摘要
为了增强智能视觉监控系统的实时性,提出了一种反馈模型。首先采用基于混合高斯模型的减背景算法进行运动目标检测;然后采用Kalman滤波算法进行运动目标跟踪;进而提出了一种反馈模型,将跟踪阶段Kalman滤波的状态预测值反馈到检测阶段的减背景操作中来约束目标检测的有效范围。最后将本文方法(采用反馈模型)与一般方法(未采用反馈模型)进行了对比实验,结果表明,本文方法具有良好的目标检测与跟踪效果,在不影响准确性的前提下,显著降低了运算量,增强了系统的实时性,对于智能视觉监控系统的现实应用具有重要意义。
A feedback model was put forward in order to improve the real-time capability of intelligent video surveillance.First,the background subtraction method based on Gaussian mixture model was adopted to detect moving objects.And the Kalman filter algorithm was adopted to perform the real-time object tracking.Then a feedback model was put forward.In the model,the results of tracking was fed back to the detection stage in order to restrict the detection regions.By experiments,the proposed method(which used the feedback model) was compared with the common method(which didn′t use the feedback model).The experimental results illustrate the good performance and enhanced real-time capability of the proposed method.
出处
《吉林大学学报（工学版）》
EI
CAS
CSCD
北大核心
2009年第S2期401-405,共5页
Journal of Jilin University:Engineering and Technology Edition
基金
广东省教育部产学研结合项目(2007B090400031)
吉林省信息产业厅专项基金项目(2007042)
