机器视觉：电脑能说明深度吗？ [英] machine vision:can computer tell depth?

问题描述

是否有人可以编写一个程序来判断某个对象是否落后于另一个？

例如这张照片可以 http://oi33.tinypic.com/14jqjom.jpg [ ^ ]

由程序读取，程序输入黑色物体在红色物体前面？

Is it possible for someone to write a program that tells if an object is behind another?
e.g can this photo http://oi33.tinypic.com/14jqjom.jpg[^]
be read by a program and the program inputs "black object is in front of the red object"?

推荐答案

是的，你可以！

但......这个能力受到一些因素的限制：

1.如果一个物体在另一个物体后面，你无法判断它是一张照片，你的大脑不知道物体（在识别的意义上），例如它们只是随机多边形，所以你和机器（能够实现三维视觉）都无法分辨。但是，一旦这些物体移动，请看看＃3

2.所以要么你的机器有一个数据库来查找像嘿这是汽车的一半，所以其余的必须被隐藏在另一个对象后面 - >车在后台！这甚至可以在2D中工作。

或

3.您的机器必须能够移动头部（只是一点点）以查看是否有更多这些对象变得可见。这是人类用来判断​​的另一个技巧，只能在3d系统上使用。



也许有一些其他的滴答声，但头部/相机移动和物体跟踪可能与对象数据库将完美地解决这个问题 - 理论上！



可能还有其他方法。大约10年前，我玩过intels opencv和电视卡，当涉及到跟踪人类时，电视上的体育是完美的训练材料，因为你可以使用人体的几何形状，经编颜色甚至公共面孔（免费和公开的）作为培训材料。而且，这些活动几乎都是以相同的距离，角度等拍摄的。所以完美的条件！



玩得开心

yes, you can!
but... this capability is limited by some factors:
1. you cannot tell if an object is behind another object if it is a photo and the objects are unknown to your brain (in the sense of recognition) e.g. they are just random polygons, so neither you nor a machine (capable of 3d vision) can tell. But, as soon as these objects are moving take a look at #3
2. so either your machine has a database to look up objects like "hey this is half of a car, so the rest must be gidden behind another object" -> car is in background! This works even in 2D.
or
3. your machine must be capable of moving the "head" (just a bit) to see if some more of any of these objects gets visible. that's another trick humans use to tell and is only possible on a 3d system.

maybe there are some other ticks but head/camera movement and object tracking maybe combined with an object database would perfectly solfe this problem - in theory!

there may be other methods. about 10 year ago I played around with intels opencv and a tv card, and when it comes to tracking humans, sports on tv are perfect "training material" because you can use the geometry of the human body, tricot colors or even public faces (free and public available) as training material. and, these events are all filmed nearly at same distance, angle, etc. so perfect conditions!

have fun

这篇关于机器视觉：电脑能说明深度吗？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！