基于MVS序列图像的油菜三维模型  

3D modeling of rapeseed based on MVS sequence image

作　　者：武静雯 薛新宇 秦维彩[2] 崔龙飞[2] 张宋超[2] WU Jingwen;XUE Xinyu;QIN Weicai;CUI Longfei;ZHANG Songchao(School of Engineering,Anhui Agricultural University,Hefei 230036;Nanjing Institute of Agricultural Mechanization Ministry of Agriculture,Nanjing 210014)

机构地区：[1]安徽农业大学工学院,合肥230036 [2]农业部南京农业机械化研究所,南京210014

出　　处：《安徽农业大学学报》2019年第6期1083-1088,共6页Journal of Anhui Agricultural University

基　　金：国家重点研发计划“智能农机装备”项目(2017YFD0701000);国家农业现代化产业技术体系建设专项(CARS-12)共同资助。

摘　　要：为实现油菜作物模型的可视化研究,给油菜作物的数字化管理提供数据基础,以感染虫害的苗期油菜为研究对象,采用MVS序列图像技术,搭建MVS技术的序列图像采集平台。根据SFM和PMVS算法获得虫害油菜的稀疏点云数据和稠密点云数据,同时,探索序列图像数量对于特征点匹配的影响。对MVS序列图像技术获得的虫害油菜三维点云数据,采用滤波、精简、Alpha-Shape曲面重建等处理,得到虫害油菜的三维形态曲面模型。结果显示,使用图像数目多和8邻域匹配两者相结合的方法可以又快又好地匹配图像特征点;在获得合适的Alpha值情况下,Alpha-Shape算法可以真实形象地表现出虫害油菜的生长状态。In order to realize the visualization research of rape crop model and provide the data basis for the digital management of rape crop, this paper took the seedling rape infected with pests as the research object and built a sequence image acquisition platform based on MVS technology by adopting MVS-based sequence image technology. Sparse point cloud data and dense point cloud data of pests rapeseed were obtained according to SFM and PMVS algorithms. Meanwhile, the influence of sequence image number on feature point matching was explored. The 3D point cloud data of pests rapeseed obtained based on MVS sequence image technology were processed by filtering, simplification and alpha-shape surface reconstruction to obtain the 3D morphological surface model of pests rapeseed. The results show that the method of combining multiple images and 8 neighborhood matching can match image feature points quickly and well. Under the condition of obtaining the appropriate Alpha value, the alpha-shape algorithm can present the growth status of pests rapeseed in a real image.

关 键 词：MVS 序列图像 虫害油菜 三维模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] US565.405[自动化与计算机技术—计算机科学与技术]