J-20 5th Gen Fighter Thread VI

[Inst]

Please, Log in or Register to view URLs content!

People can be very sloppy with terminology. I've noticed that when people say “perspective” they very often mean something else, eg. “field of view”, and that many are unaware of what “perspective” actually is. This happens even to people who should know better: Taken literally from a lens catalog of a camera maker:

[…] Wideangle lenses offer an increased depth-of-field perspective not possible with the human eye. […] What the hell is a “depth-of-field perspective”!? They probably meant “field of view”.
Perspective has nothing to do with focal length. Perspective describes how a 3D scene is projected onto a 2D canvas, eg. film, digital sensor or even your retina. Perspective describes how the objects in the 3D scene appear in the 2D image, which objects are visible and which are obscured by other objects, how big they appear in relation to each other. Focal length only influences the field of view, ie. how much of the entire scene appears in the image. Focal length does not change the relative sizes of objects in the image. Perspective is only influenced by the relative position of the objects in the scene and by the position of the canvas (viewer, camera). If you want one object in the foreground appear much bigger than another object farther away, it doesn't help to change focal length. The only way to accomplish that is to get closer to the object in the foreground. If you want both to appear as being about the same size, the only way to accomplish that is to get farther away from both objects.

Perspective is only influenced by your position, and field of view is only influenced by focal length.

You often hear things like “wide-angle perspective”. There is no such thing. People mean either “wide-angle field of view” or “close-up perspective” here. You also often hear “zooming with your feet”. This doesn't make any sense at all. Zooming changes focal length only (and therefore field of view), and walking changes perspective only. You can't replace one with the other.

So why is it still that with wide-angle photos the objects in the foreground look quite large, and with tele-photo shots the scene looks “compressed”? Doesn't this contradict what I just said? No, it's just a coincidence. With wide-angle lenses you can focus quite closely, and if you do get close to foreground objects, they will appear large. The wide-angle lens just allows you to also capture some of the background of the scene. With tele-photo lenses, you often shoot objects that are far away. It's that “being far away” that makes the scene look compressed, not the tele-photo lens. The lens only lets you capture the “right” crop of the entire scene, blocking out the foreground and concentrating on the faraway objects.

In many lens catalogs and books and on many web sites you can see a series of photos of the same scene taken with different lenses. With some of these the perspective is indeed different in different shots. Again, that's just a coincidence. What they don't tell you is that the photographer walked back to capture a foreground object at constant size while using ever longer lenses. But it's this “walking back” that caused the change in perspective, not the longer lenses.

Still not convinced? Here are a few examples:

Click on the link to see examples.

 

[latenlazy]

The lack of commonly shared flight parameters is NOT systemic bias and actually is its precise opposite. Systemic bias would instead be, say, if you're aiming to take a random sample to determine the incidence of childhood malnutrition in a population at large, but you end up setting up your sampling station in a poor village. The figure shown would be representative of the poor village, but not of the population at large, defeating your purpose.

Perhaps I shouldn’t have used the term “systemic” (a bit too quick on the rhetorical draw there) but the point is the same. Extrapolating from incomparable datasets as if they were comparable is faulty analysis. Unless you can establish a clear basis of comparison you’re doing little more than making up arbitrary facts.

As to conclusions being likely or certain, we're in an equivocation game again. You're using likely or certain to indicate that both are the same, but they're not. Nor am I even saying that likelihood or certainty matters when we're running individual video analyses; what matters is that the accuracy is statistically significant for the total sample set.

Where did you learn grammar? I said likely or certain to mean either likely *or* certain. I find either without adequate qualification to be intellectually dishonest. I’m not equivocating the two. You’re really reaching there.

What are you even talking about with regards to accuracy or statistical significance? You’re not taking random samples from a video when you do video analysis. You’re measuring distance over time. There’s no statistics involved in that kind of analysis.

The point of video analysis currently is basically to catch Chengdu screwing up.

If you start with a conclusion before you find your evidence...that’ s called seeking confirmation bias.

We know what EM-envelopes for aircraft tend to be, what we are looking for is somewhere where the J-20 is ecxeeding normal EM-envelopes. What we look for right now is a shot of the J-20 exceeding normal parameters so we can say that the J-20 does have unusual agility. Or, alternately, with enough videos, we can conclude that the available evidence suggests that the J-20 is not exceptionally maneuverable, corroborating the claims of retired PLA pilots.that its subsonic maneuverability is merely "good".

Or...you can’t get a good grasp of how the plane performs in a video or a collection of them because 1) you can’t accurately measure speed and distance without more parameters, or 2) the sample of videos you get don’t represent the full range of the plane’s performance envelope (this bit would be systemic bias).

I see what this is about now. You don’t believe native Chinese speakers when they tell you what 不错 means. Let me ask, where did you learn Chinese, when did you learn it, and how often do you use it?

