基于特征融合及自适应模型更新的相关滤波目标跟踪算法

Correlation filter target tracking algorithm based on feature fusion and adaptive model updating

针对单一特征目标跟踪算法因背景干扰、目标遮挡造成的跟踪失败问题,以及跟踪过程中每帧进行模型更新容易造成错误更新和实时性差的问题,提出了一种基于特征融合及自适应模型更新策略的相关滤波目标跟踪算法-多特征自适应相关滤波目标跟踪算法。该算法在特征提取阶段将边缘特征及HOG特征加权融合作为目标特征,加强对边缘特征的学习;在模型更新阶段通过计算预测区域与真实区域的奇异值特征向量相似度,并结合设定的阈值判断是否需要进行模型更新,通过自适应更新的方式减少模型的更新次数。在标准测试视频集下验证所提算法,并与两种经典相关滤波算法进行比较,结果表明该算法能够较好地适应背景干扰及目标遮挡问题,跟踪目标的平均中心误差减少了9.05像素,平均距离精度提高12.2%,平均重叠率提高4.53%。

For the problem of object tracking failure caused by background interference,object occlusion in object tracking algorithm based on a single feature,and the problem of error updating and poor real-time performance caused by model updating for each frame during tracking,a correlation filtering object tracking algorithm based on feature fusion and adaptive model updating is proposed in this paper.In the feature extraction phase,the edge feature and HOG feature are weighted together as the object features to enhance the learning of edge features.During the updating phase of the model,calculating the similarity of singular value eigenvector between the predicted region and the real region,and according to the set threshold to determine whether the model needs to be updated or not,and reducing the number of updates to the model by adaptively updating.The algorithm is verified under the standard test video set,and compared with two typical correlation filtering algorithms.The results show that the algorithm can better adapt to background interference and object occlusion problem.The average center location error is reduced by 9.05 pixels,the average distance precision is increased by 12.2%,and the average overlapping precision is increased by 4.53%.