CAJ | 학술논문

舰船目标自动识别通常需考虑多个特征，而复杂的特征往往需要适合的分类器与之相适应。本文借助已有的一种可组合多种特征和多种分类器的通用分类器，验证其在舰船识别中的有效性。该通用分类器将多分类问题转化为多个二分类问题，利用多个二分类器对舰船各特征进行独立识别，最后根据投票规则决定目标识别结果。本文以二分类BP网络作为多神经网络分类器的基分类器，对航空母舰、驱逐舰、护卫舰、客船、集装箱、民用货船6种船只类型进行了识别。识别结果表明，由多个二分类BP网组成的多神经网络分类器平均分类精度为89%，该通用分类器在实践中有效。
함선목표자동식별통상수고필다개특정，이복잡적특정왕왕수요괄합적분류기여지상괄응。본문차조이유적일충가조합다충특정화다충분류기적통용분류기，험증기재함선식별중적유효성。해통용분류기장다분류문제전화위다개이분류문제，이용다개이분류기대함선각특정진행독립식별，최후근거투표규칙결정목표식별결과。본문이이분류BP망락작위다신경망락분류기적기분류기，대항공모함、구축함、호위함、객선、집장상、민용화선6충선지류형진행료식별。식별결과표명，유다개이분류BP망조성적다신경망락분류기평균분류정도위89%，해통용분류기재실천중유효。
Multiple features are usually needed for ship recognition, which means that suitable classifiers are needed to cope with these features. Based on a general classifiers proposed by other researchers, we completed ship recognition and verified the effectiveness of this classifier which can combine multiple features and classifiers. This classifier changes multi-class recognition into different two-class recognition issues, and then gives voting to decide the final result. Using BP network as the base classifier, we make identification for six types of ships: aircraft carriers, destroyers, frigates, passenger ship, container and civil ship. The results show that the multi-classifier reaches classification accuracy as high as 84% for ship target recognition, the classifier proposed before is effective in practice.