序列图像中弱小运动目标检测方法

Detection method for small moving target in sequence image

通过研究利用远距离拍摄获得的飞行目标运动轨迹的图像序列,分析图像的特点,研究目标自动识别的方法。采用基于块匹配技术的配准图像,对原始图像进行配准;采用基于能量累加的背景去除技术,获得包含目标及噪声的图像序列;采用自适应阈值分割技术,得到包含目标及噪声点的二值图像;利用目标帧间相关而噪声帧间不相关的特点,根据目标的速度梯度和方向梯度,对二值图像中所有的点进行帧间相关性判断,获取目标点。实验结果表明,该方法能够有效地在飞行目标运动轨迹的图像序列中准确识别出弱小运动目标。

In order to solve the problem of detecting small moving targets in low-contrast image sequences, a new detection method is proposed in this paper. Image SNR is improved by using energy accumulation method; the back-ground of the image is removed. Then using adaptive threshold segmentation, the binary image is obtained, which includes target and noise points. Based on the correlation of target between frames, noise is removed and target is detected. Experimental results show that this method can effectively find small targets from sequential images with low SNR.