Research On Intelligent Ship Perception Video Data Compression Technology

Intelligent ship perception video is an important basis for realizing ship remote navigation,and efficient data compression is the basis to guarantee real-time and reliable output between ship and shore.According to the characteristics of intelligent ship perception video,optimizing the coding process with high computational complexity is one of the important methods to improve the performance of intelligent ship perception video compression.The main research of this thesis is the video compression technology of intelligent ship perception.Through the use of High Efficiency Video Coding(H.265/HEVC)and Screen Content Coding(SCC)technology to improve the compression efficiency,and to optimize the compressibility,the following research work is completed:The characteristics of shipborne visible light camera video,shipborne infrared camera video and shipborne radar digital video are analyzed.The redundant data in the videos are analyzed in terms of both spatial redundancy and temporal redundancy.High Efficiency Video Coding technology is used to compress shipborne visible and infrared camera video,and Screen Content Coding technology is used to compress shipborne radar digital video,and the compression performance of both is experimentally studied,and the main reasons for the compression time delay of shipborne perception video are summarized.In order to optimize the compression performance of shipborne visible camera video and infrared camera video,an optimization algorithm of intra coding delay of shipborne visual sensor video was proposed by predicting the partition structure of Coding Unit(CU)in advance combined with the deep learning.And a Coding Unit partition data set for training Coding Unit partition prediction model of shipborne visual sensor video is established.The experimental results show that compared with the official High Efficiency Video Coding testbed HM16.17,the total compression time of the algorithm decreases by about 45.49%,while the Bj（？）ntegaard Delta Peak Signal to Noise Ratio(BD-BR)increase by only 1.92%,and the Bj（？）ntegaard Delta Peak Signal to Noise Ratio(BD-PSNR)descents 0.14 d B on average.Considering the difference between Screen Content Coding and High Efficiency Video Coding in the Coding Tree Unit(CTU)partition,this thesis proposes an intra coding delay optimization algorithm for shipborne radar digital video.Meanwhile,a depth range prediction model based on deep learning for shipborne radar digital video Coding Tree Unit partition is designed.In order to further improve the compression performance of the algorithm,a Coding Tree Unit partition depth range dataset based on shipborne radar digital video is produced in this thesis for the training of the model.The experimental results show that compared with the official Screen Content Coding test platform SCM8.7,the intra coding delay optimization algorithm for shipborne radar digital video proposed can save 39.84% of coding time on average,while Bj（？）ntegaard Delta Peak Signal to Noise Ratio and Bj（？）ntegaard Delta Peak Signal to Noise Ratio only lose 2.26% and 0.19 dB.