基于帧间差分与时空上下文的自动检测跟踪算法

摘要

An automatic detection and tracking algorithm based on frame difference and STC was put forward to improve the spatio-temporal context (STC) tracking algorithm's failure to detect and track the target automatically.This algorithm transmits the rectangular box detected by the frame difference algorithm and contains the outline and position of the foreground object to the STC tracker.After the analysis and the comparison of the STC algorithm before and after improvement in the experiment, we conclude that the improved algorithm enjoys higher accuracy and achieves stable tracking of the target.%为了改善时空上下文 (STC) 跟踪算法不能自动检测跟踪目标的缺点, 提出一种帧间差分与STC相结合的自动检测跟踪算法.该算法将帧间差分法检测到的含有前景目标轮廓及位置的矩形框传送给STC跟踪器, 可达到自动检测与跟踪目标的目的, 并且提高了跟踪精确度, 降低了手动选定目标框的繁琐程度.通过实验对改进前后的STC算法进行分析比较, 结果表明, 改进后的算法具有更高的跟踪精度, 可实现对目标的稳定跟踪.