SpectroPhone: Enabling Material Surface Sensing with Rear Camera and Flashlight LEDs

Abstract

We present SpectroPhone, a surface material sensing approach based on the rear camera of a smartphone and external white LED light sources. Warm and cool white LEDs, as used for dual or quad flashlights in smartphones, differ in their spectral distribution in the red and blue range. Warm and cool white LEDs in combination can produce a characteristic spectral response curve, when their light is reflected from a surface. We show that with warm and cool white LEDs and the rear-camera of a smartphone 30 different materials can be distinguished with an accuracy of 99 %. Based on a dataset consisting of 13500 images of material surfaces taken at different LED light intensities, we report recognition rates of support vector machines with different parameters.