针对核相关滤波算法(KCF)难以应对光照变化、目标完全遮挡等问题, 提出基于改进核相关滤波的运动目标跟踪算法。首先提出基于相位特征的高斯核相关算子, 增强算法对光照强度变化的适应能力, 然后融合Kalman滤波器形成预测-跟踪-校准的跟踪机制, 结合遮挡处理提高系统在目标被完全遮挡时跟踪的准确性。在模型更新方面, 将在线更新与离线更新相结合, 提出自适应更新的策略, 利用跟踪效果较好的历史模型建立备选模型, 替代跟踪效果较差的模型, 及时纠正模型偏移、特征丢失等问题。与原始的核相关滤波算法进行对比实验的结果表明, 改进算法适应光照强度变化的能力明显增强, 当目标被完全遮挡时也能保持较好的跟踪效果。

As Kernelized correlation filter is difficult to deal with the problems of illumination changes and total occlusion of the target, a target tracking algorithm based on improved Kernelized correlation filter was proposed in this paper. Firstly, a Gaussian Kernel correlated operator based on the phase characteristics was proposed to improve the ability of the algorithm to adapt to the change of the light intensity. Then, a tracking mechanism of predicting-tracking-correction based on Kalman filter and an occlusion-handling mechanism were proposed to improve the accuracy of tracking while the target was totally occluded. In the aspect of model updating, an adaptive updating strategy was adopted. The models with better tracking effect were used to establish the alternative model and replace the models with bad tracking effect to correct the problems of model migration and characteristics losing. The experimental results show that the improved algorithm can effectively improve the ability to adapt to the illumination changes and keep a good tracking effect while the target is totally occluded.