Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 7:08 pm
by Chris Peterson
zloq wrote:A key is that these are very raw and abstract "images" and they don't represent an x/y cross-section - they represent range vertically and frequency horizontally - with brightness corresponding to the intensity of the reflection. If you try to view it too literally as an image, then it just won't work.
Perhaps I'm misunderstanding you here. This is not an abstract image with range on one axis and frequency on the other. The range and Doppler data has already been inverted into a true spatial image. That is, we are seeing this object as it actually is, in two spatial directions. That's why when you see the movie version, actual spatial structure- ridges and a huge crater- is plainly visible.
The inversion process can produce ambiguous results, which is part of what produces the noisy looking image. As more data is integrated, the ambiguities will be resolved, and a more accurate image synthesized. But what we see in this preliminary reconstruction is a reasonably accurate representation of what this asteroid actually looks like, not an abstract view.
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 7:39 pm
by neufer
Chris Peterson wrote:
In reality, if a 400m asteroid struck the Earth, the odds are that casualties would be very low.
The casualties on Earth, perhaps; it would be a rough go for the asteroid passengers, however.
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 7:54 pm
by zloq
Chris Peterson wrote:Perhaps I'm misunderstanding you here. This is not an abstract image with range on one axis and frequency on the other.
I think if you study the lower right plot in the link provided by neufer you will get the idea. In that plot, as it is formed, frequency is vertical and range is horizontal. In this case range is vertical and frequency is horizontal. This is a range-Doppler image - and they call it that because it is a range/frequency depiction. I think if you review the literature on the topic you will find many such images of asteroids and they all look like this, with the dimensions being meters in one axis and Hz in the other. If you think these images have somehow been converted to true x/y depictions - then I would be interested in any citations you have regarding the inversion you allude to - and an explanation for why we only see 1/4 of the object - in this and other such range-Doppler images of asteroids.
If the shadows were done synthetically somehow - then why wouldn't they show two halves - each shaded - as separate depictions - since we are seeing the object face on? The reason is - this is a raw depiction of the frequency/time info - just as in neufer's figure - and there is no way to distinguish location except through range (z, or time) and velocity (rotation based doppler shift, or frequency). Inversion happens once you have many, many poses - and even then it's an ill-defined, model-based inversion process.
zloq
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 8:46 pm
by Chris Peterson
zloq wrote:I think if you study the lower right plot in the link provided by neufer you will get the idea. In that plot, as it is formed, frequency is vertical and range is horizontal. In this case range is vertical and frequency is horizontal. This is a range-Doppler image - and they call it that because it is a range/frequency depiction. I think if you review the literature on the topic you will find many such images of asteroids and they all look like this, with the dimensions being meters in one axis and Hz in the other. If you think these images have somehow been converted to true x/y depictions - then I would be interested in any citations you have regarding the inversion you allude to - and an explanation for why we only see 1/4 of the object - in this and other such range-Doppler images of asteroids.
The images posted by Art show the raw data. These data are the input to a method known as Hudson inversion (look up R Scott Hudson for a list of his papers on the subject), the output of which is a spatial image. I've never seen an asteroid image labeled as you describe- meters versus frequency. Do you have an example?
If the shadows were done synthetically somehow - then why wouldn't they show two halves - each shaded - as separate depictions - since we are seeing the object face on?
We are seeing the asteroid face on, but the image produced is more complex. The time-domain data returns a set of cross-sectional views along one axis, and the Doppler data provides a set of cross-sectional views on an axis orthogonal to that one. In a similar manner to computed tomography, these separate cross sectional data sets are combined to produce (with some ambiguities) a spatial map, which is the published image. I think the apparent shadowing is a consequence of the way the inversion process works: in essence, the object appears illuminated orthogonally to the line between it and the observer.
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 9:30 pm
by zloq
Yes - there is an inversion process to recover geometry - but you can't do it with just one frame. If you have never seen a frequency/range image of an asteroid - then I don't know what to say. They are the input to the inversion process - and the output is a 3D model - not an image - though once you have such a model you can make such a rendering.
The images and animations we see now are hot off the press - and there has been no time to crunch them into a 3d model. As I said - I don't know when the first such results will appear.
Here is a link to a 2010 Icarus paper by Ostro, who was a co-author with Hudson on one of its citations. I believe it is publicly available. If not - it is Ostro et al., Icarus 207, 2010, 285-94.
The paper describes the full process of going from raw, delay-Doppler images to a 3D model, and includes the fuzzy, silhouetted 1/4 view images of asteroids that are commonly shown - and look like these new ones. I'm confident you have seen other such images depicted in the same manner - but the details may not have been included in the captions - as is the case for these new images. There are more recent inversion models based only on light curves - but it's even worse than radar in that it takes tons of poses to deduce a 3d model that is most parsimonious with the photometry. This asteroid will have models combining both radar and photometry I expect - but it will take a long time.
Note that I'm not denying that an inversion process can and will happen for these data eventually. But when it does - the output won't look fuzzy at all - it will look like a 3D shaded model. If you have conflicting information on how exactly these recent images were processed, then I welcome a pointer. In the meantime note that they are not only fuzzy, but they appeared quickly after the event, and they only show 1/4 of the surface. Furthermore, the animation suggests that the rotation axis just happens to be pointing directly at the viewer - i.e. out of the plane of the screen. That is what you would expect from raw data - and the apparent alignment is an artifact of having the x-axis be frequency. It is all consistent with being raw frequency/range images - like the ones you are unfamiliar with, but are depicted in neufer's reference and in my 2010 Icarus reference.
