SAR Automatic Target Recognition based on Decision Fusion

Abstract:
In this paper, we propose a novel automatic target recognition algorithm based on decision fusion. First, image chips are represented as feature vectors by 2-D wavelet transform and principal component analysis. Second, the feature vectors are classified by using algorithms of support vector machine, AdaBoost, k-nearest neighbor and Gaussian mixture models independently. Then, ranking based decision fusion algorithm is proposed to perform decision fusion. Final classification decisions are gained from the output of fusion algorithm. The experimental results show that with decision fusion, the probability of correct classification get an improvement of approximately 0.5% over the best individual classifier performance.