Kinship Understanding for a Family Photo

摘要

In social media analysis technology, kinship mining and understanding is an import issue. Previous works mainly focus on kinship verification between two people, while there has been less research on multiple person kinship recognition of the entire family. In this paper, we present a model to analyze the kinship of all members in a single family photo. Assume that a standard family relationship framework already exists. For an original photo, pre-processing and feature extraction are performed first. Then, we use kinship constraint rules to fill the framework all potential position with members in the photo. Next, calculate the genetic similarity information of the person to determine more specific position. At last, adjust and output the corresponding family tree. Extensive experimental results demonstrate the effectiveness of the proposed approach.