The human eye is mostly sensitive to only three bands of the electromagnetic spectrum—red, green, and blue (RGB)—in the visible range. In contrast, off-the-shelf smartphone camera sensors are potentially hyperspectral in nature, meaning that each pixel is sensitive to far more spectral bands. Now scientists have found a simple way for any conventional smartphone camera to serve as a hyperspectral sensor—by placing a card with a chart on it within its view. The new patent-pending technique may find applications in defense, security, medicine, forensics, agriculture, environmental monitoring, industrial quality control, and food and beverage quality analysis, the researchers add.
“At the heart of this work is a simple but powerful idea—a photo is never just an image,” says Semin Kwon, a postdoctoral research associate of biomedical engineering Purdue University in West Lafayette, Ind. “Every photo carries hidden spectral information waiting to be uncovered. By extracting it, we can turn everyday photography into science.”
Using a smartphone camera and a spectral color chart, researchers can image the transmission spectrum of high-end whiskey, thus determining its authenticity. Semin Kwon/Purdue University
Every molecule has a unique spectral signature—the degree to which it absorbs or reflects each wavelength of light. The extreme sensitivity to distinguishing color seen in scientific-grade hyperspectral sensors can help them identify chemicals based on their spectral signatures, for applications in a wide range of industries, such as medical diagnostics, distinguishing authentic versus counterfeit whiskey, monitoring air quality, and nondestructive analysis of pigments in artwork, says Young Kim, a professor of biomedical engineering at Purdue.
Previous research has pursued a number of different ways to recover spectral details from conventional smartphone RGB camera data. However, machine learning models developed for this purpose typically rely heavily on the task-specific data on which they are trained. This limits their generalizability and makes them susceptible to errors resulting from variations in lighting, image file formats, and more. Another possible avenue involved special hardware attachments, but these can prove expensive and bulky.
In the new study, the scientists designed a special color reference chart that can be printed on a card. They also developed an algorithm that can analyze smartphone pictures taken with this card and account for factors such as lighting conditions. This strategy can extract hyperspectral data from raw images with a sensitivity of 1.6 nanometers of difference in wavelength of visible light, comparable to scientific-grade spectrometers.
“In short, this technique could turn an ordinary smartphone into a pocket spectrometer,” Kim says.
The scientists are currently pursuing applications for their new technique in digital and mobile-health applications in both domestic and resource-limited settings. “We are truly excited that this opens the door to making spectroscopy both affordable and accessible,” Kwon says.
The scientists recently detailed their findings in the journal IEEE Transactions on Image Processing.
