图像序列中的自动点目标空时快速检测

Fast Spatial-temporal Automatic Detection of Point Target in Image Sequences

提出了一种新的序列图像中点状运动目标空时快速检测方法,该方法假设经过背景杂波抑制预处理后图像已转换为类SPGWN(信号加高斯白噪声)模型。自动目标检测时图像序列被划分为多组检测单元,每单元包含两帧相邻的图像。根据目标速度限制条件,采用在单元内以及两相邻单元间沿轨迹集成像素灰度的检测算法。在付出同等运算量的代价下,克服了二维投影检测性能较差的弱点,同时也具备可检测多速率目标的优点。理论分析和仿真试验表明了该检测方法的有效性。

A new method of fast spatial-temporal detection for moving point target in image sequences was presented, It was based on the hypothesis that images had been transformed into the quasi SPGWN （Signal-Plus-Gaussian-White-Noise） model through pre-processing of background clutter suppression, Sequences then were divided into some detection units and each unit included two sequential frames. According to the target velocity limits, it adopted the algorithm based on pixel gray level integration along tracks within an unit as well as between neighbor units. At the same cost of operations, the method abstains the disadvantage of bad performance which two-dimensional projection detection could bring forth. It also had the advantage of detecting multi-velocity targets. The theoretic analysis and simulations show its validity.