Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Mathematical representation of human visual perception of experience and its application in the depth estimates

Author LinXiangMing
Tutor WeiHui
School Fudan University
Course Computer Software and Theory
Keywords Perceptual Experience 3-D Information Depth Estimation
CLC TP391.41
Type Master's thesis
Year 2010
Downloads 55
Quotes 0
Download Dissertation

The paper presented the mathematical expressions of the human visual perceptual experiences, and put them into the 3-D depth estimation.Firstly, the human visual perceptual experience was denoted mathematically. The research how the 3-D coordinates of the points, the lines and the planes and the depth in the 3-D space are derived from the 2-D coordinates of the points, the lines and the planes in the single 2-D image has been a ill-posed problem, because one dimensional information was lost in the imaging process from the 3-D space to the 2-D world and the 3-D information can not be got only from the 2-D information. But the human visual system can finish many missions which the machines can not finish, and the answers the human visual system poses are usually vague and not precise. Inspired by the operating principle of the human visual system, the author don not think the research how the 3-D information are derived from the 2-D information is a ill-posed problem. In the paper, a mathematical model was posed to represent the human visual system, and the human visual perceptual experiences were denoted by the mathematical formulae. The mathematical model includes a imaging model whose principle are similar with the human visual principles. And in the model the simple 3-D objects (include the straight line, the intersecting lines and the plane) were imaged, and the paper gained the length, the angle and many geometrical characteristics from the 2-D information in the 2-D imaging plane. Some simple forms of the human visual perceptual experiences under some circumstances were denoted by the mathematical formulae, and the reason why the human perceptual experiences can be gained quickly was also investigated.Secondly, how depth is estimated from the 2-D information was researched. Depth estimation from the 2-D information are applied in many fields. Inspired by the depth estimating method of the human visual system, the paper posed a new method to estimate the 3-D depth. Specifically, the robot devised with a monocular camera took a sequence of photos while moving through a unknown scene. And two photos were chosen from the sequence of photos and corner-detected and matched. And for the every pair of matching points, the 3-D depths of the points whose projections are the corners can be calculated by using the novel calculating method the paper would posed. So the depth map of the 3-D scene can be built.

Related Dissertations
More Dissertations