Identification of buried landmines using Mel Frequency Cepstral Coefficients and Support Vector Machines

In this paper a new identification technique for buried landmine objects is presented. Most of the existing supervised identification methods are based on traditional statistics, which can provide ideal results when sample size is tending to infinity. However, only finite samples can be acquired in practice. In this paper, a proposed learning method, Support Vector Machine (SVM), is applied on landmine images which have two main categories. Firstly, the 2-D images are lexicographic ordered to 1-D signals, and then the Mel Frequency Cepstral Coefficients (MFCCs) and polynomial coefficients are extracted from these 1-D signals or from their transforms. Secondly, the SVM is used to match the extracted features in the testing phase to those of the training phase. Experimental results show that the recognition rate for features extracted from the Discrete Sine Transform (DST) of images contaminated by AWGN, features extracted from images plus the Discrete Cosine Transform (DCT) of images contaminated by impulsive noise and features extracted from images plus the DST of images contaminated by speckle noise achieve better performance compared with the other cases.