The neufer link is from a guy who keeps getting asked - Why do these images of asteroids always look so ugly? - and he is answering as best he can - describing the raw range/frequency images - which is what people are seeing and asking him about - as are people in this thread. I'm just trying to fill in some of the details I don't think he addressed.
zloq
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
zloq wrote:
The neufer link is from a guy who keeps getting asked - Why do these images of asteroids always look so ugly? - and he is answering as best he can - describing the raw range/frequency images - which is what people are seeing and asking him about - as are people in this thread. I'm just trying to fill in some of the details I don't think he addressed.
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 10:58 pm
by zloq
Yes - my mistake. I got your link mixed up with Brenner, who did the Goldstone observations. Emily writes a blog for the planetary society. Either way - I thought her write up and animated diagram, which you embedded in your response, were pretty good. She didn't specifically say users complained about ugliness of images - but she did a good job of conveying the abstract nature of the images - and put things like "imaging" and "picture" and "look" in quotes to emphasize this point.
zloq
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Wed Nov 09, 2011 11:56 pm
by Chris Peterson
zloq wrote:Here is a link to a 2010 Icarus paper by Ostro, who was a co-author with Hudson on one of its citations. I believe it is publicly available. If not - it is Ostro et al., Icarus 207, 2010, 285-94.
Interesting. I was looking at a different paper, which presents things somewhat differently. I agree, based on your reference, that this image appears to be a single snapshot in a distance-frequency space, meaning it hasn't yet been transformed (I took it to be not a single snapshot at all, but the resultant transform of a series of images- something that can be produced immediately). If it is a distance-frequency space map, it's interesting how effectively it reveals true physical structure (like that big crater). It isn't intuitive to me that such a mapping would do that.
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Thu Nov 10, 2011 12:35 am
by zloq
Yes - it is interesting that features do show up - even though it's abstract. But that's because a generally smooth surface with just a few features on it would have those features mapped locally onto the image - as long as each feature has a unique-ish range and velocity pair of coordinates. As it rotates the velocity will change and the z-value will change - both looking like simple rotation around an axis.
If the axis of rotation were pointing directly at earth, there would be no horizontal information - just a vertical line corresponding to latitude, and the brightness of each point on it would sum all the features across longitude for that value of latitude - assuming a sphere.
In Emily's blog, at the bottom, she makes reference to the triple asteroid 1994cc, shown here:
If this were a "real" video, then all the alignment looks coincidentally perfect. The big asteroid axis is aimed right at the viewer, and the other two asteroids are moving vertically with a weird offset. But for the moons, you only know their distance, in the y axis, and their velocity, in the x-axis - which is why they are shifted a bit from the vertical axis and move only vertically. But you can definitely see features rotating on the asteroid. There is no info on where the moons really are laterally - at least from each single image.
I finally found a direct reference to this new APOD image and more background - and this time the image is annotated with pixel resolution that mentions both spatial and frequency dimensions:
It's unfortunate that with so many write ups of this event and the images on the web, they all avoid talking about what is really depicted in the images.
Maybe not too far from now there will be an APOD with a 3D rendering of what it really looks like.
zloq
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Thu Nov 10, 2011 9:25 pm
by TheEmperor
I think this image is stupid. All the technology we have... beautiful pictures of mars and Saturn yet we can't even get a decent photo of the asteroid that's closer to the moon? Come on man how dumb do you think we are?
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Thu Nov 10, 2011 9:36 pm
by Chris Peterson
TheEmperor wrote:I think this image is stupid. All the technology we have... beautiful pictures of mars and Saturn yet we can't even get a decent photo of the asteroid that's closer to the moon? Come on man how dumb do you think we are?
This object is smaller, darker, and farther away from the camera than the examples you give. And of course, as previously discussed, this image is an early release. You can be sure that better, higher resolution images will be coming along once the various science teams have had the opportunity to do more processing.
Personally, I think it is remarkable we are able to get this good an image of a black object the size of a warehouse nearly as far away as the Moon!
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Fri Nov 11, 2011 6:19 am
by iamlucky13
TheEmperor wrote:I think this image is stupid. All the technology we have... beautiful pictures of mars and Saturn yet we can't even get a decent photo of the asteroid that's closer to the moon? Come on man how dumb do you think we are?
I think some perspective would help.
It's 400 meters across. The moon is 8700 times bigger.
It passed 200,000 miles from earth. Compared to the probes we sent to Mars or Saturn it's 1000 times further away.
A picture of this asteroid taken with the highest resolution instrument on the Hubble Space Telescope would be only 10 pixels across. Viewed at 100% resolution on your computer monitor, it would display less than 1/8 inch across.
Re: APOD: Asteroid 2005 YU55 Passes the Earth (2011 Nov 09)
Posted: Fri Nov 11, 2011 9:00 pm
by neufer
Click to play embedded YouTube video.
iamlucky13 wrote:
TheEmperor wrote:
All the technology we have... beautiful pictures of mars and Saturn yet we can't even get a decent photo of the asteroid that's closer to the moon?
I think some perspective would help.
It's 400 meters across. The moon is 8700 times bigger.
A picture of this asteroid taken with the highest resolution instrument on the Hubble Space Telescope would be only 10 pixels across. Viewed at 100% resolution on your computer monitor, it would display less than 1/8 inch across.
