图像目标检测问题
我即将开始一个不属于任何课程的大学项目。我的任务是识别任意图像上的对象（例如图像中的圆圈数量）。我将主要处理 2D 图像。我对此类图像处理没有太多经验，想知道是否有人可以指导从哪里开始。
我最初的谷歌搜索给了我一些可以构建的关键字，例如 OpenCV。我有一些关于信号和系统的课程（特别是傅里叶变换），很多计算机科学相关的课程（数据结构，算法等），数学课程（微积分，线性代数等），等等。
有人可以给我一个快速指南，告诉我要寻找和学习什么样的东西吗？任何形式的指导将不胜感激。谢谢！
I am about to start on a university project that is not part of any courses. I was tasked with identifying objects on a arbitrary image (say number of circles in a image). I will be working mostly with 2D images. I don't have much experience with such image processing and was wondering if someone could give a guidance on where to start.
My initial google search gave me some keywords to build upon, such as OpenCV. I have some courses regarding signals and systems (fourier transform in particular), lots of computer science related courses (data structures, algorithms, etc), math courses (calculus, linear algebra, etc), and others.
Can someone please give me a quick guide on what kind of things to look for and learn? Any kind of guidance would be greatly appreciated. Thanks!
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（3）
机器视觉图像处理是一个极其复杂的领域。我推荐 Milan Sonka 的《图像处理、分析和机器视觉》一书。这是一本昂贵的书，但其独特之处在于面向实用算法。
Machine vision image processing is an extremely complex area. I recommend the book Image Processing, Analysis, and Machine Vision by Milan Sonka. An expensive tome, but unique in being oriented toward practical algorithms.
阅读 Richard Szeleski 的新书 (PDF 草稿可用），了解当前技术的概述。
此处还提供了简短会议课程的材料。
Read chapter 14 of Richard Szeleski's new book (PDF draft available) for an overview of current techniques.
There's also material from a short conference course here.
一些简单的物体（例如圆圈）可以使用霍夫变换来检测。
如果您需要一些更高级的对象检测，您可以使用 Viola-Jones 算法（请参阅 OpenCV 或 Rapidminer http://spl.utko.feec.vutbr.cz/en/image-processing-extension-for-rapidminer-5）或定向梯度直方图（是行星到 http://spl.utko.feec。 vutbr.cz/en/image-processing-extension-for-rapidminer-5)
Some simple object such as Circles can be detected using Hough transform.
If you need some more advanced object detection you can use Viola-Jones algorithm (see e.g. OpenCV or Rapidminer http://spl.utko.feec.vutbr.cz/en/image-processing-extension-for-rapidminer-5) or Histogram of Oriented Gradients (is planet to http://spl.utko.feec.vutbr.cz/en/image-processing-extension-for-rapidminer-5)