基于监控视频的知识图谱数据挖掘

为了解决依靠人力难以有效从海量监控视频中发现隐藏信息的难题,提出了一种基于监控视频的知识图谱构建和数据挖掘方法:先利用深度学习算法对监控视频进行行人重识别,给每一个独立的人分配唯一ID;再将非结构化的视频数据转化为结构化的三元组格式,输入Neo4j图数据库构建知识图谱;最后,基于该知识图谱利用行人共现算法、轨迹挖掘算法、社团检测算法等对监控视频进行数据挖掘。采用知识图谱对监控视频进行数据挖掘的实验结果表明,该方法在充分利用监控视频数据上有独特优势。

To extract the hidden information from massive surveillance videos effectively,this paper presents a knowledge map construction and data mining method based on these videos:Firstly,person re-identification is carried out by using a deep learning algorithm,and a unique ID is assigned to each independent person.Then,the unstructured video data is trans-formed into structured triples,and a knowledge graph is constructed by inputting them into the Neo4j graph database.Lastly,based on them,the data mining process is performed by using co-occurrence relationship mining,frequent trajectory mining and community center detection algorithm.The experimental results show that this scheme has a unique advantage in mak-ing full use of surveillance video data.