Abstract:
It is difficult for the existing machine vision systems to sort out white foreign fibers (contaminants) from cotton because their colors are the same or very close.The laser imaging method previously suggested by our research team can detect most of the white foreign fibers in cotton, but the reflect lights of laser from most of the colored foreign fibers, especially from the ones with deep colors are so weak that the colored foreign fibers cannot be distinguished from cotton.In this paper, a new method using a double-light illumination of light emitting diode (LED) and laser for imaging was presented, which could detect different colored foreign fibers, including most of the foreign fibers with white color and deep colors.In the experiment, lint cotton and 20 kinds of typical foreign fibers with deep colors, light colors and white color were collected from a textile plant, and then the foreign fibers were randomly distributed on the cotton surface and driven by a conveyor.Using the line-laser at wavelength of 658 nm, exposure time of 106 μs and aperture of 2.8 C that were previously suggested by our team, 310 frames of the sample images were obtained by a high-speed color line-scan camera under the double-light illumination formed by 2 LED arrays with adjustable luminance and a line-laser with adjustable power.Then, the quantized differences of foreign fibers and cotton background in each frame of the image were calculated by their average pixel values in R channel of RGB (red, green and blue) color space and their average saturation values in S channel of HIS (hue, saturation and intensity) color space separately.Based on the experiment, it was found that in the R channel or S channel, with the increase of the line-laser power or the LED luminance, quantized differences of the foreign fibers and cotton were enlarged, and after the differences reached the maximum value, the curve gradually decreased from the top.Thus, by the significant difference of the targets and background at the 2 curves’ tops respectively, the white foreign fibers and other colored fibers could be easily detected from cotton in the images under the double-light imaging condition.Combining the 2 curves together, an optimized imaging method was obtained, by which the images were captured by a colored line scan camera L
301KC at the exposure time of 106 μs and aperture of 2.8 C under the double-light illumination with the LED luminance of 3326lx and the line-laser at wavelength of 658 nm and power of 7.01 mW.Finally, in a comparing experiment, 840 frames of images of 20 kinds of typical foreign fibers with different colors on the cotton surface were taken by the double-light imaging method, LED light imaging method and line-laser imaging method respectively.Then, using the Sobel edge detecting algorithm, the whole successful detecting rate was 86.7% for the double-light imaging method; using the fixed threshold binary algorithm, the successful detecting rates of the white foreign fibers and the other colored fibers could reach up to 84.1% and 93.9% respectively, which were much higher than that of the images from the illumination of individual light source of LED or laser.Next research will focus on the development of the double-light imaging method from the laboratory stage to an on-line stage and the design of faster and more efficient image processing algorithms.