Sweat latent fingermarks are prevalent trace evidence in forensic investigations, and their age estimation is crucial for case resolution. This study utilizes hyperspectral imaging to capture data from sweat latent fingermarks on common indoor surfaces (glass, plastic, and tile) at varying ages. Three preprocessing techniques (SG, SNV, and SG+SNV) were implemented alongside feature wavelength extraction to develop SVM models for age prediction. The applicability and predictive performance of the models under different conditions were compared. The result demonstrates that the SVM models, combined with appropriate preprocessing and feature extraction of hyperspectral imaging, are applicable for predicting the age of sweat latent fingermarks on indoor common objects. Specifically, the model with SG smoothing combined with SNV preprocessing and SPA feature wavelength extraction showed the best predictive performance on glass and plastic, while the model with SNV preprocessing and SPA feature wavelength extraction performed best on tile. Under optimal conditions, the models achieved RPD values of 1.823, 2.074, and 2.039, RMSEP values of 1.580, 1.491, and 1.417, and R² values of 0.755, 0.781, and 0.758, respectively, on the three surfaces.