基于目标识别的HEVC分割算法优化与实现  

Optimization and Implementation of HEVC Segmentation Algorithm Based on Target Recognition

作　　者：吕昊 郭江宇[1] 靳文兵[1] 张宾 王宇新 LYU Hao;GUO Jiangyu;JIN Wenbing;ZHANG Bin;WANG Yuxin(North Automatic Control Technology Institute,Taiyuan 030006,China;Second Military Representative Office in Taiyuan,Taiyuan 030006,China)

机构地区：[1]北方自动控制技术研究所,太原030006 [2]驻太原地区第二军代室,太原030006

出　　处：《火力与指挥控制》2022年第5期177-182,共6页Fire Control & Command Control

摘　　要：为了优化视频信号在复杂网络环境下由于网络拥塞出现的数据丢包等情况,提出了一种基于目标识别的HEVC(high efficiency video coding)分割算法的优化方法,提高了视频编码的压缩效率,并实现了算法在全国产平台上的移植运行。利用深度学习进行目标识别,确定感兴趣区域,再通过指导不同区域采用不同编码单元分割深度,提高整体压缩效率。实验结果表明,该优化方法可以有效提高整体压缩效率,并完成在全国产化平台的运行实现,实现整体技术的自主可控。In order to optimize the data packet loss of the video signal due to network congestion in a complex network environment,an optimization method of the HEVC segmentation algorithm based on target recognition is proposed,which improves the compression efficiency of video coding and realizes that the algorithm is produced nationwide. Porting operation on the platform. First,use deep learning for target recognition and determine the region of interest,and then guide different regions to use different coding units to segment the depth to improve the overall compression efficiency.Experimental results show that the optimization method can effectively improve the overall compression efficiency,and complete the implementation of the national production and chemical platform,and realize the independent control of the overall technology.

关 键 词：HEVC 目标识别 ROI 自主可控 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]