基于局部视图的三维树木模型递进检索方法

Progressive Retrieval Method for 3D Tree Model Based on Local View

针对现有的三维模型检索方法对树木模型存在着准确性低、语义不相符、检索速度慢的问题,为了提高三维树木模型的检索效率和准确性,提出一种基于局部视图的三维树木模型递进检索方法.该方法采用多层次的局部视图对树木模型进行组合描述,通过简洁高效的形式突出了树木对象的树冠轮廓、枝干骨架等关键特征;然后结合视图信息对应的树木语义实现了视图特征的提取转换,强化了描述特征对树木模型的区分性;最后采用层次递进的方式组织视图特征实现了由粗略到精细的多阶段检索过程,并进一步采用动态索引机制以提升模型的检索效率.分别在通用模型与树木模型集合中进行了检索实验的结果表明,该方法能够明显地提升树木类模型的检索准确性,有助于三维树木模型的快速检索和复用.

Retrieving 3D tree model using generic retrieval methods often encounter problems such as poor accuracy, semantic mismatch, and low efficiency. To overcome these problems and improve accuracy and efficiency of 3D tree model's retrieval, a progressive retrieval method for 3D tree model based on local view is proposed. This method adopts hierarchy local view to describe tree model, and highlights the important features of tree model(e. g. crown contour, branches skeleton) in a simple and efficient form. Then, tree semantics is introduced to extract view features, which makes view feature be distinguished more easily for different tree models. At last, view feature is adopted in the multi-stage retrieval process that achieves a coarse-to-fine retrieval for 3D tree models, and dynamic index is used to speed up the retrieval process. Experimental results show that this method achieves better performance in tree model's retrieval both for general model benchmark and tree model benchmark, thus it make 3D tree model resource be reused in an easy and efficient way.