基于颜色特征和尺度自适应的时空上下文算法

Spatio-temporal context algorithm based on color feature and scale adaptation

常规的时空上下文算法在描述目标外观时使用简单的灰度特征,而灰度特征不能很好地描述目标外观,为此提出一种基于颜色特征和尺度自适应的时空上下文算法。通过融合灰度特征和颜色特征描述目标,构建目标的时空上下文模型并计算置信图,以置信图中的最大值处作为估计的目标位置并通过尺度滤波器进行尺度更新。实验选取CVPR2013中20段视频进行测试分析,与近年提出的7个优秀的跟踪器进行对比与分析,实验结果表明,目标在遮挡、旋转、快速运动及光照变化等复杂场景中,该算法能够成功跟踪目标,跟踪性能优于其它算法。

The conventional spatio-temporal context algorithm uses simple grayscale features to describe the appearance of the target,and the grayscale features do not describe the target appearance well.A spatio-temporal context algorithm based on color features and scale adaptation was proposed.By merging the gray features and color features,the target was described,and a spatio-temporal context model was constructed.The confidence map was calculated.The maximum value of the confidence map was used as the estimated target position and the scale was updated using the scale filter.20 videos in CVPR2013 were selected for testing and analyzing,experimental results show that compared with 7 excellent trackers proposed in recent years,the proposed algorithm can successfully track the target in complex scenes such as occlusion,rotation,rapid movement and illumination change,and the tracking performance is superior to other algorithms.