Wearable ultraviolet sensor based on convolutional neural network image processing method

Abstract

The wearable sensors based on image processing possess distinct advantages such as being power-free and without complex wire connections, which are of low cost and easy to manufacture. In this paper, a wearable UV sensor made from photochromic material and PDMS was proposed to be employed in real-time UV monitoring and daily solar protection. The convolutional neural network image processing method was introduced and developed for quantifying UV intensity, and it was shown to decrease the impact of ambient light significantly. The limit of detection of the sensor was about 9 μW/cm2 and the recognition rate of the network exceeded 90% under different ambient light conditions. The CNN test was complete within 3 s. Finally, regarding applied scenarios, a UV intensity recognition APP based on a mobile convolutional neural network was designed, which displayed the real-time UV intensity by simple photting.