卡－27，第一张旋翼很奇怪，求大神解毒！

只因为拍摄感光速度是有限的。

数码摄影机感光元件是按行扫描的，
于是画面不同地方的图像实际上是在“不同时间”里“拍下来”的，

在物体运动角速度一般的时候看不出来，但是如果物体相对于镜头高速运动，
就会看到有变形。

对了，国内好多网站都有防盗链，
楼主这样直接贴图片通常是没法让大家直接看到的，

有用的能看的还是楼主最后贴的原帖网址。

多谢2、3L·······

这样子岂不是可以通过快门速度猜出旋翼速度了～

相机的快门分两种。镜头中的一般是叶快门，由几个叶子一样的片围成一个圆圈，同时打开和关闭。还有一种是底片前的快门，好像拉门一样在底片前挡着。曝光的时候就抽开，到了曝光时间就用另一扇拉门重新拉过来挡住底片。这种类似拉门一样的快门就会造成第一章照片中的效果，因为上下部分曝光的时间点不同。所以想要知道转速，更重要的是知道这个相机快门的机械速度。

楼上几位大湿的分析是纯理论的，没有事实可以证明。我拍了十多年的直升机，没拍到过一张这种变形旋翼。
我觉得第一图是P出来的，理由如下:1/如果如几位大湿分析的那样，为何只有两叶旋翼变形，而不是所有的旋翼都变形？2/看旋翼在机身上的影子就什么都无需解释了！P者漏了手尾，穿帮了。3/从旋翼在机身上的影子来看，相机快门速度很快，旋翼的动感都没有，还怎么拉出变形？

楼上几位大湿的分析是纯理论的，没有事实可以证明。我拍了十多年的直升机，没拍到过一张这种变形旋翼。
我觉得第一图是P出来的，理由如下:1/如果如几位大湿分析的那样，为何只有两叶旋翼变形，而不是所有的旋翼都变形？2/看旋翼在机身上的影子就什么都无需解释了！P者漏了手尾，穿帮了。3/从旋翼在机身上的影子来看，相机快门速度很快，旋翼的动感都没有，还怎么拉出变形？

如果第一张确保真图，那么楼主恭喜

其實是旋翼快速轉動時因為接近光速而產生客觀坐標上的物體扭曲,詳細可以參考勞倫兹变换,LZ不必大驚小怪

感觉应该是P的，很多细节都有漏洞

老鬘 发表于 2011-8-1 13:39
楼上几位大湿的分析是纯理论的，没有事实可以证明。我拍了十多年的直升机，没拍到过一张这种变形旋翼。
我 ...
如果你拍了十几年的直升机都没有这种情况，那么可以证明你用的都是叶快门或者电子快门的相机。你可以回去自己看看，你的快门是在镜头里的还是在相机里的就知道了。
只有一部分叶片变形正好符合底片前快门的特性。注意看一下，画面上水平的几个叶片变形最小，因为曝光时间是同时，而画面上不在同一横线上的叶片变形就很大，因为不是同时曝光。至于影子那是一个道理。

Yourxianda 发表于 2011-8-1 21:43
如果你拍了十几年的直升机都没有这种情况，那么可以证明你用的都是叶快门或者电子快门的相机。你可以回去 ...
分析得很专业，到位。不得不顶

搞不懂。

老鬘 发表于 2011-8-1 13:39
楼上几位大湿的分析是纯理论的，没有事实可以证明。我拍了十多年的直升机，没拍到过一张这种变形旋翼。
我 ...
有种说法，卡式共轴旋翼，容易产生乱舞，这张P的图片，是不是影射这个啊

