Pulmonary　nodules　are　a　type　of　lung　disease　that　can　involve　multiple　organs　in　severe　cases,　so　early　diagnosis　and　treatment　are　particularly　important　for　patients.　This　paper　introduces　an　automatic　classification　model　of　pulmonary　nodule　CT　images　based　on　transfer　learning.　Firstly,　data　preprocessing　is　performed　on　the　CT　images　of　pulmonary　nodules,　including　geometric　transformation　of　the　data　set　to　achieve　data　enhancement　and　improve　the　generalization　ability　of　the　model.　Xception　is　selected　as　the　benchmark　network　of　the　classification　model,　and　the　pre-trained　weights　of　the　model　on　the　ImageNet　dataset　are　migrated　as　the　initial　weights　of　the　new　model.　SGDM,　RMSProp　and　Adam　optimization　algorithms　are　used　to　optimize　respectively,　and　the　newly　constructed　model　is　fine-tuned　and　trained　with　the　preprocessed　data　set　to　realize　the　secondary　classification　of　CT　images　of　pulmonary　nodules.　The　experimental　results　show　that　the　Xception　+　SGDM　optimization　algorithm　model　based　on　transfer　learning　has　a　recognition　accuracy　rate　of　97.86%,　an　accuracy　rate　of　98.65%,　a　recall　rate　of　96.10%,　and　an　F1　score　of　97.37%,　all　of　which　are　superior　to　existing　automatic　identification　methods.　©　2022　IEEE.