Unlock new applications through real-time spectral imaging
What if your vision tool would not only show you the 2D scene as it is truly there, but also tell you what you are looking at, instantly at video-rates and beyond? The potential applications for such devices are many. In healthcare for example, hyperspectral imaging can be used to discriminate tissue, nerves, and blood vessels during non-invasive surgery to support the surgeons precision. It can help to provide more accurate identification of cancerous tissue during surgery and pathology; or be used to monitor the recovery of organs after transplantation, and much more. In global security, hyperspectral imaging can improve the ability of situational awareness systems to detect, recognize and identify instantly what is begin observed under more difficult environmental conditions.
Adimec is partnering with Imec (Belgium) to bring the capability of instant spectral imaging at video-rates and (and faster) possible to our customers in Machine Vision, HealthCare and Global Security. Imec has developed a semiconductor process in which they apply spectral filters at the pixel level on image sensors. This results in hyperspectral imaging sensors with the same size and packages as regular imaging sensors, making it possible to build cost-efficient and compact spectral imagers, detectors and devices/systems.
By applying this pixel-based filter process onto high-speed image sensors, hyperspectral imaging sensors are created that are ideal for use in push-broom and snapshot acquisitions solutions. The snapshot tiled and mosaic hyperspectral CMOS imaging sensors enable instant hypercube acquisition at high speeds. Hypercubes contain the spectral fingerprints of the objects captured in a 2D scene. Now, 2D spectral imaging becomes possible in situations with fast motion and moving objects.
Adimec is working with customers to explore the applicability of the first Imec sensors in applications. Currently, Imec provides various filter solutions utilizing a CMOS global shutter image sensor. Adimec uses our QUARTZ camera series as the evaluation platform to facilitate hyperspectral cubes with a 2 Megapixel field-of-view at 340 frames per second. The QUARTZ series facilitates real-time reflectance normalization and multi-frame pixel based averaging that increases signal-to-noise performance essential for optimal spectral fingerprint detection.
Read more about hyperspectral imaging in the next sections, or contact us to discuss how Adimec can help realizing your hyperspectral imaging needs.
How hyperspectral imaging works!
Hyperspectral imaging is a combination of spectroscopy and imaging where the light spectrum is split into narrow bands. This can be done in the UV, visible, or NIR.
In certain application, this technique has advantages over color when very small differences must be detected or when different materials must be identified.
With hyperspectral imaging, many more wavelength bands are captured than with color imaging. Also, the bandwidth is much smaller.
This results in a series of images of the object with information related to many wavelength bands. A wavelength band image is called a spectral plane.
With the signal value of the same pixel in each spectral plane that is plotted as function of the wavelength, a spectral profile is created (spectral fingerprint of that pixel in the scene). By comparing the spectral fingerprint with known materials/objects databases one can identify what is being detected.
What technology will Adimec use?
Adimec is an official partner of Imec on hyperspectral imaging. Imec sensor technology is made available with Adimec cameras by Imec and Adimec. Adimec cooperates with select customers to explore the potential of the Imec technology in applications and to provide the true accurate imaging camera solution for reliable and repeatable fingerprint extraction.
Our goal is to go beyond a sensor on an interface. In addition to eliminating non-wanted pixel and sensor specific characteristics our camera solutions will also deal with spectral response characteristics of the filter-pixel combinations, which are unique and different for each hyperspectral sensor (due to semiconductor process variations). The spectral information we provide will be suited for reliable fingerprint analysis, comparable between cameras, available in real-time and at high sampling rates.