A camera that can see around corners | David Lindell

Transcriber: Ivana Korom Reviewer: Krystian Aparta

翻译人员: Yizhuo He 校对人员: Wanting Zhong

In the future,

在将来，

无人驾驶汽车将比 人类驾驶的汽车更安全可靠。

self-driving cars will be safer and more reliable than humans.

But for this to happen,

但为了实现这点，

we need technologies that allow cars to respond

我们需要能使汽车

反应比人类更快的技术，

faster than humans,

we need algorithms that can drive better than humans

比人类更会驾驶的算法，

and we need cameras that can see more than humans can see.

和视野比人类更广阔的摄影机。

For example, imagine a self-driving car is about to make a blind turn,

比如说，想象有一辆 无人驾驶汽车将要在其盲区转弯，

and there's an oncoming car

此时有一辆车正在驶近路口，

or perhaps there's a child about to run into the street.

或是有个小孩快要跑到车道上。

Fortunately, our future car will have this superpower,

幸运的是，我们的未来汽车 将具有这样一种超能力——

a camera that can see around corners to detect these potential hazards.

一台能看到拐角后面、 察觉这些潜在危险的摄像机。

For the past few years as a PhD student

过去几年间，作为一名

in the Stanford Computational Imaging Lab,

斯坦福计算成像 实验室的博士生，

I've been working on a camera that can do just this --

我一直在研发 这样一台摄像机——

a camera that can image objects hidden around corners

它能显示藏在拐角后面的、

or blocked from direct line of sight.

或是被挡在直视 视线之外的物体。

So let me give you an example of what our camera can see.

让我来举例说明 这台摄像机能看见什么。

This is an outdoor experiment we conducted

这是我们进行的一项户外实验：

where our camera system is scanning the side of this building with a laser,

我们的摄像机系统正在 通过激光扫描此建筑的侧面，

and the scene that we want to capture

而我们想捕捉到的场景

is hidden around the corner behind this curtain.

藏在这块幕布背后的“拐角处”。

So our camera system can't actually see it directly.

我们的摄像机系统 并不能直接“看到”那些物体。

And yet, somehow,

然而，神奇的是，

our camera can still capture the 3D geometry of this scene.

它仍旧能捕捉到 这个场景的三维几何轮廓。

So how do we do this?

这是如何实现的？

The magic happens here in this camera system.

玄机就藏在这个摄像机系统中。

You can think of this as a type of high-speed camera.

你可以把它想象成 一种高速摄像机。

Not one that operates at 1,000 frames per second,

不同于那些每秒能 捕捉一千帧、

or even a million frames per second,

甚至百万帧图像的摄影机，

but a trillion frames per second.

它每秒能捕捉 一万亿帧的图像。

So fast that it can actually capture the movement of light itself.

它的速度如此之快， 甚至能捕捉到光的移动。

And to give you an example of just how fast light travels,

为了举例说明 光的传播速度有多快，

let's compare it to the speed of a fast-running comic book superhero

我们不妨把它和 能以三倍声速移动的

who can move at up to three times the speed of sound.

漫画超级英雄做个比较。

It takes a pulse of light about 3.3 billionths of a second,

一段光脉冲跨越一米距离

or 3.3 nanoseconds,

需要 33 亿分之 1 秒，

to travel the distance of a meter.

即 3.3 纳秒，

Well, in that same time,

与此同时，

our superhero has moved less than the width of a human hair.

我们的超级英雄移动的距离 还没到一根人类头发的宽度。

That's pretty fast.

看来这确实很快。

But actually, we need to image much faster

但实际上，想要捕捉到 在次厘米规模上移动的光线，

if we want to capture light moving at subcentimeter scales.

我们还需要能更快成像的摄影机。

So our camera system can capture photons

我们的摄影机系统能以 每帧 50 万亿分之 1 秒，

at time frames of just 50 trillionths of a second,

即 50 皮秒的帧率

or 50 picoseconds.

