基于MRF的自适应帧差运动目标检测

Adaptive Frame Difference Moving Target Detection Based on MRF

为了克服传统的运动目标检测算法容易受到显露遮挡、空洞以及噪声现象的影响,构造了一种基于MRF的自适应帧差运动目标检测算法。采用最大类间方差法自适应确定序列图像的初始标记场;通过帧差及"与"运算处理,消除伪运动信息;结合马尔可夫随机场理论构建了自适应马尔可夫随机场模型系统能量函数;并利用迭代条件模式算法完成标记场的优化过程,准确提取出运动目标。实验结果表明,该算法能够有效地实现运动目标检测,效果理想。

The traditional moving target detection algorithm is always affected by background emergence-occlusion ,void spots and noise. In order to avoid these phenomena,a novel algorithm of adaptive frame difference moving target detection based on MRF is constructed. Otsu method is adopted to obtain initial label field of sequence images adaptively. And the false motion information is eliminated by frame difference method and the ＂and＂ operation. Then the system energy function of adaptive Markov Random Field model is established in the combination of Markov Random Field theory. Finally, opti- mization process of label field is completed by iterative conditional mode algorithm, and the moving target is extracted accurately. Experimental results show that the algorithm is effective and the moving target detection effect is desirable.