采用多边形质心的相关滤波跟踪位置校正方法

Polygon centroid based position correction method for correlation filter tracking

为了解决多个尺度相关响应图最大值对应位置与目标真实位置存在偏差的问题,本文提出了一种采用多边形质心的相关滤波跟踪位置校正方法。首先,本文提出一种相关响应图评价指标对响应图的质量进行评价,并对当前帧跟踪结果进行可信判别;接着,对判别为不可信的跟踪结果采用多边形顶点的质心进行校正,以减少跟踪结果与目标正确位置的偏差。最后,在OTB50、OTB-2015和UAV20L三个基准视频集上对本文算法进行性能评估实验,本文算法在OTB50、OTB-2015和UAV20L的成功率曲线面积分别达到了0.625、0.668和0.429,跟踪精度分别达到了0.844、0.885和0.578。结果表明,与近年来主流的跟踪算法相比,本文算法在多种复杂场景下都取得了较优的成功率曲线面积和跟踪精度。

To solve the problem of the deviation between the maximum position of the multi-scale correlation response map and the true position of the target,a polygon centroid based position correction method for correlation filter tracking is proposed.First,we proposed a correlation response map evaluation index to evaluate the quality of the response map,and make a credible judgment of the tracking result of the current frame.Then,we use the centroid of the polygon vertex to correct the tracking result that is judged to be unreliable to reduce the tracking result and target deviation from the correct position.Finally,the performance evaluation experiment of the algorithm in this paper is carried out on the three benchmark video sets of OTB50,OTB-2015 and UAV20 L.The Area Under Curve of this algorithm in OTB50,OTB-2015 and UAV20 L reached 0.625,0.668 and 0.429 respectively,and the tracking accuracy reached0.844,0.885 and 0.578 respectively.The results show that compared with the mainstream tracking algorithms in recent years,our algorithm has achieved better success rate and tracking precision in a variety of complex scenarios.