Real-time Image Processing for Production Tracking in Manufacturing Plant

Tracking the number of work-in-process (WIP) and finished goods is a big challenge in factories. Additional workers are required to perform this task. It is a tedious and error-prone task for the workers. The production planners need to add additional finished goods to their production schedules to account for unknown mismatched production, which wastes a lot of money, workforce, and time in manufacturing. In recent years, there has been an advancement in computer vision, such as object recognition and detection. This research proposes a method using image processing to count the WIP in safety shoe manufacturing. This technique combines Yolov4-tiny as a detection engine and Deepsort algorithms along with OpenCV to track, count, and classify products for a real process in a factory. Special regions are assigned in the video frame to count the shoes and bad-quality shoes. The experiment is demonstrated on 5 videos streaming from the CCTV inside the factory. The results have shown that besides the blurry streaming, the proposed algorithm can track and classify both categories of shoes effectively.