【独家发布】拼图的形状识别和匹配

llb_321

1708

收藏 2021-08-24

悬赏 700 个论坛币已解决

问题：将一张纸随便撕成多块，随机打乱后平铺在一起形成图片，通过形状识别（小块碎纸的边缘识别）、匹配和矩阵变换，完成所有碎纸的位置调整及复原。
目前可以实现碎纸的外缘图形数据提取。

请教：如何在多个平面图形的外缘数据中，判断出形状一致的边，然后通过旋转和平移的矩阵变换（本案不涉及缩放和剪切）实现复原。重点是如何找到可以吻合的边缘，这里需要理解的是，两个图形只是某段边缘可以吻合，而这两个图形是不一样的，所以不能简单地通过图形相似度判断。

换个说法，如何通过拼图的形状（而非内容）实现自动拼图。

论坛币所剩无几，各位大侠别嫌少啊。

最佳答案

owenqi 查看完整内容

你有查过相关的文献吗，这个我感觉并不属于统计方向的问题。

扫码加我拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

全部回复

owenqi

2021-8-24 09:00:36

你有查过相关的文献吗，这个我感觉并不属于统计方向的问题。

https://www.abtosoftware.com/blog/computer-vision-powers-automatic-jigsaw-puzzle-solver

https://roy-zohar-40.medium.com/ ... part-1-35e8ea26259e

https://github.com/Kawaboongawa/Zolver

扫码加我拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

llb_321

2021-8-31 15:04:19

owenqi 发表于 2021-8-31 09:44
你有查过相关的文献吗，这个我感觉并不属于统计方向的问题。

Thanks a lot. The works they described in your suggestions are quite helpful.

You said it. it's merely an issue of personal interest. The idea of solving jigsaw puzzle using R or whatever programming language came into being while my daughter played the game the other day.
Actually, I've got an idea to segment the outer shape lines by finding the convex points, which is the key of the whole solution.

Since the jigsaw puzzle pieces are of high-precision conformed shape, the issue is not that hard to solve.

And by now, it is of my concern how to find identical but disordered order out of big data chaos.
I just wonder if there is any method of solving a precision problem in a fuzzy way.