|Ding, Fujian - VISITING SY-ISL, ANRI|
Submitted to: Proceedings of SPIE
Publication Type: Proceedings
Publication Acceptance Date: October 23, 2005
Publication Date: November 20, 2005
Citation: Ding, F., Chen, Y.R., Chao, K. 2005. Application of color mixing for safety and quality inspection of agricultural products. Proceeding of SPIE conference, October 23-26, 2005, Boston, Massachusetts. p. 5996OR-1-5996OR-13. Interpretive Summary: The Instrumentation and Sensing Laboratory (ISL) is developing a low-cost optically enhanced device that can assist inspectors or plant processors in small meat and poultry plants to conduct inspection in situ. The system would be a color-based optical inspection device. In this paper, we show that two- and three-color mixing is related to color band ratios, which are often used for class discrimination and target detections. We show that the three-color mixing technique provide better discrimination power than the two-color mixing technique for identification of the conditions of chicken carcasses and for the detection of chilling injury in cucumbers. The novel color mixing techniques for visual inspection can be implemented on visual devices for a variety of applications, ranging from target detection to food safety inspection. This paper also presents methods of selecting optimal waveband pairs from color difference and chromaticness difference indices, and to design optically enhanced binoculars for inspections based on two- and three-color mixing techniques. The simulation of color appearance is also presented. We showed that three color-mixing binoculars have potential for practical use in a processing plant environment. This will directly impact small-scale poultry processors in terms of improved efficacy of the HACCP program. This information would be useful to FSIS and researchers who are interested in developing low cost vision-based systems for inspecting agricultural products.
Technical Abstract: In this paper, color-mixing applications for food safety and quality was studied, including two-color mixing and three-color mixing. It was shown that the chromaticness of the visual signal resulting from two- or three-color mixing is directly related to the band ratio of light intensity at the two or three selected wavebands. An optical visual device using color mixing to implement the band ratio criterion was presented. Inspection through human vision assisted by an optical device that implements the band ratio criterion would offer flexibility and significant cost savings as compared to inspection with a multispectral machine vision system that implements the same criterion. Example applications of this optical color-mixing technique were given for the inspection of chicken carcasses with various diseases and for the detection of chilling injury in cucumbers. Simulation results showed that discrimination by chromaticness that has a direct relation with band ratio can work very well with proper selection of the two or three narrow wavebands. This novel color mixing technique for visual inspection can be implemented on visual devices for a variety of applications, ranging from target detection to food safety inspection.