Novel approach to high-security print inspection

Paul Boughton

The combination of a contact image sensor (CIS) with a print and pattern inspection tool greatly improves security print inspection and verification, explains Peter Stiefenhöfer.

The inspection and verification of security features is a common requirement for a wide range of web- and sheet-based products, such as bank notes, cheques, lottery tickets and software security labels. Requirements include checking the presence, position and integrity of applied features such as foil and hologram devices and base paper inserts such as security threads, as well as watermarks. 

Verifying these security identifiers is frequently conducted by comparing the image of the sample under test to a ‘golden template’ of a single known good part. By subtracting the image of the part under test from the golden template, any defects can then be highlighted by the vision system. 

Two factors determine the effectiveness of this approach: whether the test items are imaged under identical conditions, and whether there is any natural variation in those items. The combination of a contact image sensor (CIS) from Mitsubishi Electric and Flex Inspect from Stemmer Imaging addresses both of these issues.

Web- and sheet- based products such as security labels are frequently inspected using line scan cameras to acquire the image. Although these cameras produce extremely high resolution images, they require a lens to project the image onto the sensor. This means that there can be angular variations in the way that the sensor pixels ‘see’ the sample along the length of the camera. This can be particularly important when imaging holograms, since by their very nature, any slight change in orientation changes their appearance. 

Latest advances in CIS

The new Mitsubishi Electric KD series of CIS, however produces an image with virtually no distortion since the long sensor line maps the object on a one-to-one scale. Image formation in the CIS is achieved via a double-row lens array, using gradient index rod-lenses, matched to a colour tri-linear CMOS sensor. 

Each individual rod-lens captures an image of a very small region of the target. Thanks to some overlap in the individual images, a clear, sharp, combined image is produced along the narrow line of the sensor with a resolution of 600 dpi. 

Two sets of integral white LED lighting can be varied in intensity and controlled independently. The lighting is homogenous, with no shading along the length of the sensor, meaning defects can be detected even at the edge of the scan. 

The arrangement produces a quasi-telecentric image at each pixel, giving remarkable image uniformity. This is illustrated by the high consistency in the array of hologram images shown in Fig.2. Image data output is provided via standard industrial CameraLink interfaces. The images generated by a contact image sensor show virtually no distortion since the long sensor line maps the object on a one-to-one scale The CIS products are available in three different lengths, offering scan widths of 309, 617 and 926mm. Crucially for machine vision applications, these new CIS have a working distance of 12 ± 0.5mm from the front face and a machine vision standard CameraLink output.

The difficulty of the traditional ‘golden template’ comparison method is that if there is any variation in the fidelity of each sample, subtracting the image from the golden template will produce a difference. Rather than using a single golden template, CVB Flex Inspect uses the latest in machine learning research to build a complex appearance model of good examples ‘on the fly’ during a training phase. 

During this process, the tool is presented with a range of acceptable samples and learns the variability. Users have the ability to view the modes of variation of the template model to gain an understanding of the level of product deformation. 

Creating a synthetic template

Once in inspection mode, a synthetic template is created from the template model for each inspection within the limits of acceptable variation and is applied to the inspected sample to give a difference image just like a traditional golden template approach. As new variations are identified as samples are inspected, they can either be incorporated into the template model, or the sample rejected. 

CVB Flex Inspect is designed to make use of Intel multi-core CPUs with two or more independent central processing units. The software operates on sub images, or pixel blocks, within each of the images presented to it. Each of these pixel blocks can be processed independently according to the number of core CPUs in the system. 

Complete flexibility

CVB Flex Inspect is one of the many versatile imaging tools contained within Common Vision Blox (CVB), Stemmer Imaging’s hardware-independent imaging toolkit. Developed for CVB by Vision Experts Ltd, Flex Inspect’s sophisticated approach to pattern matching can be just as easily applied to images from line scan and area scan cameras. Stemmer Imaging is Mitsubishi Electric’s partner for the distribution of CIS in Europe.l

For more information visit

Peter Stiefenhöfer is marketing and PR manager at Stemmer Imaging in Germany. 

Recent Issues