cxlhs9 发表于 2011-8-1 14:09
好多稀有图片哦，可以和活捉一拼啦

可能是数码相机的问题。

Yourxianda 发表于 2011-8-1 21:43
如果你拍了十几年的直升机都没有这种情况，那么可以证明你用的都是叶快门或者电子快门的相机。你可以回去 ...
看来你是个缺乏实践的纯理论者。既然实践是检验真理的唯一标准。请问，你实践过你的理论吗？不要说你没条件去实践，现在地球人都会按快门。
现在拍飞机绝大部分是用单反相机，单反相机的快门都在机身上，这种快门以前叫布帘快门,当然现在是金属的了，开启方式又分横走和纵走。我的相机快门是纵走，我就按纵走来分析问题。
快门设定慢速度时布帘是全开，即前帘先开，后帘再跟上。快门设定快速度时是前后帘仅留一个空隙（空隙的大小，示速度的大小而定），快速扫过底片（感应器）。
你说的理论如果成立，那么当快门扫过底片时，随着快门缝隙的移动，旋翼在底片不同的位置留下了痕迹。为何到了旋翼根部又还原正常影像了？那么快门在拍到旋翼根部正常影像的时候，翼梢的影像又在哪里？谁都知道旋翼是直的。
你说的理论如果成立，那应该可以拍出长着怪臂的羽毛球或网球（或其他挥臂的）运动员了？有谁能有这种照片？
你说“至于影子那是一个道理”，没有理由,变形的旋翼留下一个不变型的影子。这也太科幻了吧？

应该发到无忌去问问呢，反正不是镜头的关系对吧

图片大好，争论很精彩！这个帖子值了！

旋翼根部形变小是因为根部的旋转速度小啊在快门拍到旋翼根部的时候因为翼梢在画面上边所以早些时候已曝光完了。
另外付科幻图几张：

老鬘 发表于 2011-8-2 20:32
看来你是个缺乏实践的纯理论者。既然实践是检验真理的唯一标准。请问，你实践过你的理论吗？不要说你没条 ...

旋翼根部形变小是因为根部的旋转速度小啊在快门拍到旋翼根部的时候因为翼梢在画面上边所以早些时候已曝光完了。
另外付科幻图几张：

貌似ccd是这样，不知道现在的CIS是否可以解决这个问题。

另外我再解释一遍为什么影子没变形。形变需要两个条件。第一就是速度，严格来讲是相对于底片的角速度。第二就是曝光时间差异，就地一张图来说就是不在同一水平线上。图片里没变形的旋翼和影子都基本在同一水平线上。虽然他们的速度很快，但是因为是同时曝光所以就是直的。

似乎现在的数码单反机都喜欢用纵向的快门，而以前的胶片及很多用横向的帘幕快门，这样变形的效果就不太一样了，呵呵。找个什么时候自己在家拍电风扇试试。

学习了，不错不错。

拍风扇小了，拍吊扇很合适！如果不是P的，出这样的效果快门应该不是很快的速度！

Yourxianda 发表于 2011-8-2 23:11
另外我再解释一遍为什么影子没变形。形变需要两个条件。第一就是速度，严格来讲是相对于底片的角速度 ...
说破天，你也是纸上谈兵。拿别人P过的照片证明不了你的理论是可以成立的。
我呢，也是不到黄河心不死，你自己拍一张变形的照片（风扇、吊扇等等）来亮亮相，以证明你的说法。
否则只是在证明：吹牛不上税。懒得再和你浪费时间了。

按你的逻辑我要说J10B存在，你也会让我去自己拍照片给你看才存在么？我没有单反也没有直升机可以拍，所以恭喜你赢了，这些都是无聊的人PS的。

推荐你Google “focal plane shutter distortion”看一下前20个链接，然后默念“这都是PS的”一百遍。你就炼成网路金钟罩了，在辩论中永远会利于不败之地。

估计是相机问题！

这逻辑。
维基百科上说的很清楚了。
http://en.wikipedia.org/wiki/Focal-plane_shutter
不错的文章：
http://people.rit.edu/andpph/text-focal-plane-artifacts-in-digital-cameras.html

老鬘 发表于 2011-8-3 08:07
说破天，你也是纸上谈兵。拿别人P过的照片证明不了你的理论是可以成立的。
我呢，也是不到黄河心不死，你 ...

这逻辑。
维基百科上说的很清楚了。
http://en.wikipedia.org/wiki/Focal-plane_shutter
不错的文章：
http://people.rit.edu/andpph/text-focal-plane-artifacts-in-digital-cameras.html

Yourxianda 发表于 2011-8-3 10:48
按你的逻辑我要说J10B存在，你也会让我去自己拍照片给你看才存在么？我没有单反也没有直升机可以拍， ...
对你论证的态度表示一下赞赏！

刚才google一下， 类似这种图很多啊

明白原理了，下次试试拍出这种效果来，尤其20楼最后那张～

捅血，光速是多少啊？ 一般直升机翼尖上的线速度不会超过音速，如果在音速附近，会产生音爆，噪极大，破坏力很强，对机后部的小螺旋桨会产生一定的影响。所以一般最大速度会控制在0.92倍音速以内。如果接光束，那，那，无语了。光束是音速的88万倍

oneway2007 发表于 2011-8-1 13:58
其實是旋翼快速轉動時因為接近光速而產生客觀坐標上的物體扭曲,詳細可以參考勞倫兹变换,LZ不必大驚小怪{:15 ...

