基于深度学习和改进证据理论的海上多源舰船信息融合识别方法

Offshore multi⁃source ship information fusion recognition method based on deep learning and improved evidence theory

为了解决复杂环境下基于单一舰船信息进行目标识别准确率较低,以及多源舰船信息高冲突时无法有效融合识别的问题,提出一种基于深度学习和改进证据理论的海上多源舰船信息融合识别方法。主要从两方面入手:首先利用深度学习高效特征学习能力实现更加准确的分类识别;然后通过改进的证据理论实现多证据体的高效正确融合。高悖论证据融合实验结果表明,相比于其他融合方法,文中方法融合结果具有更高的概率分配值。同时,在不同信噪比条件下对单模式识别以及文中融合识别方法进行测试,文中方法在噪声情况下仍能比单模式平均水平高出6.53%的识别性能。因此,利用文中融合识别方法能够提高舰船目标识别系统的识别准确率和鲁棒性。

In view of the low target recognition accuracy based on single ship information in complex environment and ineffective fusion recognition of high⁃conflict multi⁃source ship information,a method of offshore multi⁃source ship information fusion recognition method based on deep learning and improved evidence theory is proposed.This paper mainly starts from two aspects.The deep learning efficient feature learning ability is used to carry out more accurate classification and identification.Then,the improved evidence theory is used to fuse the multiple evidence bodies efficiently and correctly.The results of high paradox evidence fusion have higher probability assignment value than those of the other fusion methods.In addition,under different SNR(signal to noise ratio)conditions,the single mode recognition method and the text fusion recognition method were tested.The recognition performance of the proposed method is still 6.53%higher than the average level of the single mode recognition method in the noise environment.Therefore,the proposed fusion recognition method can improve the accuracy and robustness of ship target recognition system.