视频多运动目标提取的新方法

提出一种利用视频序列的时域和空域信息进行多运动目标检测的新方法.首先对相邻的图像帧进行差分,并利用3σ准则二值化差分图像,得到运动目标的时域信息,然后标记该二值图像的连通区域来消除噪声,结合边缘检测所得到的空域信息得到运动目标模型,最后连接模型的最外围轮廓点即得到目标的完整信息.该算法充分运用了运动目标的时域、空域两方面信息,能够准确的提取出运动目标.

A New Algorithm for Detecting the Moving Objects of Video Sequences

A new algorithm to detect the moving objects of video sequences based on time and spatial information is pro-posed in this paper. Firstly we operate the difference calculation to adjacent frames based on three- regulation, do bina-rization processing, so we can gain time information because of movement of the moving objects. Secondly sign adjacent pixels of the bilingual picture to remove the polluted pixels, then we can get models of the moving objects with the help of spatial information resulting from edge detection. At last the integral information will he obtained after connecting the out-most pixels of the models. The algorithm will detect the moving objects precisely according to using of the time informa-tion in combination with the spatial information of the moving objects.