基于卷积神经网络的激光视频压缩研究

Research on laser video compression based on convolution neural network

传统激光视频压缩方法存在压缩率小、丢失重要信息、耗时间长的弊端,提出了基于卷积神经网络的激光视频压缩方法。首先对激光视频压缩工作原理进行分析,找到当前激光视频压缩效果不佳的原因,然后采集激光视频,并进行相应的处理,采用卷积神经网络对激光视频关键帧之间的相关性进行拟合,根据拟合结果对激光视频进行有效压缩,最后在Matlab平台上进行了激光视频压缩的仿真对比测试,结果表明,在保证重要信息的条件下,卷积神经网络可以尽最大幅度对激光视频进行压缩,减少了激光视频压缩时间,激光视频压缩效果明显优于当前经典压缩方法,具有十分广泛的应用前景。

The traditional laser video compression methods have low compression rate,loss of important information and long time consumption,a convolution neural network is proposed laser video compression method.Firstly,the working principle of laser video compression is analyzed to find out the reason why the current laser video compression effect is not good.Then the laser video is collected and processed.The correlation between the key frames of laser video is fitted by convolution neural network.According to the fitting results,the laser video is effectively compressed.Finally,the laser video simulation is carried out on the Matlab platform,the results show that under the condition of ensuring important information,the convolution neural network can compress the laser video as much as possible,which reduces the compression time of laser video.The compression effect of laser video is obviously better than the current classical compression methods,and has a very wide application prospect.