多特征融合的相关滤波红外单目标跟踪算法

针对红外单目标跟踪问题,提出一种多特征的相关滤波目标跟踪算法。该算法融合了图像的卷积特征和差分特征,使用卷积特征和差分特征分别训练相关滤波模型。在跟踪阶段,对两种特征的相关滤波模型得到的响应图动态融合,利用动态融合的响应图来确定目标的最终位置,使用得到的目标位置分别更新相关滤波模型。在林雪平热红外数据集上进行了实验验证,与一些经典的跟踪算法进行了对比,表明该算法拥有更高的跟踪准确率。

Algorithm for tracking an infrared single target based on correlation filtering with multi-feature fusion

iAimng at the problem of infrared single target tracking, a tracking algorithm based on Multi-feature and correlation filtering is proposed. The algorithm fuses convolution features and differential features. The convolution feature and differential feature are used to train the correlation filtering model, respectively. In the tracking stage, the response graphs obtained from the correlation filtering model of the two features are fused dynamically. The final position of the target is determined by the dynamic fusion response graph, and then the correlation filtering model is updated separately by using the obtained target position. Experiments on the Link?ping Thermal InfraRed dataset show that the proposed tracking algorithm has a higher tracking accuracy than the conventional tracking algorithms.