基于形状相似的三维模型检索方法研究

RESEARCH ON SHAPE SIMILARITY-BASED 3D MODEL RETRIEVAL

在基于内容的三维模型检索方法中,基于形状的检索方法由于其视觉上的直观性和效果上的稳定性,得到了广泛的研究,本文提出了一种基于模型视觉投影图相似的形状检索方法,对任意模型通过坐标规范化,方向投影,提取其二维投影图像的Fourier描述符和Zernike矩描述符作为模型特征,在数据库中进行匹配检索.解决了该类方法中的投影图最优匹配的问题,并在检索效果和效率上取得了很好的结果.

In content-based model retrieval, visual similarity-based retrieval is widely used because of its visual intuitiveness and robust effect. In this paper, visual similarityd retrieval based on 2D model projective image was introduced. For each model, coordinate standardization, direction projection to a series of projective images were used, two kinds of image features were then used as the model feature： Fourier descriptor and Zernike moments. Coding the two features in an appropriate way, model feature can be used in content-based model retrieval system, reducing considerable time in finding best-matched images. Extraordinary results were obtained regarding both precision and effects.