Visual Recognition System

A visual recognition system is installed in the glass fiber creel, with its core function to realize automated and precise management of yarns and yarn packages.

1. High-Precision Positioning and Recognition

Equipped with industrial cameras and deep learning algorithms (such as the improved YOLO model), the system can accurately identify the positions of yarn packages and yarns with a positioning accuracy within 3 mm. Even if yarn packages are randomly placed or offset on the creel, their coordinates can be obtained in real time, providing precise guidance for robots to perform automatic package changing, yarn picking and placing. It replaces manual positioning and eliminates positioning errors.

2. Real-Time Status Monitoring and Abnormal Early Warning

It can conduct online detection of yarn status such as yarn breakage, hairiness and defects, and even identify tiny yarn flaws at the 0.1 mm level. Meanwhile, it monitors the remaining yarn of packages and appearance defects (such as stains and loops). Once an abnormality is detected, an alarm is triggered immediately, preventing broken yarn and defective products from entering subsequent processes and increasing the yield rate to over 99%.

3. Non-Contact and High Adaptability

Adopting optical imaging and image processing technology, the system operates in a non-contact manner with yarns and yarn packages, avoiding abrasion or contamination to glass fibers. It can also adapt to the complex production environment of the creel (such as high-speed drawing and light changes), eliminating interference through light source control and image preprocessing (filtering, segmentation) to maintain stable detection.