基于改进核相关滤波算法的目标跟踪

Target Tracking Based on Improved Kernel Correlation Filtering

针对核相关滤波算法仅使用一种特征表达进行目标追踪,使其在一些场景中跟踪效果不佳的问题,提出了一种多特征融合的核相关滤波跟踪方法。采用31维的方向梯度直方图特征、58维的局部二值模式特征和1维的灰度特征进行融合。该算法选择在特征层进行特征融合,先将方向梯度特征和局部二值模式特征并联融合,再将融合后的特征串联融合灰度特征,形成新的特征表达。在OTB(Object Tracking Benchmark)数据集上进行了测试,结果表明,该算法具有更好的跟踪效果。

A kernel correlation filtering method based on multi-feature fusion is proposed for the kernel correlation filtering algorithm using only one feature expression for target tracking,which makes the tracking effect of the algorithm in some scenes poor.The 31-dimensional histogram of oriented gradient feature,the 58-dimensional local binary patterns feature and the 1-dimensional gray feature are used for fusion.The algorithm selects the feature fusion in the feature layer,combining the histogram of oriented gradient feature and the local binary mode feature.Then,the fused features are fusing into the gray feature to form a new feature expression.The improved algorithm is tested on the Object Tracking Benchmark(OTB) dataset.Experiments show that the algorithm has better effect on tracking.