Stemmer Imaging
The third dimension of light
Image processing based on polarized light can be used to detect hidden product properties such as stresses in plastics or glass. Jan Sandvoss from Stemmer Imaging explains the possibilities of the technology.
Mr Sandvoss, please briefly describe the polarization technology
Sandvoss: To better understand how polarization cameras work, you have to look at the physical principles: Light can be described by a propagating electromagnetic transverse wave in which an electric wave and a magnetic wave oscillate perpendicular to each other and to the direction of propagation. The polarization is defined by the plane of oscillation of the electric wave. Usually, light is not polarized, i.e. all directions of oscillation of the electric waves are equally probable. If there is only one direction of oscillation, this is referred to as linearly polarized light. If the phases of the perpendicular and parallel components of the electric wave are different, the light is elliptically polarized. Circularly polarized light, on the other hand, occurs when the phases of both components are shifted by exactly 90°.
These statements apply to the entire electromagnetic spectrum and therefore also to the light spectrum, which includes ultraviolet light, visible light with wavelengths between 440 and 650 nm, near infrared light and short-wave infrared light.
Jan Sandvoss: "With a suitable polarization system, the third dimension of light can be brought to light and made usable for industrial image processing - in other words, properties and defects can be made visible that are not visible with any other method.
© Stemmer ImagingWhat is the main difference between unpolarized and polarized light?
Sandvoss: Unpolarized light consists of many waves that oscillate randomly in different directions - such as the light from light bulbs or sunlight. In industrial image processing, this form of lighting has the disadvantage that reflections in certain areas are practically unavoidable, especially on test objects with shiny surfaces. Polarized light, on the other hand, means that all the waves emitted by a light source have the same polarization and therefore correspond in the directions of the electric fields. The clever use of polarized light allows unwanted reflections to be filtered out.
When using a polarizer, only the part of the light that oscillates parallel to the optical axis of the polarizer is transmitted.
© Stemmer ImagingWhat are typical application examples?
Sandvoss: In general, various characteristics can change the polarization state of light: For example, the surface properties of objects, such as their roughness, scratches, dents or coatings, or other physical properties such as mechanical stress or birefringence, have an influence. Especially when inspecting shiny, reflective or mirror-like surfaces such as foils, metal or glass, polarization images enable improved image processing - for example, easier detection of scratches or robust reading of codes on multi-layer foils.
With a suitable polarization system, for example, it is very easy to check whether the tear tabs on card decks shrink-wrapped in film are free of defects. When inspecting with unpolarized light, such errors are much more difficult to detect and sometimes even impossible. Other examples are classic pick & place applications in which shiny, often metallic components always lead to reflections in different image areas under different illumination and polarization angles. By combining reflection-free sections of the images captured under different, sometimes virtual polarization angles to form an overall image, it is possible to merge areas that are easy to evaluate and thus simplify image processing and recognition of the objects and their position on this synthetic image.
The differences
Polarization images and mapping on the HSV colour space allow the visualization of stresses in the material, here in the structure of a ruler.
© Stemmer ImagingWhen did polarization become interesting for image processing?
Sandvoss: In autumn 2018, Sony presented a CMOS image sensor with polarization functions integrated at pixel level that make it possible to solve certain tasks. Since then, the topic has gained significant momentum in image processing. The sensor can filter light in four planes at 0°, 45°, 90° and 135° and only transmit the part of the light that oscillates parallel to the optical axis of the respective polarizer.
Some manufacturers of industrial cameras have introduced products based on this sensor: Stemmer Imaging, for example, works together with Allied Vision, JAI and Teledyne Dalsa here. As with area scan cameras, four polarization filters with different polarization angles are also implemented here, which can be captured directly. For both area scan and line scan polarization cameras, however, it is also possible to calculate other 'virtual' polarization angles by interpolating the four main directions in order to achieve the optimum alignment for different tasks.
By combining partial images with different polarization (left: 0°, middle: 90°) into an overall image, it is possible to merge the areas that are easy to evaluate and thus simplify image processing on the synthetic image (right).
© Stemmer ImagingWhat is the difference between images from polarization cameras and images from conventional cameras?
Sandvoss: The basis for polarization images is the so-called Stokes vector, which can be used to quantitatively determine and mathematically represent the polarization of light. This vector consists of four values that can be used to define the direction and intensity and thus the degree of linear, circular or elliptical polarization of electromagnetic waves. After recording objects with a polarization camera, it is possible to display images for the first three Stokes parameters. These can be used in a further step to calculate the linear degree and angle of polarization. For better visualization, these images can also be mapped to the HSV colour space, for example to make it easier to display stresses in the structure of plastic objects. The subsequent image evaluation is also simplified by this form of representation.
What equipment is required to take polarization images?
Sandvoss: In addition to a polarization camera, suitable lighting and optics are required on the hardware side in order to capture high-quality images. Some software environments, such as the 'Common Vision Blox' software library developed by Stemmer Imaging, already contain the appropriate tools for evaluating polarization images.
You might also be interested in
IDS Imaging Development Systems
3D data directly from the camera
Shifting computationally intensive processes away from the industrial PC and into the camera improves both the speed and accuracy of data processing. But what are the benefits of integrated data processing in detail?
read more...Space for microscopy and analysis
In quality control, space is often limited, so there is little room for microscope equipment. With the Doku-Kit3, Optometron offers a handy camera-monitor unit that enables self-sufficient microscopy and analysis without a PC.
read more...Stable image evaluation
Baumer is launching six LXT cameras with resolutions from 0.5 to 7.1 megapixels that combine Sony 3rd generation Pregius CMOS sensors with 10 GigE.
read more...Waterproof and compact
For use in harsh environments, Matrix Vision offers a waterproof version of the 'mvBlueCougar-X' industrial camera with the IP67C option.
read more...Light section and vision sensor combined
The light section method is an often underestimated sensor principle for presence, completeness and position checks and similar tasks. Intelligent light section sensors deliver much more precise and reliable results than other types of sensor.
read more...Machine Vision streamlined
In order to generate the required amount of light for high-quality images, image processing systems usually work with external flash controllers. A current approach integrates these control units into the camera, thus reducing effort and costs for...
read more...Multispectral imaging
JAI has introduced a new generation of prism cameras for its 'Fusion' series of multispectral imaging solutions.
read more...Installation in mobile, desktop and portable systems
The Firefly DL is Flir's first deep learning-based, interference-enabled machine vision camera with 'Neuro' technology.
read more...Camera provides 'chemical fingerprint'
Hyperspectral-based image processing makes it possible to detect organic or inorganic contaminants on objects - not only on their surface, but in some cases also inside them. An overview.
read more...