基于聚集疑似目标的快速TBD弱小多目标检测方法

提出了一种无需复杂图像预处理的快速的自适应弱小多目标检测方法,适用于有背景起伏的低SNR的噪声随机分布的红外图像序列.采用线性滤波算子,直方图排序分割,快速目标聚集三个核心算法,实现了高速的TBD弱小目标检测.根据直方图进行梯度滤波的结果自适应地进行门限分割,然后剔除随机的不聚集的高梯度值噪声,最后进行时间低通滤波,减少虚警率和漏检率.通过对比实验,表明本文提出的先跟踪后检测的方法,在具有良好检测效果的同时,速度优于需要复杂形态学滤波或中值滤波的方法.

A Fast Multiple Small Targets TBD Algorithm by Grouping Suspicious Targets

We proposed a fast adaptive small moving targets detection algorithm,which does not necessitate complex image pre-processing.The core of this algorithm is to utilize linear filter,histogram adaptive division and grouping suspicious targets.Using the time-space correlation of targets,the algorithm excepts noise points or fake targets.Finally,the result is filtered with a low-pass filter in time-domain in order to ameliorate missing rate and error-report rate.The algorithm is suitable to capture targets in ripple background with low SNR