Maize endosperm hardness, often classified as degree of vitreousness, is a relevant trait due to its connection with lysine and tryptophan concentrations and starch digestibility. A direct measurement of vitreousness is accomplished by tedious manual dissection of kernels which has generated the need to develop predictive methodologies. There are several alternatives to estimate endosperm vitreousness; from visual ranking of kernel’s translucency under a light source to hyperspectral analysis of grinded samples with spectrophotometers. In this work, we are presenting a novel, non-destructive, image-based technique developed and optimized for hyperspectral images obtained using a Near Infrared Spectroscopy (NIR) flatbed scanner. For this objective, we have developed a calibration curve using more than 1,500 maize kernels individually characterized to build the training set. The hyperspectral flat-bed scanner is equipped with a high-speed near infrared camera with a spectral range of 950-1700 nm and spectral resolution of 3.5 nm. Custom made Matlab scrips have been developed to extract spectral information along with morphometrics features of the kernels. Preliminary results show a high coefficient of correlation (r = 0.84) between manually acquired endosperm vitreousness and NIR-based predicted vitreousness using partial least square regression (PLSR) validated using the leave-one-out (LOO) method. As a benchmark, the same materials were used to create a destructive NIR calibration curve in a conventional NIR spectrophotometer. A similar correlation (r=0.91) was observed between manually acquired endosperm vitreousness and this NIR scan with substantially more labor involved. The work presented here shows a competitive phenotyping tool and analytical method that allows researchers to predict endosperm vitreousness without the need for sample preparation and therefore substantially higher sample processing throughput. Additionally, multiple physical features can be extracted from the intact kernel hyperspectral images including accurate measurements of kernel length, kernel maximum width, kernel and endosperm area, among others.