Image segmentation， as an important component of the vision system in smart agricultural farming， is widely used in the smart farming of livestock and poultry. In recent years， deep learning algorithms have been booming， and image segmentation technology based on deep learning have also made significant breakthroughs. These methods give more accurate semantic information to the segmented region， making the segmentation more accurate and intelligent， and providing stronger technical support for poultry and livestock smart farming. Through extensive collection and analysis of relevant domestic and foreign research literature， this paper first elaborates the application of image segmentation based on deep learning of poultry and livestock farming in detail， such as measurement of body size and weight， attitude estimation and behavior recognition， body condition and disease detection， precision feeding， etc. Suggestions are given on how to choose appropriate image segmentation methods based on actual performance requirements （accuracy， processing speed）， datasets， and computational resources. Furthermore， the open datasets， which are summarized and analyzed in current literature， related to livestock and poultry farming can be used for image segmentation training. This paper points out the challenges and future development trends of image segmentation technology based on deep learning in livestock and poultry farming， hoping to provide reference for the specific application of image segmentation technology in livestock and poultry farming.