基于深度学习的监控视频中船舶识别方法

Ship recognition in monitoring video based on deep learning

针对当前监控视频中船舶识别成功率低、无法进行在线识别的难题,为了对监控视频中船舶进行准确识别,提出基于深度学习的监控视频中船舶识别方法。首先对监控视频中船舶识别原理进行分析,采集船舶识别的监控视频,将船舶识别从背景中分割,然后提取船舶识别的不变矩特征,将不变矩特征输入深度学习算法中进行训练,建立监控视频中船舶识别模型,最后进行了多个监控视频中船舶识别验证性实验。实验结果表明深度学习算法可以准确对监控视频中的船舶进行识别,提高了监控视频中船舶识别成功率,误识率急剧下降,远低于当前其它监控视频中船舶识别方法,实时性要也要高于其它识别方法,是一种速度快、结果可信的监控视频中船舶识别方法。

Aiming at the problem of low success rate of ship recognition in surveillance video and the difficulty of online recognition,a ship recognition method in surveillance video based on deep learning is proposed in order to accurately recognize ships in surveillance video.Firstly,the principle of ship recognition in surveillance video is analyzed,the surveillance video of ship recognition is collected,and the ship recognition is segmented from the background.Then the moment invariant feature of ship recognition is extracted,and the moment invariant feature is input into the deep learning algorithm for training.Finally,the ship recognition model in surveillance video is established.Verification experiments of ship recognition in multiple surveillance videos.Deep learning algorithm can accurately identify ships in surveillance video,which improves the success rate of ship recognition in surveillance video,and the rate of error recognition decreases sharply.It is far lower than other methods of ship recognition in current surveillance video.The real-time performance of ship recognition is also higher than other methods of ship recognition in surveillance video.It is a fast method.Results The credible method of ship recognition in surveillance video.