Applications for multispectral and hyperspectral imaging

Multispectral imaging (MSI) and Hyperspectral imaging (HSI) have been much talked about recently, we take a look here at what the techniques are adding to the real world.

The science behind it
The human eye, and most standard vision sensors, detect light within a relatively narrow spectrum and translate this into three bands – long wavelengths perceived as red, medium wavelengths perceived as green and short wavelengths perceived as blue. Multispectral sensors are able to capture image data from beyond this range, from thermal infrared to ultraviolet, and measure spaced spectral bands. Hyperspectral sensors take this further and measure this extended spectrum in continuous bands, meaning they break the spectrum into even narrower bands, and the resulting images provide even more insight into the nature and composition of target objects. To illustrate this, MSI measures light across tens of bands while HSI can cover hundreds. Objects previously undetectable to the naked eye or traditional vision system can now be visualized and interpreted.

In the field: MSI and HSI in agriculture
Precision agriculture refers to the use of technology to improve yield and reduce waste in farming practices, and vision is playing a huge part. AIRINOV is a French subsidiary of Parrot putting Parrot SEQUOIA multispectral cameras into the sky on senseFly drones. The multispectral sensor measures how much light is being reflected from crops and analysis of green, red and infrared light results in advice on the optimum amount and placement of fertilizer and pesticide. Such techniques require little processing power as only a “yes, no” outcome is required – does the area need more fertilizer/pesticide/water or not? DroneZon recently featured a very interesting article explaining multispectral imaging in agriculture in more depth – read it here.

Hyperspectral sensors “see” an even wider range of spectral bands and are being used to monitor crops with even more detail and accuracy. Gamaya, a start-up born out of the Swiss EPFL (Federal Institute of Technology), has developed a drone-based hyperspectral camera along with a software toolkit which uses machine learning to interpret and act on what it’s seeing. Gamaya claim that, in addition to seeing that a crop is becoming diseased, data from the processed image is able to identify which disease is affecting the plant and suggest the best course of action.

Enhancing industrial machine vision
HSI has been used for inspection and quality control in food processing for some time. This non-destructive method sheds light on a variety of components within food including levels of sugar, fat and moisture, as well as surface defect detection, allowing control of items from baked goods to fish. Cubert illustrates an innovative method to use machine learning and HSI to classify herbs in real time, demonstrating their bespoke software and spectral snapshot cameras. Imec is working with Java Coffee Company, employing HSI in accurate quality control in the coffee bean industry without the need to grind beans for inspection, and, they say, at an acceptable price point.

Greater insight for medical imaging
MSI and HSI construct an image cube made up of images captured with incident light of varying wavelengths laid on top of one another. In this way, the composition of tissue and cells can be better visualized and interpreted. MSI imagery results in the ability to diagnose cancerous cells from normal cells, HSI can take this further and classify, or grade, cancerous tissue, enabling doctors to compile a more accurate care plan and prognosis. For a more universal application, an HSI sensor can detect an incorrect tablet within a blister pack, helping to ensure medical packaging is up to standard.

Active Silicon supports a range of imaging techniques
Our frame grabbers, embedded systems and camera interface boards support a range of imaging techniques in application areas including science, industry, defense and surveillance. Get in touch to discuss how our components could help you see more.