在 iPhone 上实现手写字符识别的最佳方法是什么？
在 iPhone 中实现手写字符识别的最佳方法是什么？
我的想法如下：
首先，拍一张手写字的照片。
其次，识别手写字符的图像，然后输出文本。
What is the best way to implement handwritten character recognition in iPhone?
I'm thinking as follows:
First, taking a picture of handwritten character.
Second, recognizing image of handwritten character, and then outputs a text.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
您可以使用神经网络，因为神经网络在分类方面很好，但答案取决于概率。
但是，如果您可以识别角色中的某种逻辑，则可以使用模糊逻辑，因为这样答案将比神经网络更准确！
you can use neural networks because neural networks are good in classification but the answer depends on probability..
But if you can identify a certain logic in your characters, you can use fuzzy logic because then the answer will be more accurate than in neural networks!!!
问题出在步骤2：
“2.其次，识别手写字符的图像，然后输出文本。”
这不是图像处理中一个完全解决的问题。事实上，说“没有完全解决”是表明问题非常困难的一种令人愉快的方式。说“使用神经网络”或“尝试遗传算法”是一回事，而实际实现一些可行的东西，并且会超越当前的技术水平，则是另一回事。
首先，我建议从头到尾阅读以下书籍：
字符识别系统，作者：Cheriet、Kharma、Liu 和 Suen。
这是我通常推荐的书，因为它相当最新，它涵盖了各种各样的技术，而且它也相当清楚地表明这不是一个已解决的问题。邦克有一本更古老、更昂贵的书，也是一个很好的资源。
手写识别有部分解决方案。如果您可以选择一个非常具体的应用程序，例如识别支票上写的数字，那么您就可以稍微简化问题。如今，许多 ATM 机都具有良好的 OCR 功能来识别支票金额，但同样，这是一个高度受限的问题。
如果您研究神经网络，那么我希望您在训练完成后永远不必对其进行调试。
The problem is in step 2:
"2. Second, recognizing image of handwritten character, and then outputs a text."
This is not a completely solved problem in image processing. In fact, saying "not completely solved" is a pleasant way of indicating that the problem is insanely hard. It's one thing to say "use a neural net" or "try a genetic algorithm," and another to actually implement something that works, and that would outperform the current state of the art.
As a start, I would recommend reading the following book from cover to cover:
Character Recognition Systems by Cheriet, Kharma, Liu, and Suen
It's the book I usually recommend because it's reasonably up to date, it covers a large variety of techniques, and it also makes fairly clear that this is not a solved problem. There is an older and much more expensive book by Bunke that is also a great resource.
There are partial solutions to handwriting recognition. If you can pick a very specific application, such as recognizing numerals written on a check, then you simplify the problem a bit. Many ATMs these days have good OCR to recognize the amount of a check, but again, this is a highly constrained problem.
If you investigate neural networks, then I hope you never have to debug one once it's trained.