| Virtual reality technology,rapid developing in recent years,can generate simulation environment by computer and it is a kind of system simulation tec hnology with multi-source information fused and interactive three dimensional dynamic visual and physical behavior.As the key technology of virtual reality,three dimensional reconstruction can establish a suitable mathematical model f or the actual object for the computer representation,processing and analysis.Inthe three dimensional reconstruction technology,indoor three dimensional re construction realized by mobile robot can be applied in exploring the unknow n environment,such as plants full of harmful gas,the dangerous house of dis repair and other scenes that endanger the lives.That's why this direction has been widely researched and applied in recent years.Depth map,as an universal expression for three dimensional scene infor mation,describes depth information of the scene,that is distance information,and it is an important and indispensable part in three dimensional reconstructi on.This paper demonstrates relevant research of the depth map information pretreatment and compression transmission technology in the three dimensional scene reconstruction,and by installing Microsoft Kinect camera on the mobile robot,it builds up a set of indoor three dimensional scene reconstruction sy stem based on mobile robot.The main work and contributions of this paper are as follows:(1)Firstly,this paper proposes a Kinect depth figure recovery plan: on th e one hand,it suppresses the noise in depth images through the improved ada ptive median filter;and on the other hand,it improves the FMM(Fast March ing Method)algorithm,and it fills the missing parts of depth map information by the direction filtering mechanisms.In the experiments,the image qualit y of the Kinect depth map recovered by the solutions in this paper is signifi cantly improved compared with the original image,and on the speed it also has certain advantages compared with similar algorithms;(2)Secondly,as for the Kinect transmission,this paper designs the comp ression transmission based on the special properties of the depth data,and also on the HEVC(High Efficiency Video Coding): firstly,it introduces the appr oximation three dimensional volume algorithm to suppress unstable features in time domain of the Kinect depth sequence,so as to improve the quality of r eference frame of the image generated by HEVC,and to reduce the prediction residual;Secondly,In HEVC inter-frame prediction,it directly predicts residu al operation for macro blocks that conform to the threshold,and skips from t he minimum matching module of search rate-distortion cost,in order to save the encoding time.Sequence compression coding for Kinect depth map demon strated in this paper has obtained certain advantage in the bit rate and PSNR and three dimensional reconstruction results as well,compared with the dire ct compression processing.(3)Finally,the paper designs the system process for three dimensional reconstruction based on mobile robot : it collects the texture and depth informa tion by the Kinect carried on the mobile robot and then conducts pre-treatment process as in(2)for the collected depth map sequences.After recovery,it co nducts compression coding for the sequences,and remotely decodes through TCP/IP protocol.At last,the decoded sequences will be three dimensional reco nstructed,and its effect will be displayed on the interface.Through the interfa ce,the operator can choose whether to join the color information,whether to r efresh the reconstruction cubes and choose from the cube size and update the Angle of view and other functions. |
