This study presents a deep learning-based surveillance system for early detection and monitoring of forest fires using a four-rotor UAV equipped with an NVIDIA Jetson Nano AI computer and a camera. The system integrates YOLOv8 and YOLOv5 for object detection and a CNN-RCNN network for image classification. The YOLOv8 model achieved 96% accuracy in classification, while YOLOv5n had 89% accuracy in object detection. The CNN-RCNN model, combining CNN and RCNN, achieved 96% accuracy in classification and 89% in object detection. A ground station interface was developed to receive and display fire-related data, enabling targeted intervention. The system's effectiveness is demonstrated through real-time flight tests and comparisons with previous studies, showing promising results for proactive forest fire detection and management.This study presents a deep learning-based surveillance system for early detection and monitoring of forest fires using a four-rotor UAV equipped with an NVIDIA Jetson Nano AI computer and a camera. The system integrates YOLOv8 and YOLOv5 for object detection and a CNN-RCNN network for image classification. The YOLOv8 model achieved 96% accuracy in classification, while YOLOv5n had 89% accuracy in object detection. The CNN-RCNN model, combining CNN and RCNN, achieved 96% accuracy in classification and 89% in object detection. A ground station interface was developed to receive and display fire-related data, enabling targeted intervention. The system's effectiveness is demonstrated through real-time flight tests and comparisons with previous studies, showing promising results for proactive forest fire detection and management.