NEW COLOR IMAGE ANALYSIS METHODS TO MEASURE COTTON COLOR DISTRIBUTIONS AND VARIATIONS (CI PROJECT NO. 12-227)
Cotton Structure and Quality Research
2013 Annual Report
1a.Objectives (from AD-416):
To obtain the distribution of color of a cotton sample instead of one single color grade by using image analysis methods, and to investigate if higher variation in cotton color is associated with higher variations in other cotton properties such as maturity and fineness. This research will provide new tools for better evaluation of cotton quality and aid the development of the next generation rapid test instrument for cotton color measurement.
1b.Approach (from AD-416):
Image analysis techniques will be used to explore a series of sub-areas of the area of the measured cotton samples. The size of these sub-areas can be down to pixel level. Sophisticated image processing techniques will be used to analyze the sample images’ color distributions and identify the impacts of trash and other impurities. The digital images can have very high resolutions and color levels that are sufficient for differentiating the variations among the different parts of the sample.
This is the final progress report for this project. ARS Scientists at Southern Regional Research Center (SRRC) in New Orleans, LA have completed testing 182 cottons covering a wide range of cotton color grades. In addition ARS Scientists at SRRC in New Orleans, LA have also completed testing another set of 35 cottons with prep problems. Each cotton was scanned using a color scanner, and then was measured by using an High Volume Instrument-1000 to obtain its color Rd (diffuse reflectance), +b (yellowness) values and trash content. Meantime ARS Scientists at SRRC in New Orleans, LA are also using Advanced Fiber Information System to measure other properties of these cotton samples.
ARS Scientists at SRRC in New Orleans, LA are conducting data analysis for these samples’ color distributions and variations using computer image processing. ARS Scientists at SRRC in New Orleans, LA are also investigating the possible correlations between the intra-sample variation of cotton color and the variations of other properties that can be obtained from High Volume Instrument and Advanced Fiber Information Systems.
The image processing method is able to look at the distribution of color within one scanned image of a cotton sample. The preliminary data revealed that, though the inter-sample color variation (the color variation among different repetitions of the same cotton) is usually very small, the color is not uniformly distributed as one single value in sub-areas of a image of a cotton sample. Within one image, a cotton sample’s color shows a certain variation (intra-sample variation) as some other cotton properties do, for example, the variation of cotton length. ARS Scientists at SRRC in New Orleans, LA are working to develop parameters to characterize the distribution and variation of cotton color.