基于视觉注意机制的红外弱小目标检测

Infrared dim target detection based on visual attention mechanism

提出了一种基于视觉注意机制的红外小目标检测算法.通过形态学Top-hat算子对图像背景进行抑制,并根据目标与周围背景的对比度不同生成显著图后分割出目标感兴趣区,在感兴趣区域内对每帧图片在尺度空间采用Dog算子处理提高图像信杂比,获得具有较大信杂比的点.为避免目标在帧中消失,采用PID算法跟踪目标点,在可疑目标点周围小区域内采用视网膜皮层理论(Retinex)算法对图像局部区域增强再重新分割出目标.实验证明:该算法能有效对红外小目标进行检测,算法在不同背景的图像检测性能都趋于稳定,当目标融于背景时,能很好地将红外目标检测出来.

An algorithm based on the human visual attention mechanism was proposed in this article. The background was suppressed by the top-hat transform of mathematical morphology.The ROIs (region of interest)of the image can be obtained by segmenting the saliency map which is calculated according to the difference of the target and the background around it.The difference of Gaussians (DOG)filter was used to process the image in the scale space,and the image’s signal-to-clutter ratio (SCR)could be enhanced.Considering the targets are easily submerged in the background and in or-der to prevent the targets not being detected,the PID algorithm was used to track the targets.The Retinex theory was used to enhance the local domain of the target and the target could be re-detected by segmenting the domain.Experimental results show the proposed method can detect the target well and it is stable in a variety of background.When the small target is submerged in the background,it can detect the target well also.