Abstract:

Through the analysis of the advantages and disadvantages of the sample processing methods and mainstream detection methods for the detection of flue-cured tobacco leaf impurities, a set of detection schemes that fully reflect the superiority and high comprehensive performance is proposed. The 256-band hyperspectral camera was used to obtain data information, and the RGB color space was mapped by calling the RGB band, and then converted to the HSV color space for detection of green and impurity content in tobacco leaves. The HSV color gamut range of green impurity was obtained through amounts of real experimental measurements, and the number of green and impurity pixels of the tobacco leaves to be tested was accurately given, and the proportion of green and impurity pixels in the flue-cured tobacco leaves was given. The precise labeling of green and impurity pixels of the flue-cured tobacco to be tested provided a visual detection results. Combined with the RGB tobacco leaves, the algorithm of green and impurity detection had strong interpretability. Meanwhile, the execution delay of the proposed detection algorithm was about 4 s. The flue-cured tobacco leaf green impurity detection scheme not only meets the actual acquisition needs, but also has high visualization and interpretability.

Key words: hue-saturation-value (HSV), hyperspectral, green impurity detection, machine vision, automation