【摘要】 水下目标识别在国防及国民经济中具有重要的作用。为了提高多神经网络分类器分类结果的有效性和可靠性,本文提出了一种利用多神经网络分类器输出向量来实现对各分类器进行加权的算法。舰船目标实测数据分类实验证明:基于该算法的多分类器融合技术能有效地提高目标识别的性能,同时选择适当的表决阈值又可提高分类结果的可靠性。因此,该算法在水下目标识别系统中具有一定的工程应用价值。更多还原

【Abstract】 Underwater target recognition is important to the national economy and defense. In order to improve the reliability and validity of multiple classifiers, a new weighting algorithm based on the output vectors of the classifiers is presented in the weighted voting scheme. Ship radiated-noises抯 classification experiment reveals that multiple classifiers fusion based on this algorithm can effectively enhance the performance of recognition and improve the reliability of the classification results by making choice of the voting threshold. This classification algorithm may have an important application value in underwater target identification.更多还原