基于人脸卡口数据的行人共现关系图谱构建

随着人脸识别技术的不断进步以及人脸卡口的大范围且密集的部署, 本文针对团伙犯罪案件侦察这一应用场景, 对人脸卡口数据进行深入挖掘, 探究其中行人间的共现关系, 获取所关注的嫌疑人的现实社交网络, 锁定团伙其余人员. 经过实验比对和论证, 本文使用Chinese Whispers聚类算法对行人节点进行识别, 通过Faiss加速邻接边的构建, 加速图的初始化步骤, 解决其聚类效率低下的问题. 在此基础上, 使用共现频次和Apriori算法中的置信度挖掘行人间的共现关系, 构建行人共现关系图谱.

Construction of Co-occurrence Relationship Between Pedestrians Based on Smart Face Capture Cameras' Data

Nowadays, face recognition technology keeps advancing and smart face capture cameras are widely and intensively deployed. Aiming at the application scenario of gang criminal case detection, this study conducts in-depth data mining and explored the co-occurrence relationship between pedestrians. Other gang members are locked through the realistic social network of suspects. After experimental comparison and demonstration, this study uses the Chinese Whispers clustering algorithm to identify pedestrian nodes and Faiss to accelerate the construction of adjacent edges. In this way, the initialization of the graph is sped up, which improved the clustering efficiency. On this basis, the co-occurrence frequency and the confidence in the Apriori algorithm are used to mine the co-occurrence relationship between pedestrians, resulting in the graph of the co-occurrence relationship.