Location: Cotton Ginning Research
Title: Evaluation and implementation of a machine vision system to categorize extraneous matter in cotton Authors
Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Proceedings
Publication Acceptance Date: February 8, 2011
Publication Date: April 26, 2011
Citation: Siddaiah, M., Whitelock, D.P., Hughs, S.E., Grantham, S., Knowlton, J.L. 2011. Evaluation and implementation of a machine vision system to categorize extraneous matter in cotton. National Cotton Council Beltwide Cotton Conference. 1304-1312. Interpretive Summary: The price of US Cotton is based on Agricultural Marketing Service classification. This classification is determined using High Volume Instrument machine measurements for color, length, fineness, etc. and human visual inspection for leaf grade and extraneous matter. A Cotton Trash Identification System, called CTIS, developed at the USDA-ARS Southwestern Cotton Ginning Research Laboratory in Mesilla Park, New Mexico to aid the human classer in making extraneous matter calls was evaluated by comparing it to AMS classer extraneous matter calls on samples processed with the HVI. The preliminary results showed that CTIS agreed fairly well with the classer in the identification of bark/grass in cotton samples. In all samples where the classer identified objects as bark or grass, CTIS also categorized the objects as bark/grass in the samples, but usually more objects were identified. In the future, CTIS, as a machine-vision-based independent extraneous matter-classification system, may aid the classer in assigning extraneous matter calls for classification of cotton by alerting the classer to certain levels of extraneous matter, high and low, or by aiding in the development of extraneous matter standards for calibration of machine grading systems.
Technical Abstract: The Cotton Trash Identification System (CTIS) developed at the Southwestern Cotton Ginning Research Laboratory was evaluated for identification and categorization of extraneous matter (EM) in cotton. The system’s categorization of trash objects in cotton images was evaluated against Agricultural Marketing Service classer EM calls. AMS classers were tasked in assigning extraneous matter calls (bark/grass) in images acquired by various High Volume Instruments. Soft computing techniques were used to identify EM in the acquired cotton images and categorize them into bark/grass, stick, leaf, and pepper trash categories. Classer EM calls in the HVI images are compared with CTIS bark/grass and stick categorization. Images were acquired from different HVI systems on both the upper and lower camera head, and human classer’s identified the EM by marking the objects on the acquired images. CTIS categorization of EM, were later compared to the Classer assigned EM calls. The main objective of this research is to use CTIS as an effective tool to aid classers in the identification of EM in cotton and, eventually, aid in the development of EM standards for cotton classification.