Current Issue Cover


摘 要
目的 目标建模是机器视觉领域的主要研究方向之一,主动目标建模是在保证建模完整度的情况下,通过有计划地调节相机的位姿参数,以更少的视点和更短的运动路径实现目标建模的智能感知方法。为了反映主动目标建模的研究现状和最新进展,梳理分析了2004年以来的相关文献,对国内外研究方法做出概括性总结。方法 以重构模型类型和规划视点所用信息作为划分依据,将无模型的主动目标建模方法分为基于表面的主动目标建模方法、基于搜索的目标建模方法和两者相结合的方法三大类,重点对前两类方法进行综述。结果 首先解释了每类方法的基本思想,然后总结每类方法中涉及到的问题,最后对相关问题的主要研究方法进行归纳和分析。将各个问题的解决方法进行合理的搭配组合,形成不同的主动目标建模方法。 结论 目前对主动目标建模的研究已经取得了一定进展,但是建模过程的精度和效率仍有较大的提升空间。给出了其他可行的研究方向,为本方向的研究提供参考。
Active Geometric Reconstruction Methods for Object:A Survey

kong yan zi,zhu feng,hao ying ming,wu qing xiao,lu rong rong(Shenyang Institute of Automation, Chinese Academy of Sciences)

Objective Target modeling is one of the main research directions in the field of machine vision. When modeling the geometry of an object, the data obtained from one viewpoint is often incomplete, and even large-area losts may occur. Therefore, it is necessary to obtain the information of the target from different viewpoints and fuse the information to achieve the complete geometric modeling of the target. Active object reconstruction is an intelligent perception method that achieves target modeling with fewer viewpoints and shorter motion paths by systematically adjusting the pose parameters of the camera while ensuring the integrity of the model. In order to reflect the research status and the latest development of active object reconstruction, the relevant literature since 2004 has been combed and analyzed, and a summary of domestic and foreign research methods has been made. Method At present, active object reconstruction is mainly aimed at two types of task: model-based active object reconstruction and non-model active object reconstruction. Based on the rebuilt model type and the information used during view planning, the model-independent active object reconstruction methods are divided into surface-based methods, search-based methods and the combined methods. The surface-based methods use point cloud models, triangular patch models, etc. It extracts shape information from the local model that has been obtained so far, and classifies the shape of the unknown region to determine the next viewpoint. The search-based methods use voxel models. A certain method is used to determine the candidate viewpoints, and then these viewpoints are scored by a reasonable evaluation function. The candidate viewpoint with the highest score is used as the next best view. The combined method uses both the surface model and the voxel model and it utilizes the advantages of the two methods comprehensively to provide more effective information for view planning. But the combined method has not obtained much research recently and the first two methods are mainly focused on. Result The basic ideas of each type of method are explained and problems involved in each type of method are summarized later. Surface-based methods contain problems of detection direction determination, unknown surface prediction and the next best viewpoint determination. Search-based methods involve problems of model type selection and search space determinition, undetected area prediction and the design of the evaluation function to sort the candidate viewpoints. Finally, the main research methods of those related problems are summarized and analyzed. The solutions to each problem are combined reasonably to form different active object reconstruction methods. Conclusion Researchers who study active object reconstruction have made some progress at present, but there is still much room for improvement in the accuracy and efficiency of active reconstruction. Other feasible research directions are given in the end, which provide reference for the future research in this direction.