Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Abstract Only
Publication Acceptance Date: 1/3/2006
Publication Date: 1/3/2006
Citation: Pelletier, M.G. 2006. Real-time sensing of lint quality. National Cotton Council Beltwide Cotton Conference. pp.410.
Technical Abstract: Modem cotton gins have the purpose of extracting lint (the cotton) from trash and seeds- usually the sticks, leaves and burrs that are entrained with the cotton. These modem gins include many individual machine components that are operated sequentially to form the gin processing line. Recent on-line imaging systems have been developed by which to dynamically monitor the trash content of the lint, and based upon the current level, provide a control signal to continuously adjustable lint cleaners for adaptable cleaning. This adaptable cleaning improves both fiber quality as well as the economic yield of the harvest cotton. The current state of the art is able to acquire and process images for trash content at a rate of three frames per second. This equates to roughly 6.3 square meters of cotton imaged per bale. While this is a large improvement over the previous generation systems, it still leaves a great deal of cotton un-imaged. Coupling this with the inherent high-variability within the lint, leads room for improvement if the computer vision system could increase its’ processing rate. In seeking a performance increase, the traditional approach is to add another PC and split the image-processing across multiple computers. However, a recent literature review has revealed new research that suggests that within the last two years, new programmable graphic processors (GPU) have been introduced. Of particular interest is the research that suggests that these GPU’s can be utilized for general processing at an increase in performance over a traditional program that is run on the PC’s CPU, provided that an algorithm can be developed to run in a vectorized format so-as to take advantage of the GPU’s inherently parallel architecture. In an effort to achieve higher image processing rates for cotton trash imaging; a study was conducted to determine if similar speed increases for cotton lint imaging. At the completion of the study, it was determined that a speed-up factor of 11 was obtained over the prior image-processing rate. This new processing speed improvement will allow the system to image over 76 square-meters of lint per bale over the previous 6.3 square-meters per bale. This is a significant improvement, as the system will now be able to continuously image and process all the lint that passes under the imaging system, thereby providing a significant improvement over the estimate of the trash content in the cotton bale.