Machine Vision for Translucent Materials: From Optics Design to Defect Inspection Application Note
Image processing capability integrated together, with optical/mechanical/electronics capabilities, can deliver system-level solutions for the manufacturing industry. These combinations of intelligence in image processing and application domain knowledge are named “Machine Vision”. Among various vision inspection applications, contact lens manufacturing puts up a challenging case as the samples are translucent and floating in liquid.
In order to reduce the reliance on existing labor-dependent manual inspection, the defects of contact lenses need to be inspected by an automatic system. Thus, increasing their output when the inspection is fully automated in the manufacturing lines. The goal is to build an automated inspection system to inspect multiple contact lens in a tray and reject any of them which has wrong tool code/marking, or tears of at least 100um, with 0% miss detection and less than <5% false detection.
The whole inspection system includes the imaging optics, the lighting design, the imaging sensor, the moving mechanism, the sample holding mechanism, and the imaging processing software, as shown in Fig.2.
A novel image processing algorithm and the optical system were designed to handle the challenge and to achieve the 0% miss detection rate.
- Dual light source – Darkfield and mixed field lighting for viewing different defects and coding.
- Dual or multi-contact lens positioning – To differentiate the defects of concern from static and dynamic noises and increase the reliability of tool code identification.
- Dual camera set (Full view and Angled view) – To encompass and capture all possible scattering angles of illuminated tears.
- Accelerated image processing and pipelining process – To maintain the minimum speed of 500 contact lenses per hour.
- Positioning: The contact lens might not be in a neutral position due to the use of the existing tray, which has a large well space compared to the size of the lens. This will lead to the captured image of a distorted lens.
- Noisy environment: The system has to differentiate abnormalities from the trays’ defects (such as scratches or water droplets); and in saline solutions’ defects (such as floating particles or air bubbles).