摘 要：目的 在视觉跟踪领域中，特征的高效表达是鲁棒跟踪的关键，观察到在相关滤波跟踪中，不同卷积层表达了目标的不同方面特征，提出了一种结合连续卷积算子的自适应加权目标跟踪算法。方法 针对目标定位不准确的问题,提出连续卷积算子方法,将离散的位置估计转换成连续位置估计,使得位置定位更加准确；利用不同卷积层的特征表达，提高跟踪效果。首先利用深度卷积网络结构提取多层卷积特征，通过计算相关卷积响应大小，决定在下一帧特征融合时各层特征所占的权重，凸显优势特征，然后使用从不同层训练得到的相关滤波器与提取得到的特征进行相关运算，得到最终的响应图，响应图中最大值所在的位置便是目标所在的位置和尺度。结果 与目前较流行的 3 种目标跟踪算法在目标跟踪基准数据库（OTB-2013）中的 50 组视频序列进行测试，本文提出的算法平均跟踪成功率达到 85.4%。结论 本文算法在光照变化，尺度变化，背景杂波，目标旋转、遮挡和复杂环境下的跟踪具有较高的鲁棒性。
An adaptive weighted object tracking algorithm with continuous convolution operator
Abstract: Objective In the field of visual tracking, efficient representation of features is the key to robust tracking. It is observed that different convolution layers represent different aspects of the target in correlation filter tracking. An adaptive weighted object tracking algorithm with continuous convolution operator is proposed. Method Aiming at the problem of inaccurate target location, a continuous convolution operator method is proposed to convert discrete position estimates into continuous position estimates, which makes position location more accurate. The feature representations of different convolution layers are leveraged to improve tracking effect. Firstly, the multilayer convolution features are extracted by using the deep convolution network structure, and the weight of each layer features in the fusion features in the next frame is determined by calculating the correlation convolution response, so as to highlight the dominant features. Then, the correlation filter trained from different layers is used to perform correlation operation with the extracted features to obtain the final response map. The position of the maximum value in the response map is used to calculate the position and scale of the target. Result Compared with three state-of-the-art tracking algorithms in 50 video sequences of object tracking benchmark (OTB-2013) dataset, the average success rate of the proposed algorithm is 85.4%. Conclusion Experimental results show that the proposed tracking algorithm has good performance and can track successfully and efficiently for many complicated situations, such as illumination variation, scale variation, background clutters, object rotation and occlusion.