Enhancing Depth maps from Microsoft Kinect using multiple Sensors
Enhancing Depth maps from Microsoft Kinect using multiple Sensors
复制标题
使用多个传感器增强 Microsoft Kinect 的深度图
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
Kwan H. Lee
中科院分区：
文献类型：
--
作者：
Bilal Ahmed;I. Jang;J. Yoo;Kwan H. Lee
近年来，市场上出现了一种新型的深度传感器，称为微软Kinect传感器。其主动传感技术非常适合可能需要更高帧速率的应用。然而，在对象吸收或反射由传感器投射的红外图案（即镜面或透明以及近或远的对象）的情况下，它可能完全失败。Kinect深度输出在这些对象存在的区域中包含孔。在本文中，我们制定了一个框架，以克服这一缺点。我们使用了两个Kinect传感器，它们被放置成具有不同的视角，与两个Kinect RGB摄像头协调，形成立体对。由于Kinect提供了关于不可靠深度的信息，因此仅在必要时选择性地在这些区域进行增强。通过这种方式，两个单独放置的深度传感器由于其不同的视角而能够提供一些增强。此外，我们首先对齐深度图像，并寻找仍然需要增强的区域。因此，可以使用立体匹配来改善这些区域。使用自适应算法进行立体匹配为我们提供了更快的可能性，以获得更好的深度图。
During recent years a new type of depth sensors was made available in the market known as Microsoft Kinect Sensor. Its active sensing technology is well suited for applications that may require higher frame rates. However, it can totally fail in situations where the subject absorbs or reflects the infrared pattern casted by the sensor i.e. specular or transparent and near or far away objects. Kinect depth output contains holes in areas where these objects exist. In this paper we formulate a framework to overcome this shortcoming. We utilized two Kinect sensors placed to have a different viewing angle in coordination with two Kinect RGB cameras forming a stereo pair. As Kinect provides information about unreliable depths, therefore enhancements were performed selectively only in those areas where necessary. In this way, two separately placed depth sensors are able to offer some enhancements due to their different viewing angles. In addition, we first align depth images and look for the areas that still require enhancements. Thus, those areas can be improved using stereo matching. Using adaptive algorithms for stereo matching provides us with faster possibilities to come up with a better depth map.