Error Analysis Of Projective Structure

Projective reconstruction is the first step of 3D reconstruction, also the base of 3D reconstruction and calibration etc. The precision of the projective reconstruction directly affects 3D reconstruction precision. Therefore, studying influence factors of the projective reconstruction and analyzing precision have important significance for improving 3D reconstruction precision.From the Matlab simulation, this paper analyzes the effect factors to the projective reconstruction precision. The effect factors include the image noise, pre-process, the distribution of camera and image number, choice of camera parameter and scene complication.Image and pre-processing affect the precision of corner detection and corner match to affect the projective reconstruction precision. This paper analyzes the effects of image and pre-processing on corner detection and corner match, and compares two kinds of most common pre-processing methods: Gauss filter and median filter.Then, the paper studies how camera relative position and number of images affect the projective reconstruction precision. Precision of the projective reconstruction from two images is discussed. At the same time, two improved methods on corner match using rotation match and adaptive match are provided, they enhance success ratio of corner match.Then, the paper studies the precision of reconstruction from image sequence effect, and discusses how the consistency and the length of image sequence affect projective reconstruction precision. Finally, the effect of focal length setting on camera calibration and the effect of reconstruction precision on scene complication are discussed. In image sequence aspect, this paper discusses the effect of number and length on the projective reconstruction precision, and how to choose best image number and length for system.Finally, the paper analyzes how to choice the camera focus length and scene complication affects projective reconstruction precision, and then gets the conclusions that shooting by short focus length is better than long focus length for calibration. In scene complication aspect, the paper finds that great...