基于距离面积特征的农业机械三维模型相似性评价

Similarity Assessment of Agricultural Machinery 3-D Model Based on Distance-Area

为更好地实现农业机械CAD模型的设计重用,针对经典形状分布算法在特定CAD模型领域检索性能的不足,提出一种基于三维CAD模型距离-面积统计特征的模型相似性评价方法。该算法首先在网格化的三角网格模型表面随机取点,取点过程中为保证随机取点均匀性,采用准随机数发生器Halton-Sequence产生随机数;然后计算任意两点形成有向线段距离以及模型所有面面积之间的比值,同时以距离、面积比为坐标轴构建网格坐标系,统计出现在相应网格中特征点的频次,形成模型的距离-面积特征分布矩阵;最终以Manhattan距离评价分布矩阵的差异程度,从而实现不同CAD模型相似性比较。将面积这一重要几何特征纳入评价算法中,针对相同输入模型,所提算法的检索结果中较为相近、相关的模型数量比文中对比的算法高,算法特征平均提取时间也相对较少。试验结果综合表明,所提算法能够较好地实现农业机械CAD模型相似性评价,在提升时间性能基础上,算法检索性能有较大提高。

Aimed at the retrieval performance deficiency of the classic shape distribution algorithm in the field of professional CAD models,and to achieve the reuse of agricultural machinery CAD models better,a 3-D CAD model retrieval algorithm based on distance and area distributions was proposed to achieve the model similarity assessment. Firstly,hundreds of points were selected randomly on the grid of the triangular mesh surface,during which we adopted the quasi-random number generator and HaltonSequence to generate random points to ensure the uniformity of corresponding points; then corresponding distance between two random points as well as area ratios between each surface were calculated,and the frequency of specific points was computed in the distance-area planar grid,forming a distance-area distribution matrix of a model; finally,we adopted the Manhattan to make an assessment about different matrixes to achieve the comparison of different CAD models. Area distribution is one of the essential characteristics of models. The innovation of this paper is that we put this factor and area feature into consideration. And the descriptor was distance feature with a fusion of area to implement the accurate depiction. Accuracy was higher compared with that of other methods with respect to same retrieval parts.Halton-Sequence was adopted to generate uniform random numbers and the sampling points can be reduced significantly due to this. Thus,efficiency was acceptable. Above all,test results show that thealgorithm is better in the field of agricultural CAD models compared with other shape distribution algorithms for the similarity assessment.