摘要
为了实现在静态背景下对运动目标的自动检测跟踪,提出基于改进的帧差法和Mean-shift结合的运动目标自动检测与跟踪算法。该算法改进了传统的三帧差分法,引入单高斯背景模型参与目标检测。此外,传统的Mean-shift算法,在起始帧需要手动选定目标,且选定窗口大小固定不变,不能根据目标尺寸变化而变化,从而导致失去目标。这里提出的方法先利用改进的帧差法检测目标,确定目标的位置窗口和中心,然后结合Mean-shift算法,根据是否超出设定的阈值来确定是否需要更新模板,从而实现该算法对运动目标的自动跟踪。实验表明,该算法计算速度快,具有较高的准确率。
To realize the automatic detection and tracing of the moving target in static background,the moving target automatic detection and tracking algorithm based on the integration of modified frame difference method and Mean-shift algorithm is proposed,in which the traditional three-frame different method is modified,and the single-Gaussian background model is introduced into target detection. Since the traditional Mean-shift method has some difficulties,that is,a manual operation is needed at the start frame to select the target,and the selected window size is fixed and can＇t change with the target size,the target will be lost. The modified frame difference method is used to detect the target to conform the target location window and centre,and then the Mean-Shift algorithm is combined to determine whether the new template needs to be updated by judging whether the target window and centre exceed the setting threshold. The automatic tracing of moving target can be implemented by the algorithm. The experimental results show this algorithm has fast computation speed and high accuracy.
出处
《现代电子技术》
北大核心
2016年第4期108-111,共4页
Modern Electronics Technique
