Research On Extraction Method Of Digital Elements Of Tobacco Leaf Shape

In natural state,the curly,wrinkled and folded features of tobacco leaves make the intelligent grading process more difficult.The accuracy of early digital feature extraction also has a direct impact on tobacco grading.Among them,the digital feature extraction of tobacco leaf shape is particularly crucial in distinguishing different parts of tobacco leaves.In order to make tobacco grading more accurate,higher requirements must be put forward for the extraction of tobacco leaf shape digital features.Therefore,in order to address this issue,image processing technology and semantic segmentation algorithms are used to extract the shape digital features of tobacco leaves.The specific research content and main conclusions are as follows:(1)When capturing image data of tobacco,leaves may be placed on the belt at various angles.Therefore,image processing techniques could be applied to transform the perspective of the tobacco leaves.Next,numerical features such as length,width,aspect ratio,and area of the corrected tobacco leaves are extracted.Then,the shape similarity of tobacco leaves is extracted using bitwise operations of images.Finally,a tobacco leaf tip angle detection method based on image processing is proposed to further improve the accuracy of tobacco leaf tip angle measurement.(2)Due to many areas on the surface of tobacco leaves with color features similar to the main vein,image processing technology cannot extract the main vein of tobacco leaves effectively.Subsequently,semantic segmentation algorithms in deep learning are used for main vein extraction of tobacco leaves.To address the issues of large network model parameters,intermittent segmentation,insufficient shallow feature extraction,and misclassification in certain areas,a fine segmentation network for tobacco vein based on multi-level mixed feature fusion is built.(3)To address the issue of the main vein region being partially obscured and the neural network recognizing only the trained features,resulting in a broken segmented main vein,we propose a method called convex packet scanning.This method aims to repair the segmented main vein to obtain a complete main vein and measure its area accurately.Then,an F-3MS(Floodfill Morphology Ex Medianblur Morphological Skeleton)refinement algorithm is constructed for main vein skeleton extraction,and the main vein length is measured.Finally,a geometric model for measuring the width of tobacco leaves’ main veins is constructed to measure the maximum and average width of the main veins.The study on the extraction method of tobacco leaf shape digital features confirms that the extracted digital features of tobacco leaf shape play an important role in distinguishing different parts of tobacco leaves,providing a new avenue and method for improving the accuracy of tobacco leaf grading and classification.