Generation of synthetic images for training automated visual assembly inspection algorithms
Abstract
Visual assembly inspection can provide a low cost, accurate, and efficient solution to the automated assembly inspection problem, which is a crucial component of any automated assembly manufacturing process. In this work, CAD information is used to generate synthetic images of an assembly. The synthetic images are generated to train a new multiscale image processing algorithm that is used to detect errors in the assembled product. The CAD information guides the inspection algorithm through its training stage by addressing the different types of variations that occur during manufacturing and assembly and by identifying areas within the image that can be used for error detection. In addition, since the performance of such an inspection system is heavily dependent on the placement of the camera and light source(s), new algorithms are presented that use the CAD model of the finished assembly for placing the camera and light source(s). Using synthetic images in the training process adds to the versatility of the technique by removing the need to manufacture multiple prototypes and suggesting image parameters that optimize performance. Once trained on synthetic images, the algorithm can effectively detect assembly errors by examining real images of the assembled product.
Degree
Ph.D.
Advisors
Maciejewski, Purdue University.
Subject Area
Electrical engineering|Computer science|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.