我正在尝试构建能够识别受过训练的物体的机器人手臂。我知道它可以用opencv来完成。但我不精通 C++ 编程。所以我需要用 C# 来做。首先我使用 Aforge.net 并使用颜色过滤识别对象，但这是一种糟糕的方法。所以我发现 Accord.net 有先进的技术来做到这一点。Haar 特征对象检测就是其中之一。我阅读了 Accord.net 文档。它有一个构造函数
HaarCascade(Int32, Int32,HaarCascadeStage[])
前两个整数是基本宽度和基本高度。他们是什么。什么是阶段？怎么做？
Please take a look on Accord.NET documentation. There is also the accompanying CodeProject article which explains what a stage is and even demonstrates how to create a Haar Object Detector using a Haar Cascade.
Also, please keep in mind that the current implementation can only use existent classifiers to perform the recognition. The framework does not yet offer a way to create your own classifiers at this time. However, note that it can work with any OpenCV-compatible XML definition of a Haar cascade, such as any of the ones which come bundled in OpenCV or even definitions created by other researchers.