Part detection with a 2D smart camera utilizing traditional methods : Physical and algorithmic foundations of machine vision systems

Abstract

This study investigates the physical and algorithmic foundations of 2D machine vision systems in part detection. The study begins with a literature review outlining the core components of a machine vision system, such as lighting, optics, and image processing, as well as exploring its industrial applications and benefits, while contrasting traditional rule-based algorithms with modern deep learning-based methods. The primary purpose of the research is to demonstrate the capabilities and limitations of 2D machine vision utilizing traditional algorithms when exposed to real-world complexities rather than ideal conditions and objects. To achieve this, an experimental study was conducted using a 2D smart camera and a traditional rule-based machine vision tool. Visual and coordinate data were collected from practical test scenarios involving reflective surfaces, translucent materials, low-contrast environments, and varying object heights to demonstrate the boundaries of 2D perspective and traditional edge detection. The findings demonstrate that traditional 2D systems depend strictly on clear optical gradients and that their lack of depth perception causes perspective distortions when object placement varies. In conclusion, while traditional 2D vision is highly efficient in controlled low-level settings, overcoming physical constraints in dynamic and complex manufacturing scenes requires considering more advanced technologies like deep learning-based methods and 3D vision systems.