3D Simplified Point Cloud Model Library Research Based On Stereo Vison

Binocular vision can be applied to many human imperceptibility occasions, whenpetrochemical enterprise leakage, can use stereoscopic vision for machine parts inspection andrepairing. To make the robot can quickly find specific parts,it is significant to build apetrochemical enterprise model. Because point cloud model contain sampling point position,color attribute, and can build any topology information, data simple, easy to store and edit, wecan make full use of the advantages to establish3D model.At first, this paper uses binocular stereo camera to obtain the depth of the chemicalcomponent information, namely, point cloud information, and then introduction of point cloudinformation required by the pretreatment technology and feature point extraction technology,finally establish the3d point cloud model library. In this paper, the research content and theinnovation theory mainly includes the following:1、According to the computer to obtain the depth of the image extracted parts and parts forthe complete point cloud information, and can make the point cloud image translation, rotation,scaling, three-dimensional display platform.2、Observation noise and manual removed, in addition to calculate the distance of theautomatic remove noise. For the simplified point cloud issue, according to the established8-neighborhood structure, the use of a gaussian curvature estimation algorithm for gaussiancurvature value, thus extraction can keep characteristics of the simplified method of point cloud.3、Using curvature calculation method for the rapid identification of point cloud modelboundary characteristics, using the frame theory to extract the key points of the point cloudmodel, the key points and boundary point combination for full set of feature points, andboundary fitting the feature points, obtained the boundary of the model shape characteristicscurve.4、Took a quick identification of a point cloud model of boundary feature of the algorithm,using the frame theory of feature points extraction to extract the key points of the point cloudmodel, the key points and boundary point combination for full set of feature points, and thefeature point boundary fitting, reflect the characteristics of the model shape boundary curve.