Moisture meters employed in a prototypical cotton gin control system use lead-sulfide sensors which measure near-infrared energy at wavelengths beyond the range of standard silicon-video sensors. These meters are expensive and must be constructed and installed separate from color and trash sensing equipment. Because silicon-video sensors have some capacity to measure energy in the near-infrared range closer to visible light wavelengths, a standard, silicon-chip, charge-coupled device, infrared-sensitive, black-and-white video camera was examined for its cotton moisture measurement ability. Three narrow band-pass optical filters centered at 925, 940, and 950 nm were placed individually in front of the camera lens, and digital images of cotton samples were recorded for each filter at six different moisture levels. Pixel-intensity averages were calculated for sample images. This data was compared to oven moisture tests of the samples. Multiple polynomial regression with oven moisture as the dependent variable and with various combinations of filtered intensities as independent variables indicated that the technique was significantly correlated (at the 0.05 probability level) with cotton moisture content of both lint and seed cotton. Regression on four-replication averages increased the correlation between predicted and actual moisture content in seed cotton. However, all correlations were weak. While the results were not entirely promising, they did demonstrate the ability of silicon sensors to detect moisture content in cotton, and the usefulness of further research to refine the technique.
Read full abstract