Development of an Automated Defect Detection System for Plastic Bottles in the Production Line Using Computer Vision

This study proposes a SMART defect detection system for a conveyor belt of plastic bottles, aiming to improve product quality and operational efficiency. This system utilizes computer vision, and IoT technologies to identify and address defects such as dents, missing caps, and tumbled bottles in real-time. The system was trained on a dataset of 2088 images, achieving a mAP of 83.60 %, indicating high accuracy in defect detection. Precision of $\mathbf{8 9. 2 0 \%}$ and Recall of $\mathbf{7 4. 1 1 \%}$ further demonstrate the system's effectiveness. The system successfully identified and diverted defective bottles in real-time. Future enhancements that can be considered may include the use of a robotic Arm to automate the loading of bottles onto the conveyor belt, a mobile robot to transport bins of defective bottles for disposal, and/or a dedicated housing unit to maintain consistent lighting and background conditions. Overall, the proposed system offers a robust solution for improving product quality and reducing waste in plastic bottle manufacturing. By integrating advanced technologies, it provides a practical and efficient approach to quality control.