Research On Methods Of Tobacco Grading Based On Reflection And Transmission Images

The development of tobacco leaves grading method based on machine visiontechnology has considerable significance to ensure the quality of tobacco products andpromote the sustainable development of tobacco production. In image processingsystem, the images quality and images resolution are extensively affected by theillumination conditions, therefore, the selection of appropriate light source has a greathelp for improving the quality detection of tobacco leaves and the classificationaccuracy of tobacco leaves. In addition, the maturity, oil content of tobacco leaves aretwo importment factors for tobacco leaves classification, and their quantitativeanalysis has an importment guidance meaning to the grading of tobacco leaves.Consequently, the influence that the color of light source makes to tobacco leavesimages and the quantitative analysis of maturity, oil content of tobacco leaves wasresearched in this thesis. The main works are as follows:1. An acquisition experiment device for tobacco leaves images with LED colorlight source, which have simple structure and convenient operation, was developed.This acquisition system can be used tocollect the reflection images and transmissionimages of tobacco leaves under the condition of different colors light source, andsatisfying collecting results had been obtained.2. In HSI color space, tobacco leaves images are segmented by using the OSTUmethod, and better segmentation results had been achieved.3. Three kinds of features of tobacco leaves, including shape, color and texture,which were extracted from reflection images and transmission images obtained underseven kinds of colors light source. The tobacco leaves grade models based on single-kind and multi-kind of features were established respectively.4. Starting point based on the comprehensive features which can represent theinformation of tobacco leavf images, the influence that the color of light source makes to tobacco leaf images was studied using Wilkes Λ statistics, and a selection method ofgrading image was provided. The selection result showed the grading images based oncyan light source is the best and it was proved by the the results of FDA which gradingaccuracy rate reached to96.9%.5. Starting point based on the information of tobacco leavf images, anotherselection method of grading image was also provided. The effectiveness of this methodwas also verified by FDA grading results.6. The quantitative analysis of maturity, oil content of tobacco leaves was realizedby using grey correlation analysis method.In conclusion, the research results showed that the color of source light had agreat influence on the quality and the information of tobacco leaf images. Appropriatecolor light source was chosen to the benefit of improving the grading accuracy oftobacco leaves. The quantitative analysis of maturity, oil content of tobacco leavescould be researched simply and rapidly using grey correlation analysis.