基于PCNN和改进邻域判决的红外弱小目标检测算法

Infrared Dim Small Target Detection Algorithm Based on PCNN and Improved Neighborhood Judgement

针对复杂背景下红外运动弱小目标的检测问题,提出了一种基于脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)和改进邻域判决的红外弱小目标检测算法。该算法利用侧抑制网络对输入图像进行滤波,实现背景抑制和目标增强;利用PCNN进行单帧图像分割,将可能的目标和背景杂波及噪声初步分离,确定候选目标;利用改进的邻域判决方法分析候选目标的运动特性,自适应确定判决的邻域大小,结合图像流分析提取出真正的目标。实验结果表明,该算法能够极大地减少候选目标数量,准确有效地检测出复杂背景中的红外运动弱小目标。

Aiming at the problem of detecting small infrared targets under complex background,an infrared dim small target detection algorithm based on Pulse Coupled Neural Network（PCNN） and improved neighborhood judgement was proposed. Firstly,the lateral inhibition network was used to preprocess the input image for background suppression and target enhancement; And then,PCNN was used to image segmentation in single frame for separating potential targets and background clutters,which can determine the candidate targets in images; finally,the real target was extracted by using image sequences and improved neighborhood judgement,which can verdict the size of neighboring region by analyzing the motion characteristics of candidate targets. Experimental results showed that the proposed algorithm can greatly reduce the number of candidate targets and detect infrared dim small targets under complex background accurately and effectively.