Color and image processing for output extraction of an LSPR sensor

Sensors based on the Local Surface plasmon Resonance (LSPR) are attractive due to their simple structure and good sensitivity, but the expensive optoelectronic part of the device is limiting the practical applications. There is a need for new strategies to bring the excellent detection properties of LSPR sensors to the playground of low-cost devices and materials. In this work, it is proposed a novel approach to the output extraction of from LSPR sensor whose sensing element is composed by metal nanoparticles (MNPs). Illuminated with an incident broad light source, the sensor produces a spectral transmission output where the MNPs act like a band-stop optical filter for a specific wavelength. An alteration of the refractive index in the surrounding medium corresponds directly to a shift of the filtering rejection band, which corresponds to a slight change in the colour of the light transmitted by the sensor elements. This colour change can be captured by a CMOS photo-camera, used as an image sensor. It is proposed in this paper an approach based on an automatized image processing algorithm for colour change detection, yielding to a system capable of detecting refractive index variations, avoiding the use of expensive spectrometers. The algorithm comprises three stages: (1) Region of interest detection: images are first cropped using the Otsu threshold binary image to remove the uninteresting areas in the image. (2) Image segmentation: using the watershed algorithm, the sensor elements (sample) area is detected automatically in the cropped image. The segmentation is done using the gradient image, where the watershed markers are the regions of low gradient and barriers are the areas of high values inside the image. (3) The resulted sample region is then processed to find its average or dominant LAB colour and then compare it to its corresponding sample image immersed in different mediums using the colour difference measurement CIEDE2000.