使用Kinect V2测量平面的面积 [英] Using Kinect V2 to measure area of flat surface

问题描述

我正在开展一个项目，我需要在其中加入一个面部表面积的测量结果。 Kinect正在关注的平板显示器。关于如何计算表面积的任何想法？

I have a project I'm working on where I need to incorporate the measurement of the surface area of the face of a  flat panel that the Kinect is looking at. Any ideas on how to calculated surface area?

推荐答案

这适用于任何以深度不连续性为界的平面剖面：

This should work for any planar section bounded by a discontinuity of depth:

嗯，你将从一个点云开始，它是使用映射器移动到相机空间的深度点。

Well, you'll start with a point cloud, which is depth points moved into camera space using the mapper.

你必须有一些启发式来决定哪些点在云中属于面板。  如果你可以知道飞机上至少有一个点（比如交互式或者可能要求中心十字准线与面板相交）那么你可以在那时计算
a正常（使用点邻域的线性回归）在云中）。  一旦你在飞机上有一个点和平面的法线，你基本上就有了一个平面方程。  您可以包含/排除
距飞机一定距离的所有点。

You'll have to have some heuristic to decide which points in the cloud belong to the panel.  If you can know at least one point on the plane (like either interactively or maybe requiring that the center crosshairs intersect the panel) then you can compute a normal at that point (using a linear regression on a neighbourhood of points in the cloud).  Once you have a point on the plane and the normal to the plane, you basically have an equation of the plane.   You can include/exclude all points that are within some distance to the plane.

如果您知道飞机上的所有点，那么您可以使用另一个启发式来确定面板（如果已知）或者如果它是未知的，则计算围绕这些云点的凸包（在2D中，在平面内）（当然，使用它们在
中的坐标）。  然后计算凸包多边形的面积。

if you know all points on the plane then you can use another heuristic to determine the shape of the panel (if it is known) or if it is not known, then compute a convex hull (in 2D, within the plane) around those cloud points (using their coordinates in the plane, of course).  Then compute the area of the convex hull polygon.

校准Kinect使得相机空间单位为毫米。

The Kinect is calibrated such that camera space units are millimeters.

我正在做很多填空工作在这里工作，因为你没有详细说明已知或未知的内容。

I'm doing a lot of fill-in-the-blanks work here, because you didn't specify very much about what is known or not known.

对于任何图像处理任务，你应该总是考虑尽可能多的先验信息。  你需要做的越少，越好。

With any image processing task, you should always consider as much a priori information you have as possible.  The less general you need to make it, the better.

这篇关于使用Kinect V2测量平面的面积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！