基于自适应模板更新的改进孪生卷积网络目标跟踪算法

OBJECT TRACKING ALGORITHM BASED ON IMPROVED SIAMESE CONVOLUTIONAL NETWORKS COMBINED WITH ADAPTIVE TEMPLATE UPDATING

现有的孪生网络目标跟踪算法采用边界框模板进行跟踪,在目标形变、遮挡等干扰下很容易导致跟踪漂移。在轮廓检测网络和孪生卷积网络(Siamese)跟踪网络的基础上,提出一种基于深度轮廓模板更新的改进孪生卷积网络目标跟踪算法。利用轮廓检测网络获取目标边缘轮廓,降低背景杂波干扰;利用改进的Siamese网络获得轮廓模板和搜索区域的深度特征;通过相似性匹配获得最优跟踪目标。仿真实验结果表明,所提出的改进模型能够提高目标形变、遮挡等干扰下目标跟踪性能,具有较高的工程应用价值。

The existing Siamese network object tracking algorithms use the bounding box template to track the object,which can easily lead to tracking drift under the influence of object deformation and occlusion.Based on the contour detection network and the Siamese tracking network,this paper proposes an improved object tracking algorithm based on deep contour template updating.It used the contour detection network to obtain the edge contour of object,so as to reduce the background clutter interference in the image;the improved Siamese network was used to obtain the deep features of the contour-template and the search area;the optimal tracking object was obtained by similarity matching.The simulation results show that the improved model can improve the object tracking performance under the influence of object deformation and occlusion,and has high engineering application value.