Image-processing solution to cotton color measurement problems in gin process control.

摘要

Cotton color, measured as blue (Z) and green (Y) reflectance, is an important quality factor, largely determining price, and must be measured in automated gins to optimize processing. However, cotton during ginning contains more foreign matter (trash) than cotton after ginning. Trash interferes with conventional color measurement accuracy. Improving on-line color measurement in gins is necessary.;A new instrument was designed and constructed for reducing trash effects in cotton color measurement. The instrument's illumination system included four quartz-tungsten-halogen lamps in aluminum elliptical reflectors. The instrument's sensor was a panchromatic video camera that acquired images through color filters on a rotating wheel. The camera was connected to a computer through a frame-grabber. Software was written to control the filter wheel, image acquisition, color/trash computations, and data recording. Image processing was employed to differentiate trash particles from cotton in the images. Color was calculated from the image portion judged by image analysis to be cotton.;The new instrument was compared to conventional cotton color/trash meters. First, 242 cotton samples were measured on both new and conventional instruments. Data were analyzed for measurement error, and the new instrument compared favorably to conventional instruments. Second, 78 cottons were examined for the relationship between cleaned cotton color and that of three stages of uncleaned cotton. The color correlations between cleaned and uncleaned cotton were higher with the new instrument than with conventional instruments. When predicting Y of cleaned lint from Y of lint after one lint cleaner, the root mean square error (RMSE) reduction was 13.2% (significant at 10%, F = 1.33). When predicting Y of cleaned lint from Y of lint after no lint cleaners, the RMSE reduction was 14.0% (significant at 10%, F = 1.35). When predicting Z of cleaned lint from Z of lint after one lint cleaner, the RMSE reduction was 23.9% (significant at 1%, F = 1.73). When predicting Z of cleaned lint from Z of lint after no lint cleaners, the RMSE reduction was 20.8% (significant at 1%, F = 1.60). The new instrument can improve accuracy in selecting the appropriate machine sequence in gin process control systems.