Aiming at the current problem of difficult cattle identity recognition in the su rvey and loss determination of livestock insurance industry,this paper combines deep learning framework with internet of things (IOT) technology to build cow face recognition system.Inception _Resnet_v1 is used as the deep learning network framework in the system,combined with Triplet Loss loss function,to complete the extraction of cow face features,and the identifica tion of features by calculating the cosine distance.Based on flask and vue framework,we deploy cow face detection and feature extraction and verification model,develop cow face inf ormation database,and provide identity registration and verification web service.To ver ify the feasibility of the system,the cow face recognition dataset CFID200 is collected and prod uced.Under the condition that 20% of cow faces are not seen,the accuracy and verification r ate of the cow face recognition system reaches more than 95%,which meets the need of cow fa ce identification in insurance business.