基于数字图像处理技术的叶片结构模型构建OACSTPCD

In order to study the use of image method instead of traditional four-layer vibrating screen method to establish a leaf structure feature area model,the image of each unfolded tobacco leaf was scanned offline,the B-component index threshold and the corrosion expansion closure operation were carried out,and the complete tobacco leaf image was extracted by eliminating the background.The pixel area of each tobacco leaf was calculated and sorted by 8-neighborhood extraction.One point at every interval traversed all points in the pixel area interval,and compared with the actual large,medium,small and fragment rates.The pixel area value with the smallest prediction error under different area thresholds was selected as the area feature threshold model composed of large,medium,small and fragment rates,and external verification and repeatability detection were carried out on the model.The results show that the blade structure is composed of[+∞,50880],(50880,17230],(17230,2380],(2380,0]area characteristic interval model.Through external verification and comparison with the standard four-layer vibrating screen,The errors of large,medium,small and fragment rates were 1.09%,0.64%,0.47%and 1.07%,respectively.The standard deviations of large,medium,small and fragment rates were 0.88%,0.84%,0.19%and 0.26%,respectively.The blade structure model established by image method has good stability and can predict the large,medium,small and fragment rates of tobacco leaves well.