IR Target Detection Based on Kernel PCA and Quadratic Correlation Filters

In this paper a novel approach for infrared target detection based on kernel principal component analysis (KPCA) and quadratic correlation filters (QCF) is proposed. The feature extraction for training images and detecting IR image are first implemented using KPCA, and then QCF based on the Fukunaga Koonz transform is applied to the extracted principal component vectors, the detecting sub-images segmented from detecting IR image corresponding to the output of QCF above a given threshold are considered as required IR target. The proposed method has a good ability to restrain IR target noise so as to improve detecting accuracy. Experiments on the real-world IR images show that the proposed approach is effective and efficient.