Camera/projector-based document/object capture system using structured light: Reflectance map image quality assessment and design of structured light patterns and analysis algorithms

Yang Lei, Purdue University

Abstract

The use of mobile phone camera technology in systems for capture of documents or three-dimensional objects is becoming increasingly popular. The product TopShot LaserJet Pro M275 MFP introduced by the Hewlett-Packard Company (HP) in 2011 is an example of such a system. While providing more flexibility and potential applications, such systems pose special image quality concerns due to the fact that the underlying camera and illumination technologies are inherently low-cost. A specially designed composite target and an automatic analysis tool for image quality inspection of camera-based document/object capture systems will be introduced. The method automatically finds all the components of the target and analyzes various image quality metrics. This novel composite target and the tool can be used in both the research laboratory and on the manufacturing line, and were very effective in picking out units with defects. The criteria and thresholds for final production are chosen based on visual inspection of image data from hundreds of pre-production units. Currently, our solution is used in mass manufacturing, and shows effectiveness in failing units with low image quality, as well as saving manufacturing time. Speaking of applications, 3D shape reconstruction is one of the most important topics in computer vision due to its wide field of application. Among various technologies, structured light is considered to be one of the most reliable techniques. A low-cost structured light 3D reconstruction system is built around HP Topshot at Purdue University and the automatic 3D capture software is developed. The problem of finding correspondences is addressed by the newly designed 6-symbol M-array pattern with guaranteed minimum Hamming distance of three among all 3-by-3 windows. The decoding algorithm allows successful identification of most symbols, and corrects one possible error or missing symbol in each 3-by-3 window. To shorten the working distance, two pico-projectors with their field of view overlapped are used to cover the platen which holds objects. To separate the projected patterns from two projectors, we use different colors in each binary M-array pattern. The optimal values of the two colors are determined by the pattern search method with the presence of noise, which is modeled as multivariate Gaussian noise. A spectral analysis based model of the projector-camera system will be presented. The camera sensor's responses to the projector are measured after linearization with gray balance curves. The system noise characteristics are also measured for different input colors. After being properly calibrated, based on one image shot of the object with binary M-array patterns projected on it, the system can calculate the 3D coordinates of points on object surface, and therefore obtain the 3D shape of the object.

Degree

Ph.D.

Advisors

ALLEBACH, Purdue University.

Subject Area

Electrical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS