提高海上智能监测水平, 为了实现对海上船只目标的跟踪, 文中针对典型海况环境下的海上船只目标跟踪问题, 提出了一种改进核相关滤波器(Kernelized Correlation Filters, KCF)的船只目标跟踪算法。首先, 针对KCF算法的特征, 提出船只目标跟踪临界概率的概念, 用来判断目标跟踪是否异常; 接着, 加入卡尔曼滤波模块, 用来预测跟踪目标下一时刻的位置; 然后, 对跟踪异常的目标设计目标跟踪异常处理模块进行处理; 最后, 针对4组典型的海上目标跟踪场景, 通过实验验证了算法的性能。实验结果表明: 文中算法在海上船只大幅度晃动、跟踪目标被遮挡、目标出界、目标尺寸变换等复杂情况下, 跟踪准确率和速率比原KCF算法分别提高17.23%和7.86%。满足海上目标跟踪精度、实时性、适用性等方面的要求。

To achieve tracking of marine vessel targets and improve maritime intelligence monitoring levels, this study proposed a target tracking algorithm with improved kernelized correlation filters (KCF). First, per the characteristics of a KCF algorithm, the concept of ship target tracking critical probability was proposed for evaluating whether target tracking was abnormal. A Kalman filter module was added to the algorithm to predict the position of the tracking target at the next moment. A target tracking exception processing module was needed when tracking was abnormal, and an exception handling strategy was adopted. Finally, four sets of typical target tracking scenarios at sea were used to verify the performance of the algorithm through experiments. Under scenarios of large-scale shaking of the ship, tracking target occlusion, target out of bounds, and target size change, the result shows the accuracy of our algorithm to be 17.23% higher than that of the original KCF algorithm, and the rate is increased by 7.86%. Thus, our algorithm meets the requirements of ship tracking accuracy, real-time performance, and applicability.