Previously, the use of automated optical inspection (AOI) on production lines often led to situations of missed detections and false positives. While this addressed the critical issue of reducing missed detections, the cost incurred from false positives was prohibitively high. Quality inspectors were required to identify non-conforming products flagged by AOI machines and manually reevaluate them for acceptance. This approach led to drawbacks such as production line slowdowns, high labor inputs for reinspection, and ran counter to the goal of automated production. GodShine's Automated Optical Inspection Rework Machine, utilizing AI model training and inference, achieves the dual objectives of minimizing missed detections and false positives.

Product Features

Rapid AI Reclassified Setup with Minimal Samples Utilizes a one-class learning architecture that requires only a small number of defect-free samples to establish the AI reclassification model. This allows for quick deployment and application on the production line without the need for extensive defect sample accumulation.

One-Click AI Deployment Features a user-friendly interface designed with UI/UX aesthetics in mind. The intuitive design allows for one-click deployment, automating the AI reclassification process. This eliminates the need for manual data sorting and preprocessing, simplifying the workflow and saving time.

Advanced Image Acquisition and Estimation Captures partial images of PCB boards and compresses them using wavelet packet analysis. Convolutional Autoencoders (AE) extract features from these compressed images, and Self-Organizing Maps (SOM) create topological features. The system then compares these topological features to detect differences between test images and defect-free samples.

Generative AI Model Optimization Implements a feedback loop where inspection images are processed, and AI results are compared with manual judgments. This comparison allows for continual refinement of the detection model, ensuring ongoing accuracy and improvement.

System Components