Fast3D Reconstruction Based On Black And White Code Of Structured Light

It is one of the current hottest research topics of the computer vision to reconstruct objects’3D shape from2D images(i.e.,3D reconstruction), and it is also the important research content and ultimate goal of the computer vision. The traditional method is finding correspondence among two or several views to restore3D surface information accurately, which requires tremendous computational power and with low efficiency and bad real-time performance. It becomes an even harder problem to reconstruct object model rapidly if dynamic objects exist in the scene. It cannot satisfy the demand of the emerging market such as the virtual reality. So it is necessary to find the method to reconstruct the real-time3D information of object surface.when using the traditional method based on black and white coding structured light, we should normally project multiple patterns for reconstructing only one image, so the operation is relatively complex and it is with bad real-time performance. This paper proposed a real-time3d reconstruction method based on black and white stripes coding structured light, adopting integrate static reconstruction with dynamic reconstruction. At first according to classical black and white decode method, get initial three-dimensional information of the object. Then begin with the first image, by finding out the corresponding feature points of adjacent two images and adopting relation formula, decode the stripe serial number of the next picture rapidly. In other words, the follow-up3D reconstruction of each image only need photographed one image for decoding to realize3D reconstruction, so it greatly reduces the operation’s complexity and the time of decoding.This paper aims to study fast3D reconstruction technique based on black and white stripes coded structured light, the main job content is followed as:1. Choose the black-white stripe structure light and design a rapid acquisition method by rotating object, use one-shooting method to take images. Simple and convenient operation makes fast reconstruction possible.2.The existing camera calibration technique is mature, so this process is adapted form Zheng You Zhang’s classic calibration method. For projector calibration, we use extracting two kinds of nodes and twice fitting straight-line (Horizontal straight-line fitting for stripe border feature points and longitudinal straight-line fitting for black and white corners of the planar checkerboard pattern image) then solving intersection of these two types of straight-lines to extract projection feature point fast and accurately, then use plane cross-ratio invariability.3. Reconstruct static image, as the first frame reference information for subsequent reconstruction. Firstly extract the stripe border feature points on the basis of gray information of black and white stripe structured light which is projected on the object surface, then draw calculation principle according to the understanding of the classical black and white decode method, and decode stripe serial number rapidly and accurately, thus figure out the three-dimensional information of the first static image using3d reconstruction principle.4. Reconstruct Dynamic image. First extract stripe border feature point of the acquired image of the object moving continuously, since we have obtained the3D information of the points of the first frame static image, we can decode the information of the second frame image from the first frame image according to the stripe matching method between adjacent images proposed in this paper, finally figure out the three-dimensional information of the second image according to3d reconstruction principle. By analogy, decode subsequent images through matching the points’information of adjacent two pictures, thus get the3d information of the object from a frame to another.Finally, the experiment realize object’s3d reconstruction based on black and white coding structure light, the experimental results show that it demands the lower cost of equipment, the operation is simple and the effect is good, have basical real-time performance. The method in this paper has a certain contribution for preliminary stage of industrial robot research.