项目名称: 复杂场景点线光流三维重建模型的建立及鲁棒性分析
项目编号: No.61462062
项目类型: 地区科学基金项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 陈震
作者单位: 南昌航空大学
项目金额: 47万元
中文摘要: 基于光流的三维重建技术是计算机视觉研究中的热点问题。现有的光流重建算法多是基于点光流或线光流的单模态方法,其重建结果常常无法全面地描述目标物体的宏观结构和表面细节信息。 本项目主要研究基于点线光流的三维重建算法。拟在复杂场景下,通过成像原理的理论与实验分析,分别定义点光流和直线光流的概念和表达形式;在透视投影模型下,描述二维点、直线的运动与其三维运动和结构之间的相互几何关系,构造基于点线光流的三维重建模型;分别使用数学方法和实验方法对构造的三维重建模型进行鲁棒性证明与分析;选择快速和优化的求解方法,确定目标物体相互遮挡情况下的计算策略;设计客观的评价体系,对三维重建模型和求解方法进行定量比较和定性分析。 本项目的研究将有助于推动光流重建技术的深入发展,使得一个重建算法既能得到目标物体的表面细节,又能重建目标物体的拓扑结构和运动,并能有效地保证重建算法的鲁棒性和计算精度。
中文关键词: 点线光流;三维重建;模型;鲁棒分析;复杂场景
英文摘要: Reconstructing the 3D structure from the image optical flow is the major area in computer vision. The most existed reconstruction approaches are belong to the single-mode method based on point optical flow or line optical flow, which could not describe both the macroscopical structure and the surface detail of the target object overall. The main research of the project is studying the 3D reconstruction strategy which combined the point and line optical flow. Firstly, it defines the the concept and the expression of the point and line optical flow respectively through the theoretical and experimental analysis of the imaging principle in the complex scene. Secondly, it descibes the mutual geometry relationship between the 2D motion and the 3D motion and structure of the point and line with the perspective projection model, constructs the 3D reconstruction model based on the point and line optical flow. Thirdly, it proves and analyzes the robustness of the 3D reconstruction model by the mathematics method and experimental method respectively. Then, it chooses the fast and optimized algorithm to confirm the computing strategy when there are the motion covered problems. Finally, it designs an objective evaluation system to compare the presented approach with the existed methods quantitatively and to analyze them qualitatively. This project would be helpful to promote the further development of the 3D reconstruction technology based on optical flow. The multi-mode method could make the 3D reconstruction algorithm obtain surface details, and also reconstruct the topological structure and motion of the objection, and could ensure the robustness and accuracy of the reconstruction algorithm effectively.
英文关键词: Point-line optical flow;3D reconstruction;Model;Robustness analysis;Complex scene