一种融合有效卷积操作子和颜色直方图的目标跟踪算法  

作　　者：郑耿峰 

机构地区：[1]福建省特种设备检验研究院,福建福州350000

出　　处：《科技创新与应用》2020年第36期73-76,共4页Technology Innovation and Application

基　　金：国家工信部2018年智能制造综合标准化与新模式专项项目(编号:GXSP20181001);国家科技部质检公益性行业科研专项项目(编号:201410139);福建省质量技术监督局科技项目(编号:FJQI2018030)。

摘　　要：基于有效卷积操作子的相关滤波跟踪算法能够自然融合不同特征,并且通过模型降维,达到良好的跟踪性能。而基于颜色信息的模型虽对光照变化效果不好,但能很好地对形变进行处理。考虑到上述两种算法在特性上可互补,文章提出了一种基于颜色直方图模型和有效卷积操作子的方法,采用加权求和的方式求得目标最终位置。该算法能够提高目标跟踪的鲁棒性,提升跟踪精度和速度。最后,采用数据集的方法验证了该算法的有效性。The correlation filtering tracking algorithm based on effective convolution operators can integrate different features naturally and achieve good tracking performance through model dimensionality reduction.Because the algorithm relies on spatial layout of the object,it is sensitive to deformation.The model based on color statistics is exactly complementary,it can handle deformation well,but not good for illumination changes.This paper presents a method of integrating effective convolution operators and color statistics histogram model.The final position of the target is obtained by weighted summation.The algorithm can improve the robustness of object tracking and improve the precision and speed.Finally,through making datasets,the performance of the improved algorithm is tested and evaluated.

关 键 词：目标跟踪 相关滤波 有效卷积操作子 颜色直方图 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]