基于深度图像的不规则零件检索技术研究

Research on Irregular Part Retrieval Based on Depth Images

提出了一种基于深度图像的三维零件模型的检索技术,通过比较模型在三个方向上的投影图像来进行零件本身的相似距离计算。该检索算法首先对零件模型进行方位规范化,然后将其投影到三个正交平面上,对投影图像利用不变矩和极半径傅里叶变换分别提取图像特征,其联合特征向量可以很好地进行图像比较,最后加权三个图像之间的相似距离得到三维模型的相似距离。该算法充分利用了机械零件自身的特点,有较好的检索性能。

A method for retrieval of 3D model of components was proposed based on depth images, in which the similarity between two components was computed through comparisons of their cotresponding projected images from three directions, This retrieval algorithm proceeded first on orientation normalization, then the model was projected on three orthogonal planes, afterward the feature vector which can differentiates images well was extracted with the way of Moment Invariants and Polar Radius Fourier Transform, and at last the similarity between two models was obtained from the three similarities of images with different weights. This method adequately utilizes the characteristics of mechanical components, so that it has good retrieval performance.