M.D. Watson

Clustering Analysis for Cotton Trash Classification

Raw cotton may contain various kinds of trash, such as leaf, bark, and seed coat particles. The content of each of these trash categories is useful information for finding more efficient cleaning processes and predicting the quality of the finished products. This paper addresses the importance of using chromatic and geometric features of trash for trash description, and presents three different clustering methods that automatically classify trash based on the feature measurements. Compared with the geometric attributes of trash, such as size and shape, color attributes are less changeable during harvesting and ginning of cotton and are therefore more reliable and descriptive in categorizing trash. Three clustering methods—sum of squares, fuzzy, and neural network—prove effective for trash classification. Sum of squares clustering and fuzzy clustering require iterative computations and generate comparable classification accuracy. Neural network clustering yields the highest accuracy, but it needs more computational time for network training.

Chromatic Image Analysis for Cotton Trash and Color Measurements

The U.S. cotton classification system has been undergoing significant changes, moving from human classing to the use of precise instruments. Along with this trend, the current research is an effort to develop a new computer vision system to measure detailed trash and color attributes of raw cotton. The system primarily consists of a color ccd camera, xenon flash light, and customized software. In this paper, we introduce a new trash and spot identification method, multidimension thresholding, and the methods for characterizing size, spatial density, shape, and color of trash and spots present in cotton samples. We report on the trash and color measurements of twelve cotton samples, including statistical data and distribution curves, and we compare the results from this system with those from other instruments such as the Spinlab and Motion Control hvi machines and the Minolta Chroma Meter CR-210. Finally, we investigate the influence of trash and spots on cotton color values.

INVESTIGATING NEW FACTORS IN COTTON COLOR GRADING

The objective of this research is to further investigate factors that may have significant influence on cotton color measurement but are not considered in the current cotton color grading system. These factors include the redness content (a) in cotton chroma and the presence of yellow spots and trash particles in cotton. The study is based on the color data of the USDA physical standards for U.S. upland cotton and a number of selected samples measured by the imaging colorimeter (CTC), developed in the previous research, and a Minolta CR-210 colorimeter. There are three major results in this study: first, the a content makes up 10% to 33% of the chroma, varying primarily with the major color category (white, light spotted, spotted, tinged, and yellow stained). Within the same category, a is less variable than yellowness b. An approximate +a range for each major color category is determined. Second, CTC is less sensitive to the presence of spots and trash particles in the sample than CR-210 because CTC has a much larger viewing area. Third, the influence of spots and trash on the cotton color measurements depends on their sizes and colors in the sample. A change in cotton color data made by spots and trash may lead to a change in color grade. The computational removal of these regions from the sample image in CTC is effective in minimizing the effects of spots and trash. In addition, the paper introduces a new color diagram built on the measurements of reflectance-redness (Rd ∼ a), which identifies the useful role of a in cotton color grading.

COTTON COLOR MEASUREMENTS BY AN IMAGING COLORIMETER

Color grade is one of the major components of the current cotton classing system. This paper presents the application of an imaging colorimeter to color measurements of the USDA standards for U.S. Upland cotton and cotton samples with a wide range of colors. Tests on the reproducibility and spatial uniformity of the color imeters data are reported, and the results are directly compared with results from other colorimeters and human classers. Methods for creating new color grade dia grams in the CIE L*a*b* system are explained. A new concept suggests using primary and secondary color grades for cotton samples. These two grades can be determined from the L*∼b* and L*∼a* diagrams created by the imaging color imeter. The secondary grade, using the red-green attribute, can help reduce the uncertainty of color grading when the color measurements are near a boundary between two grades in one color diagram.

Cotton Color Classification by Fuzzy Logic

This paper describes the application of fuzzy logic to cotton color grading in an attempt to improve the acceptance of machine grading for cotton colors. Cotton color grades are a number of classes in the (Rd, b) color space. Adjacent color classes have blurry and overlapping boundaries, making crisp-boundary methods ineffective for cotton color classification. Fuzzy logic is specialized to deal with uncertainty and imprecision in the decision-making process, and thus offers a new approach for grading cotton colors. In this paper, we present the procedures for constructing a fuzzy inference system (FIS) using fuzzy logic to classify major classes of cotton colors, and the preliminary results to demonstrate FIS effectiveness in reducing machine-classer disagreements in color grading. The results from the Fis show great consistency for multiple year of cotton color data.

