CAJ | 학술논문

针对视频复杂运动事件进行了分析, 对用于自然语言处理的格语法理论进行了扩展, 增加时间格结构, 并采用扩展后的格框架对复杂事件进行了标注, 并将数据存储在数据立方体中, 最后采用MDFP-growth方法对多维概念格进行关联规则的挖掘。在实验部分, 对比了扩展后的格框架与扩展前原方法对复杂事件标注的结果, 并比较扩展了时间格后的四维数据〈PRED, Ag, T, Loc〉和扩展前三维数据〈PRED, Ag, Loc〉挖掘其关联规则中系统运行时间, 产生规则数对比结果及视频检测的准确率和召回率对比。实验证明, 提出的方法能更加准确地标注出复杂事件, 并具有更高的处理效率。
침대시빈복잡운동사건진행료분석, 대용우자연어언처리적격어법이론진행료확전, 증가시간격결구, 병채용확전후적격광가대복잡사건진행료표주, 병장수거존저재수거립방체중, 최후채용MDFP-growth방법대다유개념격진행관련규칙적알굴。재실험부분, 대비료확전후적격광가여확전전원방법대복잡사건표주적결과, 병비교확전료시간격후적사유수거〈PRED, Ag, T, Loc〉화확전전삼유수거〈PRED, Ag, Loc〉알굴기관련규칙중계통운행시간, 산생규칙수대비결과급시빈검측적준학솔화소회솔대비。실험증명, 제출적방법능경가준학지표주출복잡사건, 병구유경고적처리효솔。
This paper analyzed complex motion events of the video, extended the case grammar theory of natural language processing, </sup>and increased the time case structure. Then it used the extension method to annotate video complex events, and stored data in the data cube. Finally, it used the MDFP-growth method to mine association rule for multi-dimensional case concepts. </sup>In the experimental section, this paper compared the results of the extended case framework and the original method of annotating a complex event, and compared the four-dimensional data extended the time case to the three-dimensional data mining association rules in the system. Finally, numbers of rules between the two and the precision and recall rate of video detection</sup> had been compared. The experiments show that the proposed method can be more accurately annotated a complex event</sup> and the higher processing efficiency.