边缘滤波后的多边形匹配 [英] Polygon matching after edge filtering

问题描述

基本上我要做的就是检查两张图片之间的区别。

现在我已经完成了4方式已经（他们给我一个相当不错的估计值）我正在努力做另一个第五种方式。 （这只是一个有趣的项目，我试图学习它atm）



所以这就是我所拥有的：

< a href =http://img809.imageshack.us/img809/5138/edgecomp.png> http://img809.imageshack.us/img809/5138/edgecomp.png [ ^ ]



这是同一部手机的2张不同图片。



现在我想做的是比较这两个形状/多边形或整个图像是否匹配。现在我可以逐个像素地比较它，但很快你就会意识到为什么这不是一种可行的方式，因为如果我拍摄同一部手机的照片但是从另一个高度（和/或角度）手机的照片边将绘制略微不同的多边形。我尝试比较每个多边形的点数但没有任何成功。



有什么想法吗？

解决方案

如果你有一个连续线你可以使用A *搜索算法替换它，这样你会找到任何一个空行：

WPF：A * search [ ^ ]



然后你可以使用Douglas-Peucker或simular算法来简化这条线：

Polyline Simplification [ ^ ]



现在它只是一个找到类似的形状。



PS：假设图片是Soble或Canny边缘检测？

Hi,

Basically what I am trying to do is check the difference between two images.
Now I have done it 4 ways already (and they give me a pretty good estimated value) I am struggling to do it another fifth way. (it is just a fun project, I am trying to "learn" it atm)

So here is what I have:
http://img809.imageshack.us/img809/5138/edgecomp.png[^]

These are 2 different pictures of the same cellphone.

Now what I would like to do is to compare these two "shapes"/"polygons" or the whole images if they match. Now I could just compare it pixel by pixel but soon you will realize why that''s not a way to go, because if I take the photo of the same phone but from another height (and/or angle) the phone''s edges will draw a slightly different polygon. I tried comparing the number of points of each polygon but without any success.

Any ideas?

解决方案

If you have a continuos line you could replace it by using an A* search algorithm, That way you would find any hole lines:
WPF: A* search[^]

Then you could simplify the line by using Douglas-Peucker or simular algorithm:
Polyline Simplification[^]

Now it would just be a metter of finding simular shapes.

PS: Assuming the pictures are Soble or Canny edge detection?

这篇关于边缘滤波后的多边形匹配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！