3-D image recognition for distorted and cracked objects: a proposal of six-surface-image integrating method

Abstract

Thus far, the pattern-matching method has mainly been adopted for visual recognition, applicable only to objects which can be put into patterns. Objects which have individually variable sizes are difticult to process with satisfactory results. For example, superconductive products are difficult to put into patterns because of their expansion, shrinking and cracking in the baking process. This is one of the reasons why the development of these kinds of objects still cannot be automated. To solve this problem, especially for automatic manipulation of these kinds of products, this paper proposes an image processing system including the six-surface-image integrating-method. Avoiding individual differences, the system can perform the reasoning and recognition of bending, twisting and cracks of a brick-shaped object through integrating the image information of its six surfaces. The effectiveness of the system proposed in this paper was confirmed by the results of experiments.