Abstract
The present invention discloses a defect detection method for plastic injection molding product, comprising the steps of: collecting product (102a) samples with normal and abnormal classes; collecting image dataset from the collected product samples, and enhanced with data augmentation (102b); annotating the image dataset (102c); applying data augmentation to the annotated dataset by a computational algorithm (102d) to increase the variation of the dataset; training a model (104a) using the generated datasets to prepare a defect detection model; localizing and labelling target region of the products; recording frames with defective products and detection information (106b); and monitoring the production performance in real-time (108a) using the detection information.
| Original language | English |
|---|---|
| Patent number | HK30080291 |
| Filing date | 28/02/23 |
| Publication status | Published - 28 Apr 2023 |