（1. 中国烟草总公司重庆市公司烟叶分公司，重庆 400023；2. 西南大学工程技术学院，重庆 400716）
(1.Chongqing Tobacco Branch of China National Tobacco Corporation, Chongqing 400023; 2.School of Engineering and Technology, Southwest University, Chongqing 400716)
Abstract: [Objective] The stage discrimination of tobacco curing is of great significance for improving the curing efficiency and quality. It is urgent to achieve automatic discrimination of tobacco curing stage, reduce human influence, and improve discrimination accuracy. [Method] Firstly, the image processing technology was used to equalize the image, then the image blocks with more background information were cut out, and the image blocks with the most effective information were select. Secondly, the color features of the flue-cured tobacco map blocks were extracted, and the HSV three-channel values of each block were extracted to obtain the i item channel values. Finally, the threshold-based screening method was used as the feature processor to classify the stage of flue-cured tobacco leaf according to different h, s and v values, and the judgment results are compared and verified with the actual situation. [Result] The image equalization processing technology combined with feature extraction algorithm based on HSV color space was used to distinguish the curing stage of tobacco leaves, and the overall accuracy reached 90.64%, and the accuracy of stage 3 and stage 4 reached 100%, the effect is very good. [Conclusion] Using image processing technology combined with the color characteristics of the image itself can effectively distinguish the curing stage of flue-cured tobacco, which has practical significance in judging the curing process, guiding the adjustment of curing parameters, improving curing quality, reducing material waste and curing costs, and has broad application prospects. This study provides a direction for subsequent research in bypassing large arithmetic algorithms such as deep learning but improving the effectiveness of practical applications.
Keywords: tobacco curing; HSV color space; feature extraction; the image processing