Research On3D Reconstruction Based On Multiple Surround Camera Array

Vision-based processing and visual recognition is a significant research direction in thearea of mobile robot. As the technology is advancing rapidly, and the complexity of workingenvironment along with the task requirement is escalated, relatively simply two-dimensional(2D) visual information is struggling in fulfilling the practical requirements. In thesecircumstances, an increasing number of researchers have been devoted to three-dimensional(3D) research projects. Compared to conventional3D geometric modelling, image-based3Dreconstruction enables the construction of photorealistic3D models simpler and quicker.3D reconstruction techniques have progressed prominently regarding the accuracy andeffectiveness of its performance. However, in daily mainstream applications,3D reconstructiondevices demand high requirement of reconstruction range. Aiming at this issue, this papercarries out a4-camera array approach, by analysing image acquisition, image preprocessing,image calibration and3D reconstruction processes in details. By presenting the reconstructionexperiments of3ds max scene rendering, this paper introduces an algorithm with enhancedreconstruction performance of simple/complex realistic scenes without needing to adjustinternal parameters. The research on the4-camera3D reconstruction modelling is briefed asfollows:Firstly,An introduction to the fundamental3D reconstruction theories, and identificationof current states and issues of3D modelling.Secondly,A comprehensive illustration of major3D reconstruction methods and themainstream in midst.A average threshold edge is presented to detected edge,which is good forhigh precision of within all the camera parameters.Doing SIFT registration of school labbuilding simulation experiment,analyzing the effect of changing threshold.Thirdly,An introduction to the structure and process of3D reconstruction systems, andexperimental analysis of relevant approaches. Being the core of this paper, put forward afour-mesh camera method for three-dimensional reconstruction, greatly improving theregistration accuracy.Improved array can withstand more due to camera position andthree-dimensional object surface features of the effect of image registration.Finally,An introduction to the hardware and software platform, and an explanation ofexperiment environment, data and demos, with the result analysed eventually.