基于计算机视觉的Affordance理解研究综述  

Survey of Affordance Understanding Based on Computer Vision

作　　者：信建佳 王立春 尹宝才 XIN Jianjia;WANG Lichun;YIN Baocai(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;Beijing Key Laboratory of Multimedia and Intelligent Software Technology,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部,北京100124 [2]多媒体与智能软件技术北京市重点实验室,北京100124

出　　处：《北京工业大学学报》2024年第7期872-882,共11页Journal of Beijing University of Technology

基　　金：国家自然科学基金资助项目(61876012)。

摘　　要：利用基于计算机视觉的Affordance理解研究行为者和周围环境之间的交互属性,对指导机器人导航、抓取具有重要意义。因此,全面、深入地综述了基于计算机视觉的Affordance理解研究现状。首先,对近年来提出的方法依据研究方向进行归类,综述不同方法的思路和特点;然后,对多个常用的公开数据集进行介绍,并对不同方法在这些数据集上的性能进行对比分析;最后,阐述基于计算机视觉的Affordance理解各类方法的优势与不足及未来的发展趋势。The interaction properties between actors and the surrounding environment are studied by Affordance understanding based on computer vision,which is significant in the fields of robot navigation and grasping.Therefore,the current research status of computer vision-based Affordance understanding was comprehensively reviewed.First,the methods proposed in recent years were classified according to the research direction,and the ideas and characteristics of different methods were synthesized.Then,several public datasets were introduced,and the performance of different methods on these datasets was comparatively analyzed.Finally,the advantages and disadvantages of various methods in computer vision-based Affordance understanding and future development trends were expounded.

关 键 词：AFFORDANCE 计算机视觉 人-物交互 语义分割 深度学习 机器学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]