A cost-efficient hyperspectral imaging (HSI) technology has been developed that could enable new artificial intelligence (AI) applications to be introduced into mobile devices, allowing for potential use in smart home systems. The small VNIR hyperspectral camera can show the difference between raw and ripe avocados. Courtesy of VTT Technical Research Centre of Finland. The hyperspectral camera, developed by a team at VTT Technical Research Centre of Finland, uses very-near-infrared (VNIR) wavelengths to detect and label materials and properties of different objects within an environment. AI is used to interpret the environmental spectral data within the VNIR range. VNIR wavelengths (600 to 900 nm) exceed the red color visible to the human eye and are typically filtered out of standard camera images. However, they can be detected even by low-cost mobile phone cameras. “VTT’s technology has a simple optical path, making it compatible even with the very compact and low-cost optics used in mobile cameras, which is not possible with other spectral imaging technologies,” said researcher Antti Näsilä. "This is a huge advantage because it enables very cost-efficient mass production for these hyperspectral camera sensors." The sensor core is a tiny tunable MOEMS wavelength filter a few millimeters in size, integrated with small camera optics. Courtesy of VTT Technical Research Centre of Finland. Spectral data taken from the camera's images can be used to generate information related to food safety and product authentication and could be used by mobile applications created for sensor data interpretation. Such a sensor could be integrated into everyday surroundings to make them more intelligent and incorporated into home systems and appliances, mobile devices, robots and autonomous vehicles, which need to interpret visual information in order to function securely. “In the future, an increasing share of vehicles and systems will become autonomous, and the need for reliable visual camera information for automated decision-making will increase. Adding the third spectral dimension to images could provide more safety and security for autonomous systems relying on machine vision and artificial intelligence to make decisions based on visual camera data,” said researcher Anna Rissanen. Researchers believe the cost for the new VNIR range hyperspectral camera sensor hardware would be only about $150. VTT’s tunable filter technology, with mass-producible MEMS technology, could be integrated with a camera sensor without significantly increasing its cost or size. High-volume production and calibration methods could lower the cost of the sensor further. Researchers believe that the core component — a micro-opto-electro-mechanical (MOEMS) chip — could cost less than one dollar. Currently, most hyperspectral imagers cost from thousands to tens of thousands of dollars, making them impractical for consumer applications such as a smart refrigerator that measures food freshness. Other spectral imaging technologies that aim toward mass-producible volume scaling in order to lower the final product cost typically process fixed wavelength filters directly into individual camera sensor pixels — an approach that usually requires expensive telecentric optics. During the next few years, VTT plans to commercialize cost-efficient HSI technologies in cooperation with companies operating in the field. VNIR hyperspectral imager: Making the invisible visible. Courtesy of VTT Technical Research Centre of Finland.