捅血，光速是多少啊？ 一般直升机翼尖上的线速度不会超过音速，如果在音速附近，会产生音爆，噪极大，破坏力很强，对机后部的小螺旋桨会产生一定的影响。所以一般最大速度会控制在0.92倍音速以内。如果接光束，那，那，无语了。光束是音速的88万倍

老鬘 发表于 2011-8-2 20:32
看来你是个缺乏实践的纯理论者。既然实践是检验真理的唯一标准。请问，你实践过你的理论吗？不要说你没条 ...
相机底片属于平面，
所以成像是以平面为基准的。
对于圆周运动而言，相对于平面，运动速度是一条正弦或者余弦曲线————这是基本的原理。
同时，对于相机而言，相机是感光元件，也就是说，有光会覆盖无光，如果你排过著名的“煤块上的煤矿工人”就知道，首先背光拍煤块，这时候因为煤是黑色的，在胶片上不会感光，然后再对工人二次曝光，这个时候，因为工人的背景是黑的，已经感光的煤的影响不会再次感光，而工人肖像上有光，恰好在煤块的上次未感光的部分感光，两次曝光叠加，就是一个工人在煤块上的肖像————相机是靠光记录影像————这是第二个基本原理。
分析第一张照片，首先，指向机头前方的叶片，由于相对底片的位移是垂直的，所以景深起主要作用，考虑到相机的距离和景深以及光圈的关系，判断是高光，高iso，小光圈，慢速快门。
然后考虑上面的叶片，由于我们知道螺旋桨旋转的方向，所以判断叶片的尖头部分先曝光，而靠轴的部分后曝光。
根据上面说的两个原理，首先如果不是快门问题，那么变形的几个螺旋桨应该扫过一片阴影，如同第二张照片，但是考虑基本原理二，由于背景亮度大，背景会对影像产生覆盖，所以叶片虚，但是不厉害。
考虑相对速度的关系，我们判断该照片是四周先曝光，然后中心后曝光，这样导致越靠近画面中心曝光时间越靠后，典型的叶片快门...至于机身上的影子————那是最后曝光的位置...
＋＋＋＋＋＋
＋解答完毕＋
＋＋＋＋＋＋
二次及多次曝光技术，属于摄影里相对中档的技术，一般而言，掌握到光圈、快门和iso之间的关系，就能拍出绝大多数不错的照片————同样，常用的一般单反都有后帘闪光功能，这是简单的二次曝光技术的应用————还属于初级技术，掌握这个就可以拍个夜景或者溶洞什么的。
著名的二次和多次曝光照片，出了我们中国人创造的“煤块矿工”照片外，比较体现技术的还有美国人用一只闪光灯闪20多次，勾画全境地面的照片，哪张照片里闪光灯光斑分布在画面多个位置————虽然实际上只用了一次B快门，但是也属于多次曝光技术。
＋＋＋＋＋＋＋＋＋＋＋＋＋＋
由于叶片快门多用于镜头而不是机身，所以也叫镜间快门————您没接触过也算正常，
但是凭这个就判断是Ps，就有点过分了...
要是给你几张移轴相机拍的照片，看看那变态的景深分布...
＋＋＋＋＋＋＋＋＋＋＋＋＋＋
您作为贵宾，拍了十多年直升机，用的都是机身快门的单反吧，
至于传统的拍摄手法和偏门相机...您要像本菜鸟一样业余选手没事只敢看看纽约摄影教材的还算说的过去，但是作为专业人士————不了解这些，确实...

