones however easily blends into the sky. Am I saying no camera would do it? No, that’s obviously wrong. But the post I replied to used a very vague wording which implies a majority of cameras would do it, which is definitely not the case. Happy to provide real photos to inspect this weekend (assuming it’s not still raining) at 100, 200, 300, and 350ft at whatever megapixel you’d like (up to 22mp) and focal length (from 8-1600mm, 35mm equivalent) so you can see what I mean.
- kbenson 8y ago
  > In reality, your 8” is already way overshooting it. Given the body area of most these drones, I'm not so sure. I'd best the visible cross-section is pretty close to 64 sq. inches from most angles, if you include the small amount of area from the arms, but ignoring rotors. > But the post I replied to used a very vague wording which implies a majority of cameras would do it I don't read what I said that way. I said a digital camera, as in there exists a camera or type of camera that would work. I mean, there's some pretty crappy digital cameras out there, if you want to go really cheap. > Happy to provide real photos to inspect this weekend (assuming it’s not still raining) at 100, 200, 300, and 350ft at whatever megapixel you’d like (up to 22mp) and focal length (from 8-1600mm, 35mm equivalent) so you can see what I mean. Sure, I would be interested in see how it looks in practice. I purposefully limited myself to a little over 8 megapixels (4k) as I assumed video, and that's fairly easy to get with consumer grade hardware, which sets a nice precedent. If cheap consumer grade hardware can do it, a government solution with possibly special purpose hardware would have no problem. Knowing how a decent still camera views it at full res with little or no zoom would be a good way to look at is, as I think that would approximate a high-end video system, or a mid-range one in a couple years. I get emails on replies, so I shouldn't miss it even if it's a few days later.
learc83 8y ago
>Should appear to be a quarter the size of the moon to the naked eye at 300 The moon is illuminated by the sun, so not really a fair comparison. Take a drone and paint it the color of the sky and it's going to be a lot more difficult than “a quarter the size of the moon” would imply. A drone can also fly much higher than 300 feet, and it's not a solid object. But I've written a fair bit of object detection code though, and I'm certain a decent camera could do it fairly accurately. False positives are the problem though. Birds, airplanes (farther away), insects (closer to the camera), clouds. All of those things are much more common than drones and so even a very small false positive percentage is going to cause this system to be completely unreliable. That’s all without even mentioning licensed drones. How will the system handle not sounding the alarm when it detects a licensed drone? All drones will have transponders I assume? So now the camera needs a radio, and access to a national database of transponder codes. Assuming drones become more popular how does the camera know which transponder code belongs to which drone? Signal strength? This will be different depending on environment. Signal direction? You need multiple receivers for this. How do you stop a drone from emulating another drone’s transponder code? A challenge response system with a public/private key pair? And the most important question yet. What do you do when you spot an unauthorized drone? If a cop spots a car he can chase it. Does each camera have an interceptor drone? How does the interceptor drone bring down the unauthorized drone without causing any collateral damage? What happens if there are multiple unauthorized drones in an area? How do you trace the unauthorized drone back to the owner? Unlike a car, the operator isn't in the vehicle. Do you just allow the cameras to hand off monitoring to other nearby cameras until it lands and send cops to deal with it? How do you build a camera network with enough cameras to provide continuous monitoring, taking into account trees and buildings. All of that just to say yeah it’s possible, but it’s a lot harder and a lot more expensive than it seems. Much too hard and much too expensive to feasible anytime soon. Unless drones prove to be a lot more popular and effective than truck, car, and suitcase bombs, than I think they will, no one is going to pay for all of this. That’s not to say that we won’t have more licensing and registration for drones, but I don’t think it’s going to prove very enforceable or very effective at anything more than reducing the number of nuisance complaints. Maybe it’s worth it just to do that, but I’d need to see real data first. That's also not to say that some kind of automated drone defense won't almost certainly become more popular in very specific high security areas like the White House.