Abstract: Litchi is one of Guangdong Lingnan characteristic fruit industry and one of Guangdong top 10 speciality crops.Although modern production technology of litchi has been greatly improved in recent years，climate conditions in Guangdong are easy to cause pests and diseases，which has a serious adverse effect on yield and quality of open-air growing litchi in Guangdong，which hinders development of litchi industry and income of growers.Therefore，it has great practical significance to promote development of litchi brand industry and increase income of litchi growers.Research and application of image recognition technology for litchi diseases and insect pests were carried out by using convolution neural network algorithm.It mainly included two aspects：one was to discuss image recognition technology for litchi diseases and insect pests；the other was to explore construction of a new type of meteorological defense system of diseases and insect pests based on Guangzhou intelligent agrometeorological service platform.In other words，it not only provided Litchi growers with intelligent identification service of litchi images，but also provided them with knowledge of diseases and insect pests as well as prevention suggestions.It also constructed“personal space”for growers to realize targeted early warning，and gradually built a personalized litchi base disease forecasting and early warning model to explore litchi disease forecasting and diagnosis combined service model.Therefore，innovation of this paper was to explore how to realize image recognition technology and to study stereoscopic defense measures.