我靠，好多小白，看不惯了，我用高中物理也知道是因为圆周运动分解到平面上，对着镜头部分的角速度大

90度两边角速度小，所以变形的都是在中间。。。

还贵宾

libo_qaz 发表于 2011-8-3 22:04
我靠，好多小白，看不惯了，我用高中物理也知道是因为圆周运动分解到平面上，对着镜头部分的角速度大

90 ...
人家讨论的是弯曲...不是虚影...
这个超过高中知识了...

这种弯曲的旋翼，我拍出来过
仰视，拍的是从头顶飞过的直升机，回去一看，5片旋翼都有点变弯了

老鬘 发表于 2011-8-3 08:07
说破天，你也是纸上谈兵。拿别人P过的照片证明不了你的理论是可以成立的。
我呢，也是不到黄河心不死，你 ...
这个网页有专门的解释
http://people.rit.edu/andpph/tex ... igital-cameras.html

The focal plane shutter artifact in certain digital cameras

Andrew Davidhazy
Imaging and Photographic Technology
School of Photographic Arts and Sciences/RIT
The image of the propeller of an airplane in flight arrived to my in-box along with a question as to what might have caused the anomaly in the reproduction of the whirling propeller blade made by a camera that was one built into a cell phone. I don't know at this time the make and model of the camera but will report it when I find out.



Digital cameras record images in many different ways. Some use a separate mechanical shutter to allow light to affect the individual photo-sensors for a predetermined period of time. Others turn the sensors on electronically and simultaneously for a predetermined period of time. Yet others extract information from the sensor by sequentially transferring data starting at the top of the frame until they get to the bottom.

It is this latter type that is represented by the photograph referenced above. The scanning nature of the image acquisition step by this type of camera can be simulated by the manner in which a flat-bed scanner scans material placed on its platen.

Manufacturers are careful to point out that users should keep the cover closed while the scanner is scanning. One reason for this is that if the subject being scanned moves then an imperfect reproduction of the original will be the result. In summary this means that the scanner can not deal with moving subjects and reproduce them accurately. For accurate reproduction a simultaneous record of the subject must be made. For this there are area arrays such as built into higher end digital cameras like in DSLRs and most consumer grade cameras.

But when it comes to digital cameras built into phones the situation is more variable. These cameras are most often equipped with CMOS, as opposed to CCD, sensor arrays and these lend themselves to line-by-line extraction or progressive of image data.

Although neither approach is guaranteed to be used by all digital cameras, since both types can be made to record images sequentially, the camera that took the above photograph was equipped with just such a method for image extraction. Sometimes this is called a "rolling" shutter approach.

To demonstrate how the artifact is produced I decided to use my Canon flat-bed scanner as the recording instrument. I built a simulated propeller by adding 4 spokes of masking tape to a clear CD protective disk. The central hole of the disk was just right to accept a pen through its central hole. I taped the pen to the disk. I located a drawing aid that happened to have a hole just the right size to pass the pen through. This allowed me to twirl the pen which rotated the simulated propeller just above the glass surface of the scanner's platen.

After conducting some preliminary "twirls" it was obvious that the system would work. But it needed refinements in set-up and understanding of the recording process.

The scanner was used only in the PRESCAN mode, and the image produced copied from the screen. The reason for this was that in the SCAN mode the process was so slow that it became hard to make the propeller disk rotate at a steady rate.

Directional arrows were added to the scanner platen because it was not at first obvious that the resulting prescan image would need to be flipped left to right. The "camera" was really looking at the rotating disk from below and so left-right orientation was inverted from that point-of-view. Ultimately a series of  records were made and assembled into a visual technical report that can be seen in the images below.



This is essentially the manner in which the experiment was set-up and the results of rotating the simulated propeller at a fairly rapid rate as the flat bed scanner was acquiring a prescan.

The next set of images was acquired by rotating the disk at a number of rates starting off slowly and gradually increasing the rotation rate. It is obvious that as the rate increases the resulting artifacts "develop" into the reproduction acquired at the fastest rotation rate I was able to achieve and which are shown in the previous illustration



Finally, below once again are the original photographs that were sent to me marked up with the rotation direction of the propellers (they all turn in a counter clockwise direction when viewed from the front and so clockwise when viewed from the rear!) and the scanning process used by the camera to capture image data in a sequential manner takes place from bottom of the sensor upwards (when camera is held so long frame direction is horizontal) and thus from top of the image towards the bottom when viewing the image on the camera screen.

