LAUSR

In the field of infrared search and track system, we find the prior art usually has poor robustness and adapts to a single scene. Inspired by the human visual attention mechanism, an effective small target detection method that can detect both bright and dark targets under multiple interference with low false alarm rate is presented in this paper. First, a saliency histogram map that roughly highlights the salient regions is obtained by frequency residual. Then, a multiscale histogram of oriented gradients difference measure map is constructed to enhance the target signal. Next, this map and the saliency map are multiplied to be a fused feature map. Finally, the targets can be obtained by using the adaptive threshold. Experimental results on four groups of test images demonstrate that our method can double the local signal to background ratio gain. Simultaneously, the receiving operating characteristic curves demonstrate both the effectiveness and robustness of this method.