It is an important study direction in the area of smart farming to explore the combination of new progress in the field of pig farming with artificial intelligence. Among them， how to automatically identify the individual identity and behavior of pigs is a key and hot issue to be solved in the current pig breeding industry. This article summarizes the existed methods of using deep neural networks to identify the individual identity and behavior of pigs based on the progress of computer vision and deep learning models in human recognition. The problems in the existed methods are analyzed， and the key study directions in the future are proposed. Five aspects urgently needed to be developed are as follows： （1） the methods of accurately extracting the features of pig’s identity and behavior under the conditions that pig’s behavior cannot be controlled and the key parts of pig’s body are often contaminated； （2） the deep learning models based on computer vision that dedicate for pigs to recognize the identity and behavior based on the specific features of pigs； （3） the studies on multi-task deep learning models that can recognize pig’s identity and behavior simultaneously； （4）the studies on general-purpose pig behavior recognition methods based on basic postures and movements that are applicable to multiple scenarios； （5） the studies on the deployment methods of pig identification and behavior recognition based on edge computing.