摘要
由于自然或人为因素,文物经常以破损的碎块形式呈现,将诸多不规则碎块准确拼接使文物复原是一项耗时费力的工作。为此,该文提出一种基于断裂面信息的文物碎块自动拼接方法,该方法包含匹配和配准两个阶段。第一阶段,根据断裂面轮廓线分割出断裂面,基于快速点特征直方图搜索匹配点对,并调整轮廓线上点的权重,得到匹配关系。第二阶段,提出一种由粗到细的配准策略,采用基于主成分分析(PCA)的粗配准方法获得初始位置估计,然后应用深度最近点神经网络(DCP)做进一步调整。实验结果表明:该文配准方法的配准成功率较其子方法分别提升了2.22%和18.06%,平均绝对误差仅为0.9202 mm,能够应对轮廓线破损情况,完成断裂面较为完整的文物碎块拼接。
Due to the natural or human factors,cultural relics are often presented in incomplete fragments and it is a time-consuming and laborious task to accurately assembly many irregular fragments together to restore the original appearance of cultural relics.Therefore,an automatic reassembly method for cultural relics based on fracture surface information is proposed,which includes two stages of matching and registration.In the first stage,the fracture surface is segmented according to the contour curve,the matching point pair is searched based on the fast point feature histograms(FPFH)feature,and the contribution of the points on the contour line is adjusted to obtain the matching relationship.In the second stage,a coarse-to-fine registration strategy is proposed.The rough registration method based on principal component analysis(PCA)is used to obtain the initial position estimate,and then the deep closest point(DCP)neural network is used for further adjustment.Experimental results show that the rough success rate of the proposed registration method is 2.22%and 18.06%higher than those of the sub-methods,respectively,and the average absolute error is only 0.9292 mm.It can deal with the damage of the contour curve and complete the splicing of cultural relic fragments when the fracture surface is relatively complete.
作者
耿国华
姚文敏
周明全
刘杰
徐雪丽
曹欣
刘阳洋
李康
GENG Guohua;YAO Wenmin;ZHOU Mingquan;LIU Jie;XU Xueli;CAO Xin;LIU Yangyang;LI Kang(College of Information Science and Technology, Northwest University, Xi′an 710127, China;National and Local Joint Engineering Research Center for Cultural Heritage Digitization, Northwest University, Xi′an 710127, China;Virtual Reality Application Engineering Research Center of the Ministry of Education, Beijing Normal University, Beijing 100875, China;College of Mathematics and Computer Science, Yan′an University, Yan′an 716000, China)
出处
《西北大学学报（自然科学版）》
CAS
CSCD
北大核心
2021年第3期397-403,共7页
Journal of Northwest University（Natural Science Edition）
基金
国家自然科学基金重点项目(61731015)
国家重点研发计划(2019YFC1521103)
陕西省重点产业链项目(2019ZDLSF07-02)
国家自然科学基金青年项目(61701403)。
关键词
断裂面
轮廓线
快速点特征直方图
主成分分析
深度最近点
fracture surface
contour curve
fast point feature histograms
principal component analysis
deep closest point