 

[latenlazy]

This is incorrect. You can test it yourself. If you have a zoom lens, shoot a photograph of an object, centered to minimize barrel distortion, at your longest focal length. Then shoot one of the object at your longest focal distance, from the same position as you shot it from the telephoto, at your shortest focal length. Then crop the wide angle until its FOV is the same as that of the telephoto picture. You should see about the same damn thing.

Did I say crop anywhere in my comment? No. I said if two objects *take up the same area in the frame*, but one is shot with a wide and one is shot with a long, as in if I took a picture of a bottle with 35 mm and a 100 mm and in both pictures the bottle took up the same area in the frame. Read more carefully.

 

[Inst]

See, it's the perspective point that really gets me. The point is, let's say we put a 4 foot cube about 20 feet away, facing you from below, so you can see both the near and far upper edge. The ratio of the near edge to the far edge is going to be the same, whether you use a wide angle lens or a telephoto lens, because it's cutting through field of view.

Likewise, if you have a J-20 flying parallel to you (and this is the easiest case), you can use the ratio of the wing edge closest to you relative to the wing edge furthest from you to determine distance. Say, if the former is 40 pixels long, while the latter is 38 pixels long, this ratio of 20:19 will be the same no matter the focal length you are using. In other words, the length decreases by 5%, so using a method of similar triangles, we know that the aircraft is roughly 250 meters away from you. This is, of course ,using an idealized and simplified case.

 

[Gloire_bb]

Anti-stealth radar still doesn't solve the problem of accurately tracking the target. Without missiles that can home in on stealth aircraft or 5th-gen interceptors of their own, anti-stealth radars simply give commanders a great view of the aircraft that kills them.

Sounds almost like cheap(ok, nowadays expensive too) blockbuster's bad guys.
They see them, but you only can say stupid stuff. Tactics are prohibited, and stealth a/c seem to be covered by plot armor.

 

[latenlazy]

See, it's the perspective point that really gets me. The point is, let's say we put a 4 foot cube about 20 feet away, facing you from below, so you can see both the near and far upper edge. The ratio of the near edge to the far edge is going to be the same, whether you use a wide angle lens or a telephoto lens, because it's cutting through field of view.

If you shot the image *at the same distance* using the wide and long then the relationship between the close and far edge would be the same. If you shot the image closer with the wide and further with the long so that the picture took up the same portion of the frame the relationship between the close and far edges in each image would be different.

Likewise, if you have a J-20 flying parallel to you (and this is the easiest case), you can use the ratio of the wing edge closest to you relative to the wing edge furthest from you to determine distance. Say, if the former is 40 pixels long, while the latter is 38 pixels long, this ratio of 20:19 will be the same no matter the focal length you are using. In other words, the length decreases by 5%, so using a method of similar triangles, we know that the aircraft is roughly 250 meters away from you. This is, of course ,using an idealized and simplified case.

The preservation of that ratio only holds true if the long and wide lense shots you’re comparing were shot at the same distance from the object. It does not hold true if the distance at which you shot the same object were different. This is *why* you need distance from the camera or focal length to get depth information. Without knowing the relationship of distance between the camera and object, you have no way of scaling the depth between a closer part of the object from a further part.

 

[Inst]

Correct, because the ratio changes depending on your distance. Say, if the ratio were 80:72 instead, you'd see that it's a reduction of 10% in length, suggesting that it's an increase in distance of 11% from the near to the far edge. Using the same wingspan of 12.88 as above, you'd end up with 115.92 meters distance to the aircraft. Which is why I'm saying you can use perspective distortion to determine the distance to the aircraft.

What we're trying to do, essentially, is to establish an E-M diagram of the J-20; i.e, use perspective effects and distance to determine airspeed and altitude, then note its maneuvers at a given point in its flight envelope.

 

[latenlazy]

Correct, because the ratio changes depending on your distance. Say, if the ratio were 80:72 instead, you'd see that it's a reduction of 10% in length, suggesting that it's an increase in distance of 11% from the near to the far edge. Using the same wingspan of 12.88 as above, you'd end up with 115.92 meters distance to the aircraft. Which is why I'm saying you can use perspective distortion to determine the distance to the aircraft.

What we're trying to do, essentially, is to establish an E-M diagram of the J-20; i.e, use perspective effects and distance to determine airspeed and altitude, then note its maneuvers at a given point in its flight envelope.

And what I’m saying is you don’t have the information you need to get a good measurement of anything in these videos (even setting aside the sampling problem). For example, if a J-20 in a video did a turn that looked like a very narrow flat parabola you don’t know if in fact that’s actually a much wider curvature turn or if the turn really was just that narrow because you don’t have good depth information. If you tried to get that depth information by trying to compare the size of the plane at different points of the turn even assuming your camera was holding steady enough, if the plane was really far away as it’s likely to be, that means very small differences or mismeasurements between the size of the plane at two different points can imply very large differences in depth, given how much distances relationships are flattened for distant objects. This is a pointless exercise.

 

[latenlazy]

And also, you’re not actually doing any of this stuff, so I’m not sure how that’s at all a relevant defense of the analysis you’ve provided.

 

[Deino]

Guys ... this discussion is going off-topic !!!