基于改进背景预测和流水线的弱小目标检测

针对复杂背景下液体中弱小目标的检测,提出了一种基于改进背景预测和双层流水线的算法.该算法首先对单帧图像进行背景预测处理,并初始化双层流水线管道;然后对第1层流水线管道中的图像进行交叉差分并二值化,将差分后二值化的图像传送到第二层流水线管道的顶部,更新第2层流水线管道;最后采用逻辑与运算和形态学开运算去除噪声,获得真正的目标,而且应用该算法对弱小目标序列图像进行了验证.实验结果表明：与传统最大化背景预测相比,改进的最大化背景预测方法预处理时间减少了55%,且双层流水线结构算法比单层流水线结构算法在处理时间上减少了0.5s以上.

Detecting weak and small targets by modified background prediction and pipeline

A new method to detect weak and small targets on melt against complex background was proposed,which was based on modified background prediction and double-pipeline.At first,the images sequence was processed by modified background prediction algorithm and the double-pipeline was initialized.Then cross difference and binary images in the first pipeline were constructed,which result images were transported to the top of the second pipeline and refreshed the original images.Lastly,the true target was detected combining with the logic and algorithm and open operation of mathematical morphology to remove noise,and the verification was constructed for images sequence of the weak and small target by the above algorithm.Experimental results show that the preprocessing time of this method is less 55% than the time of the conventional maximum background prediction,and the total time of double-pipeline algorithm is reduced to less 0.5 s than that of single-pipeline algorithm.