Font Size: a A A

Active Stereo Vision Method Research Based On Structured Light

Posted on:2013-11-18Degree:MasterType:Thesis
Country:ChinaCandidate:S LiFull Text:PDF
GTID:2248330395490045Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
Computer vision simulates human eyes, understands and estimates the surrounding throughthe2d images. Furthermore, it can acquire the geometry information of the3d target object in thespace. The3d reconstruction of the target objects is an important research content of computervision. The traditional3d reconstrution, such as binocular vision, has the disadvantages of lowmatching precision and difficult to match in less textured area. And structed light, which is therepresentative of active vision, also has the disadvantages of low resolution and inefficient. In aword, the concentration of computer vision is achieving high-precision, robust andhigh-efficiency3d reconstruction.The binocular stero vision has the problem of low matching precision and difficult to matchin less textured area. This thesis presents research on these issues, bringing in structed light, anddesigns the active stereo vision system, which can weaken the surrounding influence on thetarget object, improving the matching precision, robustness and real time. The research content isas follows:1. Analysing the multicolour vertical stripes and the multicolor pseudo-random arraystructed light pattern. Describing the design principles, characteristics, advantages, and thereasons to choose these structed light.2. Bringing the multicolour vertical stripes pattern in stereo matching, using2-normconcept, comparing a certain width and vector normalized wave forms of the characteristicpoints nearby in the left and right images. Seeking th similar wave forms along the extremelyline, the characteristic points on the similarest vectors are the matching points.3. The multicolor pseudo-random array has randomness and window feature, andevery window in the pattern has an absoulte position. Using these properties, this thesis putforwards the multicolor pseudo-random array stereo matching method. Take the centroids ofthe array gridding as the characteristic points, and every characteristic point has a single code word in the array. This method take advantage of the coded structed light, can acquire thematching results quickly and accurately by only one shot.In the end, this thesis reconstrut the target objects in the active structed light stereo visionsystem.The results demonstrate the feasibility of the proposed methods, and analyzed the finalresults at the same time.
Keywords/Search Tags:encoded structed light, stereo vision, image rectification, structed light pattern stereomatching, 3-d calculation
PDF Full Text Request
Related items