The economic and ecological importance of palm trees has made it increasingly important to develop automated and efficient methods for their analysis, especially in agricultural resource management. This research integrates deep learning models with geographic information systems to develop a robust system for palm tree identification using aerial imagery. State-of-the-art convolutional neural network (CNN)-based models, including EfficientNet, DenseNet, InceptionResNetV2, and Vision Transformer (ViT), were implemented to analyze aerial images. Following the training and testing, the model EfficientNet was found to perform at a much better accuracy of 98.2% precision, 98.8% recall, and an F1-score of 98.5%, making it the most suitable for real-time use cases. DenseNet also came out with a 94.3% precision, 94.5% recall, and an F1-score of 94.4%. InceptionResNetV2 was one of the more competitive models, providing an accuracy of 96.2% precision, 96.5% recall, and an F1-score of 96.3%. Vision Transformer, although having a novel self-attention mechanism, scored lower, with 85.8% precision, 86.7% recall, and 86.2% F1-score, indicating that it is unable to capture the local spatial features that are so important for palm tree detection. These results highlight the significance of model architecture in achieving optimal performance.