Location: Food Quality Laboratory
Project Number: 8042-44000-009-02
Start Date: May 01, 2013
End Date: Apr 30, 2016
There are two major efforts under Objective 1. The first is to develop and refine procedures and algorithms for detecting and quantifying mycotoxins, and assessing quality parameters of interest on the whole-surface of samples and the extent of damage caused by Fusarium, heat, frost, black point and insect invasion. The second is to build data fusion algorithms that can enhance the synergy of multiple detection systems using the technologies previously developed at BARC that would facilitate the development of a system for simultaneous acquisition of reflectance, fluorescence and Raman images. This will involve the development of image processing routines that identify the infected/damaged kernels in representative samples of intact and damaged kernels. It will be necessary to identify multispectral wavebands and develop detection algorithms and image segmentation procedures for whole-surface inspection of cereal kernels which can be utilized for multiple detection screening for safety and quality concerns. Objective 2 focuses on rapid evaluation of the extent of damage and presence of contaminants at trace levels in cereals. This addresses the need of the cereal industry for evaluation or inspection of cereal safety risks using tools for rapid detection of contaminants at sub-pixel resolution. It also expands sensing capabilities to trace levels. Suitability of algorithms for multi-frame image super-resolution will be tested using data acquired with the optical spectral systems at BARC. These algorithms gain additional information from the sub-pixel spatial shift in the multiple images of the sample. The increased resolution will be investigated for improvement of accuracy and limit of detection of preliminary models. Combination of “superresolution” and data fusion of the different optical spectral techniques will be carried out for further enhancement. Quantitative assessment of spatial and spectral quality will be performed on superresolved fused image.