Trash in cotton refers to non-fiber particles such as leaf, seedcoat, bark, grass, dust and other foreign matters. Trash content in cotton is a strong consideration in the present cotton grading system, because the presence of trash degrades yarn evenness, yarn strength and fabric appearance and causes problems in textile processing. The methods that have been used for assessing trash content in cotton may be divided into two basic groups: geometric and gravimetric. The geometric methods estimate the trash portion in a sample according to sizes of particles, while gravimetric methods evaluate trash content by trash weight. The classer's grade (USDA) is the most commonly used geometric method by which a classer compares trash contaminants in a cotton sample with those in the standard samples. The HVI trashmeter is a replacement for this visual assessment method using the video image technology. The typical gravimetric devices are the Shirley Analyzer (mechanical separation of foreign matter from fiber) and the MicroDust and Trash Analyzer (aero-mechanical separation). The AFIS-T (Advanced Fiber Information System, Trash module) uses the aero-mechanical technique to separate a fiber sample into fractions, and an electro-optical sensor to measure particle size in each fraction.

The HVI trashmeter is a very efficient trash measuring instrument, and the result is correlated to the classer's grade [9]. However, current image analysis techniques used in the HVI trashmeter limit its data to the count and the percent area of trash particles. It lacks an ability to provide information about detailed particle size distribution and trash classification, which is extremely useful for process optimization and prediction of cleaning behavior during processing [9]. Since the trashmeter employs a black and while video camera and a simple image thresholding technique [6,7], trash mis-identification, such as surface shadow areas, cannot be effectively avoided, thus undermining the accuracy of trash measurements.

We have been conducting a research project to develop a new image analysis system for comprehensive, accurate and fast cotton color and trash analysis (CCTA). In this paper, we focus on the explanation of a new thresholding method, multi-dimension thresholding, for trash identification, and the methods for characterizing size, shape, color and density of trash particles. We conducted a trial test to compare the results obtained from this system with those obtained from Spinlab and Motion Control HVI machines, and to analyze the influence of trash particles on cotton color.