A retrospective study included 352 COPD patients treated at Tianjin Chest Hospital from January 2019 to December 2023, of which 71 had non-infectious acute exacerbations (AECHOCP) stratified into training (n=211) and test (n=141) cohorts. The aim was to develop a radiomic model of habitat imaging from chest CT to identify non-infectious AECHOCHP. Whole-lung CT scans were divided into three subregions: habitat 1 (emphysema/bullae), habitat 2 (bronchovascular bundle), and habitat 3 (lung parenchyma). Radiomic traits were optimized using least absolute shrinkage and regression, logistic regression (LR) and support vector machine (SVM) models combined habitat traits with clinical factors. The habitat model achieved predictive power with AUC of 0.853 (LR) and 0.897 (SVM) in training, 0.800 (LR) and 0.807 (SVM) in test. Multivariate analysis confirmed total habitat score and GOLD stage as independent predictors (p < 0.001). The model provides an objective imaging biomarker to quantify the heterogeneity of COPD.