捕捉到移动的光子。

So we take this ultra-high-speed camera

接着我们将这台 超高速摄影机

and we pair it with a laser that sends out short pulses of light.

和能发出小段光脉冲的激光组合起来。

Each pulse travels to this visible wall

当每段光脉冲传播到 这面可看见的（白色）墙时，

and some light scatters back to our camera,

就会有光线被散射回 我们的摄影机这儿，

but we also use the wall to scatter light around the corner

不过我们还用这面墙 将光线散射到拐角后面、

to the hidden object and back.

被挡住的物体上， 然后再散射回来。

We repeat this measurement many times

我们通过多次重复 这种测量过程，

to capture the arrival times of many photons

来捕捉从墙上各个位置

from different locations on the wall.

散射的光子的到达时间。

And after we capture these measurements, we can create

得到这些测量结果后， 我们就能生成

a trillion-frame-per-second video of the wall.

这面墙每秒一万亿帧的一段影像。

While this wall may look ordinary to our own eyes,

在我们的肉眼看来， 这面墙或许再普通不过了，

at a trillion frames per second, we can see something truly incredible.

但以每秒一万亿帧的帧率， 我们却能看见神奇的景象。

We can actually see waves of light scattered back from the hidden scene

我们能看见从被遮挡的场景 散射回来的光波

and splashing against the wall.

在墙上溅开。

And each of these waves carries information

每段光波都携带着

将其散射回来的 被隐藏物体的信息。

about the hidden object that sent it.

So we can take these measurements

我们将这些测量结果

and pass them into a reconstruction algorithm

输入到一段能重建图像的算法中，

to then recover the 3D geometry of this hidden scene.

来重现被隐藏场景的 三维几何轮廓。

Now I want to show you one more example of an indoor scene that we captured,

我想再展现一个 捕捉室内场景的例子，

this time with a variety of different hidden objects.

这次被隐藏物体的种类更丰富。

And these objects have different appearances,

这些物体的外观各不相同，

so they reflect light differently.

所以它们反射光线的方式也不同。

For example, this glossy dragon statue reflects light differently

比如，这座富有光泽的龙雕像 与这颗镜面迪斯科球，

than the mirror disco ball

或是这座《掷铁饼者》雕像

or the white discus thrower statue.

反射光线的方式是不同的。

And we can actually see the differences in the reflected light

我们可以把视频帧堆叠在一起，

by visualizing it as this 3D volume,

将被反射的光线 可视化为三维立体，

where we've just taken the video frames and stacked them together.

从而观察到它们之间的区别。

And time here is represented as the depth dimension of this cube.

立方体的宽度则代表了时间。

These bright dots that you see are reflections of light

这些亮点是从迪斯科球的每个镜面

from each of the mirrored facets of the disco ball,

反射回来的光，

scattering against the wall over time.

随时间的推移被分散到 墙面的各个角落。

The bright streaks of light that you see arriving soonest in time

这些到达时间最早的亮条纹

are from the glossy dragon statue that's closest to the wall,

是被离墙面最近的 龙雕塑反射回来的，

and the other streaks of light come from reflections of light from the bookcase

其他的亮条纹则是从书架和

and from the statue.

《掷铁饼者》所反射回来的。

Now, we can also visualize these measurements frame by frame,

我们还可以将这些测量结果 一帧一帧地

as a video,

以录像的形式播放，

to directly see the scattered light.

以更直接地观察 这些被散射回来的光。

And again, here we see, first, reflections of light from the dragon,

这次，同样的， 我们最初看见的光反射是来自

离墙最近的龙雕塑，

closest to the wall,

followed by bright dots from the disco ball

接着是来自迪斯科球的亮点

and other reflections from the bookcase.

和其他来自书架的反射光。

And finally, we see the reflected waves of light from the statue.

最后，我们看见的是被 《掷铁饼者》雕塑反射回的光波。

These waves of light illuminating the wall

这些光波像是仅能绽放

are like fireworks that last for just trillionths of a second.

