Infrared dim target detection method based on the fuzzy accurate updating symmetric adaptive resonance theory

Abstract

Motivated by the human visual system, we propose an infrared dim target detection method that is based on a fuzzy accurate updating symmetric adaptive resonance theory network. From the bottom-up perspective, the regions of interest (ROIs) are extracted using a difference of Gaussians at multiple scales and our designed ROI model in a saliency map. From the top-down perspective, five feature categories are extracted using the ROI model, which are used to train the proposed Fuzzy AUSART network. The well-trained network realizes the true identification of all ROI candidates. The results of the receiver operating characteristic (ROC) curves verify that the proposed method can better adapt to different circumstances and targets in our experiment. The average detection accuracy of the Fuzzy AUSART is improved by 15.4%, and the average F1 index of the proposed method is higher than six typical comparison methods by more than a factor of 2.48.