A methodology for automatic detection and improving Datamatrix code decoding rate in industry

Abstract

Datamatrix codes play a crucial role in enhancing productivity and efficiency in manufacturing processes. Ensuring accurate identification and decoding of these codes is paramount. However, challenges such as scratches on products, uneven color distribution, and strong reflections all negatively impact code readability. To overcome these challenges, this paper proposes an advanced method employing deep learning models Yolov8 and image enhancement techniques. Firstly, the Yolov8 model is utilized to detect and extract regions of images containing Datamatrix codes into smaller patches. Next, image enhancement techniques are applied to enhance the image of the Datamatrix code. Implementing this method leads to a significant improvement in code read rate, reaching up to 94.5%. These results illustrate the effectiveness of the proposed method in achieving more accurate and reliable Datamatrix code decoding.