Measuring the short fiber content of cotton

A selection of cotton samples is tested with the Suter-Webb Array, AFIS (advanced fiber information system), and HVI (high volume instrument) methods. Short fiber contents as measured by these different methods show significant differences and high variations. The calibration level is one of the major factors causing these differences. Provided all other conditions are the same, a shift of 0.01 inch in fiber length calibration can cause an approximately 0.37% absolute value change in measured short fiber content based on the average of the test data. Based on the results from AFIS tests and computer simulation, sample nonuniformity, which is a characteristic of cotton fibers, contributes a major portion of the variation of the measured short fiber content.

Color Grading of Cotton Part I: Spectral and Color Image Analysis:

Spectrometer measurements of the color of cotton based on CIE standards are investi gated and compared with high volume instrumentation (HVI) methods. Additionally, color imaging and spatial interpretation are examined with the intent of demonstrating the importance of color uniformity, trash content, and yellow spots on classer color grading. The result is that agreement between HVI color grading and the cotton classer can likely be enhanced by including the CIE color space redness parameter α*, trash content, and yellowness variability.

An investigation of the sampling bias of the beard method as used in HVI

The beard method is used for sampling cotton fibers to generate fibrograms from which length parameters can be obtained. It is the sampling method used by the Uster High Volume Instrument (HVI™). A fundamental issue of this sampling method is its bias since the mathematical computation to obtain length parameters is quite different. There have been different assumptions regarding the bias of the beard sampling method. In our experiments, we have seen discrepancies in measurements that cannot be explained as length‐biased or unbiased, especially in the short fiber region. We report a fundamental research, including experimental and theoretical analysis, and computer simulations, that reveals the bias due to this sampling method. We find that the beard sampling method as used in HVI is not completely length‐biased; fibers sampled by using this method are similar to the original fibers except in the short fiber region. Short fiber content (SFC) of the sampled fiber is lower than that of the original fiber, and this difference is inherently introduced by the sampling method.

A comparative study of the effects of cotton fiber length parameters on modeling yarn properties

Fiber length is one of the key properties of cotton and has important influences on yarn production and yarn quality. Various parameters have been developed to characterize cotton fiber length in the past decades. This study was carried out to investigate the effects of these parameters and their combinations on yarn properties. Linear regression models with different numbers of fiber length parameters and their combinations were developed for predicting ring and open-end (OE) spun yarns’ properties. The R2 and Mallows’ Cp plots of the models were compared for model selections. The results indicate that, for predicting a yarn property, a model usually involves more than three length parameters to achieve better prediction when considering the R2 and Cp values. This may be because only one single length parameter cannot sufficiently represent fiber length characteristics. The results also show that the variations in fiber length distributions play important roles in predicting yarn properties, such as strength and irregularity. The best prediction models for the properties of different yarns (ring, OE) include different combinations of length parameters. Not all yarn properties can be well predicted by linear regression models with length parameters: other fiber properties (strength, micronaire, etc.) need to be included to further improve the models.

An investigation on different parameters used for characterizing short cotton fibers

The quantity of short fibers in a cotton sample is an important cotton quality parameter. Short cotton fibers have detrimental impacts on yarn production performance and yarn quality. There are different parameters for characterizing the amount of short fibers in a cotton sample. The most widely used parameter is short fiber content (SFC). However, SFC has a significant shortcoming of very high measured variation. An investigation was carried out to compare the short fiber parameters to find a parameter that has lower variation and predicts yarn properties similarly as SFC does, including SFC defined at 0.5 inch length, SFC defined at 16 mm length, lower half mean length (LHML), floating fiber index (FFI), floating fiber percentage (FFP), and relative short fiber content (Rel. SFC). Based on our experimental data, we found that LHML had the lowest variation, was highly correlated with SFC, and predicted yarn properties similarly as SFC did. Therefore, LHML is a suitable alternative to short fiber content.