Real time shape recognition by fusing range images and intensity images

本文提出了融合距离图像和强度图像的方法。融合被认为是实现智能机器人系统三维形状识别的重要技术。通过融合，可以构建鲁棒实用的三维形状识别方法。首先，讨论了每幅图像的特征。选取平面和圆柱区域作为距离图像提取的特征。这是因为距离图像通常是稀疏的，很难从距离图像中鲁棒地提取复杂的特征，而且这些特征通常包含在真实物体中。对于强度图像，选择边缘信息。利用这些特征，提出并制定了以下融合方法：(1)从距离图像中提取平面和圆柱形区域作为融合的预处理；(2)通过融合平面区域和边缘来测量三维边缘；(3)通过融合边缘来更新圆柱形区域的参数。它们是用最小二乘法表述的。将融合方法应用于多机器人分布式感知问题。假设每个机器人都有一个距离图像传感器和/或强度图像传感器。应用所提出的融合方法测量三维特征。此外，利用融合结果对各机器人的航位推算进行校正。

This study proposes methods to fuse range images and intensity images. The fusion is thought to be an important technology for 3D shape recognition of an intelligent robot system. By the fusion, robust and practical 3D shape recognition methods can be constructed.Firstly, characteristics of each image is discussed. Planar and cylindrical regions are selected as the features extracted from a range image. This is because a range image is usually sparse and complex features are difficult to extract robustly from a range image, and additionally, these features are often included in real objects. For an intensity image, edge information is chosen.. Using the features, following methods of fusion are proposed and formulated : (1) extraction of planar and cylindrical regions from a range image as the preprocess of the fusion, (2) measurement of 3D edges by fusing a planar region and an edge, (3) update of parameters of a cylindrical region by fusing an edge. They are formulated using least square approach.The fusion methods are applied to the problem of distributed sensing of multiple robots. Each robot is assumed to have a range image sensor and/or an intensity image sensor. 3D features are measured applying the proposed fusion methods. And additionally, dead-reckoning of each robot is corrected using the fused results.