防止纹理扭曲的倾斜摄影模型轻量化算法
Lightweight algorithm for preventing texture distortion in oblique photogrammetry model simplification
王彦海 1张宇昊 1李成 2陈树平 2李恩阳1
作者信息
- 1. 三峡大学电气与新能源学院 宜昌 443000
- 2. 中国南方电网超高压输电公司梧州局 梧州 543002
- 折叠
摘要
针对倾斜摄影模型网格简化后出现纹理扭曲的问题,基于二次误差度量算法(quadric error metrics,QEM),提出一种防止纹理扭曲的倾斜摄影模型轻量化算法.该算法依据模型中各顶点的拓扑连接关系对顶点进行分类,对不同类型的顶点制定相应的折叠策略,并引入顶点起伏度因子、纹理形变因子和顶点一环领域内三角形平均面积对QEM算法中的边折叠代价进行改进.实验结果表明,改进算法可以在高简化率下避免出现纹理扭曲现象,有效保持了模型的几何特征和纹理特征,并且相较于传统QEM 算法,改进算法处理的模型最大误差、平均误差和均方差均分别至少降低了43.55%、52.50%和21.61%;在90%简化率的情况下,相较于带纹理模型的简化算法,改进算法处理的模型最大误差、平均误差和均方差分别降低了34.90%、10.61%和12.31%.
Abstract
In response to the issue of texture distortion that arises after mesh simplification in oblique photogrammetry models,a lightweight algorithm is proposed to prevent texture distortion based on the quadratic error metric(QEM)algorithm.This algorithm categorizes vertices based on their topological connectivity within the model,and then formulates appropriate folding strategies for different vertex types.It also introduces vertex fluctuation factor,texture deformation factor,and average triangle area within a one-ring neighborhood of each vertex to enhance the edge folding cost in the QEM algorithm.Experimental results show that the improved algorithm effectively avoids texture distortion phenomena at high simplification rates,preserving both the geometric and texture features of the model.The improved algorithm reduces the maximum error,average error,and mean square error of the processed model by at least 43.55%,52.50%and 21.61%,respectively,compared to the QEM algorithm.At a simplification rate of 90%,the improved algorithm reduces the maximum error,average error,and mean square error of the processed model by 34.90%,10.61%and 12.31%,respectively,compared to the simplification algorithm for textured models.
关键词
倾斜摄影模型/二次误差度量算法/网格简化/纹理扭曲/模型轻量化Key words
oblique photogrammetry/quadric error metrics/mesh simplification/texture distortion/model lightweight引用本文复制引用
基金项目
国家自然科学基金(U22A20600)
国家自然科学基金(52079070)
防灾减灾湖北省重点实验室开放资金(2022KJZ07)
出版年
2024
NETL
NSTL
维普
万方数据