Color recognition-based computer vision technology refines traditional manufacturing factory processes

Abstract Currently, various smart manufacturing technologies are in operation in high-tech industries. However, this is still out of reach for many traditional manufacturing industries, particularly those in developing countries. Therefore, this study is dedicated to using new technologies to improve the process inspection in high-manpower production lines in a low-cost and efficient manner. In this experiment, the assembly process for 67 engines was continuously detected. Each engine has 11 locking points. After cleaning the data, 1249 coordinate values were recorded. Based on the monitoring results obtained during the experimental process, the probability of successful color recognition was extremely high. First, it was found that an air tubing pipe connected by pneumatic tools could effectively replace object and motion detection as the basis for judgment. Second, it was found that doing so could also avoid shadows caused by tools or actions, thereby reducing the chance of misjudgment. This discovery not only reduces the number of cameras used, but also more accurately detects whether the screws are locked. Moreover, the hardware specifications required for color recognition are not in strong demand, and the recognition rate can reach 1 μs. In the future, it can be widely used on various tools by changing the finger sleeves or marking the tool with a ribbon without hindering the operation. In the future, applying cost-effective and easy-to-maintain smart manufacturing methods to traditional manufacturing production lines will also effectively reduce labor shortages caused by factors such as the gradual decline in the workforce and the impact of disease and provide a safer and more efficient working environment.