面向无人机影像的目标特征跟踪方法研究

摘要

An object feature tracking model oriented to UAV images is proposed. Firstly, corners in the region of interest which determined based on the object detection result are detected by Shi-Tomasi corner detection method. Feature corners are extracted and updated in real-time to assure the quantity of the detected features meet the tracking requirement. Finally, the pyramid Lucas-Kanade algorithm is used to track object features. The simulation results show that the proposed model is easily implemented and has good robustness. Besides, it performs well in object tracking and the real-time requirement can be fulfilled.%提出了一种面向无人机视频影像的目标特征跟踪模型，首先根据目标检测结果在目标感兴趣区域中采用Shi-Tomasi算法进行角点检测，提取目标特征角点并在跟踪过程中对特征角点进行实时更新以保证角点数量满足跟踪条件，最后通过金字塔Lucas-Kanade算法对目标特征进行跟踪。实验结果表明，该模型鲁棒性强且易于实现，在获得良好跟踪效果的同时可满足实时处理要求。