一种运动背景下视觉注意辅助的目标检测方法

针对运动背景下序列图像中前景目标难以准确检测的问题，提出了一种视觉注意辅助ViBe算法的目标检测方法.首先利用提出的“记忆窗”随机抽样一致性算法估计出背景运动模型，再将补偿后的帧图像送入视觉显著性辅助ViBe算法进行目标检测，其中背景更新因子由图像的二维熵和显著性共同决定，同时，显著性特征也被用来对“鬼影”效应进行滤波抑制.实验结果表明，在缺乏运动前景目标的先验知识的情况下，本方法能够有效地解决通常方法对运动背景中的目标难以甚至无法检测的问题，并且具有较高的鲁棒性和检测效率.
It is difficult to detect the foreground targets accurately in the sequence images in the situation of moving background. According to the characteristics of the targets in moving backgrounds, a fast visual saliency-aided ViBe method was proposed. The “memory window” random sample consensus algorithm was first proposed to estimate the background motion model. Then, the compensated frames were delivered into the detection algorithm with adaptive background updating factor determined by the two-dimensional entropy and saliency of the image. The saliency was also used to suppress the “ghost” effect. Experimental results show that, in the absence of prior knowledge of the moving foreground targets, the proposed method can effectively detect targets in the situation of moving background, and it also has good robustness and high efficiency.