Vision is an important means for human to see the world and understand the world. About 75% of the world's human interest is derived from the visual system. This shows that the amount of visual information is huge, and it is proved that human beings have a high utilization rate of visual information.

　　According to the different ways of information acquisition, handheld 3D scanner can be divided into two categories: contact and non-contact. Using the contact probe directly touch the surface of the object, according to the geometric structure of the measuring device coordinate measuring head; non contact is mainly based on the principle of computer vision, 3D information acquisition target from image captured by the camera.

　　Computer vision is the use of computers to achieve the visual function of the objective world of the three - dimensional scene perception, recognition and understanding. The main research object of computer vision can be summarized into two. They complement each other. The first goal is to build a computer vision system to complete all kinds of visual tasks. That is to make what is function with various types of visual sensors (such as CCD, CMOS camera device etc.) image scene, and perception and recovery of objects in the environment of 3-D geometric properties, structure, motion, attitude and position of each other, identification, description, explanation of the objective scene. And then make a decision. The main goal of this paper is the technical mechanism.

　　Application of portable 3D scanner in many industries, and optical, mechanical and electronic, automatic control, data acquisition and processing, image processing and other areas of knowledge integration, the three-dimensional hand-held scanner is a high-tech product has great application value.

　　After obtaining the point cloud data of the sample points, it is necessary to describe the model with polygon, curve and surface. The triangular mesh model is used as the 3D reconstruction method, and the triangulation of regular data and scattered data is studied.

　　According to the structure light scanning rule data, propose a heuristic search algorithm based on contour correlation; and of multi joint scanned scattered data first cut to obtain cross-sectional contour, and then sort the line, then building a triangular mesh in the two order line between.