Image interpretation and object recognition in manufacturing

Some of the characteristics and peculiarities of the vision problems in the manufacturing domain are examined. Work in developing and experimenting with a framework for intelligent interpretation of images in manufacturing applications is presented. The framework encompasses the issues surrounding the knowledge representation, reasoning and control, and maintenance of consistency, and incorporates three types of knowledge: knowledge about objects, about relations, and about reasoning and problem solving. Knowledge representation in this system is described, and the issues of reasoning, control, and determining model-to-image transformations are addressed. An example that involves determining the identity, position, and orientation of a part is provided. A brief review of related work in knowledge-based image interpretation and CAD-based vision is included.<>