Accurate and timely Land Use and Land Cover (LULC) mapping plays a crucial role in various applications, including environmental monitoring, natural resource management, and urban planning. Although satellite imagery continues to serve as a valuable tool for regional-scale analyses, its inherent spatial and temporal constraints often impede the realization of detailed classification tasks. Unmanned Aerial Vehicles (UAVs) equipped with multispectral (MS) sensors offer a cost-effective, high-resolution alternative, enabling the acquisition of fine-scale spatial heterogeneity. This study proposes an object-based LULC mapping approach using MS UAV imagery and advanced machine learning and deep learning algorithms to enhance classification accuracy. Three models, Extreme Gradient Boosting (XGBoost), Convolutional Neural Networks (CNN), and AutoGluon, were implemented to evaluate their performance in multi-class LULC classification. A high-precision Normalized Digital Surface Model (NDSM) was generated through photogrammetric processing and combined with 68 object-level features, including spectral indices, geometric, and texture attributes. Based on F1 scores, XGBoost outperformed the other models with a score of 0.95, followed closely by AutoGluon (0.93) and CNN (0.92), confirming the strong effectiveness of tree-based ensemble methods for LULC classification tasks. Although AutoGluon exhibited slightly lower accuracy, its user-friendly interface, automatic model selection, and minimal manual parameter tuning requirements rendered it an accessible framework for efficient LULC classification. The findings indicate that the integration of MS UAV-based imagery with AutoML and deep learning techniques facilitates high-quality mapping in complex landscapes. The incorporation of NDSM data further augmented classification performance, enhancing accuracy by 8 to 11% for concrete roofs and approximately 8% for roads.