23 - CAD-Based Vision Systems in Pattern Matching Process

Abstract

This chapter presents strategies concerning vision task programming using a computer aided design (CAD) database. It describes the practical applications for 2D, 2D 1/2, and 3D vision models, which demonstrate that integrated vision system are promising. Precise, robust, and flexible learning can be achieved, allowing high recognition rates. Thus, offline programming techniques, applied first to NC machines and robots, can be extended to machine vision systems. The complexity of the scene, the CAD database structure, the time allowed for processing, and the task definition will orientate the matching process differently in each case. This chapter proposes generating vision oriented models while learning on the CAD workstation and building CAD oriented models during recognition on the vision system. The complete simulation of the vision programming steps such as sensor placement, CAD model learning, and part recognition on a vision workstation enable the learning process to be optimized globally. The learning process can be long, but it is carried out offline. On the contrary, the recognition process is designed to be executed on the production line in real time. This chapter presents the structure and operation of the integrated vision systems, and focuses on CAD and vision model matching. It surveys CAD and vision models, and proposes strategies to be applied in the learning and recognition processes. It also provides practical applications and other CAD-based vision applications that are not concerned with mechanical engineering. This chapter closes with an account on future developments.