In traditional agriculture, the method of identifying pests and diseases that relies on experience and visual observation often fails to respond in time when facing new foreign pests, resulting in delayed prevention and economic losses. To solve this problem, this study proposes an intelligent agricultural pest and disease detection model based on transfer learning, combining modern computer technology, image processing algorithms and big data analysis, and using multispectral imaging technology to monitor crops in real time. The model is first pre-trained using a large data set, and then the features are transferred to the crop pest and disease detection task for fine-tuning to optimize the pest and disease identification accuracy of small samples. The research results show that this method can effectively improve detection efficiency and reduce labor costs in complex environments, providing scientific decision-making support for agricultural production. This innovative technology provides a solution for intelligent pest and disease detection in modern agriculture and promotes the digitalization and intelligent development of agriculture.