高斯平稳随机场中运动小目标识别算法

Algorithm of Moving Point Targets Detection from Gauss Stationary Random Field

提出了高斯平稳随机场中运动小目标的识别算法,对平稳随机场的图像序列进行训练,由最大似然估计法估计出各对应点处的概率均值和方差,根据3-σ原则对运动图像序列的各帧进行概率域值化处理,对图像序列进行差分多帧叠加,在叠加帧上根据双向链表,由链表的深度进行轨迹判决,试验证明该方法的有效性。

This paper proposes an algorithm of moving point target detection from Gauss stationary random field. It trains the images from the stationary random field and acquires the probability mean and variance by MLE, then does probability threshold process to every frame according to 3 o principle and accumulates the difference of multiple frames. According to the depth of the Bi-direction chain in the sum frame, it does track detection. The experimental results show that the method is effective.