基于SCNN的红外弱小目标检测算法研究

Research on the algorithm of infrared dim small target detection based on SCNN

提出了基于结构化特征卷积神经网络(Structural Feature convolution neural network,SCNN)的红外弱小目标检测算法。通过将红外弱小目标结构化特征引入CNN网络中,去除CNN网络的池化层、改变卷积扩展性、并加入分层融合机制,克服了CNN会损失小目标本身的信息和无法检测弱纹理小目标的问题。实验结果表明:本文提出的方法相比同类方法具有更高的检测率,并且对于不同场景具有较强的鲁棒性。

An infrared dim small target detection algorithm based on structural feature convolution neural network(SCNN)is proposed.By introducing the structural features of infrared dim small target into CNN,the pooling layer of CNN is removed,the convolution expansibility is changed,and the layered fusion mechanism is added to overcome the problems that CNN will lose the information of dim small target itself and cannot detect dim small target.The experimental results show that the proposed method has higher detection rate and stronger robustness for different scenes.