基于快速形态滤波和类间方差的弱小目标检测方法

A DIM SMALL TARGET DETECTION ALGORITHM BASED ON FAST MORPHOLOGICAL FILTERING AND INTRA-CLASS VARIANCE

针对地空武器系统背景较为平稳、实时性要求高、目标灰度极性与对比度变化小等特点,提出了一种快速形态滤波和类间方差自适应门限分割相结合的弱小目标检测算法。采用扁平结构元素作为形态滤波结构元素,在水平、垂直两个方向进行开、闭运算。通过对运算结果的处理,解决目标极性反转问题。在膨胀、腐蚀过程中引入最值的权值,对算法进行优化,大大减少了背景抑制处理时间。采用类间方差自动门限分割,使目标提取不受目标与背景对比度变化的影响,从而实现对目标的快速、准确标记。

A dim small target detection algorithm which combines fast morphological filtering and intra-class variance adaptive threshold segmentation is put forward aiming at the characteristics of ground-to-air weapon systems such as fairly stable in background,high requirement in real-time property and small changes in gray-scale polarity and contrasts.The algorithm uses flat structure element to be the element of morphological filtering structure and performs opening and closing operations in both horizontal and vertical directions.It resolves the target polarity reversal problem through processing the operation results.The weights of extrema are introduced into dilation and erosion processes to optimise the algorithm,in this way the time of background restraint processing is reduced greatly.The use of automatic intra-class variance threshold segmentation enables the extraction of targets not to be affected by the contrast variations of target and background,therefore the rapid and precise marking on the target is achieved.