基于时空模型的尺度自适应视觉跟踪*

针对视觉跟踪过程中因目标尺度变化跟踪精度低的问题,提出一种基于时空模型的尺度自适应视觉跟踪算法(Spatio-temporal context – scale adaptive,简称STC-SA)。在颜色属性空间下提取目标颜色直方图特征；再通过时空上下文学习获取置信图中概率最大位置；最后利用颜色直方图进行相似度匹配和自适应方法修正跟踪框尺寸以达到最佳的跟踪效果。实验选取Benchmark中五组具有明显尺度变化的图像序列进行测试,STC-SA算法的跟踪成功率最高达到91%,验证了STC-SA算法具有较高的跟踪精度和跟踪实时性。

Scale adaptive visual tracking based on spatio-temporal model

To solve the problem of low effectiveness of visual tracking with the target scale change, an adaptive visual tracking algorithm is proposed based on spatio-temporal model Firstly, target characteristics are analyzed in the color naming space and the target color histogram is extracted. Then the maximum probability of confidence map is obtained via spatio-temporal context learning. Finally, the similarity of color histogram is matched and the size of tracking window is adjusted with adaptive method. Five groups of scale-change image sequences were selected and tested in the Benchmark. The experiment results show that the tracking success rate of STC-SA is up to 91%. STC-SA algorithm has high tracking accuracy and real-time performance.