Skip to main content
ARS Home » Research » Publications at this Location » Publication #149531

Title: VISIBLE/NIR IMAGING SPECTROSCOPY FOR ASSESSING QUALITY AND SAFETY OF AGRO-FOODS

Author
item Chen, Yud
item Kim, Moon

Submitted to: Near Infrared Spectroscopy International Conference Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 6/20/2003
Publication Date: 11/1/2004
Citation: Chen, Y.R., Kim, M.S. 2004. Visible/NIR imaging spectroscopy for assessing quality and safety of agro-foods. In: Proceedings of the NIR Publications-NIR2003, Cordova, Spain. p. 67-68.

Interpretive Summary: This paper was prepared for an invited presentation at the 11th International Conference on Near-Infrared Spectroscopy, which was held at Cordoba, Spain, May 5-10, 2003. This paper presents an overview and research progress of imaging spectroscopy, commonly called hyperspectral imaging, for assessing quality and safety of agricultural products. Imaging spectroscopy acquires images in which a spectral signature is associated with each spatial element or pixel. It combines the advantages of imaging and spectroscopic techniques. While hyperspectral imaging provides important spectral and spatial information, it does not have the capacity for rapid on-line, real-time data acquisition and processing. However, hyperspectral imaging data can be used for the selection of several spectral bands, which, when combined, subtracted, or mathematically manipulated, often provide important information for a variety of effective applications. The optimal bands and algorithms can then be implemented in multispectral imaging systems for rapid or real-time prediction or inspection of many food commodities for quality and safety. This information is of great interest to the researchers who are interested in extending the application of near-infrared spectroscopy from point measurements to mapping whole object surface for assessment of agricultural products for quality or safety.

Technical Abstract: Imaging spectroscopy, commonly called hyperspectral imaging, acquires images for which a spectral signature is associated with each spatial element or pixel, resulting in a 3-dimensional data cube of information, including 2 spatial and 1 spectral reflectance or fluorescence dimensions. It combines the advantages of imaging and spectroscopic techniques. Textural and morphological information of the whole object or region is combined with color and compositional absorption (or in case of fluorescence, re-emission) information at the pixel level. While hyperspectral imaging provides important spectral and spatial information, it suffers from the incapacity for rapid on-line, real-time data acquisition and processing. However, hyperspectral imaging data can be used for the selection of several spectral bands, which, when combined, subtracted, or mathematically manipulated, often provide important information for effective applications. The optimal bands (up to 4 bands) and algorithms can then be implemented in multispectral imaging systems for rapid or real-time prediction and/or inspection of many food commodities for quality and/or safety. In this lecture, the advantages of hyperspectral and multispectral imaging techniques as research and practical application tools are presented. Aspects of several systems, including hardware, signal and image processing, optimal bands selection, and data classification algorithms, will be discussed. Examples of on-line applications of these techniques for real-time detection of quality and safety of agro-foods will be given.