Harvester-Based Cotton Fiber Quality Sensor

Abstract

A multispectral sensor for in-situ cotton fiber quality measurement was developed. The sensor consists of a camera, four optical filters, a halogen light source, a fiber sampling device, and an image collection and processing system. The optical filters for the camera were selected based on the multiple linear regression analyses with the 100-nm wavebands and wavelet coefficients. The cameras image of cotton fiber was a measure of fiber reflectance at the selected wavebands. A first prototype of the sensor had been evaluated by measuring micronaire of standard lint samples. Results indicated that fiber micronaire has a very strong relationship with the sensor output at the 1550- and 1600-nm (R2 =0.98). A ruggedized prototype of the multispectral fiber quality sensor was built for installation on a cotton harvester. An automatic filter wheel was added into the sensor system. The filter wheel is controlled by software to change optical filters of the sensor so that images at selected wavebands can be acquired automatically. Seed cotton samples were collected and measured using the sensor. Images of the seed cotton samples were acquired at three near infrared wavebands and one visible band. The image at the visible band was used to exclude the pixels that represent anything other than cotton fibers before determining fiber quality with the NIR images. After being measured by the sensor, the samples were ginned and the lint quality (HVI and AFIS) was analyzed. Models were developed by comparing the measurement results from the sensor and from HVI and AFIS analysis. The results of this research showed a close relationship between the NIR reflectivity of seed cotton and the measured micronaire value, indicating promising results as research continues. This sensor can be used with GPS on a cotton harvester for cotton fiber quality mapping. It also has the potential being used for segregating cotton based on fiber quality during harvesting.