针对视觉目标跟踪算法中存在的快速运动、尺度变化、形变和遮挡问题, 提出基于图像签名算法的视觉目标跟踪算法。该算法以相关滤波算法为基础, 通过多种特征构建目标的外观模型, 提高了算法的跟踪精确度和稳健性; 为了解决严重遮挡情况下的目标重定位问题, 利用图像签名算法计算图像的稀疏显著性区域, 获取候选目标的位置, 通过分类器对候选目标进行重排名, 实现目标重定位; 采用尺度池策略和自适应模板更新策略, 解决跟踪中的尺度变化问题和跟踪漂移问题。利用标准数据集测试所提算法的性能, 结果表明, 所提算法在跟踪成功率和精确度上均优于传统的相关滤波算法, 能较好地解决快速运动、尺度变化、形变和遮挡情况下的目标跟踪问题。

Aiming at the problems of fast motion, scale variation, deformation and occlusion in visual target tracking algorithms, the visual target tracking algorithm based on image signature algorithm is proposed. The proposed algorithm is based on the correlation filtering algorithm. The target appearance model is constructed with various features, and the precision and robustness of the proposed algorithm are improved. In order to solve the target relocation problem under the condition of severe occlusion, the image signature algorithm is used to calculate image sparse salient regions and to obtain the position of candidate target. The candidate target is re-ranked by the classifier, and the target is relocated. Scale pool strategy and adaptive model updating strategy are used to solve the problems of scale variation and tracking drift in tracking. Standard data sets are used to test the performance of the proposed algorithm, and the results show that the proposed algorithm is superior to traditional correlation filtering algorithms in terms of tracking success rate and precision. The proposed algorithm can solve the target tracking problems under the conditions of fast motion, scale variation, deformation and occlusion.