摘要
本文分析了空域视频场景中运动对象实时检测、跟踪系统的模型。提出了一种在运动背景下实时检测与跟踪视频运动目标的技术。该方法首先进行背景的全局运动参数估计,并对背景进行补偿校正,将补偿校正后的相邻两帧进行差分检测。然后利用假设检验从差分图像中提取运动区域,利用遗传学方法在指定区域内确定最优分割门限,提取视频运动对象及其特征;最后利用线性预测器对目标进行匹配跟踪。在基于高速DSP的系统平台上的实验结果表明该方法取得了很好的效果。
A technology of real-time detection and tracking for spatial video motion objects based on moving background is presented after analyzing the system model. At first the global motion parameters of the moving background are estimated in order to compensate background. Then the capable region of the moving object existing is acquired using the difference of the consecutive frames and hypothesis test, and the threshold value for segmentation is acquired using genetic method. Finally an algorithm for matching tracking is used after object characters are acquired. This technology is proved valid and efficient through experiment based on high-speed DSP system.
出处
《信号处理》
CSCD
北大核心
2005年第2期195-198,135,共5页
Journal of Signal Processing
