South Korea has 63% of its national territory covered by forests, playing a crucial role in carbon sink management for climate change mitigation. However, existing forest type maps rely on manual field surveys or visual interpretation, resulting in low accuracy and inconsistent quality. This study proposes a method for automated object-based tree species classification by combining Airborne LiDAR (ALS) data and high-resolution aerial orthoimages (RGB, NIR) with AI technology. The study area covers approximately 207 km² of the Garisan Forest Management District in Hongcheon, where five machine learning algorithms (Random Forest, XGBoost, CatBoost, LightGBM, SVM) were compared, and ensemble techniques (Voting and Stacking) were applied to maximize classification accuracy. The Stacking Ensemble model, comprising Random Forest, LightGBM, and CatBoost as base models with Logistic Regression as the meta-model, achieved the highest performance with an F1-score of 0.9330. Additionally, a three-step post-processing procedure using open-source library reduced data volume by approximately 14.58-fold, ensuring practical applicability. This study establishes a workflow for automated tree species classification over large areas and is expected to serve as a fundamental resource for precise forest management and carbon sink monitoring.