At present, most of the methods for identifying the relationship of the single parent (father-son relationship, father-daughter relationship, mother-son relationship, mother-daughter relationship) , however, these methods are not effective for identifying father-daughter or mother-son with large age difference and different sex. The best way to solve this problem is to deal with the relationship between children and parents (the relationship between parents) . At present, there are few methods to identify parental relationship, and there is room for improvement. In order to improve the accuracy of identifying the relationship between children and their parents, a model based on metric learning and correlation analysis is proposed. Based on the biological genetic relationship between children and their parents, this paper designs a multi-linear parallel network which can fuse the genetic characteristics of children and their parents, and makes use of discriminant measure learning and canonical correlation analysis to gain advantage in data processing, in order to improve the recognition accuracy, the feature information which is helpful to identify the blood relationship between children and their parents is extracted from facial features which contain many kinds of human identities. Experiments show that this method is more effective in identifying the relationship between children and their parents.