Food safety is of vital importance to people's lives. With the improvement of living standards， public concern about food safety has also increased， and the urgent need for reliable and effective information response has arisen. Existing intelligent question-answering systems， such as those based on text vector libraries， have made some progress in knowledge reasoning， problem handling， and semantic relationship recognition， but still have some shortcomings； question-answering systems based on knowledge graphs face inherent challenges in terms of graph construction cost and information recall rate； while question-answering systems based on template matching are poor in question generalization ability and context understanding. This paper proposes an innovative knowledge graph-based intelligent perception question-answering system based on large language models (LLM). Using large language models to improve the shortcomings of the template matching method， combining text vector libraries and knowledge graphs at the same time， and introducing high-dimensional vector semantic search technology into the knowledge graph retrieval process to improve the knowledge graph retrieval recall rate. This innovative approach significantly improves the overall performance of the question-answering system and the user's experience.