Intelligent grading and motion prediction dynamics of tobacco

One of the most important yet ancient technique in the tobacco industry, is the grading of flue-cured tobacco. Factors to be considered include overall color, blemish, damage, texture, leaf length and ripeness. It is therefore imperative that leaves from different plant positions (different groups) i.e. primings, lugs, cutters, leaf, and tips must be kept separate because they have important different chemical characteristics. The efforts of this research project are directed towards restoring the objectivity in the grading of flue-cured tobacco in Zimbabwe, by means of using Machine Vision and Artificial Intelligence during the grading process. MATLAB software was used to implement elements of AI specifically, Computer Vision System and Machine Learning. An artificial grading expert system was designed by means of supervised machine learning particularly Support Vector Machines. This was achieved by training a data sample of flue-cured tobacco according to human expert knowledge of tobacco grading. A test subject is then tried and the system is evaluated and it was found that the system is 93.1% efficient which showed that automating the grading system in Zimbabwe is not only feasible but it is also more economic.