摘要
提出了一种新的运动目标检测与定位方法。针对位移变化较小的运动目标,先对运动序列中所有相邻两帧图像作绝对值差分运算,然后再将绝对值差分结果进行累加,从而得到累积绝对差图像。利用交叉熵分割法对累积绝对差图像二值化,并结合形态学方法去除噪声,求取出目标的运动区域。对运动序列的首帧和尾帧进行差分运算并二值化,为了去噪,将首尾帧差图像与累积绝对差图像进行逻辑与运算,确定出目标在首尾图像中的位置。实验结果表明了本方法的有效性和鲁棒性。
A new motion target detection and orientation method is proposed, This paper accumulates absolute difference between sequential two frames in motion sequences and gets accumulative absolute difference picture（AADP）,changes AADP into binary image by cross entropy and denoises it by morphology method,makes difference operator between first frame and last frame and binarizes it,at last,makes logic and operator between this difference image and AADP and gets target position in first image and last image. The results of experiments show that this method is robust and effective.
出处
《计算机与现代化》
2005年第8期75-77,共3页
Computer and Modernization
关键词
累积绝对差图像
交叉熵
形态学方法
运动区域
accumulative absolute difference picture
cross entropy
morphology method
motion region
