Maneuvering target with nature background displays several characteristics in videos, for example, features of targets are unstable, deformation is severe and the discipline of moving is hard to predict, which make traditional video-based tracking methods lose effects. In order to solve this problem,a novel detecting based method is proposed in this paper. The effectiveness of this algorithm doesn't depend on the relationships between frames. Firstly, wavelet transformation is used to process the current frame of the video. Secondly, fusion work is done with high frequency images. Thirdly, the fusion result is strengthened and a threshold is applied to achieve ultimate detection. Finally, the video tracking is accomplished with every frame detected. The result illustrates that the proposed algorithm performs better in detecting and tracking maneuvering target in nature background especially when the targets are blocked by the cloud and fog.%空中强机动性目标在视频中表现出运动形变剧烈、运动规律难预测且背景特征不明显不稳定等特点,传统的检测跟踪算法不能有效地对其实施检测跟踪.为解决上述问题,提出一种不依赖于帧间运动特性的基于提升框架检测的空中强机动视频目标跟踪算法.首先,利用提升框架实现图像的小波变换,得到目标的高频分量信息.然后,对高频分量进行融合,并提出了目标特征增强的方法,完成单帧图像的目标检测.最后,实现视频序列的实时跟踪.实验结果表明,该算法能够实时、准确地对空中背景下的强机动性目标进行检测跟踪,有效地解决了目标运动形变剧烈、某些帧中存在遮挡而造成后续帧失跟等跟踪困难.
展开▼
