基于照片的社交关系可视化方法

Photo-based Social Relationship Visualization Method

随着数码摄影设备的普及和多媒体技术的发展,照片在线上线下社交活动中发挥着越来越重要的作用.本文提出一种基于照片的社交关系可视化方法,结合社交媒体信息揭示出社交照片中隐含的社交网络,并分析照片人物间的朋友推荐潜力,以此设计了一个探索社交关系网络的可视化系统SPVis.SPVis利用人脸识别技术识别社交照片中的人物,并获取识别对象在微博上的社交数据(关注数据和文本数据等);接着,利用LDA模型发现文本主题,以微博文本相似度来度量识别对象间的相似度;最后,通过可视分析识别对象的关注数据和相似度,挖掘对象间关系的特点,分析对象间关系更加紧密的潜力.本文使用娱乐明星数据集进行用户实验和案例分析,实验结果验证了方法的可行性和有效性.

With the popularity of digital photographic devices and the development of multimedia technologies,photos play an increasingly important role in online and offline social activities.This paper proposes a photo-based visualization method for social relationships,reveals the social netw ork hidden in social photos by combining social media information,and then analyzes the friend recommendation potential of the characters in the photos.We have designed a visualization system for social relationship netw orks w ith photos,called SPVis.SPVis uses a face recognition method to identify people in social photos,and obtains social data(e.g.follow ing data and text data)of the identified objects on Weibo;then,it employs the LDA model to discover Weibo text topics and calculate Weibo text similarity to measure to the similarity betw een different persons in the photo;finally,based on the recognition data of the identified persons and their similarities,w e analyze their social relationships and the potential of friend recommendations.This paper introduces the entertainment star data set for user experiments and case analysis.The experimental results demonstrate the feasibility and effectiveness of our method.