Human-robot collaboration can be a powerful tool for increasing productivity in production systems by combining the strengths of humans and robots. Assembly operations, in particular, have shown great potential for utilizing the unique abilities of both parties. However, for robots to efficiently perform assembly tasks, components and parts must be presented in a known location and orientation, which is achieved through a process called parts feeding. Traditional automation methods for parts feeding, such as vibratory bowl feeders, are limited in their ability to accommodate variations in parts design, shape, location, and orientation, making them less flexible for use in human-robot collaboration. Recent advancements in machine vision technology have opened up new possibilities for flexible feeding systems in human-robot assembly cells. This paper explores the application of the vision system in the collaborative robot ABB Yumi and its ability in object detection. In this case, the characteristic of the vision system was determined experimentally by changing the light intensity on the test rig. The system was validated, if the angle of incidence of light affects the stability of the vision system. The results of the study demonstrate the efficiency of vision system in collaborative robot and provide insights into its industrial application.