Design of UAV Single Object Tracking Algorithm Based on Feature Fusion

Considering the real-time performance and robustness of the object tracking algorithm, this paper proposes an improved tracking algorithm for single object tracking based on UAV. The algorithm uses Tiny-YOLOv3 for preliminary detection, and the detection results combine histogram of orientation gradient(HOG) and RGB histogram to extract the features. We use histogram matching to find the highest similarity between the detected candidate object and the object to be tracked to achieve the purpose of tracking. Different tracking strategies are designed when the object is stationary and moving. The experimental results show that the algorithm improves tracking accuracy and robustness while ensuring real-time tracking.