Hand Movement Detection in Collaborative Learning Environment Videos
Hand Movement Detection in Collaborative Learning Environment Videos
复制标题
协作学习环境中的手部运动检测视频
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Callie J Darsey
中科院分区：
文献类型：
--
作者：
Callie J Darsey
数字视频中的人体活动检测目前吸引了大量的研究兴趣。这个问题对于具有大量人类活动、照明噪声和结构噪声的视频数据集来说尤其具有挑战性。与推进数学和工程学校外学习（AOLME）项目相关的视频数据集具有这些挑战。ALOME录像已用于研究“野外”人类活动。本论文主要探讨利用颜色与光流的手部运动侦测。探索性分析考虑了从应用于运动和颜色的阈值创建的组件上的问题组件。该方法使用补丁颜色分类，时空补丁的视频，光流直方图。该方法进行了验证，从15个AOLME视频剪辑提取的视频补丁。该方法实现了84%的平均准确度和89%的平均受试者工作特征曲线下面积（ROC AUC）。
Human activity detection in digital videos is currently attracting significant research interest. This problem is especially challenging for video datasets that have a lot of human activity, illumination noise, and structural noise. The video dataset associated with the Advancing Out of School Learning in Mathematics and Engineering (AOLME) project has these challenges. ALOME videos have been used in the study of human activities “in the wild”. This thesis explores detection of hand movement using color and optical flow. Exploratory analysis considered the problem component wise on components created from thresholds applied to motion and color. The proposed approach uses patch color classification, space-time patches of video, and histogram of optical flow. The approach was validated on video patches extracted from 15 AOLME video clips. The approach achieved an average accuracy of 84% and an average receiver operating characteristic area under curve (ROC AUC) of 89%.