An inspection approach for glass containers can be as unique as the containers themselves, with the difficulty being to optimise the inspection for multiple defect types. As lighting and optical geometries are tuned for a target defect, it may become harder to identify other types of imperfections present in the same container. This is the limitation of standard lighting and monochromatic imaging techniques that effectively require users to strike a balance between primary and secondary defect detections, or in many cases, to invest more time in machine set-up as jobs and detections change.
Colour provides critical information
Applied Vision’s Volcano Sidewall (SW) inspection system helps plants to achieve their goals for greater accuracy, speed and reliability in sidewall and dimensional defect detection. The use of colour illumination provides manufacturing facilities with a customisable solution when faced with process-induced defects and the ever-growing complexity of containers. The Volcano imaging stations are fully programmable in terms of intensity, colour selection and pattern geometry. The system can perform a wide range of inspections including opaque defects such as inclusions, transparent defects like blisters, bubbles and tears, as well as dimensional defects including lean, filler lean, height, diameter and thread inspection. The system also employs dedicated cameras utilising circular cross-polarising filters for detection of internal stresses.
Volcano SW incorporates high-quality, programmable light-emitting diodes (LEDs) as a light source, multiple independent arrays of high-resolution colour cameras, and fully calibrated imaging geometries to inspect glass containers. Using advanced software and processing algorithms, a single image can be utilised to perform virtually any type of detection.
False reject rates are reduced by machines that recognise a glass container on the narrow edge of tolerance as good packaging that should not be recycled. It is now possible to automatically detect acceptable variation limits that only a human inspector could once perceive.
Inspection accuracy and efficiency
The Volcano SW is a self-learning solution built to inspect the most complex glass containers at line speeds. Time, money and material can be saved when machines tolerate a level of variation not exceeding tight thresholds for dimension, colour and pattern inspection regardless of heavy embossing, colour or container thickness.
It is the combination of capable hardware and software architecture that unlocks operational benefits for container manufacturers. Inspection accuracy and efficiency are improved by Volcano SW technologies that alleviate the burden of plant engineers and machine operators. Capabilities include:
- A bottle geometry tool that learns, locates and registers the sidewall in the captured image automatically. The benefit is that established inspection sequences are saved and recalled quickly. A library of algorithms is used to inspect for defects in these processed images.
- Sentinel software that highlights regions of the captured colour image that deviate from the established statistical model for a glass container. This tool looks for candidate defects, learns what is normal, and comes to understand what is abnormal by training on a large number of containers.
- A ‘blob classifier’ algorithm that helps plant personnel to see what types of defects are being produced and how often. Defects are identified, categorised and graded based on user-applied criteria that can be adjusted without taking the machine offline.
As a ‘no-touch, no-turn’ solution for plants striving for a contactless inspection, Volcano SW eliminates handling of containers that can lead to breakage and limits production speeds. Machines that provide a high degree of imaging precision and defect classification accuracy without the need to reorient containers can save manufacturers both floor space and maintenance costs since there are fewer moving parts to manage.
Today’s machine vision systems can perceive obscure anomalies between glass containers – subtle imperfections that the human eye can masterfully distinguish but machines once struggled to conceptualise. Supported by self-training algorithms and expanding AI capabilities, these systems can better decide between acceptable and unacceptable variations, further reducing the number of false positives. Throughput is also improved by machines that can perform multiple optimised detections in a fraction of the time once required. Making it all possible are multispectral illumination solutions that use colour to help container manufacturers operate on schedule and within budget.