It is as shown from the previous illustratins that one has to keep in mind the scanning direction. As stated earlier and as shown in an experiment illustrated later, within this camera the scanning direction is from bottom to top and thus at the subject from top to bottom because within the camera the image is inverted and reversed left to right.

It is the "inverted" nature of the optical image that makes the scan at the subject effectively happen from top to bottom such as shown in the left hand picture. The prop blades naturally align with the scanning direction of the photo-sensor. In the left photograph they travel horizontally from bottom to top at the subject and line up with the horizon. For the right hand picture the photographer turned the camera 90 degrees and so the propeller blades now preferentially align in the vertical direction since the scanning action is now perpendicular to the horizon.



I then did a quick test to determine postivley the scanning direction in a particular phone camera. The iPhone. I knew that the propeller blades when viewed from the rear turned clockwise. I checked this as I stepped off the plane on which I had just flown myself. I made sure to note the position of the camera lens for each of the photographs so I would know what the camera orientation was for each.

Note that for the picture on the upper right the blades form the "tines of a fork" pattern typical of the blades moving in the opposite direction of the scanning line. In the picture at lower right (camera "upside down so to speak) they are traveling in the same direction and in the picture on the left they are moving to the right while the scaning action must be taking place from right to left.

The "fork" shape to the blades is a result of the fact that the blades and the scanning action are moving in opposite directions. Due to this more blades can pass by the scanning line then when they are moving in the same direction. Effectively the relative speed between the blades and the scanning line is greater when they are moving in opposite directions than when in the same direction.

Since the scanning line moves at a consistent pace more blades are able to pass over the line in a given a=mount of time when they travel counter to the direction in which the scannin line moves then when the two are moving in the same direction. It is all a matter of relative speeds!





So there you have it. Another photographic anomaly, long associated with what is called focal plane shutter distortion, has resurfaced in the digital realm. In this latter case, however, the device that mimics the moving slit of the focal plane shutter is effectively much finer than the rather large and slow moving slit in the mechanical focal plane shutter. This accounts for the relatively excellent sharpness of the rotating blades of the propeller.  Distortion remains, however.

One of these days I will make a record with a focal plane shutter of the propeller and have a direct comparison. At this time, alas, I don't have such a comparison but hope to make it soon!

If you found this interesting or useful or you want to discuss this feel free to send  me an email at Andrew Davidhazy,  andpph@rit.edu


A set of Qustions and Answers was sent to me by Konstantin Othmer,  kon@coremobility.com,  and are included here with his permission. They help with some of the concepts discussed above.

A friend and I were thinking about this effect in a video he took and we found your excellent web page: After reading the page, I still didn't grock it, so I wrote my own Q&A to figure it out; I've convinced myself I now "get it". Not that your explanation wasn't excellent - I'm just a bit slow! Your page is really outstanding! Thanks for the great write-up and really cool experimental results with the scanner. Konstantin

Q: Is each line the same blade or a different blade?

A: Trick question! At the top and bottom, each line is one blade - first on the way up and then on the way down. In the middle, there is one line where the blade goes the same direction as the scan, and multiple lines on the other side. The multiple lines are each one blade, but the single line is all four blades!

Q: How fast are things moving?

A: A blade is at 1800 RPM, or 30 RPS. Assume scan is at 30 FPS. With four blades, each blade appears at every point in the frame once, and you should see four lines on the side where the blade and the scan are going the same direction, and eight lines on the side where they are going the opposite direction. Assume clockwise blade and top to bottom scan.

On the right side, blade and scan are going the same direction.
Blade 1 intersects scan from 0 to xx
Blade 4 intersects scan from just before 1/2 to just after 1/2
Blade 3 intersects scan from xx to 1
Blade 2 never intersects the scan on the right hand side!

On the left hand side, blade and scan are going opposite direction:

Blade 4 intersects part way, probably 1/8th down, drawing left to right, downward sloping line to the center point
Blade 3 intersects in the  2nd part of the way down, probably about 1/4th, downward sloping line to center point and stop
Blade 2 intersects starting at 1/2way point, makes very thin line to center point
Blade 1 intersects before the ¾ point and scribes thin line sloping down to right
Blade 4 intersects just before the end and scribes an upward sloping line all the way across

In this case, the scan intersects three blades on the right and five blades on the left! On the right you have three lines, two that go all the way across and one that goes to the center. On the left you have five lines, two that go all the way across, and three that go into the center.

