基于特征轮廓的灰度图像定位三维物体方法

讨论了一种基于特征轮廓的从三维灰度图像确定三维物体位置和姿态的方法，该方法首先建立物体的三维网页模型，检测模型上的特征点，并建立该物体的特征轮廓模型，然后检测输入图像中物体上的特征点，形成特征轮廓，并与特征轮廓模型相匹配，就可得到该物体在三维空间中的姿态；最后使用最小二乘法对物体进行精确定位，实验证明，该方法在物体遮挡情况下不是很严重时，可以快速精确地从灰度图像定位三维物体。

FEATURE-OUTLINE-BASED 3D OBJECT LOCALIZATION FROM AN INTENSITY IMAGE

Discussed in this paper is a fast algorithm to localize 3D object from an intensity image in all six degrees of freedom. The feature points distributed on the 3D model surface of an object are detected to create a feature-outline model for fast alignment between the 3D model and its real intensity image. The localization algorithm involves detecting feature points in a real image, creating feature-outline from the feature points, searching for candidates in the feature-outline model, verifying the candidates by matching corresponding feature points, and refining localization with least-squared algorithm. The experiment results show that the localization algorithm can localizes 3D object from intensity image fast and precisely when occlusion is not very serious.