Dynamic Object Detection and Tracking Model on UAV for Surveillance Applications Using Adaptive Yolov8 with Spatial–Temporal Attention-Based Feature Pyramid Network

A significant approach for dynamic object detection and tracking systems on UAVs for surveillance applications is introduced. In order to begin the process, input videos were gathered through standard datasets. Next, gathered input video frames are fed to the module of object detection and tracking, whereas an Adaptive You Only Look Once V8 (YOLO V8) with Spatial Temporal Attention-based Feature Pyramid Network (AY8-STAFPN) is used for accurately detecting and tracking the objects. Furthermore, the performance is improved by optimally tuning the network parameters of AY8-STAFPN with the help of Random Integer-dependently Dark Forest Algorithm (RIDFA). Thus, the numerical validations evaluate the developed model’s performance by comparing it with other conventional models to ensure its supremacy.