Q: Is the blade faster or the scanner faster?

A: At 1800 RPM, the blade moving is twice as fast as the camera scanner at (30 fps). It will appear very thin when its going the opposite direction. Imagine the blade starting at 12 and turning clockwise. When the blade is ¼ of the way around (at 3), the scanner is 1/8th of the way down. So the blade and the scanner only intersect for the very top piece after which time the blade outruns it. The next blade is not far behind, and when the scanner is at 1/8th of the way down, the next blade is at noon, having already drawn the left part of the line which is sloping down since the outer most part will hit the scanner first, and the scanner is going down.

The faster the propeller spins, the more lines you will get. Each blade "draws" a horizontal line as it passes through the scanner "window". It passes through with sine wave speed - fast on the ends, slower in the middle, and then fast on the ends again.

QQ:You would think the lines would be much wider on one side than the other, and a lot fatter in the middle than on the tips, but they aren't. Not sure I get why that is. Maybe I don't understand it after-all!

The easiest way to grock this is to imagine looking through one gap in a horizontal blind, and imagine a propeller spinning on the other side. You'll see a dot move from one side to the other - taken across time, it draws a line.


and this is a photograph made by Loret Steinberg illustrating the effect on her winshield wiper as she was driving on the NYS Thruway!





NOTE: When this article was originally prepared I had wrongly assumed that props generally turned clockwise when viewed from the front and the scanning action in the camera was from top to bottom at the sensor (bottom to top at the subject). Both were incorrect assumptions but the analysis was still OK.

Below is the message that prompted the "review" of propeller rotation direction:

I am very interested in your explanation of the focal plane shutter artifact as it relates to the digital images of the aircraft propeller. Your simulation device is very clever and helps in the understanding of this digital phenomenon I am a radiologist so I have an interest in this from the standpoint of patient motion artifact created by breathing or cardiac motion during digital chest imaging. I am also a pilot which drew me to your paper since it was quoted in an aviation forum. The engine nacelle seen in the pictures is of a mid sized, twin engine, turboprop aircraft -- probably a Beechcraft King Air series. As presented, the pictures are of the right engine as seen by the right seated co-pilot who is behind and inboard of the propeller arc. Counter to your statement in the paper,  almost all aircraft engines actually turn in a counter clockwise (CCW) direction when viewed from the front. (There are only a few European engines that turn in the opposite, CW direction-) So, the camera would have to be looking at a disc rotating in a clockwise direction since the camera is behind the arc. The only circumstance that could confuse this scenario is that the picture was actually made of the LEFT engine by the pilot in the left seat and then the image was reversed prior to printing, but that would also have reversed the presentation of the artifact. I am interested in your thoughts with the above in mind. Best regards Crispin Spencer, CSPENCER@ATLANTIC.NET

快门同步率的问题，前面有版友解释的很清楚了。我是来围观贵宾讲大道理的

这种现象不少见

http://blog.timmatsui.com/2010/08/17/canon-5dmkii-hdslr-tips-the-camera-shutter/

http://www.cameralabs.com/forum/viewtopic.php?t=21083&sid=f8593ceeab6b691ba37514c7a974b3a9

http://www.theaerodrome.com/forum/aircraft/24857-ot-spinning-propeller-question.html

http://chdk.wikia.com/wiki/Talk:Samples:_High-Speed_Shutter_%26_Flash-Sync



快门同步率的问题，前面有版友解释的很清楚了。我是来围观贵宾讲大道理的

这种现象不少见

http://blog.timmatsui.com/2010/08/17/canon-5dmkii-hdslr-tips-the-camera-shutter/

http://www.cameralabs.com/forum/viewtopic.php?t=21083&sid=f8593ceeab6b691ba37514c7a974b3a9

http://www.theaerodrome.com/forum/aircraft/24857-ot-spinning-propeller-question.html

http://chdk.wikia.com/wiki/Talk:Samples:_High-Speed_Shutter_%26_Flash-Sync

直升机的螺旋桨挥舞起来本来就有变形么。
只不过是垂直于镜头的，用长镜头拉近以后轴向变形明显，平行于镜头的不明显罢了。