万亿分之一秒的烟花一样 照亮了这面墙。

And even though these objects reflect light differently,

尽管这些物体 反射光的方式不同，

we can still reconstruct their shapes.

我们依旧能重建它们的形状。

And this is what you can see from around the corner.

这就是你能从 被遮挡的拐角后看到的图像。

Now, I want to show you one more example that's slightly different.

现在，我想再演示一个 稍微不同的例子。

In this video, you see me dressed in this reflective suit

在这段录像中， 我穿着这件能反射光的连体衣，

and our camera system is scanning the wall at a rate of four times every second.

我们的摄像机系统 以每秒四次的速率扫描这面墙。

The suit is reflective,

因为这件连体衣能反射光，

so we can actually capture enough photons

所以尽管摄影机 并不能直接拍到我，

that we can see where I am and what I'm doing,

我们却能通过捕捉足够多的光子

without the camera actually directly imaging me.

来反映出我的位置与动作。

By capturing photons that scatter from the wall to my tracksuit,

通过捕捉从墙面被散射到 连体衣，再被散射回墙面，

back to the wall and back to the camera,

最后回到摄像机的光子，

we can capture this indirect video in real time.

我们能实时地捕捉这段间接录像。

And we think that this type of practical non-line-of-sight imaging

我们认为这种实用的 非视距成像技术

could be useful for applications including for self-driving cars,

能被应用于自动驾驶，

but also for biomedical imaging,

还有生物医学成像这样

where we need to see into the tiny structures of the body.

需要观察身体微小结构的技术。

And perhaps we could also put similar camera systems on the robots

也许我们还能将类似的 摄像机系统安装到

that we send to explore other planets.

被送去探索其他星球的机器人上。

Now you may have heard about seeing around corners before,

也许你曾听说过这种能 观察到拐角另一侧景象的技术，

but what I showed you today would have been impossible

但我今天所展示的内容

两年前还无法实现。

just two years ago.

For example, we can now image large, room-sized hidden scenes outdoors

比如，我们现在能实时地 将室外的房间大小的

大型被遮挡景象可视化，

and at real-time rates,

and we've made significant advancements towards making this a practical technology

我们在提高此技术实用性的方面 也取得了重大进展，

that you could actually see on a car someday.

将来你可能在 汽车上看到这种系统。

But of course, there's still challenges remaining.

当然，我们依旧 面临着诸多挑战。

For example, can we image hidden scenes at long distances

比如，我们能否在使用 低功率，确保对人眼安全的激光

where we're collecting very, very few photons,

采集极少数光子的情况下，

with lasers that are low-power and that are eye-safe.

将远距离的景象可视化？

Or can we create images from photons

或者，不同于 只被反射一次的光子，

that have scattered around many more times

我们能否通过被散射多次的光子

than just a single bounce around the corner?

将景象可视化？

Can we take our prototype system that's, well, currently large and bulky,

我们能否将这种 目前还体积庞大的样机

and miniaturize it into something that could be useful

缩小到能被用于

for biomedical imaging

生物医学成像的大小，

or perhaps a sort of improved home-security system,

或是将其应用于更先进的 家居安防系统，

or can we take this new imaging modality and use it for other applications?

亦或是将这种新型的成像形式 应用于其他领域？

I think it's an exciting new technology

我认为这是一项令人振奋的新技术，

and there could be other things that we haven't thought of yet

我们也许还能将其应用于其它 现在还想象不到的

to use it for.

情景之中。

And so, well, a future with self-driving cars

也许一个自动驾驶汽车 被普及的未来

may seem distant to us now --

在我们看来还很遥远——

we're already developing the technologies

但我们实际上已经在发展

that could make cars safer and more intelligent.

能使汽车更安全， 更智能的科技了。

And with the rapid pace of scientific discovery and innovation,

随着科学创新与发明的高速发展，

you never know what new and exciting capabilities

你永远都想象不到还会有 哪些新型的，令人振奋的技术

could be just around the corner.

出现在不远的拐角处（将来）。 （一语双关）

(Applause)

（掌声）